0:05
Good morning, or good afternoon, or good evening, wherever you are, or wherever, whenever you're listening to this.
0:14
This is day three of the 2022 online split workshop, and we're delighted you're here.
0:23
Um, first two days, we've gone through the first date coin, gone through trajectories and let me start up my screen share here.
0:34
Um, first aid we went through running trajectories with the high split model. And yesterday, we started doing the dispersion calculations.
0:42
And today, we'll continue on with some more advanced applications of those dispersion calculations.
0:53
And I'm giving you a brief intro today.
0:57
Let's make sure my screen is catching up, OK.
1:02
As we've mentioned, we have a workshop webpage where we're posting information that we, we'd like to give you, and, uh, we update that each each day, with links to the videos, and, and downloads and things like that.
1:20
In fact, here's a picture of at least a recent version earlier this morning.
1:26
And you'll notice that, no, for each day, we have links to the videos and links to the different handouts.
1:36
And, actually, for today, you'll notice that there is some downloads that you you could do, or not necessarily required, but could be helpful if you do.
1:48
Um, we found, and I'm Erin, one of the programs, and we've created a hot fix for you, and so, for different operating systems, Windows, Mac, and different flavors of Linux.
2:05
You can download a zip file, and what you would do, let's see, is you could download the zip file, here's what it would look like on Windows unzipped, uh, three files, there's three executables, and then just copy them over into the high split exec directory.
2:28
And when you do that, you can overwrite the existing files there.
2:33
Um, we made it a change, sort of at the last minute that we thought this would be helpful for people and that change actually was helpful but it created an inadvertent problem that we didn't didn't expect.
2:48
Um.
2:49
We actually will be using, I think, at least the stat main, uh, program today, and so, um, It might be helpful, actually, if you did, did this.
3:04
If you have time to do this, you're just going to download those zip file for your operating system. I think it will work on Windows 11 as well.
3:14
Um, and unzipped them and then copy those three files into the high split exec, directory overwriting files that are there.
3:29
See here, So today we'll start out with Pollutant Transformation and Deposition, and then we'll go into a section on air concentration uncertainty.
3:42
Then a section on source attribution methods, um, and then we'll start with some sort of specific applications, wildfires, smoke and dust storms which we'll continue on tomorrow with radioactive pollutants and volcanic emissions.
4:04
Here is the more detailed schedule for today.
4:10
Um, as you know, we've been asking people, too, I'm asked the detailed questions about the model in the high split form.
4:21
If you can, it's much easier for us to answer them there, and in fact, people have been doing that. I noticed just wrote this screenshot, a couple.
4:33
There's a section in that forum for each of the of the workshop sections.
4:39
And here's one example of some questions that were asked in one of the sections.
4:45
And what's nice about this is that we can give you a more detailed answer. You can upload things like screenshots.
4:53
But maybe more importantly, other people can see that question and see the answer.
4:58
And you'll see that a lot of people are looking at these questions and answers, the number of views, So, it can be really helpful, excuse me.
5:08
Um, can be really helpful for, for others as well.
5:14
If you want to post a question, you do have to register.
5:17
It just takes up a minute or two.
5:20
But if, even if you don't register, Excuse me, cold here.
5:26
Um, even if you don't register, you can still log it, still look at the questions and answers in the forum.
5:36
Drink water.
5:44
OK, so here's a quick recap of the logistics.
5:51
We are posting the handouts, we have the, I think something's about to post these slides and I'll try to add them as a handout soon. The goto Webinar interface.
6:00
And as we've mentioned, excuse me, if you do get little behind, which is easy to do, it's OK to view the, view, this, the session as a demonstration, then go back and do this either on your own, if it's interesting to you.
6:20
Excuse me, again.
6:27
I've had a cold these three days, and I've been lucky enough not to be Copping, but today may not be quite as as lucky, OK?
6:35
So, now, let's begin the section on Fluting transformation and deposition, and give me a second.
6:45
I'm going to share a different screen now.
7:00
All right.
7:03
So, um, as I said today, we're going to be looking at Pluton transformation into position first, somebody to go to the tutorial and open up that, um, that page.
7:18
And the first thing I'm going to do is, the net, my working directory.
7:25
It's something that you that you can do from time to time to kind of give yourself a fresh start.
7:30
That's under the advanced option, and I'm going to click cleanup working.
7:36
And when it does, you do that.
7:38
It takes your existing working directory and copies it to a different name, Kind of fixing a number at random.
7:45
I'm going to go ahead and do that.
7:47
You can see that in my high slipped directory, now. There's this new working directory. called Working 900.
7:53
I've done this once before, and then if you click on the work directory, now.
7:59
There's hardly anything in it, Um, and, uh, So, and let's see, what happened to my High split Windows here, you second here Seems like.
8:21
Give me one second mmm hmm, You start this up again.
8:33
Without Screen.
8:41
I see, OK, could be when you do that actually a closest heisler down that maybe what would happen, OK?
8:51
And so now what we'll do is we're going to start up here, and you can see when I opened the haislip back up.
8:58
Um, the working directory not populated with, sort of the basic default items that it needs in the in the working directory.
9:09
All right, so we're going to start with concentration set up run.
9:15
And let me actually open up a little cheat sheet here in Toronto.
9:21
Organize my screen a little bit, little bit difficult to do, as I know, he's probably dealing with the same issue of screen space.
9:37
Um, so the first thing we're going to do is load the cap tax control file and the cut checks setup file.
9:50
And so, I'm going to retrieve that from the workshop tutorial, Files, Directory, detergent, Isles, ... control.
10:06
And you might remember yesterday I got into trouble I wished I tried to up to retrieve a setup file, into look into the control menu and create a bunch of problems.
10:16
So don't don't do that. So we're gonna upload the kept X Control that text.
10:23
OK, and I'm going to save and I'm also going to go to the Advanced Menu, Configuration Setup.
10:31
Concentration, Retrieve.
10:33
And I'm going to retrieve the cap tech setup, um, out of that tutorial files directory kept at setup.
10:42
OK, OK, I'm going to save.
10:46
And now let's go back to the concentration setup. Ron, we're going to make.
10:51
Uh, a few, a few small changes.
10:55
At the beginning, we're just going to change the runtime to 25 hours.
11:00
Um, but this particular section, we don't need to do the whole.
11:05
The whole run, and it will make the simulations go faster.
11:11
And by the way, when you retrieve those files from the tutorial Files Directory, the one thing that might be different on your computer, it will be the location of the meteorological files.
11:23
Um, depending on where you put your tutorial and where you put high split, that's one thing you might need to adjust.
11:30
So, and if you do need to adjust it, what you do is, you would clear the meteorology, and then you add the mythological files and then you would navigate to where your tutorial isn't where your cap Text Directory is. And you would add that file?
11:47
In my case, it actually worked, But here, you can even see it more explicit, um, description of that path.
11:58
So, that would be the one issue that could come up if you, if you load those control files is set, control files from the tutorial Files Directory.
12:10
And, by the way, you know, if I split can't find that data, then it pretty much crashes right away, you know, if it can't find the method, it can't really do Hardly anything.
12:21
That would be a fairly quick, and hopefully it's a, uh, descriptive error message that you can't find the, can't buy them a tape.
12:32
OK, so we'll save that, and the other change we're going to make is we're going to go to the advanced configuration setup concentration menu, and we're going to change the number of particles that will be, or computational points, that we simulated for 50,000 down to 20,000.
12:51
This will make the runs go faster and won't really sacrifice too much in the way of, of accuracy, and go save out.
13:03
And, um, so the first thing we're going to do, um, we're going to create a baseline calculation.
13:10
Um, without any, um, chemical reactions are or deposition that we'll be able to then compare against the calculations we do with fee, um, with those features enabled.
13:28
So we're going to go, concentration, Brian, model, And we are going to use the setup file, so I can click that, Yeah, these runs will take some time.
13:39
We are using 20,000 computational points, so it's doing 20,000. So, you know, trajectories.
13:46
It is adding that's just the AAA dispersion component to each computational point.
13:53
So it's creating this cloud of pollutants as it's moving downwind.
13:59
Yeah?
14:02
Let's give it a second here.
14:17
Getting getting closer.
14:21
OK?
14:31
OK, it seems to have finished on on my machine, hopefully it will be finishing on your machines as well, either already or soon.
14:39
So, will exit out of this menu and well, now, create this space.
14:47
Sort of maps, that will then be able to use to compare.
14:51
So concentration, display, concentration, contours, and we're going to make just a few changes here, We're going to make the concentration multiplier, 10 to the 12th, E plus 12, and also the deposition multiplier E plus 12.
15:13
We don't need that, right now.
15:17
You don't have to do that right now. We can do that.
15:21
Let's do that later when we need to do that.
15:24
And we're going to label that.
15:27
I put this PICO grams because we're we've remitting grams per annum multiplied by 10 to the 12th so that will get answers and PICO grams.
15:36
And, the other thing we're going to do is, um, increase the zoom to 80%, sort of to focus it on the area that we want, focusing on.
15:49
Let's go ahead and execute the display.
15:53
And you can see if we go, um.
16:00
To the end of the, I think we have now put every three hours, And if you go to the and the eighth section, you'll see a concentration pattern that looks something like something like this.
16:17
So, now that's, that's going to be our base calculation and, um, actually what I wanted to do is call this, but base.
16:31
And you can either rename that plot dot HTML in your working directory or you can just do it.
16:39
Again, just do it again, it might be easier.
16:43
So we call it pod base.
16:45
And I think we see the neck the same.
16:48
Same answer, if you want to just to not do it again and rename, um, in the working directory what you would take is you do ...
16:56
plot, I'd HTML and rename, It is just plot base and I kept rename it the same thing.
17:05
You would rename it, you know? Like based on HTML if you want?
17:09
Um, OK, the other thing we're going to do is we're going to be looking at these message files and Comparing them and so, we're gonna take that message file, and we are going to rename that message Base, that text.
17:26
And, um, we could look at that really quickly, I'm opening it up in, in Notepad and, um, what you can see is that.
17:41
We've emitted about 201, 210,000 grams over the three hours.
17:51
That's the Tracer release and pretty much we have throughout the simulation that same 200 and 10000 grams of of gluten eat.
18:07
Um, and you can see that.
18:13
A distribution among the different layers in the, in the model, you know, you know, 10 or 15% is in the lower 6 or 7 layers or something like that, but also additional material up, ayer as as well.
18:28
Close this down.
18:30
OK, so, the first thing we're going to do, Dan, is, is we're going to try a chemical reaction and, um, these chemical reactions that we, that we can put into high split are relatively simple.
18:49
And, um, what we, what you can do fairly easily in high split is to add in linear chemical reaction, where a chemical is converted to another chemical, in a linear fashion.
19:05
And many reactions in the atmosphere are linear.
19:10
Um, simple reaction of a certain chemical, with the hydroxyl radical going to another chemical, this is a linear reaction, it basically just depends on the concentration of the pollutant, and then the rate is some constant times you know some other thing.
19:30
The concentration of another pollutant in the atmosphere if you know that concentration of the other pollutant then it's A Um, it's effectively a linear linear reaction, and if you, um, let's see, I think I have a, uh, quick slide here.
19:49
Might have the slide here, let's see, give me one second.
19:57
one second, me.
20:03
Mmm hmm.
20:07
There we go.
20:09
And I'll bring it over to the right.
20:16
To the right screen, I made a few slides about this. Let's see if I could find the site I wanted to show you, OK?
20:27
If you think of a chemical reaction, simple reaction of going from one chemical, C C chemical, one to chemical two, and the rate being some constant times, the concentration of chemical one, um, you can write that as a a first order differential equation where D, C D T equals minus K, C one concentrates going down because it's reacting away.
20:55
And the solution two, this differential equation looks something like this is exponential with that rate in the end time in the in the exponential term.
21:10
And if you're looking at what the concentration of this reactant, or excuse me, the product is, it's going to be equal to the initial amount of this minus, the amount that will cut reacted the way. That's, that's the amount that got changed into C two.
21:29
And so, C two, it can be shown to be, look, looking like something like this as a function of time. And it's also the masses are the same.
21:40
The same expression, um, you know, for very small values of time, this exponential can be.
21:52
Especially in a simpler way.
21:54
Um, and then 1, 1 thing that you do have to keep in mind when you do these kinds of reactions in, in high split is that, we're not really keeping track of, know, all of the chemistry, but we're just saying that one chemical is being converted to another chemical at a certain rate.
22:13
So, imagine if you're going from something like sulfur dioxide into sulfate which is a common reaction in the atmosphere, um you have to take into account the the difference in molecular weights.
22:26
So, what's happening in that reaction is the S O two is being oxidized and two oxygen molecules are being added to that two oxygen atoms are being added to the to the molecule.
22:42
And so for every mole or, you know, every gram of SI O two that is converted, you actually end up with a heavier version of that of that molecule and the ratio of molecular weight, just 1.5.
22:57
So when you do this calculation in high split and we're only looking at the mass of pollutants, you have to multiply by 1.5 and, you know, depending on what reaction you're doing and what chemicals you're doing.
23:11
You have to take that into account, and we let you do that in, in the model.
23:17
All right, let me stop there.
23:19
That's what this section here is talking about.
23:25
And it does give the example of the sun, too, as well.
23:30
OK, so to demonstrate this, we're gonna then open up the.
23:37
So you need pull this over sorry, I'm Abby. I wish I had the bigger screen.
23:45
Oh, all right, we're going to quit out of this display menu, and we're gonna go back to the set up run, and we're going to click on this please: step, position, and grids.
24:01
Ah, and you want to go back to this data intriguing here.
24:06
Here we go. So what we're going to do now is, we're going to define two pollutants.
24:11
So we, because we need to define this, this pollutant that's going to be a product of this reaction.
24:17
So, we go up to the top here, and turn that into the two, and I'm going to save me, actually, open it up again.
24:27
Um, and, opening up Pluton number two, I'm going to click on that radio button, and we're going to call this gluten too, and we're not going to emit any of that gluten.
24:45
It's just going to be something that the PMC agents didn't react into.
24:50
And, um, zeros, as we've mentioned, just mean, if we hadn't been at anything, you would be starting it at the beginning of the simulation, you gotta put something, They are.
25:01
Even if you're not eating anything, let me save that.
25:05
Let me save that.
25:07
And, um, the next thing we have to do, by the way, this tracer that we use is PMC H Tracer actually did not react and actually did not really deposit.
25:21
That's why it was a good tracer experiment for this dispersion field field experiment.
25:27
So this is kind of a hypothetical exercise just to show the the functionality of this in the high split model.
25:37
The actual chemical would not have reacted in this way, If it had, it wouldn't have been a great tracer to use in our experiment, because then the concentration will be changing due to reactions and maybe deposition, things like that, Um, which we would have to factor in, An ad would add more uncertainty into the into dispersion evaluation.
26:05
OK, so we're gonna save this, and the other thing we're gonna do, is we're gonna go into the Advanced menu, configuration setup.
26:13
Concentration.
26:14
And we're going to pick a new menu that we haven't picked before, this inline chemical conversion modules, which is menu number 10.
26:23
And we're going to select the radio button that says, convert species one to species two at 10% per hour, and you could change right here, if you wanted to give us a different amount.
26:42
Let us go ahead and and do that 10%, and can say, You'll note that.
26:48
um, what we're actually doing here is going to basically change the dimension of the mass description on a particle from 1 to 2.
27:00
So instead of having just one species on a computational point, we're now going to add a another species onto that same computational point, and, um, the model will actually change this to two.
27:20
Automatically, you don't have to do that here.
27:26
And when you do something like that, you've got to be a little bit careful because, and actually, you know, this is a good example, the SO two to SO four.
27:38
Example is interesting, one, because S O two will be largely in the gas phase, but sulfate would be largely in the particle phase, and those two, um, chemicals, because they're in different phases in the atmosphere.
27:54
We actually behave somewhat differently, and, but by treating them on the same particle, you are losing a little bit of accuracy, because, um, the fate of those two species, uh, would not necessarily be the same in the atmosphere.
28:14
So you can put them on different verticals too.
28:18
We'll go into that, but that is something that you can do with.
28:24
With some chemical reactions, if you want to do it that way.
28:28
So let's go ahead and save.
28:31
Save this, and Save.
28:34
And let's run the model.
28:41
And we are going to use that.
28:45
Set up Advanced Options, that we used.
28:49
I should have started there, model running before I went into that explanation.
28:55
Um, no, there are.
29:01
First of all, you know, if you're doing a lot of, a lot of chemistry in your simulations, you probably aren't going to want to use the high split model.
29:08
And there are other, you know, really excellent models that can use things like ..., geo scam and other models that do all sorts of non-linear chemistry.
29:21
And if you're doing photo chemistry, trying to estimate the ozone concentrations or gas to particle conversion of different different species to, to get particulate matter, um, know, Heisler can do some relatively simple chemical calculations.
29:42
We're showing you them them here.
29:45
We actually some people have taken high split and added some early complex, and even non-linear chemistry Ariel's Stein who gave us the welcome on the first day, the director of the High split, director of the air resources Laboratory.
30:00
For his PHD thesis, he added a, a complete photo chemistry mechanism. You know, enter.
30:08
I split and turned it into being able to do sort of non-linear chemistry myself.
30:14
I've added a bunch of chemistry infer mercury into the model that other people have done other things as well. So, you can do it.
30:22
Um, But for most of the off the shelf type of applications, um, you can do simply relatively simple things.
30:31
And if you want to do really complicated things, you might want to use another another model.
30:36
Alright, so let's exit out of here and let us go ahead and, um, display the results.
30:43
And we're going to do two things here because we have to see concentration display, concentration contours.
30:53
Now that we have to two pollutants, we're gonna create a map of results for both, and you can see now, the display now is populated with These are two different choices that we could make.
31:11
So, um, The first thing we'll do is we'll create results for pluton one, which is this PMC CH, we could call it lot PMC agents, call it plot one.
31:26
Um, we don't need to see the results.
31:29
Each time, right now, we're just going to create these maps. So you could leave that on if you want.
31:34
But I'm just gonna, it won't automatically show up on my screen after it's done.
31:40
Um, we're gonna kick, make sure that we're picking this PMC H one for this first set of maps, we are going to No added this concentrated multiplier which we had already done.
31:53
I think in the last last step of PICO ring arms to make PICO grams and also do this 80% and let's go ahead and execute the display.
32:05
And it actually did it and it didn't show up right away because.
32:13
We told it not to not to put you on it but we did create this plot one type one dot HTML then we're going to do it again um and create a plot for pollutant to set new pollutant that we made so we're going to change the output file name to Pollutant to and we can go ahead and put the view on this time so we can see it.
32:39
But now we're going to collect select Fluting too.
32:44
And um.
32:47
Everything else looks right and we'll execute display and then of course now we get.
32:53
Um.
32:56
A result for Putin to make infinite goes down and we'll look at you know how that looks compared to some of the other um, results.
33:09
Actually, if you look at the plot base versus gluten two, there are differences, and actually I can go ahead and open up one, You go to the end.
33:24
This is, uh, all right.
33:30
I won't go to the end.
33:34
I'm looking to you see there are differences between pollutant one, Pollutant two, Luton base, but you'll notice that the area of the plume is this is the same.
33:47
And that's because we kept things on the same particles.
33:50
So there's no way or the blues to be different in this case.
33:58
Different shape overall shape, but the concentrations have changed as you as you might extend.
34:07
OK, now let's look at the results, little more.
34:11
More detail, um.
34:19
Let's see here, let's go back to the data entry.
34:23
Um, If you look at the, the maximum concentrations.
34:36
And we can, what we can do is look at the maximum concentrations in that last.
34:44
Or in each, in each case, if, um, what we've done here, as I'll show, you, give me a second here previous.
34:54
Let's go to this, sorry, um, for each of these cases the plot base flipped pollutant one and pollutant To.
35:04
You can, you can take out the maximum concentration, um, in each of the three hour averaging period, um, simulation results, so for, for the plot base calculation, in the first three hours, the maximum concentration was 48 thousand.
35:27
And you can see here 48,000 for the pollutant for the PMC AH pollutant one. If I go back to that first.
35:38
First time step.
35:39
The maximum concentration, 39 thousand and for the Putin too.
35:48
Go back to the beginning, the maximum concentration was 8300, and we've done that.
35:55
You can do this, but we've done that for each of the of the simulations.
36:01
And you can see what's happening here, is that the maximum concentration of one RPM CH is going down now will go down in general because of the dispersion, But it's also going down even a little bit faster, because it's reacting away.
36:22
And Pluton, too, is, um, you know, sort of increasing, but it's also going down because of the dispersion, but it's not going down as fast because it keeps being replenished.
36:35
And I think I made a little, You click on this results here, made a little map, a little graph of this, um, the base calculation, The maximum concentration definitely just goes down, this, it is yellow, um, symbols and it gets going down.
36:54
Because as the gluten is moving further and further away from the, from the source, it's disbursing out and the cottage maximum concentrations are getting lower and lower.
37:05
When we look at the the these other concentrations at the PMC Asia, polluting the water pollutant to what you see that the PM CH is actually now going down faster.
37:17
In fact, by the end of the simulation is hardly any left at all.
37:22
And the pollutant number two, these orange bars.
37:28
It keeps getting replenished, so even though the simulation is you know trying to disperse it and make it more dilute, it does keep getting replenished. Now, if you went out further, this would get to be lower and lower and lower.
37:40
And this is a linear um scale you know a lot of times an in-house But we have these logarithmic scales, these, you know you sort of see things that are 2 or 3 orders of magnitude lower.
37:53
But if you actually plotted them in a linear scale you wouldn't even see them at all.
38:00
Um.
38:02
Let's see.
38:04
You can, um, do a more more complicated versions of these things, and in fact, there's a, um, a file that if it's in, if you've, if you've enabled that chemical conversion, it's called Kim, right?
38:23
Dot text and figure here.
38:29
It's just an ASCII file that has four entries in it.
38:34
Um, it says that Pollutant one is being transformed into gluten too, at this rate per hour, and the ratio of molecular weights is one that so, this is what we just did, know, if we wanted to do, um, A Oh, sorry, that's, uh, This is an image, sorry. I'd have to open up a notepad to to edit it here. Let me do that.
39:07
It's live here.
39:12
Let's see if I can find kim rate dot text in our working directory. There it is, right there.
39:19
It was created when we when we invoke dysfunctionality the first time.
39:27
You also can create it yourself. If I open this up in Notepad.
39:32
Um, now I could edit if I wanted to, you know, it defined more pollutants and more complicated scheme.
39:38
You know, one goes to two, and one goes to three, or one goes to 2, 2 goes to three things like that.
39:46
I could do so. I could change that. I could change the rates. I could change the, you know, this ratio of molecular weights.
39:53
So you could add it's many of these sort of linear reactions and pollutants as you as you want to.
40:02
Um, it's fast.
40:05
Otherwise, um, don't say, OK, OK, And as I said, no.
40:17
If you're trying to do a very complicated chemical simulation like photo chemistry upfront to estimate ozone concentrations, things like that, you're probably not going to use the high split.
40:30
Let's go to the next, um, section and, um, need to go a little bit faster here. This is going to be of dry deposition of, of gasses.
40:43
And let me open up cheat sheet here. Let me get out.
40:49
Yes.
40:51
Um.
40:54
And we're going to.
41:03
Load the kept X control.
41:06
Um, and kept tech setup again.
41:10
We don't have to, but it is a little bit easier in a way because we may see pages and, um.
41:20
The main change we made actually was to invoke that chemical reaction.
41:25
Um.
41:26
So to save time, I think I will just go ahead and un clic that.
41:30
So, configuration setup, concentration.
41:34
And I'm going to go back to the default, that was the one change we made.
41:40
two, do that last simulation.
41:45
Save.
41:47
And so, in order to.
41:51
We've already done it.
41:52
And if you aren't, if you did retrieve these, again, you would want to change to 25 hours, Concentration setup run 25 hours, and you'd also want to change, too.
42:03
The reason is that of concentration to the police numbers are 20,000 articles, OK?
42:15
And it's telling you you could do the base calculation again if you already, if you hadn't already hadn't already done it, all right.
42:24
So in order to do this deposition, we have to make a few changes, let me save this.
42:31
You gotta go to the concentration, um, set up run pluton deposition and grids, set up and um.
42:43
Let me go down here.
42:45
Keep yourself some clues as to what were do we see little similar. What we're going to do is we're going to go to the grid.
42:57
Great number one and what we have to do is we have to add another vertical level.
43:05
Um and Chris I think talked about this yesterday Chris Lochner.
43:12
If you want to do deposition, and if you want to get the deposition output in your concentration grid, you have to add this surface layer into the, into your grid. If you didn't do this and you, but you still invoked the deposition, the model would do the deposition.
43:32
but you wouldn't get any output of it.
43:34
So, again, these concentration grids, or just kind of a bookkeeping, um, sort of a bookkeeping function, You know, if you define them, and if you define them in certain areas with certain grids, sizes, et cetera, et cetera, certain levels, then you'll get the model will give you the results for that for that concentration grid.
43:56
If you don't, the model will still go on its merry way.
44:00
It's just give you the results out. So, this is a common mistake, if you forget. You know, you would you put in this deposition, and you forget to put it as zero layer and you wonder why you're not getting out the the answer. So we save.
44:16
Save this.
44:18
And Save.
44:22
And, um, we're going to save this as deposit.
44:26
Um.
44:28
Control, dot tax is we're going to use this now for each of their.
44:37
Next section's Let me save as deposit.
44:44
Control text save.
44:48
Say, but let me save the advanced the ration set up concentration save as Ooops deposit.
45:02
Setup.
45:05
Next, say OK, so now we are going to make a few more changes to get the dry deposition velocity, and try to position, um, phenomenon into the model, but go to setup, run.
45:21
And oh sorry.
45:23
Aye.
45:26
And so on.
45:28
Concentration setup, I'm not my model.
45:31
We're going to open up the pollutant deposition in great setup. And now we're going to open up a new menu that we haven't done before. We looked at the pollutants.
45:39
Um, oh actually you know what.
45:42
You do want to make a change here make that a pollutant once I should've.
45:48
Probably, um.
45:53
I should have probably, we retrieve that control file, let's save this.
45:58
And let's save again as deposit control dot text, save, concentration set up run.
46:09
So we go to, uhh, deposition.
46:14
and what we do here is, this is a fairly complicated menu and there's a lot of things, it will, we'll do, you can put in here.
46:25
We're gonna go through some of these things today, but not all of them, um, what we're going to do first is we're just going to specify on this second line here.
46:37
It deposition.
46:38
velocity, um, and, um, by doing that, we're telling high split that it's a, it's a gas.
46:50
And we have this particle diameter being zero, so it's, it's, it's, it's a gas and that the gas has a deposition velocity, zero point zero one meters per second.
47:02
Which is a pretty fast, which is about one centimeter per second, pretty fast deposition velocity.
47:08
And we're not going to add, do any of the other, um, any of the other parameters at this time.
47:17
And note that you can kind of click on some of these buttons, and some of them work in different ways by sort of pre populating some of these, uh, fields with sort of chief default values.
47:34
Depending on where you are in the GUI and what you've done, it might or might not give you exactly what what you want.
47:40
You're better off putting in the numbers that you want for the simulation that you're doing.
47:46
So let's go ahead and, and save this, and save.
47:52
And we're going to go ahead and run.
47:56
I'll save this, Analytical ahead and run the model.
48:08
I didn't talk about the deposition velocity.
48:10
The deposition velocity is essentially, um, the deposition flux, um.
48:24
Equals the concentration of pollutant times deposition, um, velocity.
48:31
and you can write that in the same kind of for the first order Depletion, um formalism or you know that same way as we did with the chemical reactions And you can show that they, the time constant equivalent to sort of the inverse of the reaction rate um, or the reaction rate is equal to the deposition velocity divided by the height of the surface layer.
49:01
And the dry deposition happens in high split when the pollutant is in that first layer above the ground.
49:11
We call it the second meta logical data lake level, because the first logical data level is the surface, it has all the surface variables, and that second level is actually the first level above above the ground.
49:25
And, um, the reason why we create these time characteristic times for each of these different processes for the reactions and for the different deposition, is because we ended up in high split suppose you had all these things going on. At the same time.
49:46
Suppose you had chemical reactions occurring, and dry deposition occurring, and what depositional occurring.
49:54
Well, it would matter, then, which one you did first in the calculation, if you, if you did one and then depleted the mass of, of your chemical.
50:07
And then did the next one, there would be less mass to do that next phenomenon.
50:14
And so, it would then matter which order you did them in.
50:17
And so, that's called, operator splitting are that's one piece, some people call the operator splitting in chemical models, and first part models are models in general. And you don't generally want to do that, you know, you don't want to have your unit for dry deposition. Results depend on whether or not you did the chemical reaction first or the director position first.
50:36
So what we do is we create these time constants for each phenomenon that's happening and we sort of do them all at once.
50:45
And so all of them kind of factor into the into the depletion of the chemical, and if more than one of them is happening at the same time, they all kinda get their fair share of of what might what might be being depleted.
51:01
All right.
51:01
So the, um, chemical.
51:07
This has completed.
51:11
So let's let's exit and let's make a plot here.
51:18
Display, concentration, contours, and we are going to.
51:26
Call this one plot, Dry.
51:35
Um, we're going to, now, now that we have these two different levels, now you can see this is populated with, now these two different levels.
51:47
We're going to show both of the levels.
51:52
And we're now going to add that deposition multiplier to make that also in PICO grams.
51:58
The other thing we're gonna do is we're going to just, we're gonna say total.
52:01
So instead of bonding, the, the incremental deposition each time, we're going to show the total deposition, um, up to that point on each in each of the.
52:20
In the output maps, actually we're just gonna show the total deposition at the end in the output map.
52:28
Um, and again, we'll have 80, 80%, I think that's all we want to, Did you hear?
52:37
Let's go ahead and execute the display and what you'll see in this display is the first.
52:49
eight frames are, this is the concentration, the concentration between 0 and 100 meters for these three hour chunks.
52:59
Then the last one, now, is that total deposition.
53:06
Um, and here so you see deposition. PICO grams per meter squared at ground level.
53:11
Um, You're seeing these these values um.
53:19
See if this is more or less what?
53:26
What we expected, yes, I think, I think it is the maximum layer 3.9.
53:32
Tinnitus six PICO grams per meter.
53:35
Squared OK.
53:37
Um, let's see, what else I'd like to point out here um, and if you went through um, and took the maximum concentration in each of the.
54:00
each of the three hour sections, like we had done before, you would get these.
54:08
These answers.
54:08
And just so, you can see, that the maximum concentration really has decreased as as one might, as one might expect.
54:18
And if we look at the message file, um, open it up here.
54:24
I could have open it up from there, um.
54:30
Are working directory, or are you? I'll do it actually from.
54:36
Do it actually from the, me quit.
54:41
Advanced view messages.
54:44
Because one way to look at it, um, what you'll see is that, by the end of the calculation, instead of having 281, I think it was 201, 210 thousand grams, We have 174,000 grams, and there's actually less material.
55:10
I think in the, in these lower layers because it has been depleted from that on the lower layers especially computed from the the first layer of the ground.
55:23
Um, now dry deposition is fairly complicated.
55:31
There's another way that you can do deposition in, in high split write a position.
55:40
And by the way, dry deposition is, I think that everybody is sort of familiar with web deposition when it rains it in.
55:46
But will, will, you're looking at that universe very shortly.
55:49
Dry deposition is when the pollutant comes in contact with the ground, or a tree, or a building, and basically gets absorbed, either absorbed or absorbed, into the, onto the surface. And, let me show you a quick.
56:09
picture about that.
56:13
I wanted to just talk about this for a second.
56:16
Ah.
56:19
See.
56:26
Oops, see, I didn't.
56:30
1, one more second, here we go.
56:36
The one ARR that over, um, when we do, you can do, step is dry deposition by specifying the deposition.
56:54
Velocity, like we did, we just explicitly gave it as in some for some chemicals.
56:59
These are sort of the tabulated in it.
57:01
You can get, kind of approximate values of deposition, lusty if but, but, of course, it will depend on the mythological conditions, you know, how fast the wind is going with the surfaces like, and, and there isn't much more sophisticated way that you can do dry deposition, and it's called the resistance method.
57:21
And what this method does is it, parameter arises, the dry deposition phenomena into a series of resistances.
57:29
And first one, it says the pollutant has to kind of make it down to near the ground, and that's the so-called aerodynamic resistance, and this is a function of the stability and the mixing and the kind of phenomenon that would make it either easier or harder for the pollutant to come down to the ground.
57:49
And then there's this other, um, resistance is called this quasi laminar sub Layer resistance and that is the layer right near the ground if you can think of if you can, imagine the wind blowing over the ground. We know that right?
58:02
at the surface of the ground, the wind velocity, it has to go to zero, and then as you go above the ground, the wind velocity gets, you know, more and more and more, but right above the surface, maybe in the first maybe mm or or two, There is this called so-called quasi Laminar Sub Layer.
58:20
It's the way it is just starting to get going.
58:23
It's pretty slow, and it's actually, in many cases, looks like laminer flow.
58:29
And it turns out that as opposed to turbulent, and it turns out that that resistance, that layer is kind of difficult for pollutants to go through because there's not any turbulence in that layer.
58:46
And so, it turns out that actually this quasi laminer flow resistance can have to be very, very important.
58:52
It depends on the surface roughness, um, and things like that.
58:58
And then there's a so-called surface resistance and this says that even if you've made it down to the ground and through this quasi laminar sub layer, then you have to when you, when you reach the surface, you have to stick or be absorbed.
59:12
And this will be dependent on all sorts of things. Like what the nature of the surface is, the nature of the chemical.
59:19
Whether it's a particle, you know, or a gas.
59:22
So there are these very complicated schemes that you can define that, uh, we'll give you these answers and that actually, let me, let me stop showing you this.
59:36
That actually is what you see in the concentration setup Ron, pollutant deposition in grid setup deposition. That's some of these other parameters that we're not going to really set.
59:49
But these you know, like it or weight and diffusivity, ratio and Henry's Law.
59:57
We'll let you then calculate the dry deposition of a gas for example with this more sophisticated method.
1:00:04
Um, when you put that value in there, it overrides everything and just gives you the debt right a position that at that amount, OK.
1:00:17
The next thing we're going to do is do dry deposition of particles.
1:00:22
See that here let me go to the next, ah, age and open up.
1:00:33
Data entry here, and here, what we're going to do, is define a particle rather than a gas.
1:00:46
And let me go to the concentration.
1:00:51
Set up, Ron, gluten deposition, and grid setup deposition for species one.
1:00:56
And what we'll do, then, is we'll go ahead and populate this first line, which now is telling high split that that it's a particle, so we're putting a particle diameter in microns 1000000th of a meter.
1:01:11
A particle density and a particle shape factor.
1:01:15
The shape factor is when you have a particle that's settling.
1:01:19
if it's a sphere, it settles sort of the fastest then the one shape factor could be higher up to for something which is but different aerodynamic shape and doesn't settle as fast.
1:01:35
And we'll go ahead and leave that, um, hard wired deposition velocity in there, because we're gonna go ahead and do that this time as well.
1:01:51
Let me save.
1:01:52
And the one thing I wanted to mention is that when you do particle deposition, another factor is is computed in the modeling and that's the settling velocity.
1:02:08
So a gas molecule doesn't really settle, but a particle does sort of fall through the through through the air, and depending on how big it is, and its density, and its shape, it will, um, it will settle at different, um, different velocities.
1:02:31
In fact, this settling velocity here is, is, depending on the particle diameter, the gravity, particle density, minus the air density.
1:02:41
I think this is the viscosity, um, of air here.
1:02:46
Um, we see very quickly, a picture about particle dry deposition, settling velocity.
1:02:58
On the bottom axis here is particle size in micron.
1:03:01
So this would be one micron, 10, 10, microns, 100 microns tenths of a micron 1, 100 micron.
1:03:09
And what you can see is that an opera here is a deposition of velocity, but you can see is for large particles the deposition velocity is basically dominated by the settling velocity, and for very small particles.
1:03:25
Um, it sort of depth deposits, more like a gas and it's not really affected by that.
1:03:33
The settling velocity there's this saddle in the middle here.
1:03:36
And what's happening here is that the particles are having a hard time making them through the quasi laminar sub layer resistance.
1:03:45
Um, they are too small to fall fast through that layer, but they're too big to be affected by Brownian diffusion, and make it through that layer, Whereas, when it gets smaller and smaller and smaller, they are able to be effected by Brownian diffusion of molecules and atoms on the air hitting them.
1:04:10
And then, they can then sort of diffuse down to, to get to that layer.
1:04:15
And this is interesting, this is why, um, sub micron particles are so dangerous for, for health.
1:04:22
Um, it's the same thing. Sort of happens a little bit in your lungs.
1:04:27
Big particles either fall out before they get to you, or they get captured in, in your nose.
1:04:35
You know, or, in, uh, you know, before they get to your lungs, if it's a really tiny particle or a ton of a gas, a lot of times, you can just breathe it in and breathe out. Hopefully, might not absorb.
1:04:49
If you're lucky into your lungs, but these small particles, it kinda make it all the way down into the into your lungs and, and then, um, kid gets stuck there.
1:05:05
OK, let me get out of out of this.
1:05:11
And, we're going to go a little bit faster now.
1:05:14
Um, let's go ahead and run Model C, Let me save.
1:05:24
Do you need to go?
1:05:25
What's faster here? Concentration.
1:05:28
one model.
1:05:36
Again, I'm sorry, I should've, Didn't run in the model, well, I'm just doing that.
1:05:41
Explaining?
1:05:45
Um.
1:05:49
And, one thing that's happening with this particle deposition is that not only are the particles dry depositing from the lower layer, but they're also settling through all the layers and fight by defining this, um, point 0 1 settling velocity and deposition. Lastly, we're actually hardwiring in that settling velocity.
1:06:19
Normally, it would be calculated from the parameters that we put here, But here, it's going to sort of hard wire in that settling velocity.
1:06:28
So, what's going to be happening then is particles will be falling down at a certain rate. In all every every particle in the simulation will be falling down at a certain rate At that's excuse me, at that same rate.
1:06:41
And let's go ahead and make a map of that.
1:06:48
Oops, sorry.
1:06:51
Then you display concentration, contours, and let's call this part.
1:07:05
Um, and I don't think we need to do any thing else.
1:07:14
Let's go ahead and execute the display.
1:07:19
And, um, as we go through, we see this, it's similar kinds of concentrations.
1:07:27
And the last one shows us this.
1:07:32
Uh, deposition, concentration deposition, amount, and if we look at the message file, what we see, um, let's see, advanced you messages, it's opened this up here quickly.
1:07:55
Oops, we'll get into that edge.
1:08:05
Um, at the end, we actually have a lot less mass, even more, less mass than when we did the dry deposition of of gasses and, um, and that's because even though the deposition velocity was the same zero point zero one, we had material falling down from higher levels with that settling velocity. Making it into that low, lower level, near the ground.
1:08:32
And so we ended up then, with more in that lower level, and then more could deposit.
1:08:38
That's why that particle deposition ended up being more than that.
1:08:42
Ah.
1:08:43
Then the gas just deposition, and I'm not going to do this next section: I'm running running short on time.
1:08:52
Instead of having different the mash on, it's the same particle be depleted. You actually can deposit.
1:09:01
A certain fraction of the particles themselves is another way to do the simulation.
1:09:07
Um, it's an OK way to do it.
1:09:11
You get sort of very similar results, but, what happens if you're losing particles, and then, so that if you're having a long simulation, you may end up in a situation toward the end of the simulation.
1:09:23
You don't have enough particles and, in your calculation to fill out your concentration grid. And so this made, there may be some drawbacks too.
1:09:33
To that. End. Let me, let me try to speed up here even a little bit more.
1:09:38
Um, um, and, by the way, you're seeing these higher concentrations with the particle deposition because of this phenomenon of the pluton falling down from the higher levels into the lower levels.
1:10:04
Alright, let's go to the next section where we're gonna be looking at what the position of gasses and, um, here, um, let's see, yes.
1:10:18
We don't need to do this same stuff again.
1:10:23
And here the temperature of the gas I'll say ... deposition.
1:10:26
There's two kinds of wet deposition that you could imagine.
1:10:33
one would be if the pollutant is in the cloud itself.
1:10:37
And then the cloud starts raining, then pollutant that's in the cloud droplets that then fall out of the cloud will be deposited when they get in when they finally hit the ground.
1:10:50
And the other kind of light deposition that happens is, um, well you have pollutants below the cloud and a droplet is following falling through this layer of pollutant.
1:11:05
The pollutants can be absorbed into the water droplets as the following.
1:11:11
And so, for a gas, it can either be absorbed into the, you know, the water trapped within the cloud or into the falling water droplet if it's below the cloud. And that is, the parameter that's used for that is called a Henry's Law.
1:11:26
Coefficient is sort of a simple treatment of that.
1:11:30
But no, it's telling you how much will be dissolved in the in water, if the gas has a chance to be dissolved in water.
1:11:40
And, um, here we define the same kind of sort of first order time constant, which is a function of.
1:11:51
Um, the henrys law constant and some other other parameters.
1:11:57
And let's go ahead and set this up.
1:12:02
So you go into the concentration, set up run.
1:12:08
Pluton Deposition, liquid setup.
1:12:10
So now we're going to go back to defining a gas.
1:12:19
And we're going to take out that dry deposition.
1:12:23
If you just want to look at what deposition now and we're gonna add in Henrys Law value, um, this is a relatively low value.
1:12:34
Some chemicals have a much higher absorption into water and let's save and let's go ahead and run the model before I start saying anything else.
1:12:50
Mission Model, yes, Sure. Yeah, Wet deposition flag needs to be sent to yes, it doesn't, thank you, Sunny.
1:13:01
That's a very good catch those radio buttons.
1:13:07
Let me, let me go back.
1:13:09
That has caused a lot of confusion in the past for me, And other people.
1:13:14
Let me let me go up and show this, again.
1:13:19
Um, those radio buttons at the top can be helpful, but they don't actually do anything in the model.
1:13:28
They can help you populate some of these fields.
1:13:33
If I click on yes, actually, it didn't do anything.
1:13:37
If I click on, if I click on Particle, then you see it's adding in some values that I click on.
1:13:44
Particle, and then, what the position is added in these values for for Particle. So, it can help you.
1:13:50
But, it actually doesn't do anything to override whatever those values are. But thank you very much, Sonny.
1:14:00
I have gotten confused on that myself.
1:14:03
Yeah. You're welcome.
1:14:05
And let me go ahead and put this back to three, and then 0, 0.
1:14:16
2 points: zero. And save.
1:14:21
Save.
1:14:23
Save. And let's run the model.
1:14:31
Sorry, I'm running a tiny bit late here, I'm going to try this speed up.
1:14:38
I think this is why we were thinking we would start this section last night, to get the first part done, and I'm sorry, we weren't able to do that.
1:14:48
I'm just having some computer issues, and, uh, thought it was better not to do that.
1:15:01
All right.
1:15:06
Now this was a, um, early low henrys law constant.
1:15:11
and so, we're not going to get a lot of what the position, and you know, of course, but deposition will only happen when it's raining.
1:15:21
No, or where it's raining and when it when it's raining.
1:15:23
And so obviously, no, you have to have lane or precipitation happening in your meta logical data.
1:15:35
And as we all know, meteorological models aren't perfect at predicting when it's going to rain. So that is, you know, with uncertainty in these kinds of results.
1:15:45
Let's go ahead and exit out and, um.
1:15:52
So we'll make this map display concentration, contours.
1:15:59
We'll call this one, Bought Wet Gas.
1:16:10
But, yes, I think everything else, is the same that we've been doing.
1:16:19
Um, let's execute the display.
1:16:25
And what you'll see here is that the concentrations you know, are there are different, you know, then what we have, you can't totally tell him that looking side-by-side but um, you look at the last one in the deposition.
1:16:39
Now you can see it only wet deposited when it rained, this turned out there were these two periods of rain that happened during the simulations in the middle optical data that we, that we were using.
1:16:55
And so this is showing you there was a bigger a rain event here and also there was higher concentration of pollutant earlier on, so that is why it was a greater.
1:17:08
Amount.
1:17:10
Um, OK.
1:17:12
Let's see, if anything else, If we looked at the message file, um, we're seeing that, actually, there was not very much deposited instead of 200 and, you know, very little, actually, this deposited, that's because, Henrys Law constant that we, tech was very small.
1:17:30
You know, some, um, pollutants have henrys law, constant orders of magnitude greater than that, that three that we put in there, and that could show some bigger, um, bigger deposition.
1:17:46
OK?
1:17:49
Let's go to the next section and this is the last section.
1:17:53
We're simply done with this.
1:17:57
We're going to do wet deposition of particles and let me put out of here.
1:18:04
concentration set up run.
1:18:09
We don't need to reload these things.
1:18:11
We're just going to be changing a few parameters here, um, and so now, what we're going to do is we're going to, again, set up this deposition in grid set up And we're going to go ahead and we'll click on this particle gas and particle.
1:18:29
I toggle those two radio buttons and it did go ahead and put these values, inform me, or I could have just type them in.
1:18:35
This is a particle five microns in diameter, density of six grams per cubic centimeter and the shape factor one, because that's saying it's a sphere, we're gonna do something special here.
1:18:46
We're gonna put in a very small, hard wired value, um, of deposition, velocity of settling velocity because what we'd like to do, it's just see the wet deposition only.
1:19:03
So this is going to override it, and with a negative small negative number.
1:19:07
Basically turn off that settling velocity.
1:19:12
And let's go ahead and put in some values of, I just toggle these buttons.
1:19:18
That's, there is a useful thing for these buttons, it does help, you kind of populate the menu here with some values.
1:19:27
And here we have the in cloud scavenging parameter and the below cloud scavenging parameter and we haven't both being the same. Sometimes, they can be different. This is a big uncertainty. I think, in, in all models, how to parameter rise this split deposition, especially for part of particles.
1:19:50
Possibly, there's not any, you know, any worse than, then, other models in doing this, but but not under the models are essentially perfect doing this, um, so let's save it.
1:20:03
Um, but save it.
1:20:07
Let's go ahead and, and save and run the model.
1:20:25
So the same thing is happening.
1:20:26
The protocol could be in the cloud, and you might know that cloud droplets actually form around atmospheric particulate, um, cloud condensation, nuclei.
1:20:39
And so the atmosphere, particularly, that's in cloud often, you know, is fairly effectively deposited.
1:20:49
When when that clock starts to rain, um, as the raindrops are falling through the layer below the Cloud, um, no particle could be intercepted by the the droplet, but it's not quite as easy, because as the droplets falling down, it's kind of creating a a flow field around it. It's almost kind of pushing things away a little bit.
1:21:13
Anyways, it's not quite as efficient as one might think to be captured by, by the following year, although we all know when it rains, the air does seem cleaner when completing it. And so, obviously, some things are definitely captured.
1:21:28
Let's exit and let's display concentration contours, and we'll make a plot here plot.
1:21:37
Ooops, typing.
1:21:42
Plot, what P, execute display and if we go to the blast, values here are actually greater.
1:21:55
Or let's see, are we seeing that the greater?
1:21:57
Let's see if we have a lot, right, aren't.
1:22:04
Right.
1:22:05
But gas, yeah, for the gas we had, um, 14 peak grams per meter squared.
1:22:14
Maximum, and for the, for the wave particle, we had much higher, um, maximum concentrations of deposition.
1:22:26
So in this case, this was a much bigger deposition, although, um, the total amount deposited out of the of the model message file here.
1:22:41
See how much we have at the end, um, still, not a lot we started at 201 and we're down to about 200, so it wasn't a lot.
1:22:51
Um, wet deposition can be an effective way of getting stuff out of the air, but you know, it has to rank and it has to remain where the pollutant is, and it has to rain.
1:23:01
No, ideally, and Putin has to be below, either in the cloud or below the cloud. So stuff above the clouds won't be, won't be deposited.
1:23:11
OK, I think, I'm sorry that I've gone over time here, and I think I'm going to now, and we won't go through these exercises, You can certainly do these yourself. And let's go ahead and go to a break. Let's go to I am.
1:23:29
I guess, I'm gonna go to a 16 minute break.
1:23:32
Why don't we come back in in 16 minutes, at 20 minutes to the hour, which would be 10 40 on eastern Daylight time.
1:23:39
But wherever you are, 20 minutes to the next hour, OK.
1:23:44
Bye, and we'll see you then.
1:23:51
OK, we are back from our break, and they'll now continue on with E, workshop. And let's see.
1:24:03
We are going to do a section now.
1:24:08
Section 12 of the of the tutorial, and doctor Alice Crawford, will be giving that section.
1:24:20
And Alice, alright, but it's OK with you.
1:24:23
I'm going to change the presenter over to you. Give me one second.
1:24:34
See.
1:24:36
Change presenter.
1:24:38
Aye.
1:24:42
Turn off my cam, there we go. And I could see the screen that you want us to see. So everyone, OK, OK, thanks, Alice, take, take it away.
1:24:55
All right, good morning.
1:24:58
So we're gonna start off right away getting into your Section 12.1. And we're going to start off doing the run because these ensembles take a little bit of time. So we'll start the run, and then waiting for it to finish. We'll have a little bit more of a conversation of what we're going to be going through in this section. So go ahead and I'm gonna start with the reset because I've been playing around.
1:25:24
But the model are going to go to the setup run and retrieve our control and setup files that we have been using.
1:25:39
I'm going to save them. That's the control file. Now, look at the setup file.
1:25:57
OK, and then what else we're going to do, is, just to make these friends a little bit more.
1:26:07
Oh, thank you!
1:26:12
I just sat out.
1:26:17
Alright, to make these friends envelop, ran a little bit faster, we're going to just change a few things.
1:26:22
So one thing is we're going to go into the Grid setup, kinda look at our concentration grant.
1:26:30
And we're going to change our sampling start time to the 26th, Sarah Secs.
1:26:41
And we're also gonna rename this to cullet Ensemble.
1:26:56
And, we're also going to.
1:27:05
We're going to set the starting height to 200 and that's because if you remember back from the section with trajectories, we did something similar.
1:27:17
So this is basically going to shift your meteorological data by one grid point and so it's going to do it in the vertical as well.
1:27:26
So, you don't want to start to close to the ground, because then when you sit down weren't sure underground basic All right.
1:27:35
And then we're also going to change our render ation 2 49 hours.
1:27:43
So we did all of that.
1:27:46
And then what we're going to do is concentration, special runs ensemble meteorology.
1:27:55
And we're going to use our setup file and yes, we're going to do that. So, now this calculation is going to start and it's actually going to perform 27.
1:28:07
Highest split runs with the meteorology has shifted in various directions fell pickets, I think it's starting.
1:28:23
While we're waiting for that turbine, will go back and we're kind of just an overview of what we're gonna go through.
1:28:30
Section 12, which has talking about uncertainty, is an air concentration, and probably throughout this whole workshop, we've been talking about uncertainties and the results, and, no, that's just.
1:28:44
When you work with turbulent flows, you just deal with uncertainty.
1:28:49
It's built into the system, and here in this section, we'll get into a little bit of how you can use ensembles to characterize that uncertainty and, you know, kind of try to think of it as a as a feature of, of working of modeling with turbulent flows.
1:29:07
So people have started, you know, constructing Ensemble's, we have numerical weather prediction ensembles now.
1:29:13
Um, I will, so we'll talk about that a little bit here, and then we'll actually go back to talking more about particle number. So, this is something that we've talked about at various times throughout this workshop, as well. And there's been, every time we talk about it, maybe a little bit differently.
1:29:32
So if you feel like you already understand it, then you might be able to kind of, no tune out during this section. If you feel like you're still confused, it will give a little bit different perspective.
1:29:44
And maybe that will click, because certainly, I think, with the Lagrangian Models setting, the correct particle number that you want, as well as the resolution of the concentration grid can be something that people sort of struggle with.
1:30:00
The physics ensemble we'll probably just talk about. We actually won't do the runs for that.
1:30:05
That goes back into a little bit of what was touched on in.
1:30:12
See, Section nine, where we do an ensemble with different physics parameters, Asians, two, just do in a more consistent way, see how does that affect the runs?
1:30:25
Then the Muslim tell meteorological data. We'll do some of those friends if we have time, that's looking at different datasets and then computing the, again, for the ... too.
1:30:37
And looking at the different statistical evaluations of those.
1:30:41
And then really the ensemble reduction techniques will probably just discuss kind of the theory of it to give you an idea of the ensemble's, like how people are using ensembles and constructing them.
1:30:56
And I think this is running.
1:30:59
Let's see.
1:31:04
Do I have something? I'm Alice. Yeah, I'm not sure if, did you change your particle numbers to 5000?
1:31:14
Maybe I might not have.
1:31:17
And so, it might be younger than you Sorry about that. Now, that's all right. And I think I've done this before, and then I actually have to restart, has split all over again. So that's fine.
1:31:33
Um.
1:31:36
Right.
1:31:42
Yeah.
1:31:54
Mmm hmm.
1:31:55
five else?
1:32:10
OK, now, I'm going to do this right away before I forget. So, we're going to, and sorry, if people following along were following me and, and also not changing the particle number.
1:32:20
So, we had 50,000 particles, and if we want to do 27 runs, that will take quite a long time.
1:32:28
So, we are going to change that, and I think we haven't changed to 5000, but I'm actually, let's write 5000. Yeah.
1:32:37
Yeah, I'm just thinking, if I want to run with even smaller numbers, sometimes, starting a little bit later, I'm going to change it to one thousand, which will change our results.
1:32:46
I'll get slightly different plots, then, are in here, but they should be similar.
1:32:53
And while you're in there, you can change the date core dump files to not not to be, yeah.
1:33:02
So ... just puts your heroes.
1:33:07
Um.
1:33:15
So we don't want any particle files, OK. And then I'm gonna go back to the concentration and change those things again.
1:33:30
And that changes to 49.
1:33:32
And then we also need to go in here, and set this 2, 26.
1:33:44
Ah!
1:33:47
And changed this to ensemble.
1:33:52
Then the other step, that I'm going to do, identi last time, as I'm going to save this file, I'm going to call it Ensemble control dot text.
1:34:06
I'm also gonna go save my.
1:34:11
And setup file.
1:34:14
I'm going to call that ensemble setup taxed.
1:34:18
So that way, if I made any more mistakes, we can just achieve those, OK.
1:34:24
Then we got back to special brands, ensemble meteorology.
1:34:38
OK, I'll give that minutes.
1:34:41
It hasn't.
1:34:45
Can go check, this double check here.
1:34:50
Thank God to my Working Directory.
1:34:55
Yeah.
1:35:01
It finished ensemble number one so it is OK and started good to have I am.
1:35:11
So now, test.
1:35:16
I'll just mention so this ensemble that we're doing, it's just using one meteorological data set and kind of shifting it, which actually gives you a pretty good idea of what kind of uncertainty you can expect from the meteorology. But, you know, most of us know that. nowadays The numerical weather prediction models themselves are creating ensembles. So Noah has a high resolution ensemble forecast system with which has 10 members.
1:35:43
And we also have our global ensemble forecast system, which has 31 members, the RFI, which people might be familiar with.
1:35:52
it, just like an ECM WF free analysis product also has a 10 member uh, ensemble, which stay extra. That's how they get their uncertainty information. And it's a little bit coarser than the actual error five products of the ... product just like hourly.
1:36:10
And I think, like a third of but degree, 30 kilometers.
1:36:13
And so at the ensemble member, as are, they, of course, are temporal resolution and spatial resolution.
1:36:21
But you can drive high split width, know, each member of the NWFP model ensemble.
1:36:28
And it's a it would be a little bit more rigorous way to provide uncertainty information, uncertainty in the meteorological Data Centers, NW, the ensembles are supposedly, to find, to provide, um, you know, well constructed ensembles of the state of the atmosphere.
1:36:48
Currently, I mean, you can download the RFI from the Copernicus website, and we do have a converter.
1:36:53
You can find some information about how to download at the Get Hub site, and we do have a converter for error five and datasets.
1:37:02
Um, that is quite popular, uh, the h-ref and the GFS, we do not archive those, those are a little, that could be a little bit harder for you to get currently, but maybe avail easier to get in the future.
1:37:15
Because we do, we are using them internally for a lot of a lot of work.
1:37:22
OK, so, OK, remember 11?
1:37:25
So when we got along pretty well, to give you an idea, like the years are, it says from the comparing the gas output, the global ensemble forecast system, which has 31 members to the TFS quarter degree that the client is the quadratic or HDFS, and then the thinner lines are all of those members. So this was like looking at a volcanic eruption of ....
1:37:56
The meteorological conditions around that, see, can see, you know, the wind direction, wind speed. Do you guys see this at all?
1:38:05
The Go to meeting thing. Is that interfering with the slide?
1:38:09
Are there?
1:38:12
No, it's not, you don't see it. Oh, I didn't see it. Yeah, yeah.
1:38:15
OK, So, Windspeed, and these are profiles that were taken at this particular time, actually, using that profile program we used earlier, but a different planning program.
1:38:28
And, uh, so you can see that, you know, there's quite a bit a wind shear, the wind direction with a little Dent.
1:38:35
No. We have to realize that, the zero right that wraps around.
1:38:43
This isn't really, like, a huge change, and it's a pretty continuous change in wind direction, but, still, you're getting a lot of wind shear there, um, and the, the eruption, you know, reached about this height around 300 millibar as R, so it's about 9 or 10 kilometers high.
1:39:05
Usually they do read about the tropopause and then kind of kept there.
1:39:09
So this kind of gives you like just a snapshot of what you might expect from the variability that you would get from an NWFP ensemble.
1:39:17
And it would definitely effects here your dispersion model results right when you are getting these changes in wind direction and wind speed.
1:39:26
Um.
1:39:29
Now we're also going to look.
1:39:33
So now we're at 14, all right.
1:39:35
So the other thing we're gonna look at, as we're going to look at, like once you have this ensemble, you have 27 members are 31 members, or even just 10 members, Know what are you going to see with all those outputs? Are you gonna look at our 10 numbers and examine like every, single one?
1:39:55
So there's various ways to combine those, to get useful information, um, and we're going to talk about a few of those, So, um.
1:40:12
one thing is that we are going to, there's a program, which we're going to use, actually, has finished, and it's called con Prod, and you can access it through the GUI as well And it basically combines the different ensemble members, too, um.
1:40:37
To produce different information. so this kind of gives you an overview of the different it'll produce all these different files. And each one has a different set of information in it. So one of them has just the ensemble mean. one of them is the variance. one of them is called the coefficient of variance, which is basically the standard deviation divided by the mean.
1:41:00
There's also something the probability of concentration and the concentration at percentile and we'll just take a minute to drill down a little bit into what those things actually are and that will become another slide.
1:41:17
So, the probability of concentration, that was the one that was kind of on the left, this has it can go by a couple of different names.
1:41:26
one is the probability of the concentration exceeding a certain amount.
1:41:30
Um, more accurately sometimes we just call that the ensemble relative frequency of exceedance.
1:41:37
Because sometimes the samples are not what we call well calibrated so the ensemble could say, well, there's a 90% chance, know, that the battle exceeds the probably the concentration, but in reality it's only like a 50% chance.
1:41:54
This is also sometimes called agreement and threshold level.
1:41:58
The way it's calculated is, you pick a concentration, or, in this case, actually, we did a mass loading, and then you count how many ensemble members are above that threshold at each point.
1:42:10
And then, so each point becomes the number of ensemble members above that threshold.
1:42:16
Then you divide by the total number of ensemble members, to get a percentage or a frequency, um, so this can be used like it can be useful for risk based approach.
1:42:28
Like, if you would know, that's some concentration, like a lot of times, and chemical releases, you would have, like, a chemical sheet that would tell you what are the critical concentration thresholds for human health. Right? Like if you exceed this concentration, like people might get a little bit sick. If you exceed this concentration, then people will get very sick.
1:42:50
and so you might want to know the probability of exceeding those concentrations and, um, then you can get people like these probabilities, right? So this was one for a volcanic ash eruption, actually, which looked at. These are different probability of exceedance levels. So this was observations.
1:43:11
This was just one member as kind of a, an example and then you've got there's hardly any members that are exceeding the five grams per meter squared.
1:43:20
There are quite a few that are exceeding that zero point zero two grams per meter squared.
1:43:27
OK, so the other thing that we look at is called Agreement and Percentile Level, and thus actually outputs, a concentration, or, in this case, a mass loading.
1:43:38
And in this case, you're picking a probability frequency threshold.
1:43:41
So suppose you want like, let's say, 50%, because that's what we're going to look at.
1:43:46
So in this case, the way you calculate it is at each Pixel.
1:43:49
You are ordering the points, from least, to greatest, And then you are finding the value for, which we just picked 50% right side, or at the points, from least, to greatest. And I'd find the value for which 50% of the ensemble members had a value smaller than that.
1:44:07
Um, and so this right here that just has the median or the 50% level.
1:44:13
So these are concentrations. And here's the concentration viruses zero point two.
1:44:20
So 50% of the members here had concentrations smaller than that and 50% had concentrations larger than that.
1:44:28
So if you would look at the maximum wind like this, if you look at the basically 99% or 100% level, that would give you the maximum concentration or mass loading value that any of the ensemble members produced.
1:44:44
Now, OK, you might wonder like how are these two things are related at all? They are related, and this just kind of shows you that how.
1:44:53
So, um, we have here, this first test, agreement and threshold levels, so you get probability values of exceeding a zero point two gram per meter threshold.
1:45:09
And then I applied a 16% probability threshold to that.
1:45:15
So basically, I just colored all the areas that were about 16%, rather the probability white and all the areas below that black, so then we get this outline.
1:45:26
So, if someone decided that, you know, 16% was their threshold for risk, they were like, well, um.
1:45:37
No, this is, it.
1:45:40
Again, it goes back into this.
1:45:43
The risk based approach has kind of are assuming that the end user kind of knows, um, how much risk they're willing to tolerate. Like, if it's a really, if it's a high concentration, there, like, the consequences of gotta going into this concentration, and I really, really, really bad, than they have, like, a low risk tolerance. So they might say, well, if any of the ensemble members, if there's even, like, a slight chance that it's going to be that concentration and I'm not gonna go there, then you'd pick a really low probability threshold. and you'd get a larger area that you'd have to avoid, right on.
1:46:17
The other hand, if they're like, well, the consequences are sort of bad, but, no, they're not so bad, maybe it's just more maintenance on your airplane, or maybe it's just like, um, no. You have to inspect your, if you had some.
1:46:34
You'd have to like, know and inspect something or, or maybe there's health risks, you know. If you have to evacuate people, there are health risks to evacuating people. That you have to consider.
1:46:44
Um.
1:46:46
So then you might set your probability threshold a little higher.
1:46:51
On the other hand so you can get the same contour if you do the applied.
1:46:55
So this was the applied percent agreement on percentile level.
1:46:59
And where, you know, 16% of the members had a higher concentration and 84% had a lower concentration, and then we applied that to zero point two grams per meter squared to that, like a fat threshold and then you end up with the same area. So, that's kind of how they are related to each other.
1:47:19
All right, And how are we doing? Oh, I think we're done. Alright, great.
1:47:26
So we get a little message that it's completed.
1:47:29
Are going to continue Now. We can actually look at some of these things.
1:47:36
Um.
1:47:43
Alright, so to create those files, that I was just talking about, are going to go to display ensemble, create files.
1:47:53
And see that's fair Menu tab um.
1:48:06
And we're just going to create those.
1:48:09
Now, if we go to our working directory, here are all the files I created, So it created theirs.
1:48:23
Probability of X zero.
1:48:28
These are like the median threshold, so the prob the 50.
1:48:32
It's actually like the median that we were talking about, where it gives you the content, the median concentration, of the ensemble.
1:48:39
And the c-max files are a probability of exceedance says and the way it picks which concentration it's going, it's like 10, um.
1:48:52
It's 1% of the maximum value, 10% of the maximum value, and then the probability that you're going to have the maximum value.
1:49:04
And these files have the same format as the ... files, so they are binary files.
1:49:10
And they can be plotted, actually, with the same utilities that the CDM files are plotted with.
1:49:19
But they do have their own special display, because if you just use the regular display, some of the labeling on the maps will not be accurate.
1:49:30
So, let's take a look at, um, Chris Best to look at the 50%.
1:49:42
Yeah, fair to us, and we can actually.
1:49:47
Huh.
1:49:53
Let's try that.
1:49:58
Then we can, You can look at other things, too, I think.
1:50:14
OK, and then, well, let's login.
1:50:18
Let's look at the probability of exceeding as well.
1:50:30
There we go.
1:50:32
Taken, say this as the probability of exceeding 10% of the maximum concentration, and again, wow since you very well with calculating, calculating the limits of the map.
1:50:48
Alright, we're not going to spend too much time on that Because we wanna look at the box plots.
1:50:55
So.
1:50:58
X plots.
1:51:00
So we're going to go back to the sampler.
1:51:03
Um, Templar 5, 10, so the.
1:51:17
Just enter the location of that, and then we execute the display ads.
1:51:25
Oh, I know, I didn't do.
1:51:28
I think I didn't.
1:51:46
Mmm hmm.
1:51:50
Alice?
1:51:52
Yeah, Krishna. We weren't sure when you went back and started again with the 5000 particles.
1:51:58
If you change the sampling: start time, or this simulation, start time, um, you might have changed the simulation start time, the second time around the first time you did it, right?
1:52:13
And, it's possible. Look at, look at the start time, part-time. I think you started at the, at the time, you wanted to start the sampling.
1:52:23
So, sorry, I didn't quite catch it when you did it, but I've been really challenging here today and no, no, no, no, no, no.
1:52:32
You wouldn't leave mistakes and open.
1:52:34
Thank you.
1:52:36
Well, all throughout this.
1:52:38
So, you could use the pre calculated results in the workshop files, or you could just show results from the tutorial images.
1:52:51
You know what, I mean that Um, that's the main point that shouldn't be doing the runs.
1:52:56
It's not, it's not We OK, mark a thing, as I didn't change the starting time, and, and so, Basically, The sampling time is not lined up with the I think you did change the starting time.
1:53:19
You should have left it at the start of the tracer experiment.
1:53:23
Um, you change that starting time to that What you wanted to do it started at the sampling time Millimeter hmm. Yes. What happened? Yeah? that is what happened, so I should not change this number of a catch and release Happens, that's when I am playing to start. Right. Exactly.
1:53:46
Hopefully, some of you did not follow me when it's doing that, and you're getting better results.
1:53:52
So, the box plot, said, Look, something like this.
1:53:55
Um, and so you'll see the highest concentrations where on the 18th and, ah, he asked the other says like the different members, so that the numbers are like all sync up to 27.
1:54:13
Um, and then.
1:54:22
You can look at there.
1:54:28
Yeah. So that's that's just one with like that.
1:54:31
The measurements kind of painted on for an example. So you can see it looks like there's a little bit of a time lag.
1:54:37
Maybe that's not being captured by the Just kind of.
1:54:42
I mean it's not a it's just kind of a poor man's ensemble or So it's not capturing really at the time lag.
1:54:50
Um, but you could make a good case that if you would kind of shifts the arrival time of the plame over a little bit then it would capture the range of concentrations that are being measured pretty well.
1:55:05
All right.
1:55:05
So, now we're going to shift gears a little bit and at least for a little while, sour I think people have been noting that, you know, and there, and 12.2 actually changed.
1:55:22
This used to be called turbulence ensemble and I changed the name to uncertainty due to particle number, because I thought that was a little more descriptive. But I did not change that name here at the header.
1:55:32
So it might be a little confusing. A few.
1:55:35
If you click on that and then you're, you're getting this one but they actually are, this is the crux section that you're supposed to be in.
1:55:42
Um, And so, you know we've been talking a lot about number of computational particles that we use.
1:55:51
And also like concentration grad and the resolution of the concentration grid that you can use Ands.
1:56:02
Um, this one is kind of a nice way, because you can use less particular capability to understand if you're if you're using enough particles, it will give you an idea of high quality, if I used, um, if, how much variation you can expect in the results due to the particle number you're going to use. And so let's let's get started.
1:56:29
So we're going to clean up our last one.
1:56:33
So there's a little nice little utility, um.
1:56:42
That's an special runs ensemble cleanup.
1:56:49
And if he does say, well, it will basically remove your ensemble files from your Working directory so that you don't get them confused. Because sometimes, you know, if you, if you do one ensemble ran, you think, oh, it's just going to overwrite them, but then you might have a different number of ensemble members on your next run.
1:57:09
So we're going to do that.
1:57:10
And then we're going to, Actually, before I forget anything, I'm going to go back, and I need to change my starting time back to what it should be.
1:57:24
It should be 9, 25, 17.
1:57:36
Then I'm gonna remember, I had saved that, I'm going to save that, like that.
1:57:44
Um, and then, does not sweat at the festival, like I'm actually going to check myself, right?
1:57:51
So, I want to look at, Yeah, my working directory.
1:58:06
And.
1:58:10
I'm gonna look at my control file that I'm gonna use, and I'm gonna check can against the one in the tutorial.
1:58:19
So.
1:58:30
Yeah, I think that looks pretty good.
1:58:34
Alright, so I'm gonna go ahead and.
1:58:58
Save that, Then we're also gonna go to.
1:59:07
Look at that.
1:59:12
Yeah, it's really long that artist.
1:59:20
Probably the main things are the particle number.
1:59:26
I have, OK X, I had changed the 2002 trial, we were actually using 5000.
1:59:32
Um, You might want to check the particle input and output files, just double check.
1:59:43
Yeah, so.
1:59:49
Alright, so let's look at.
1:59:54
Um.
1:59:56
Particle files, OK, We have the set, I'll save that. And then the particle number, we're going to change again.
2:00:04
So, we're going to change that to 18 52. Which is kind of a weird number.
2:00:09
But, it's sad, because we're going to, again, we're going to run 27 variations, that's just kind of, um, there's no reason for that number, just to match the 27 variations we did last time.
2:00:25
And.
2:00:28
Because they're using 5000 use, like 850 to 2, Um, so basically, you'd add all the simulations up, You would get back to your 5000 particle run.
2:00:46
And then concentration special runs An ensemble, and now we're going to do the turbulence run.
2:00:58
Yeah.
2:01:01
Gonna start. Now what's happening with the turbulence fan as it's basically changing the random seed? So.
2:01:12
The way that we're going to go into a little bit, again, from a slightly different perspective, but how highest split, no models, the turbulent motions of the atmosphere?
2:01:27
Um, and it's, so I'll go here, I'm gonna go little out of order.
2:01:34
So, you might remember this equation. It's a stochastic equation, which means it has a random component to it.
2:01:41
So, this is like the fluctuating, but this is just an example of what you might get for a fluctuating velocity tranches 10 particles.
2:01:49
So remember they and when you use this equation you end up with The PDF of your of your turbulent velocity is what the Gaussian wets, it's been shown that in a turbulent flow, it's close to Gaussian, it's not exact. And then it has this autocorrelation. So the velocities are correlated over a certain timescale the grantee and timescale.
2:02:15
And so you can see that these I know they are. They have these fluctuations, but there is, if you look at the auto correlation time of time, they would be correlated over that time.
2:02:27
So, basically, if if I would start, I start this, any I get these velocities for these particles.
2:02:33
If I would start each one at a slightly different place, right, they would have slightly different paths.
2:02:39
So that's basically what the turbulence and samples doing. It. Starting each simulation with a slightly different random, with, so that each one will have a different sequence of random numbers.
2:02:52
So what does that do?
2:02:54
Well, let's go back again.
2:02:57
So this is just, that sounds like a little toy model, but it gives you an idea, like this first one has no dispersion, it's just a trajectory.
2:03:07
The second one has like a small sigma, which is basically the width of your Gaussian distribution. So that means every time it's like picking that random number to generate your fluctuating velocity, that fluctuating velocity is not very big. So the spread in your in your particles stays fairly small.
2:03:25
Then, just one on the side, it has a large sigma so you can get much larger turbulent velocities. And so you get much more spread out there, right?
2:03:35
So, you know, that's what we're talking about. I think Chris was giving them on about how we calculate that.
2:03:43
You know, how much turbulence is is there.
2:03:47
So what does this have to do with particle number?
2:03:49
Well, so this one, all of these had the same exact sigma, but they were just different numbers of particles. So, if you can kind of do this fake scenario.
2:04:00
We're like, OK, if I released one kilogram of mass, and I'm only going to use 10 particles because I know the simulations have ran really fast.
2:04:09
So, each of these particles now is carrying, is representing about 100 grams of mass, and yeah, like, this is what I got right?
2:04:18
Now I increase to 100 particles. Now, each particle has 10 grams of mass on it.
2:04:23
And this is what I got.
2:04:25
And now I Go up to one thousand particles.
2:04:29
So now, each particle only has one gram of mass, and I get this.
2:04:33
So, if you would look at concentration that you would calculate, you know, the way we calculate the concentrations.
2:04:41
And we'll talk about this more as, um, know, you're just counting the amount of mass and each grid square that you've defined, and then dividing by the volume of that. And also, if your time averaging, there's also a time average time averaging component to that. But here, we'll just keep it simple.
2:04:59
So, you see it, I mean, here, the, you know, all your mass would be in these two print squares, and, but, still, these are the same, like, most of the masses in these two great squares. Most of the masses and these two grid squares.
2:05:13
So, these would all give you very similar results.
2:05:20
You know, clearly the more particles you have, the more you're able to resolve the underlying distribution that the particles are representing.
2:05:29
Um, maybe to talk about that another way.
2:05:33
Here's again: that's when it has 100 particles.
2:05:37
It's kind of representing if you had like two short releases and they're each represented well by kind of a Gaussian distribution.
2:05:44
So, this is 100 particles. This is a thousand particles, they're both representing the same underlying distribution of mass.
2:05:52
It's just easier to figure out what that distribution is when you have more particles.
2:05:58
Um.
2:06:01
And, You know, the way that there are other ways to kind of get the mass distribution besides just counting the mass of the particles in the current box. There's something called a kernel density estimator that people use a lot, and that that's kind of saying that, well, I don't think they've been developed for a long time.
2:06:22
They're actually not used that often because it's just another complication, unusually, know, the width, the method that we use for calculating the mass distribution works well enough.
2:06:35
And it's pretty simple and easy to understand. So, the KTEA basically spreads, instead of the particle just representing one little thought of mass, it kind of spreads it out and gives us a little bit of a distribution.
2:06:49
Um.
2:06:53
OK, that's still running so, ah, sir, ***, oh yeah, there's one more thing I wanted to show.
2:07:03
12.
2:07:14
So, this is a figure from a paper that kind of illustrates this idea.
2:07:22
With the real data, This is actually the ... experiment again, But we use a lot.
2:07:28
And the top figure is showing, um, the highest split results with different amounts of particles, and also different concentration grid sizes.
2:07:40
So, then, if you can see the little circles, those are actually the data.
2:07:47
Alright, so, those are some measurements to compare to.
2:07:53
So you can see, you know, if you're running with 5000 particles and a pretty small concentration grid, you got under sampling.
2:08:01
Um, so you're just not able to resolve that does the crack distribution with the number of particles and each grad square. You can resolve that by either adding more particles or by making your concentration grid, coarser.
2:08:19
So, like, Run E, this one right here, that one has a zero point two five resolution grid, um, whereas these are, forget what they are, They're pretty fine, zero point zero five.
2:08:36
Alright, so these are, these are zero point zero five, by zero point zero five, 25 meters, and the vertical, so that's a pretty fine concentration grid.
2:08:44
And then he shipped 250,000 particles, you know, you get something that looks even smoother.
2:08:52
Then the bottom plots, we won't get into, but they're just using the same run, but using a different method of estimating the distribution.
2:09:03
So there was a question about you know the relation of your concentration. Good resolution to the or meteorological credit resolution and there's not really.
2:09:13
It's not like a direct correspondence.
2:09:15
So you could have a really korus meteorological data and still do a fine concentration grad.
2:09:24
And you can think of that as like this particular one it's at the beginning of a release, and the plume really does have kind of a Gaussian profile. And and there's even models called Gaussian plume Models. Which take advantage of this, right?
2:09:40
Like, if you have a pretty steady wind direction, and the winds aren't really changing, they're just flowing one direction, and you have a release, the concentration profile is going to be very close to Gaussian.
2:09:54
And so, a Gaussian plume model basically just provides an analytical solution, which says: Um, but just a Gaussian and the concentration very smoothly.
2:10:04
And you don't have to know worry about the resolution. You just get a smooth concentration field.
2:10:12
Ah.
2:10:14
So you wouldn't say that, you know if you just because you're using um no one degree meteorological model with a Gaussian plume Model.
2:10:27
The concept the resolution of the Gaussian plume model isn't going to be any coarser than if you use like a zero point twenty five resolution meteorological model.
2:10:38
So with high split, it's somewhat similar. It's the resolution of your concentration grid is not really tied to the resolution of your meteorological data. It is tied to how many particles you use and there isn't one more limitation, but just that highest split has a one minute time step. So, if you have, had people, if you tell people there's no limit, large, you don't tell people that, because then, they know that.
2:11:02
So, no, if you try to do something like five meter concentration grid, there's a one minute time step, and you're gonna start, the particles will basically skip over the concentration grid in those.
2:11:18
And just that one time step, and you'll get something that doesn't really make any sense. So, you do have to consider that.
2:11:24
You know, usually highest, but as used on the order of zero point zero one, to create concentration, grants are very common, even this zero point zero five.
2:11:35
It can be pretty common.
2:11:39
Aye.
2:11:41
Let's see how we're doing on our run.
2:11:52
OK, so it has to go all the way up to 27.
2:12:05
Sure.
2:12:10
Now if you are using really coarse meteorological data and you're doing like long rates, chance for it.
2:12:17
You know, doing a, having a course or concentration grants, you're basically doing more spatial averaging.
2:12:24
Right?
2:12:24
And so when you do more spatial averaging, and if you do a longer time averaging, um, your results might be more accurate, or they might compare better to, like, a longer averaging time on, on your measurements. Like, in those ..., most of the measurements, I like a six hour, time average, right? So, it's easier to get something like a six hour time average, right?
2:12:50
Than it is to get like, a, um, like a five minute time average or even a one hour time average, because there can be at displacements and, and plume arrival times a little bit and then in space and in space.
2:13:06
Um.
2:13:11
Right.
2:13:12
All right, so, while we finish, we're going to wait for that to finish, but why don't we go ahead and look at, raise thoughts anyways, and so, forget something like this.
2:13:29
And, so, what does this telling you as, basically, how might share your estimate of concentration, can change depending on, you know, though, um, the uncertainty during due to the particle number that you picked.
2:13:48
So, if you did a completely If you did a simulation that was completely the same, but just with a different random seed, so different sequence of random numbers, then you would know, you could expect these kinds of differences in your concentration. Now, one thing to note is that, you know, it's just this bar over here, and he has like, almost exactly like the ones we did before.
2:14:11
But if you look at the X axis, the Y axis, the scale is different.
2:14:19
So these really aren't covering, you know, as much of us arrange as as the ones.
2:14:27
That's the meteorological data and I I did this, Fine, We're not gonna do it here, You can do it if you want, but if you repeat the exercise using 50,000 particles in each simulation, you know, then you get what looks like this.
2:14:42
So, the 50,000 particle range has probably, you know, pretty good for, for this particular, um.
2:14:54
And this particular set of parameters, which has included, includes the concentration grid size.
2:15:04
Now, besides using this capability with the, with this turbulence ensemble, which says.
2:15:14
We're going to look here at the executable, so this has just executable, highest, but far, it twice C F underscore var.
2:15:25
You can also, um, you can change the random seed.
2:15:33
I, either using ... equals four changes how the random numbers aren't generator, so you would get different random number as if he is fat, or you can actually now we have a nameless Parameter can't feed.
2:15:46
So in that case, it doesn't you can't set doesn't the GUI.
2:15:49
So what you have to do is go to your working directory, and you'd have to open your setup file.
2:16:00
Um, so we can open our ensemble setup and Rick had put in.
2:16:13
As seed there, as we could add, feed equals , um, And if you do that, you'll, you'll do around that.
2:16:27
famous here, other Iran, but with a different random seed.
2:16:31
And then you can, that could be a way that you can kind of try things out to see if your a particle number is good.
2:16:41
Cause another thing to mention is a lot of those little kind of toy diagrams I was showing you there, assuming that all the particles have the same amount of mass on them. But in fact, a lot of your high split runs. If you do variable emissions, if you do different particle sizes.
2:16:58
If you have deposition, then the amount of mass on each of the computational particles is not always the same, right? Because if you have a computational particle, that's been kind of close to the surface. And it's been depositing suddenly losing some of its mass to deposition. That's going to have a different mass than a ... particle that's remaining aloft.
2:17:20
And also if you, if you have particles that have Matt like, have gravitational subtle Yeah, if you have particles with different particle sizes and you're really saying I can volcanic eruptions me do this a lot.
2:17:34
Where, 10% of the particles, you know, are have a small size, unlike 50% have a larger size, and they said that the computational particles themselves will represent different amounts of mass.
2:17:56
All right.
2:17:59
Still going, oh, it's completed, all right.
2:18:04
So now, we are going to do the same thing that we did before.
2:18:09
I'm going to create a file.
2:18:18
And if I go to our working directory, now see thus, the same files there again.
2:18:26
And.
2:18:33
Think this time, we'll just add the box.
2:18:45
Alright, Satirists Mai, box plot.
2:18:56
All right, So now we're gonna, we still have quite a bit of time left.
2:19:03
So weevil, the physics ensemble.
2:19:06
I think, we'll just talk about, again, and then we'll probably go through some of these runs, actually.
2:19:14
Um.
2:19:17
So the physics ensemble, basically as a way to automate, automate, trying out different premature assassins and high split.
2:19:30
So these were mostly talked about and I think chapter nine And there are different ways of determining the mixing layer height.
2:19:39
There's different ways of determining, um, know, how large the turbulent velocities can be.
2:19:48
Um, there's different ways of even, you know, here's the hana with a variable.
2:19:52
Timescale is even different ways of calculating that Lagrangian and timescale which controls how long the turbulent velocities are correlated over.
2:20:05
Um.
2:20:09
So that's what kind of automatically compute a lot of the most popular options, and then it would give you an idea, you know, how much that variability is going to affect your results.
2:20:23
So we look at the box plots again, and I think that's one.
2:20:30
Yeah.
2:20:35
Yeah, so those are the results there.
2:20:42
And, know, this, this is an interesting thing because we add, you know, I think, as Chris said, these are all we put the default says what we think will be best for most simulations.
2:20:59
But you never really know, like all of these are pretty reasonable things to do.
2:21:05
So you know in essence, creating an ensemble from them is the reasonable thing, to do, um, if you have, you know, the computational power to tax, I do an ensemble.
2:21:18
Uh.
2:21:24
All right. So the multiple meteorological data, I think this is kind of a good exercise to go through and it doesn't take too long to do.
2:21:33
So basically, we're gonna run high split, and five times for these different meteorological data sets. And.
2:21:46
me to ask, first, we're going to reset.
2:21:51
And we're going to retrieve flowers, um, files again.
2:22:07
ARR?
2:22:26
Alright?
2:22:27
And, then we're going to, ah, change the name.
2:22:34
And we're going to call it ... 2001.
2:22:46
I asked her going to change. The protocol number, again, said that.
2:22:55
Mexican a change, that's a 5000, just Tourette a little faster.
2:23:05
All right.
2:23:07
Um.
2:23:14
So, basically, we're intuits, we're going to each of these runs.
2:23:18
Kind of when you do the same gets, even though, you know, if you if you do it five times, then sort of remember by the end what you're doing.
2:23:26
And then, we're going to use this datum again, and we're going to look at the actual, um, statistical results for, ah, the various members.
2:23:43
Since we only have five members we can kind of compare, uh, what like how each one does. So, yeah, I don't know if y'all remember, if you don't look at them again, But there's a program called ...
2:23:57
which is accessible through the GUI, that will compute various statistics, including the correlation coefficient, Commonweal Cora Smear Enough Parameter, which is something that compares the distributions of the model to the observed figure of merit and space fractional bias.
2:24:17
Then, we compute something called a rank, which says, some of these things, the Normalized Correlation Fractional Bias, Figure of Merit and Space and Common Core smirnoff parameter, and basically the rank, the best strength you can get us for. The worst you can get is zero.
2:24:37
And, so, well, we'll look at the results for each one of those law, so, compute, look at the mean, and, and look at the results for the statistical results for the mean.
2:24:50
Um.
2:24:54
And, then, the section after that, we may or may not get to, but basically, it's just a way to automate. What we do in this one.
2:25:02
So there's like a little script that will kind of automatically, um, shower the statistics that were the prettiest something that's.
2:25:14
I'll kinda make ethanol table for you here, right, so this gives you the correlation.
2:25:22
Fractional bias figure, maritain space for each one.
2:25:27
All right, so that finished. So let's take the next one.
2:25:33
So for the next one, I remember the only two things weren't going to change, or the meteorological file.
2:25:39
And I don't remember, but I'll just to air 40. So Arif 40 was supposed to be.
2:25:49
Oh, yeah, it's OK. And then we're going to change the name of the output file.
2:26:04
Hmm.
2:26:07
Yes.
2:26:09
Oh.
2:26:12
Sorry, OK, before we do like ow change the name back, FF changed already.
2:26:20
OK, so, before we do that, we need to actually calculate the the statistics so we need to go to the utility is convert to datum.
2:26:32
We need to get tea.
2:26:35
Um.
2:26:39
The file with the measurements.
2:26:42
Ah.
2:26:44
And we need to think change I want to we want to be able to find the outputs that we're going to stat A.
2:26:59
Yeah, somebody want to call others.
2:27:02
We're trying seven UW.
2:27:06
So first, we create a datum file.
2:27:08
Can we compute the statistics?
2:27:11
I think that's So there, we have first set of statistics, Um.
2:27:27
OK, and now we're going to, Yeah, Rhonda era, 40.
2:27:45
OK, sure, so we're waiting for that to finish.
2:27:50
Now, I'll just, the last section is about ensemble reduction techniques.
2:27:56
And I'm not gonna, we're not gonna cover that here, but I did want to just share that's a picture from one of my favorite books, which I think it's always kind of sticks with me when I'm thinking about the construction of ensembles. So this is from Statistical Methods and Atmospheric Sciences by Daniel S Whelks.
2:28:18
Um, and this is just a very, you know, simple schematic of the evolution of an ensemble. Say, no.
2:28:26
He's saying, this is kind of like, at the beginning. This might be the initial time.
2:28:31
And he doesn't, This axis is supposedly like the best guess forecast, and the evolution of that, and there's ensemble members.
2:28:40
Say, you know, at the beginning of the time, you might have kind of more tightly clustered because you have less uncertainty. Hopefully, hopefully you've been assimilating your observations. And, you know, you kind of know where things are, but there's just a little bit of uncertainty.
2:28:54
And then as he ran forward in your forecast time, no.
2:28:58
The uncertainty is going to get larger, and hopefully that will be captured by your ensemble members. So you'll get a larger spread in your ensemble, as well, that is correctly capturing the actual uncertainty.
2:29:10
So what can happen in ensembles that they're not constructed well as And they could be they could not spread out as much as they're supposed to.
2:29:20
So they might stay kind of tightly clustered that means they're kind of overconfident.
2:29:26
So you get these results, rather ensembles, Like, well, we're 90% Sure, but then when you actually analyze how well the ensemble is doing, it's not, you know, when it says it's 90% short.
2:29:37
It's actually only 50% probability of that thing occurring. You can also have the opposite where your ensemble spreads out like too much. So you're getting this huge spread on the ensemble.
2:29:49
Um.
2:29:53
There's, you know, the actual uncertainty is smaller. So in that case, Sarah ...
2:29:56
called Under Confident, it thinks there's more uncertainty, then, than there actually is, um, and so there are ways, and we're not gonna get into them but there's there's ways you can kind of evaluate your ensemble, to see how well it's doing to see if it's overconfident or if it's under confident. And if the spread is really capturing the uncertainty well and there are ways in which the ensemble reduction techniques have kind of like curating your ensemble as you go along.
2:30:24
So at each step you might like um Remove some ensemble members that aren't Performing well or if they're kind of I guess what they say has been stunned and like. I guess I'm just curious You know at some point They say this test last part Kind of illustrates What could happen if you raise some kind of bifurcation point in your system. And you know, a lot of your ensemble members go this way. And then a few of them go this way.
2:30:52
And what could happen, in that case, you know, is sometimes, if the actual observations are over here, um, but the ensemble as nicely saying it's over here, then there's, it's over confident that, that just this is the one that occurred. So you might want to actually remove some of these ensemble members, because the probability that that's over here is actually much smaller than the ensemble, is saying.
2:31:20
Let's see, OK, we're done with that one.
2:31:27
And now we're going to yell.
2:31:30
Keep forgetting, right? We gotta go to our datum first.
2:31:32
So, utilities convert to data, going to change this to be.
2:31:48
Error, 40, create the data file, and then compute the statistics, and they're going to be a little bit different from what's here, because they used a few less particles.
2:32:06
OK, let's do our next one.
2:32:13
Next, I'm supposed to be NAR.
2:32:22
There it is.
2:33:12
Alright, I ran pretty fast.
2:33:21
Just now, Dinar.
2:33:38
And now, now I join the other wharf.
2:33:52
Says Our fourth plan.
2:34:05
And.
2:34:13
Yeah, that pretty much brings us to the end of the section after we finish this up.
2:34:29
So hopefully after you've finished, finished this section, you have like a little bit of an idea of, know, how to deal with uncertainty, what you're never gonna get rid of, if there's always gonna be uncertainty and these calculations.
2:34:47
And, you know, how do you communicate that?
2:34:51
How do you create.
2:34:58
Information that end users can, that would be useful, two end users.
2:35:07
All right.
2:35:15
Totally, convert data.
2:35:32
OK, and then the last one thank Esther.
2:35:38
nine kilometer Wharf.
2:35:46
Hey, guess what Changed the name first?
2:36:40
Then, I guess, so, we're gonna break for lunch after this, and then we'll come back.
2:36:48
10 thing will be talking about source attribution methods, And that's usually a very popular topic. So.
2:37:23
Can I sell luck?
2:37:38
Oh, are waiting.
2:37:41
Shankar ...
2:37:41
Forum topics have been going on.
2:37:48
If you guys haven't checked out The Forum, even if you don't have a question to ask, you know, it can be really nice to just go and look and see what other people have been asking about and get some really good questions.
2:38:08
Alright, Brentano with that.
2:38:12
Just need to get one more datum file.
2:38:19
And the Swindlers the Wharf 0 9.
2:38:26
All right.
2:38:28
So, now, if we look at our working directory.
2:38:38
Said half, I should have others, um.
2:38:47
I have a letter.
2:38:48
So here's all our, uh, statistical results for each one of these.
2:38:57
And then we also wanted to look at the mean.
2:39:03
And so, because.
2:39:09
Ah, here we go, Ensemble, OK.
2:39:13
The reason why we're naming them the way we were, I said, that we can use the ... utility to create the mean.
2:39:23
So what we need to do, first is go in here, and change that, just to highest split, too.
2:39:30
So basically, that will transfer that information to this other part of takeaway, said, that, when we want to create the files. So now it has, as input, file has split two. And the comparable utility will look for all the files that start with high split two and then have a dot and then something after them and assume they're all part of the same ensemble.
2:39:55
So I'll create the ensemble files.
2:39:58
Chris, I get an error, there's, like, five minutes left, right?
2:40:03
Let's see.
2:40:05
When a dead.
2:40:08
My working directory.
2:40:11
OK, I'm actually going on.
2:40:13
I'm going to delete all of these.
2:40:15
Let's just start.
2:40:26
Delete all of these.
2:40:31
OK, I'm not trying to do this. I can create files.
2:40:37
five. But let's see if it.
2:40:39
I don't know where that arrow is coming from, but it looks like it created the files.
2:40:43
They're only a kilobyte.
2:40:46
So, couldn't be, there, is nothing in them.
2:40:55
Let's look and say Why don't we hire?
2:40:58
Can you Matt?
2:41:00
So if we look at the mean.
2:41:05
OK, so, probably, um, Yeah.
2:41:15
It didn't create those files properly, probably have zero concentrations on them and that's why it won't match them.
2:41:24
All right. Well, with only a few minutes left.
2:41:30
No, Try to troubleshoot this. And maybe I don't think we'll come back to their spouse. Maybe Mark give you an update after lunch as to what I did wrong.
2:41:43
And then, but what you should do is get the mean of those ensemble things. You can look at the statistics for them, the mean as well.
2:41:51
And in this case, you know, it's a nice example because you're like look, the mean, you know it did really well.
2:41:56
And so that doesn't it's not necessarily the case 'cause the ensemble was just kind of no, kind of randomly put together. But it does show that a lot of times.
2:42:10
It's not always the case, but a lot of times it is the case.
2:42:14
Um, all right, so that concludes I think, this section.
2:42:21
And Mark, I think I'll hand it back to you.
2:42:24
OK, thanks very much, Alice.
2:42:29
Um, don't feel bad.
2:42:32
I think, actually, all of us, that's exactly the same, same problem And it's something that, you should know, any of us can rubbery about. It's easy to click on the wrong thing and I certainly have done it about a million times myself.
2:42:51
And so the point is that Alice explained it extremely well about these ensembles, is that it's a way to examine the uncertainty in your in your results. I'm going to change the presenter back to me.
2:43:07
Yeah.
2:43:09
Let's see, um, and um, there is a lot of uncertainty.
2:43:17
And I split and there's also a lot of uncertainty in, in, in any, any model.
2:43:22
And Alice did a great job of explaining showing you how you can kind of look at the different uncertainty by looking at the different meteorology, looking at different physics, looking at different particle numbers and things like that.
2:43:37
And I think one of the things that you can do when you're doing high split simulations is that often you don't know what parameter to use.
2:43:49
And this is true for any of these like deposition parameters or anything else like that, but you can, you might know the range that it could be and you can then do simulations, um, a couple of different 2 or 3 different simulations.
2:44:05
With numbers in the perimeter.
2:44:09
In that range, and see how big of an error change it makes in your results. If it makes a really big change in your results, and it's a really uncertain parameter, then it's something you should really worried about.
2:44:22
On the other hand, in some cases, um, there's a parameter that you don't really know very well, and yet when you change it, it doesn't make a big difference in your results.
2:44:32
In which case, you know, the results are not that sensitive to that parameter, and you don't have to, no, lose too much sleep over that.
2:44:41
So, it all depends on, you know, what, what parameters, and what inputs and outputs here you're looking at.
2:44:49
But, it's important to basically consider the uncertainty in with high split, as with, with any other model.
2:44:57
OK, so let's go to a break. Let's go for one hour break, and we'll come back at 1 0 PM Eastern Time.
2:45:04
Um, and we will see you that, OK.
2:45:10
Um, bye-bye.
2:45:17
Welcome back from our, quote unquote, lunch break, which may not have been a lunch break for you, depending on what time zone you're in.
2:45:27
And so afternoon, we're going to start with a section on source attribution.
2:45:37
And this section is going to be taught by doctor Chen Thank chai.
2:45:42
Good thing, I am going to change the presentation mode over to you.
2:45:51
Year end.
2:45:52
I'm good.
2:45:53
Turn off my webcam and we can see the screen that you'd like us to see with the tutorials. So you're all set.
2:46:02
And take it away. Can we hear you, see?
2:46:07
Can you, can you talk?
2:46:09
We can't quite hear you yet.
2:46:11
Um, let's see.
2:46:15
Let's see.
2:46:18
Let's see if you're muted.
2:46:21
It doesn't look like you're muted by the system, maybe something with your headset.
2:46:29
Um, not quite yet, not quite yet.
2:46:35
Hmm, hmm.
2:46:38
Um.
2:46:41
We will figure it out.
2:46:42
I'm sure, very shortly, we still can't Here, you, um, Let's see here.
2:46:54
Let me see if I can't Mute you and then unmute you. I just tried that.
2:47:02
Um?
2:47:06
What about clicking on the audio tab in your Yeah, that's what I want to say. OK, I'm afraid is also, thank you Fred.
2:47:16
Clicking on the Audio tab, in your Control panel, um, there could be some Settings thereabout, which microphone you'll be using.
2:47:26
Yeah.
2:47:44
Yeah.
2:47:45
There we go. OK, yes, we can. Yes, OK. All right, Thank you, Fred. Thank you.
2:47:53
OK, I'm going to turn off my Mike intent thing. Take it away.
2:47:59
Sorry about that.
2:48:01
Technical difficulty, and I'm going to talk about the source attribution.
2:48:08
It's a little bit different from the other applications.
2:48:13
But I believe a lot of you how P, you'll see, has faded tax trajectory to try to find the source term.
2:48:22
Not a source term by the source location. I think it's all related to the source attribution.
2:48:29
But because that back trajectory, messer, you, uh, I'll turn off my camera so that.
2:48:40
And you can also estimate the strengths of animation. You can only get the location and also different trajectories. A lot of times, omega U of our different answers is very uncertain. So, running the dispersion model poverty, it's much better choice.
2:48:59
And here, we are going to just introduce several different methods to estimate the sauce. Strands are location and sometimes the temporal variation is it can be a combination of all of those. So, let's go to the sections certain important first.
2:49:20
And here, we should start from the cap tax released to experiment And first cultural.
2:49:29
Set up a run to retrieve the control file cap texts control dot T X T dash will be given day.
2:49:42
to tolerate the drug trade and you'll go tool, Klaus, so you can find the control.
2:49:51
Well, you can save that for the setup fellow. Go to the concentration menu.
2:50:00
Retrieved from the same directory kept house set up.
2:50:09
OK, Save.
2:50:12
Yes, let's reveal all the options here.
2:50:18
And, uh, We're going to shorten the run duration to the log, 19 hours.
2:50:30
As a start time.
2:50:35
Let's see, what they have here is ... icon to file 17.
2:50:44
And this start runtime be reduced to a 19 tyrol, and then we have the Evolve mythologize data's out here.
2:50:57
So.
2:51:05
So, they can run the model.
2:51:09
Just run the model.
2:51:11
You see the setup file.
2:51:15
So, after the model is completed, we are going to, your cell, come more to that, Um, you know, the tail it here to just can't thing.
2:51:24
I didn't save the setup run before you started.
2:51:32
I'm running.
2:51:35
Think. Sorry about that, But I'll call back this.
2:51:40
Run goes pretty fast, so, Yeah, I believe so.
2:51:46
OK.
2:51:57
Can opt in, I guess.
2:51:59
Let's see.
2:52:02
Sorry, I missed you. See, I didn't install that run in a beta of that, that arise, so sorry.
2:52:17
OK, I can stop that.
2:52:20
So go to configuration.
2:52:30
Retread login.
2:52:33
All right.
2:52:37
Yeah, save, and make the run.
2:52:46
OK.
2:52:51
Should be pretty fast, and after that is completed, we can just convert, has to be the results to the location at one station.
2:53:10
Then I'll go into yours of merriment ivory Israel versus call a cab test tool, mayor dot TXT.
2:53:21
A part of that file was renamed to the ... or ...
2:53:26
one and we just wait for the edge run to be completed.
2:53:33
And before that.
2:53:37
Let me summarize. what was going to be dong in this experiment.
2:53:42
This one is just running there, the Model four.
2:53:47
And, uh, so, the ... it's going to give you that ocean factor from the source location.
2:53:55
So, if you can do a cathode relationship between the concentration and uh emission, you can use that dilution factor tool infer your emission rate.
2:54:14
If you know the the concentration, the concentration or from the measurement, like the all our measurement we are going to use.
2:54:26
So the basic idea is they are like a salsa receptor relationship between the source is the emission and the receptor of the station. We measure the concentrations.
2:54:46
So that's all very, like, IE.
2:54:49
If you only have one merriment, you can just use the dilution factor and divided that by the dilution factor, the measurement bandwidth, that origin factor, then you can get the emission rate.
2:55:12
Or the run is not as.
2:55:18
It's not as fast as I hoped, kind of stuck.
2:55:35
So, OK, I see me reset to make it run again.
2:55:58
Um.
2:55:59
Make sure to set the runtime to 19.
2:56:03
Yeah, I think, yeah. Good, OK?
2:56:07
Oh, OK. That, Yeah.
2:56:23
That's only a change we need to do, Right?
2:56:27
And run the model, OK.
2:56:49
Well, assuming that the rank is OK, then they are going to.
2:56:55
OK, is saying that right now the speed is about right.
2:57:02
So.
2:57:07
And the way we can use our has sped the simulation results to to get the concentration valueless at the measurements size.
2:57:17
And by using the ...
2:57:18
Converter Station, you can get all the information you need.
2:57:25
And here, the, the master we are going to use is to compare Average Mayor, the concentration, and the average calculated the concentration.
2:57:39
And they can use that to adjust a emission rate.
2:57:55
So, well, the rank is still cooling.
2:58:00
Can you see?
2:58:04
Let's say, ... here is actually the insanity animation.
2:58:12
67 kilogram per hour.
2:58:16
So, suppose the ratio of CAD tel 1 to 1, racial between the average concentration and outrage from the measurement, and from the segregation.
2:58:29
But below, what we actually get is, it's about three times difference.
2:58:34
So, that means, using the same poll: racial, poverty is not going to give you the best result, because that's prefacing verified.
2:58:44
And another thing is about, uh, this method is basically adding all the measurements together, and then T the average, so the large value is always going to be dominant.
2:59:06
So right now is seven papers around the 80% complete?
2:59:35
Yeah.
2:59:46
OK, OK, so here we follow the procedure here and converted to the Datum, yes.
3:00:05
Utility convert to home.
3:00:09
And the first we need to find the, sorry, I already media measurement data, It's going to be in ... directory.
3:00:26
And here are the three all of our measurement data.
3:00:32
So, create that data file and then you can compute the staircase status between the model results and the measurements.
3:00:53
Oh, here, I made a mistake here, It's, uh, should use the conversion factor because self stagg, ..., PICO Gram, 1.2.
3:01:11
And to the coercion.
3:01:21
So, here is a little bit different results out. I'll tell you, you can say, this is a factor of three.
3:01:30
Close to our factor of 2.5.
3:01:32
So, that means, uh, evaluating this measure, poverty jaco until all are hazmat are the emission.
3:01:42
And.
3:01:51
So the next rung, whale, kowloon, too.
3:01:59
To do it is. Great.
3:02:03
Yeah.
3:02:10
We are going to like a change, share, the starting time, and trying to modify this around a little bit, and trying to see how the results are going to be.
3:02:22
Burst, set up the runs.
3:02:27
And they change the Starting Timeframes, 17, UTC, 2, 15, UTC.
3:02:39
And two houses, same ending time, the runtime is going to be $21.
3:02:51
So, that's only two changes we're going to make, and they need to call to the pollutant emission menu and change the emission rate from the rational 1 to 1, that's a unit, a mission.
3:03:12
And we also need to change the ... from three hour to 21.
3:03:19
Save that, save it, save the changes, and run the model again, OK?
3:03:36
So nice assume is coined to run as a spectator. Then we are going to get seminar statistics.
3:03:48
And from there, poverty, we are going to **** a better, actually, a better estimate of the aviation.
3:03:57
Uh.
3:03:59
So, the average measure, the concentration that's changed is still two sultans, 300 around that value by the, because we change the aviation to you'll need a nation.
3:04:12
And, uh, the emission duration is also change this hill, the hospital simulation results are going to change.
3:04:26
And by applying the same racial to the national emission rate, you are going to probably get a badger, a nation estimate.
3:04:36
Let's see whether that's a case for this run.
3:04:47
Yeah.
3:04:49
ASHRAE this the results, he is not suggesting you how to make such changes because the first roommate didn't pay yourself a sat does release time three hours.
3:05:04
And the the second neuron is X rays, is the temporal information is.
3:05:14
he's arbitrary off OK, less to the same notated, hey, come were to a datum, and we already have the ...
3:05:26
files in the same conversion rate and we can create data compared to statistics.
3:05:39
So it's, it's about like a two solids, it doesn't change, but this one is zero point zero four the average concentration, and based on this one, what we can get here is.
3:05:58
So use the calculator.
3:06:05
Tools, oh, sorry, divided by points, say all four.
3:06:14
So, for this runway, we get 57 kilogram per hour.
3:06:23
So that is a little bit different from the example here, I don't know the reason for the difference, but it's close. And.
3:06:36
So, there's been saw, this method is, The parade nod, not good. Because, when we have the eighth alcatel temporarily information, we get worse results than the other one. Is a different, 24 hour release, is different from the actual release time?
3:07:02
OK, it's just introduced the very crude racial, messer, to SMEs, aviation.
3:07:12
So that's something to think about. This is a long way to go.
3:07:18
How law crude estimate but is not good.
3:07:22
And that's up.
3:07:25
Fall follow this saves that because the results are not exactly the same. So I'm not going to say because I have previously saved those.
3:07:37
And, uh.
3:07:41
Scott thank.
3:07:44
Somebody pointed out in the chat that maybe you used a different interpellation and nearest neighbor rather than bilinear interpolation in the data menu. That's a possibility why it was a little bit different. Not a big deal.
3:07:58
Hmm, hmm, hmm, hmm, OK, that might have been the different eras, labor, OK, that's possible, but OK, oh, that's cool, but it's not important, but that's why you got a little bit different.
3:08:09
Yeah, well, that's a pregnancy. Yeah, thank you for pointing that out.
3:08:15
Thanks to the person that. That's right.
3:08:20
Yeah, that's really helpful, and yeah, a lot of times.
3:08:25
I know, it's so, you, You need to follow. Exactly. Otherwise, you're going to look at that a little bit, different results, and all the parameters that are going to have some effect for the results.
3:08:38
And, uh, so.
3:08:44
Close, Harry soon.
3:08:46
They may just, uh, OK, so, maybe just go save it, stat.
3:08:57
Sue? Us.
3:09:01
Yeah.
3:09:04
Well, I think, uh, a parish save it to the husband, working directory.
3:09:21
Yes.
3:09:23
Sounds forward.
3:09:28
Cho.
3:09:31
Test.
3:09:35
Same name.
3:09:45
Yes.
3:09:48
OK.
3:10:05
OK?
3:10:12
OK, for this section is Backward ... Forward.
3:10:17
Because there's transport model is actually a linear, so is if you run the model backward, it should give you the same result, same dilution factor.
3:10:34
And what do you need to do?
3:10:36
It's just change the data for the model run. Just change the duration to enact our time.
3:10:44
So, as a model, well, Ryan, from the end the time to look backwards for the ..., specify it and here is just to echo trying to run of our around and around trying to see whether the dilution factor arbitrary.
3:11:05
Same and we are going to see.
3:11:11
And the first retrieving the prairie saved because of it didn't change the onus on you to do that here And, uh, this is trying to set up.
3:11:40
So I'm going to start from the 17 hour, 17 UTC.
3:11:52
And the leaks brand duration is plentiful our ton of ours.
3:12:00
For the pollutant emission.
3:12:06
Is still with the emission rate as Yonatan ....
3:12:12
By the way, change the aul ourself, a mesh into one over here, save for the concentration grade.
3:12:26
Now we are going to make the change is the average hour from Sri. Sri, all of our tool every one hour.
3:12:42
OK, it's uh.
3:12:48
And we also need to, change is the sampling.
3:12:55
Star time to 17 started from over 70.
3:13:05
Save a ****.
3:13:08
OK Sou, Sou, and uh save to run the model.
3:13:26
And while the run is going, we're going to just check what we need to do next.
3:13:35
Is similar to the previous section by the V here of a ... Tool, just a line station.
3:13:44
So, stations stray 1 6.
3:13:47
So, instead of log, can learn to datum. I'm going to yourself ...
3:13:53
station and the V T input the station a longitude latitude information.
3:14:05
Yes.
3:14:11
OK.
3:14:17
Yeah, Can a lot of times, I think I, Had a, People may forget to Do, just change the Conversion factor.
3:14:30
And You are going to a, if you don't change it, because the ... de France, you can get very bad statistics.
3:14:37
And the value is going to be off by more order of magnitude.
3:14:47
OK, the last step.
3:15:03
So after you, you'll gather Simulation Dong, and you'll you'll sell your ... station. You are going to get attempts serious at that station.
3:15:13
And here's, uh, the text file for that tempts areas.
3:15:20
And you can have the column off on.
3:15:23
The first is the fractional tase and here's the year month, day over information and that's the ... starting time and then that's very common.
3:15:39
Von ...
3:15:40
for Column was uh, Nick, the ending time and the lasso, Larry last column. It's, uh.
3:15:50
Duh, predicted concentration.
3:15:56
At the location?
3:15:59
And they lack of time in the role, different roles represented different time.
3:16:14
OK?
3:16:17
And after this runs down, we're going to convert, the model run from the forward or unclog, the backward or run?
3:16:27
and that backward, Ryan's going to star to this station and we are going to check, sorry.
3:16:37
The dilution factor at the source location.
3:16:40
So it's now going into account correspond to each other is.
3:16:48
The follow the running is we are going to look at the dilution factor from the source location to the station and the backward raise the star from the station, and we're going to check that that ocean fact as a source location.
3:17:06
In Syria, the dilution factor should be all same.
3:17:15
So, their house, I'd run almost known.
3:17:25
95%.
3:17:30
Yes, OK.
3:17:33
Then Go Utilities, come learn to Station.
3:17:42
And OK.
3:17:45
Here it changed the conversion factor.
3:17:47
two: training tools, 15.
3:17:52
The HRC input.
3:17:55
Nice.
3:17:57
Sorry, 1 6. 1, comma 3. -82, 2.
3:18:10
See the other things we need to change.
3:18:20
OK, so, we can extract data.
3:18:26
..., as expected, seminar to this survey upon the 7 and 7 peak value from 72 AC on September 26, so that's similar to the value of C in the tutorial.
3:18:49
And they can plot the data.
3:19:07
OK, then we need to configure the wheeler's calculation, Open the set up.
3:19:17
Can say the start time till, that the peak value will ...
3:19:22
time is, uh, it is 30926.
3:19:31
Going to be 0 8.
3:19:36
I'll run through iterations, because that's another, and in time of the far around.
3:19:41
So that's the total runtime is going to be reduced front till 15 or so that's going to the, the starting time of the father. Brian.
3:19:56
And starting location is going to be the station.
3:20:02
3 and one sakes, claimed to be 14. I'm sorry.
3:20:10
My nearest payday tool on to two, OK.
3:20:20
Sure.
3:20:22
Right.
3:20:25
And we need to click on the iterative back radial button. So the runtime is going to be inactive care menu. We're going to run a backlog dispersion model.
3:20:41
Here.
3:20:44
Yeah, 15 hours.
3:20:49
Starting location, they made that change, and change the starting location to that, and the concentration grid.
3:21:05
Reverse the assembly and star time and the start time.
3:21:09
It says, start time of sampling, 2, 26, 0, 8.
3:21:21
And, the sampling stop time had 0, 9, 25, 17.
3:21:35
See ...
3:21:39
of salmon in tiny says, hero, hate, OK?
3:21:45
Sink.
3:21:51
Save it, and run the model.
3:21:59
It should be a little bit faster, because the runtime reduction to 15 hours, OK, so after that it's dong, we're going to use the same Converter Station utility, to get the attempt serious as a source location.
3:22:21
Then we're going to compare the dilution factor using the forward and backward run to see whether the same dilution factor from the source location at that time source release time, and the station as a measurement of time at the peak time, to see whether television factor. Same.
3:23:03
Yes.
3:23:19
73%.
3:23:34
So, that's revealed the forward crowd here.
3:23:37
We have the peak value and this station is 3 1 6 as a surrogate upon 7 7, and knowledge drive from the station straight line six going back.
3:23:57
And the ending time, delusion Saboteur, should be L same. But actually what we got is different.
3:24:06
Now, see whether it be, have the same results here.
3:24:14
OK, can we use the utility converter to station?
3:24:24
And, again, put a social location.
3:24:29
You rename it, too, Bank.
3:24:35
Been named as some location, as, as 0 0 1.
3:24:41
Certainly 9.9.
3:24:46
For extra data.
3:24:54
OK, so, it's, uh.
3:24:58
Same as before, as was shown in the tutorial is the ending time is no point six salmon so they're self ....
3:25:09
Sorry, they're not exactly same, but if you look at the peak value, they are pretty similar.
3:25:17
So that means though, there might be a little bit of time off.
3:25:23
And the forward or backward in theory, you should get the same dilution factor, but because of the the numerical scheme for the forward and backward, they are going to cause some differences.
3:25:44
So we can also plot.
3:25:48
Theater.
3:25:53
Actually probably ratio compares supplemental data between the two.
3:25:57
Here is just Eritrean Turtle, Canada dilution Factor, uh, from the Forward and Back or to compare them.
3:26:09
Close, create.
3:26:18
I think our original that I had hurried, toss up out. The reason for why we didn't get a exact same dilution factor from the ..., There are a lot of reasons for their numerical schemes. And I think they're probably going to be major reason.
3:26:38
And they are on paper, you're discussing about the difference between a forward around the backwards. If we'll have time poverty, you can have a look off that.
3:26:50
But there are maybe a lot of different reasons factors have play to cause that difference.
3:27:00
OK, so, we move forward to the next section.
3:27:06
The previous toy examples: waiting all day, exotica source location, and we have the station measurements.
3:27:17
But if it, A lot of times, survey, how, there is some, ..., we don't know the exact location.
3:27:27
So there are some Mazur to do that here.
3:27:31
Uh.
3:27:35
Gary, just like, let's make the ramp first before we discuss a little bit.
3:27:45
For example, they are going to use as we all were duration sampling data, and that's shortell non-zero Values.
3:27:53
First step is to configure the menu to predict the concentration if the source location, or Gnome cellular Concentrations setup menu.
3:28:09
Says, Star time too.
3:28:12
It is 32 year old 9 0 1.
3:28:19
So note that here from here we are going to generate not the OSA ASAP.
3:28:29
Cap tax measurement data. And this time period is told, what's chosen because of the interesting mythology conditions.
3:28:42
So.
3:28:48
So, we are going to change that tool, The beginning of September is going to be ...
3:28:56
0901 and, uh, call back to the forward rice 60.
3:29:11
Can they also new to the change of meteorology file?
3:29:23
And change that to 60 hours.
3:29:27
Yeah.
3:29:32
Going to clear and add metadata to.
3:29:39
... should appear.
3:29:47
That's a global data, like a Fall, 19 eighties. Re sample, a whole month, 0 9.
3:29:58
And then they go to the ... Menu.
3:30:11
Said the emission to tZERO, because thus, we're going to **** a pre process our staff to fulfill your mission.
3:30:24
So don't worry about that part.
3:30:28
P replace, Elvis up, Preprocessing step.
3:30:36
Terrill.
3:30:45
Say 10 thing.
3:30:47
You have to set the Set the emissions of hours of admissions to 60.
3:30:51
I think, OK?
3:31:01
OK?
3:31:03
Yeah, I think, Yeah, I Need to set the emissions start. Time as well.
3:31:09
Sorry, I may be wrong about that.
3:31:13
Mmm hmm.
3:31:15
83 oh 9 oh 1, I think at the beginning of the month.
3:31:22
Maybe it's OK with just the zeros though.
3:31:25
Hm, hm, hm.
3:31:26
I Yeah, that's yeah That's starting to erode perceive whether it's going to change.
3:31:36
So the grid is going to be is centered around the fault here 1. 73, West.
3:31:53
Can tear up onto file, tear up onto a file, can define when you called profile name as seen on.
3:32:16
OK?
3:32:20
OK, I think, Mark, you're right. And I probably need to change that, too.
3:32:26
Seeing on, OK, let's see.
3:32:34
They're all one.
3:32:44
Tero.
3:32:47
It is 0903.
3:32:53
Wow.
3:32:56
And changed already, I mean, average, 2 or 3 hours a day?
3:33:06
K is two things, uptime, sorry, Look at the sampling stop time.
3:33:15
There you go.
3:33:19
OK.
3:33:23
Thank you.
3:33:25
And the recall back to the Emission one here is ...
3:33:31
and, uh, it is three Sarah and I might be the same as the starting time, but it's better to be exact as to the ...
3:33:49
sale sale.
3:33:54
Because we'll be running the model is coarser resolution about going to.
3:34:03
To change the particles, we are going to use goto Advanced Configurations, that off, save it first, Configuration and Setup Concentration, and go Tool, Manual for.
3:34:22
And change the Parcel number, reduce that to full sell 5000 to speed up the calculation.
3:34:33
Save.
3:34:36
Sue?
3:34:39
Null, we're going to press the Special Run.
3:34:45
Geolocation.
3:34:52
And false, sad.
3:34:57
I'm going to import.
3:35:01
The measurement data is going to be a hybrid set called the ... for this test.
3:35:09
Broths kowloon to be in a cap tax directory.
3:35:17
And here's our hurdle matters. Got Cascade.
3:35:24
And similar shades ...
3:35:27
first today, choose the numerator here.
3:35:37
I'll explain later on to make the run the model.
3:35:44
OK, Truong, hypothetical Environment Data Numerator, and X a kilt.
3:36:04
Complete contain you.
3:36:09
So it's going to generate 29 control files because of the amendment has been denied measurements.
3:36:18
I'll be 42.
3:36:20
To be can go to the working directory to see, yes, we're house. Picker control files.
3:36:29
That's just alternating.
3:36:34
No?
3:36:39
So let me see.
3:36:40
Yeah, Now, show you that backwards and from the location.
3:36:50
OK.
3:36:56
All right.
3:36:59
So, look backwards.
3:37:05
So, they are going to run that Expression model, ... Care.
3:37:16
It's a ...
3:37:22
finish line each They are going to be a 29 locations OK, so one way of 18 I'll explain a little bit about this geolocation, right.
3:37:56
No, So on the right, you'll see a text file.
3:38:02
Click here, that's the habits, ethical measurements data we used for the title of our generation.
3:38:11
So, you can see, is seen the data format, you have the year, month, day, and it's starting all over duration, latitude longitude, off, the amendment location, and you also have the Merit Measurement and Value.
3:38:25
And the last column is a station ID for those.
3:38:30
And, here, the reason we call that, a hybrid sylacauga, because, we ran a model for that way.
3:38:39
Just a second poll tool, to see you, lots of merit and merriment, according to the fourth.
3:38:46
Uh, the concentration predicted this way. And, this here, for this one, the exact.
3:38:56
Release rate is going to be 300 Sultan ..., plus three kilo Gram per hour.
3:39:08
Yeah.
3:39:10
The Maryland, after the pre processing for each measurement is going to start sink.
3:39:21
Each measurement is going to make a unique run, Trump's America and location backcourt.
3:39:30
So the first round, remember me, your numerator.
3:39:36
That means we actually use the measurement value as a source strengths, totals up backward runs, and later on, we're going to change that to the denominator.
3:39:50
So, that means. No more ..., right?
3:40:02
OK, still waiting, OK, OK, Let's go back to a tutorial here.
3:40:18
So, as Tara said, by the preparation, complete Bill Clinton to replace the ...
3:40:27
default, stand for, they still have that, Millennial All Parents Inc.
3:40:37
So, welcome to the ... S 20 day.
3:40:45
It's because we have some something to be every three or so.
3:40:53
... Gates, Clinton means that we include all the time-stamps.
3:40:58
Take all the sampling three hour, the whole duration of six hail verse, and they put them together.
3:41:15
It's Crystal, a campaign. Your microphone is. Your microphone is not coming through. Perfectly, I wonder if you can adjust it.
3:41:25
A tiny bit is very good at the beginning, and then, in the last 5 or 10 minutes, it started to get very not as good. I'm not quite sure why.
3:41:37
See, OK, Is that better? Yesterdays, thank you.
3:41:43
Millimeter.
3:41:46
Yeah.
3:41:56
OK.
3:41:59
OK, It's run. It's a little bit slower.
3:42:12
See? Either way.
3:42:17
It's all generated in the working directory.
3:42:34
Yeah.
3:42:47
Lost could be running still 14.
3:42:54
Mmm hmm.
3:43:10
Totally, OK.
3:43:15
Wonder why it's so slow?
3:43:43
OK, I think Maybe after this run is dong.
3:43:50
Scapes the next day it will be ..., because it's the ... of the same procedure.
3:43:59
I'll just present the results here.
3:44:06
See?
3:44:19
OK, 16.
3:44:28
Yes.
3:44:38
Maybe, yeah.
3:44:42
Can I save the particle after I reduce the particle numbers you made that didn't save it?
3:44:50
OK, you might have retrieved it again, and maybe you didn't change it the second time, I'm not sure.
3:44:59
Um, it could just show it also shows up like that, just show the results and the ....
3:45:10
What actually happens here it's half terror off their operations done.
3:45:17
And we can aggregate all the mm 29, it's also together and, uh, after the X Q, we can get the contours that baldi.
3:45:32
They are going to be combined together and we can plot the contours.
3:45:38
See, here we are going to get the results of the first order for the rights of all the numerator runs. And you can pretty much identify the high values here.
3:45:54
It's that's the actuary, not presume the source location because the ... location to generate the measurements.
3:46:04
If they change that ensemble threatens to you saw the first of the ...
3:46:12
going to get another settled runs and actually this run.
3:46:20
The purpose of this right is a little bit different it's because this one.
3:46:28
It looks at where the use of merriment workshop, but actually if you look at the equation emission rate is going to be proportional to Merriman or dilution factor.
3:46:43
And, uh, so, if you use the dilution factor divided by the measurement, you are going to add up.
3:46:56
The EMR sulfate automation, so, data using the ...
3:47:01
and adding all everything together, you are going to look at the estimate of the EMR software donation.
3:47:09
And, actually, the results show that we can estimate the emission rate today, sway and it's going to be 1 Charlatans 700, not too much different. From the solid an Intel, we expect to get the exact solution.
3:47:27
And the first one is kind of how you look at the source location.
3:47:31
Because, OK.
3:47:36
It has high valuables of the environment has been motivating them. They are there, but if you add them together, it's not going to be make some physical sense for the emission rate.
3:47:52
OK.
3:47:55
I'll stop the run here, Maybe I'll just say Reset.
3:48:37
OK.
3:49:20
So, let us continue with the next section, OK.
3:49:29
Maybe before.
3:49:31
Yeah.
3:49:39
Before lead to an ischemic does reset Emmanuel here.
3:49:49
We're a trailer and save the location control.
3:49:59
So set up, very true.
3:50:05
Close.
3:50:18
From the tutorial, OK?
3:50:24
Files, Geolocation, Control.
3:50:40
OK?
3:50:51
Yeah.
3:50:58
Yeah.
3:51:00
OK.
3:51:03
Go to Setup Menu, True.
3:51:16
Set up.
3:51:20
OK, so.
3:51:26
Yeah.
3:51:30
Notice with you.
3:51:38
So the model is going to run a star from September first.
3:51:45
For duration of six hours using the Global Data RP 98309.
3:51:53
And uh, follow the same approach, Atlas, the trajectory matrix, open source location menu.
3:52:05
Change it to astray, OK.
3:52:12
How's the first line to pay 40?
3:52:23
, it's going to be a lower left corner of the domain.
3:52:31
Then the second day is Opera right off day.
3:52:37
Amen!
3:52:46
45 miners seven day.
3:52:51
And the third location, colon today find the resolution of the upgrade.
3:52:58
It's going to be zero point five degree.
3:53:02
This is going to be a 7.5, just tell Terraform file to pull it right off the first location.
3:53:15
See location menu.
3:53:21
Send him.
3:53:25
OK.
3:53:31
OK?
3:53:35
No concentration inquiry?
3:53:40
It's, uh, because somebody that's starting time zero Y, zero Z.
3:53:48
Too low tariffs trade policy.
3:53:54
Can innovate change the output file to source receptor matrix?
3:54:14
Yes, is it check?
3:54:24
Yes?
3:54:30
An old man knew.
3:54:35
All right.
3:54:37
Change that back to one, UNH ... should be fine.
3:54:49
two Say, so now apparently the concentration selloff.
3:55:04
Commercial multiverse, caught hill, manual number 10 and check the.
3:55:16
Second radio parsing is going to restructure the concentration credit to their source receptor format.
3:55:31
So.
3:55:34
See, that's all a change, OK, And also openly submenu four right now, a change that to 500 Sultan particles because we're going to how law.
3:56:01
Got a lot of different rights from 187 locations, so here is ..., change to 300,000.
3:56:14
I'm going to kill the particles of their cert. Hey, all verse.
3:56:22
five: hundred thousand To 200,000, yes.
3:56:30
OK, Sue?
3:56:35
Yeah?
3:56:38
Still, going to yours as fashoda Matrix, right?
3:56:57
I'm just going to set up Continue.
3:57:04
Please rise, little bit slow. Hopefully, we can get it done.
3:57:14
Want to run completes, going to your targets, Clay? And the share.
3:57:19
I think we should take all the participants, it should, and all that original release.
3:57:36
OK, I think for this ...
3:57:39
path for the next section, you'll need to download a fixed style me.
3:57:49
It's executable file in order to those they sell this section starting on fire.
3:57:58
For this, uh, software receptor matrix, Rhonda, the exception property is not true, even if you didn't update that executable file.
3:58:13
OK, let's go back to the tutorial.
3:58:24
OK?
3:58:29
Yeah.
3:58:33
Yeah.
3:58:35
The matrix right here. Actually, they find the domain care, and they have a resolution, I'll say, on the file.
3:58:44
Yeah, this is going into a generator 17 times 11, it close to the 187 sauce locations.
3:58:56
Because, for this application will assume we don't know the exact location of the release.
3:59:04
Salvia, just let it run from each location on your ....
3:59:12
There they are able to generate, uh, TCM matrix for each role you will help.
3:59:24
The results are from one: you'll need the right soft ...
3:59:30
salad and therefore, the each receptor the receptor means the event, the location at certain time, correct cart.
3:59:45
The results from the issue you'll need to run, and then you're going to save all the rights to a single file that a single file, sealed the sub matrix.
3:59:56
And that matrix, you can extract the results from a particular source location.
4:00:07
And, uh, after extracting that line, you can use your ... program such as the ... plot to cat.
4:00:20
The concentration contours, and you can also look at the receptor wheel.
4:00:27
That means if you use the same quality program, it is not a concentration, you can see it's actually give you all the information from each source.
4:00:41
Well, the location: each location is a potential source for that receptor soliz calcio contribution toward the receptor: Each source. And you can use that information for the self attribution applications.
4:01:07
So after that run install, they can just use your today to track the source bill, ....
4:01:17
And for the death matrix, it is going to be.
4:01:23
The way that you can gather all the source information and all the receptor.
4:01:39
Yeah.
4:01:47
Today's session is Connecticut, as an exception, and that exception is a continuation for this one.
4:01:56
OK.
4:01:58
Go To.
4:02:12
So, after we get all the results of the kind of trade. Because believe me how the ...
4:02:20
available, and the way you can compare, think of all the different fronts.
4:02:32
Calculate statistics to see is also compared to from one source location to the other and if the one source location has the best statistics.
4:02:50
And then, it's unlikely to be the real sauce vocation.
4:03:05
As I mentioned, you'll have to update your staff May, therefore, this step to generate this. Start to say a plot, or.
4:03:52
Yeah.
4:03:59
Yeah, Well, they are waiting.
4:04:06
Before it to the next, Run Stubby can make a call back to reveal what we did previously.
4:04:15
Paul Saxman, sort of important, one, we actually just a single model or a single model right. And we compared the average calculated the average American to too cruel to estimation for the emissions.
4:04:32
And the next one, next section, searching plus three.
4:04:39
We start the run this backward runs from the location and time.
4:04:46
Have two options to make that runs live. Have the measurements fabulous? No, return?
4:04:53
And this way, we can estimate the release location.
4:04:58
If they use the ...
4:05:01
measurements, that means the measurement says, denominator.
4:05:08
They put all the rest together, they can average share to catalog the emission rate pretty accurately.
4:05:16
Think I would break as 1700 gram Oliver and the latter released recently Accessories Holton Slowly Subleasing, a factor of two.
4:05:29
And for the matrix runs working right now, me, how 187 and 4 to rise up from a matrix of different source locations.
4:05:43
And we use statistics between that model runs from each location, and the observations, the measurements as field observations.
4:05:57
They can use data to find the best statistics to infer the likely release location.
4:06:05
Actually, they found that that is a pretty good way to estimate a dislocation.
4:06:13
If we already found the release location, we can use.
4:06:18
That particular reduced run because here we have 187 forest ..., a dislocation line, one of the 187, because you'll see that around rates also still use the racial master class emission rate.
4:06:40
It's also not bad.
4:06:46
So.
4:06:48
OK, Sir, Price of credit is below.
4:06:58
So here, Hopefully.
4:07:15
Well, actually, I, for the next two sections, we're going to use.
4:07:23
Different approach is, OK.
4:07:29
Yeah, Sections 13.6, it's going to like, have six different rice fields.
4:07:42
Different release time.
4:07:45
Each release segment is 12 hours, can feel held the first round.
4:07:53
Thank all the. Time, all our cell phone, followed by.
4:07:59
Another two hours after that, we are going to help fight with a friend, You'll need Atomization runs and similar to the Matrix run, we can construct a matrix, recall that the coefficient Matrix, all transfer coefficient Matrix. So OK.
4:08:24
Take care of a different column and all.
4:08:31
Is it's kind of a matrix transpose here and each column represents.
4:08:41
Thank you. ...
4:08:43
results for column one is from number one, and the column two is thrown round number two.
4:08:54
And the issue rolled represent the prediction from that run to a certain location at a certain time measurement.
4:09:05
And the last column is the measurement results.
4:09:12
So Pat, I mean, all those run so you can get the coefficient Matrix.
4:09:19
And one easy way to estimate the source is, uh, To use the next singular value decomposition Messer because of a relationship between the dilution factor and the source hmm, it's a matrix operation.
4:09:41
You can get the receptor, and they, you know, the receptor, all the measurements and the Iranian, the highest frame, although, you have the aleutian factors, and then you can, it becomes all jabbar problems.
4:10:00
So you can solve yourself location, but it's not a square matrix.
4:10:06
So you'll have to yourself singular value to conversational master total estimator your source, OK?
4:10:19
So the difference between this, and the matrix primary running is here we are trying to resolve the temporal variation of the animations.
4:10:38
After every ... sales Singular Value decomposition massacre.
4:10:46
A better way to solve this problem is actually using our cost function to minimize, I'll call its function.
4:10:54
You can gather the source estimate.
4:11:01
Here, it's all hall, full, that column. You can define your cloud function.
4:11:13
Yes, represent the source at each time segment and you know, order to make the problem. Lao pull a spell.
4:11:26
So you need to, how well prior information, ourselves from the patient theory.
4:11:31
And they feel our prayer, how our prayer, estimate, and how to use that information.
4:11:39
And it's going to be retained by the uncertainty of the prior information.
4:11:45
And then you'll have all the information from the ....
4:11:49
Here represent a measurement into all the receptor.
4:11:54
The seal for a script, H, represents the highest paid model, and Always, observation, and then also going to be how to consider the following real application: you have to consider the sovereignty of some element.
4:12:13
So, here U T is a variation of Merriment certainty.
4:12:24
And, uh.
4:12:29
Tied to the minimization. You can use a different list of mathematical function, and the, like. Can get better source term, estimated using, that is called function, approach, than the thing. About it with a conversation.
4:12:57
All right, so long.
4:13:17
It can thing, I wonder if I wonder if you had some other runs going, um, some, some earlier sections, and you need to kill the high split executable, not just the wish application, I don't know. Maybe some difference there?
4:13:43
I can differentiate wrong, they lie along that lines? Right, right.
4:13:52
Because they only show up as sublease applications, I see, OK, OK. Nevermind, Sorry.
4:14:04
It used to be they showed up as ICS standardized to show up as the executable, and now they don't. Thank you, Windows. Open. Sorry.
4:14:19
Millimeter, OK, so.
4:14:24
So, I'll go through the tutorial. And just to.
4:14:30
maybe kind of goes through all the steps, but L.
4:14:37
It also may have all the results, and goes through all the steps.
4:14:46
Sure.
4:14:50
So, that's the explanation for the trick.
4:14:55
For the matrix I talk about, about trying to generate and, uh, care, R, set out all for this run.
4:15:13
Structure.
4:15:20
OK?
4:15:40
OK, I think ... receptor.
4:15:45
You'll probably know how to do that part.
4:15:50
Want to know how the results are, You can just follow the next steps.
4:16:00
To extract the result from a single source all history, the result of the truss single receptor and, uh, so let us go to the next one to see all the manuals you are probably going to see?
4:16:28
And that's why it's a client or process.
4:16:34
Strategic Soul.
4:16:38
After you get a 137 different simulation results, you are going to process that data.
4:16:52
And you can plot the correlation from each source.
4:16:57
Share you can get the best correlation coefficient from dislocation that saturate data the ... location you should get.
4:17:11
And they feel, only look at this, is the simulation result from this location?
4:17:16
So you can use you can use the racial mentioned, before the aberration barrel concentration and the average calculated concentration.
4:17:34
And to gather the emission explanation for this one, we can calibrate equally result in 1700 compared to the actual ones resultant.
4:17:46
Yeah.
4:17:50
This last inlet tatar is shown scatter plots. It means.
4:17:59
If you want to plot the scatter plot, so you can see the results are pretty cool, they feel that just ..., like five times the emission rate, you do, you are probably going to get worse space all total.
4:18:17
The the average concentration and merriment until you get to to one by increasingly automation, but you'll probably overestimate at ... locations.
4:18:30
So that's just what it says is using. The ratio between the ...
4:18:35
is also reliable, Messer for the South Hershey tuition so poverty, you need to model a way to do that.
4:18:52
OK.
4:18:54
Yeah.
4:18:57
Or solving the coefficient matrix. That's what I just mentioned here.
4:19:04
It's, uh, we're going to rely on him the model from a different, uh, it's different to release a segment.
4:19:17
So you'll see on the data we can estimate the temporal variation, because assuming we don't know the exact release. So, they, haven't ... generated, is the coefficient matrix.
4:19:36
A simple way to deal with, the final result, is, you will see a similar conversation for this tutorial.
4:19:50
Yeah.
4:19:52
Has paid for the GUI interface, with you, the option tool to run this.
4:20:06
Here is how you set up the different brands, such as the Cable User Control Tool to start calculation and tell all of our software, each other and the short end, each duration, by title hours.
4:20:27
And after that, is done.
4:20:34
Similar to before, we'll do a generator matrix, and, uh, dimitrios aids claimed to be saved, no Excel format, so that you can heal us the final results, too.
4:20:55
This?
4:21:04
OK, I think I'll pop or stop this and show you how to tool the other parts maybe house the results already.
4:21:16
Let's try this, because it's almost impossible to Wait until the rhinestones.
4:21:33
Sorry about that.
4:21:37
Different problems.
4:21:42
OK?
4:21:59
I can think there is a section in the tutorial called Results, and organized by the section, and I, we've had this issue.
4:22:12
It's a common issue in previous workshops as well, so, um, possible, To find the results, there, but. Yeah. I think. I try to gather results to see him.
4:22:39
OK, so be housed.
4:22:45
Started in palm full care.
4:22:50
Please remember.
4:22:58
Yeah, and there are results for Section 13.4, and also 13.6 in the tutorial results, uh, directory, OK.
4:23:55
OK, let's try to just leave every kind of useful result.
4:24:04
OK?
4:24:18
So, what I'm trying to do here is, I'm trying to copy the results to the working directory, so that we can continue.
4:24:43
Copyright.
4:25:17
OK?
4:25:20
Yes.
4:25:23
You did the forward one.
4:25:26
Thinking about it, like, if we can use this, we can actually extract both of them from that, OK?
4:25:35
I'll try first.
4:25:46
So Here, I see Sue, and, uh.
4:26:03
OK?
4:26:09
So assuming that the run is already down can be, we can use.
4:26:25
Display.
4:26:28
Sorceress, after you.
4:26:32
Menu.
4:26:36
Yeah.
4:26:38
OK, here's how to retrieve the location or control file, and that's helpful, too.
4:26:50
Chuck.
4:26:53
To have the control the following information?
4:26:57
Yeah?
4:27:03
Yes?
4:27:07
Yes.
4:27:29
Yeah.
4:27:37
OK.
4:27:40
That's true whether that is going to work.
4:27:46
OK, here, So There's a menu.
4:27:56
Uh, OK.
4:28:03
Yes, OK.
4:28:09
Might not work for this one, because X Factor has taught in, Oh, OK, At first: try, Yes.
4:28:21
I'm not sure you copied that file over from the results. But maybe you did.
4:28:28
OK?
4:28:42
OK, Hello?
4:28:50
They're fine.
4:29:24
OK, care.
4:29:34
Yes, sir. That's what you're supposed to look at.
4:29:43
Still.
4:30:01
So that's the actual text. From that matrix took at the source location here. This is the source location.
4:30:12
And you can actually use the receptor to look at the receptor: Dario.
4:30:28
Right.
4:30:37
That's going to be for a 4, 7 day receptor 44.
4:30:45
Sounds like me.
4:31:18
Because these claims That The big matrix, so it's.
4:31:25
Explain.
4:31:36
Oh, that's the same.
4:31:42
Power.
4:32:12
OK, I think.
4:32:38
Yeah.
4:32:51
Yeah.
4:32:54
Hi, Mark.
4:32:56
Yes, so so it's almost time I'll show it to the test slash desktop here.
4:33:03
I think it'd be OK to stop, I think, what you've explained a lot, about, how we do the source attribution and the explanation that were really clear and good. It is numerically computationally.
4:33:17
It can be difficult. You're trying to get a lot of information.
4:33:21
So why don't we stop here and take a break?
4:33:23
And, um, I think that we've been able to go through a lot, a lot of what we wanted to go through here, so all, if it's OK with you, I'll take the control back up.
4:33:35
I'll go into a 15 minute break, because that seem OK.
4:33:41
OK, I'm sorry, I like, I didn't work on it. No, No, actually it's, it's, it's totally fine.
4:33:50
Um, you can be using, and you've used it for a long time to create these kinds of results. So, it is a very powerful tool. And, it doesn't always work perfectly when you do a live demonstration, like, we're trying to do back down.
4:34:07
So, don't, don't worry about it and, let's go into a 15 minute break, I'm going to take the presentation back.
4:34:18
Neat.
4:34:20
Let's see, D here.
4:34:22
OK, so we'll go into a, um, a 15 minute break, and we'll come back at five minutes after 3 0 PM in Eastern Time, or you know, in about five minutes after the hour, or whatever time zone you are in.
4:34:39
And but we'll come back, and we'll be talking about off our smoke and dust storms. And then we'll wrap up for the day. OK, we'll see you in 15 minutes.
4:34:49
Bye bye.
4:34:52
Break, and we're now going to go into a section called, Dealt with Wildfire Smoke and then Followed shortly after bias exception on dust storms.
4:35:05
And the section on wildfire smoke is going to be taught by doctor Hinshaw Kim Honcho, aye, welcome, and I'm going to, um, transfer the presentation over to you.
4:35:22
OK, can you hear me? Yes, I can. Yes, we can hear you.
4:35:28
Good question here.
4:35:29
Let's see, I'll turn off my camera, right, Perfect.
4:35:39
We can see the screen, I think that you want us to see, OK, good, good, the.
4:35:51
Welcome back, everyone. So in this session, I'll talk about our efforts, right.
4:36:01
Look, Flow, simulation of the failure surface.
4:36:07
So tell us a bit history.
4:36:10
Over our smoke appropriately, sheath dam, and the similar thing where the fire because there are some kind of a long history and actually we had a radical shift towards the ... death. And the other words, if they bear out is that 2007 and it was really my T is the tooth effort.
4:36:40
And it was a collaboration between the regulatory and also the ...
4:36:52
for the, uh, satellite data and also envision the information world.
4:37:01
Provided by the using the USDA Fo, the Forest Service produce Khayelitsha mother.
4:37:12
Oh, Fight is the most easy to collect data this species has been.
4:37:19
It was started in 2007 and has been operating for the almost 15 years and it has been affected by the North and south right now it's cracked up to and actually it is almost to retire actually instead of the tire sometime next to math. So.
4:37:48
Is a very big tradition year for us, and we are trying to develop the new Shift 10, Uh, using the ...
4:38:00
Got better fire smoke for kit feet and actually something, Oh, something.
4:38:11
I like presented today isn't that reality.
4:38:16
We didn't include the new things, your material here yet.
4:38:22
So today, I didn't, both through the, uh, Using the our traditional methods for the fire smoke over the simulations, using Headspace and then introduce our new approach to what we have developed the flow. They fit the Earth and whether the our future direction for the fire simulation in the ... using the .... So, if you have any interest for the fire simulation, but give us more feedback than the, that really the only half of the tool was that the direction of flow can use his death.
4:39:06
Oh, what they are, they presented today that basically pulled the fire.
4:39:14
Fire simulation, they did the early part where someone is using the how to learn the three-d. effect.
4:39:23
If you didn't know the information on fire, and actually there is a mostly a very straightforward, because you didn't have to shoot them.
4:39:34
For the fire simulation, we really don't use the term much complicated chemical reaction, and actually death still makes sense, and the very efficient code. Right? There any sense village from fire is a primary, and there is also some secondary reactions about the theory can turn the transport and that the person though defy upfront with FDA.
4:40:07
But this person mother showed that you could, Jennifer, for the simulation. And it is a very similar to what we already DFO.
4:40:16
The voting Workshop training period, just the worst thing, new either in this generation.
4:40:23
We need to use Firestore code that, any task that that had the information for the initial.
4:40:31
Actually, we already know how to set the initial standing for that, the growth in the Middle East. And if the using the ... times, could there be another way of providing the information we can provide more detailed information on.
4:40:51
So we have more control over it, but that is not that right, but the second part is there.
4:40:59
It's a more complicated one to run. The pre read this before the buyer simulator, we need to know the above, the information about the fire itself.
4:41:10
But basically, what we need to know if we need to know the location. And then we need to know the .... ..., how we locate those by any chance will be the only if the issue. And that is one over where, it has been, how did the, prior to the fire 30 for a long time, and at the end, the deputy theory, we will meet the? What is the amount of a fire?
4:41:38
Any English as the reason for that, kay.
4:41:41
Oh, oh, we need. So we need to keep that in our mind, because we need the information about the location type and amount, and also what type of a fire and their faith. In general, we have interest into two types of a fire in the one that generates a wildfire till the forest fire.
4:42:03
And to get to the inflammation group cases, I think the amount for the wildland fire, if they did not plan to one, so it is really hard to get the ... and the ... Entity using.
4:42:18
The satellite data is the most popular way, and the most testable way to get those information and the **** of smoke, OK.
4:42:30
Then I just mentioned we use the arithmetic shift, provided from the Nest, if to get the fire detector, that means it can provide the location of a fire And the amount of ambitions to be used for use case. For a third, these emission, neither the to get the how much of the image of the list from a certain location. Both it really depended on that. Characteristic, what would that do? Interviews? We need to know the why do you get the fuel loading and what table vegetation here, and either some deteriorating condition and that was the effect amount of a fire engine factor uploader.
4:43:14
Both the fire can be provided in either way. And in this case, we use the tariff Bruce ...
4:43:22
and the height of the ... emission. We also could the tax rate to the amount of data. It is a key. So, based on that that ship calculation, we can calculate the Boolean field, those flow. Then we can calculate how much of the depth can be lifted the vertically And then using that information, We were able to run the smoke simulation.
4:43:54
Oh, And that to get those high, to be revolutionary, we used a brief dedication. And.
4:44:04
You think that we have with the community, the more of sounds. So, I will define shortly.
4:44:10
And, therefore, the prescribed fire Oh, Daddy, there's some additional fire with that for the controller. And, although the fire we have more information, we know that you get at the location. And we know the amount to over the area.
4:44:29
We want a thon, and we're not sure about the emission amount about the also again to those K emission mother. We could get that information.
4:44:40
So, our audience, if there has been using good information to get the fire simulation and then, Oh, oh, for the future and, the alleviate it as a More options for those.
4:44:58
Feel nice and also, we think, that the most popular, they're like observe that product for the fifth to the fire ready to power, and the most of the satellite. Observation provides this information. And the FFT information have been used very intensively in the k-th. If not only for the ... also for the other ... Mother's Day, I do think that the FTP and therefore, it can kind of preliminary as a therapy can be treated with this information.
4:45:37
How to use F R P will be included.
4:45:40
But isn't that part of the hands-on training today?
4:45:46
And also, I will integrate if the finally, leave it out. You approach of ours, the system. It's a high speed of faith, the ancient .... So we, all that we have with that, They thought, the optimization technique in the previous session, the presented by .... So we utilize that information to get better information about the emissions.
4:46:14
So, in our future, in action.
4:46:17
And I will explain demos that there weren't at the end of take advantage the presentation.
4:46:26
And let's go back to the extra training session.
4:46:29
So, 14.21, Fire Smoke, and I mentioned, we used the HMS information to run the simulation.
4:46:40
And I will present how to use that information and let's just start the simulation first and I will explain the details of that simulation, though. We've made of retrieval control fire and the fire, but, one thing you should keep in mind the default is to fire, fashion, our control fire.
4:47:05
It's not in the two data files, actually, from the status and the retrieval.
4:47:14
You do need to go to the tutorial and smoke directory.
4:47:20
And then you can find the smoke control shift thief.
4:47:25
Oh, Case Dave and.
4:47:32
Configuration setup, six regions, it turns out the same graph to be to smoke, OK?
4:47:45
Bit.
4:47:48
But, Yeah, that's just the start to learn first.
4:47:55
Oh, wait!
4:47:57
Sure, if you catch it about Oh, yes.
4:48:01
We can see that the percent country is if the really quickly and they are almost at around the 30% already, but don't have any experimental data.
4:48:14
And let's see, what the real configuration DESlock?
4:48:24
Kay?
4:48:27
So for the current, so far, actually think that's typical of the concentration simulation, and there's not really sure.
4:48:42
So, for the image integration is, If they think they can see that the emission rate, these are one, and the image integration to Stephanie to both the emission rate is not directly use date for the simulation, because we are using different input format of the core of any times. And then, when you see the mom.
4:49:15
The first, there is a configuration for the number of particles at the end, and you can see that if buyer is it, specify the 13th tech to fire from the tourists smoke further, and that is the one that we can check from the configuration.
4:49:37
Yeah.
4:49:49
I see it.
4:49:57
I don't need to know.
4:50:04
OK, here, number six, emission recycling or into five, and from this an annual, we can specify which ... we use for the simulations.
4:50:19
So, when this option is on it, try to find the anytime first, Know, and then even though the, the other number of steps in the front, either from the previous menu, what is really inside in that text file. And.
4:50:40
So this is another simple task to file, and where it came about at this anywhere, anytime for how we generate. This is where and how to utilize, it will be difficult for me tomorrow, intersection at 7 16.
4:51:00
So, I just do the very short introduction, and just to use the data that the basically refers to Orion, neither does the, uh, the header for the data. And it's not really used for the actual simulation, And for this, the one, we have a just, and the third that I needed to somebody else for input and we can have multiple lengthy from epsilon. But right now, it has just the 1 9. Oh, they've kind of, it has information about the time for, basically, the 2011.
4:51:37
The 10th from the AT&T, and the follow, the 72 hour. So village of emission is at 72 hours, and then we specify the location of the source, the rough, ready to the tooth, and then we have a height.
4:51:57
And then we have a city more effective for the lake area, and shift, I mentioned about the, we need to know the three inflammation of the buyer to get to have a fixed stimulation. And if that is the location height and emission amount. So we can specify the location from the ticket longitude.
4:52:19
And also, we can specify the, the amount of emission by the rate, the, this one is there any new image, and you need to bury it the hour.
4:52:37
And for this physics simulation, actually, it was in the graph, but the converted into a microgram by multiplying 10 to the sixth, so pick. Oh. It was handled in the Amazon Fire. The data we want to compare with the observation data or the victim in the micrograph of treating mirror.
4:53:03
That's why we have this number here and we also need Arial.
4:53:10
Area for the bonds and in this case, we are using the H in that data set.
4:53:16
And in this event that we have train your client server, fire detection and usually recall this H in the therapy that we assume that one of the text and cover the one should be one kilometer.
4:53:36
The area. And we assume that 10% over that area is really running the, although, the training prior to the total amount to be assumed.
4:53:47
And then at the end, we have his plate and the number of what the tenant provides is it through the roof K, the emission mother and that this number is to use the forward from. Nice. So one thing that can be easily confuse, these are what we use the forward They're the Airflow knife, when we have at the high inflammation and the sheath.
4:54:14
The hit the least inflammation.
4:54:16
And, uh, if the ship is just the zero, either, and use the information from the heights.
4:54:25
But when we specify that this number is the highest, they will use it. The data, and using the updated ... from nice information, though.
4:54:39
For this case, we're just using just the one, Daniel, the emission information, so it will be the day.
4:54:44
We put this number into the, so.
4:54:56
In here, we actually, we can specify that, the emission rate and the hour of any sense. So we can use it, This even if the number here in the ..., but we did.
4:55:10
then method, we cannot use that the magic calculation using K So, one of the other advantage of using it, anytime, we can kind of activate that they feel free.
4:55:24
the kitchen creation, and I think there's information, So these are Oh.
4:55:34
Sure, Any time.
4:55:37
Alright. We already have T two.
4:55:42
Fifth to see.
4:55:46
Display the output.
4:55:50
And the data, There's not much Yeah. Actually, I try that, because they should have both.
4:55:57
There are two things I changed. I put the micro Chrome for the labor, and I also changes.
4:56:09
The number for the contour.
4:56:14
Prompted the Photoshop and the nice small steps.
4:56:20
Reacting?
4:56:31
Oh, it has.
4:56:35
OK, so, there are many left, and, Oh, they didn't know.
4:56:45
Oh, look consistent consideration at the beginning of simulation. Oh.
4:56:50
I can one thing, Ah.
4:57:02
So, when you took the step of a simulation the simulation example, though, that's it from the April 16 0, 0, time, and, and then.
4:57:28
OK, and then the actual curve.
4:57:31
When you check The Imitation Game, the settled that, the, uh, therefore, for anything related to 18 hours later.
4:57:42
It wasn't because it, for the observation data. We have the observation data.
4:57:50
We want to compare after we have our simulation output started from the 0 0 0, if the data to match the starting time for the better competitor.
4:58:05
Though, as I mentioned before, when we started this simulation, oh, you can see the percentage accomplishing what, though?
4:58:19
30%, because if we're almost give them the 18 hour, could take an hour, we know there are no emission the list. So, that's why we said in this way, and again, if the way we have with nothing in the plan.
4:58:38
Clear. Either, OK?
4:58:41
We could see the thumb and they begin to increase Andrew.
4:58:52
We call the theater, the east of ...
4:58:57
Shear and the mixture from the.
4:59:02
We could compare things along the way, the actual error.
4:59:08
Ah, image it from, the better I then did, you can see the location of the ..., both though the direction of the transport, below me, they're very well simulated using ....
4:59:26
This, a Shimmering output can be.
4:59:32
Well, it's a compelling story, though, Observation, and let's do that.
4:59:41
Oh, from the utility convert to the station.
4:59:48
OK, that's number one.
4:59:53
We are using the smoke of that being that she don't fire, we adjust the teeth and we extract.
5:00:04
The simulated failures out over death. She Delphi for certain locations remained deadlocked, the ... mail and the latitude and the longitude.
5:00:19
Oh, and then we can check the data, though, we can see that it's a total number beta.
5:00:35
Peter, we can see the software. So, we can plot them on immediately, and, but, if the red line is the, simulated, the output for the rotation over that may have patient. And, then we can compare them to be the observation data, the mayor of PM 2.5.
5:00:59
And then try again, though, you can see the simulation of Codenvy.
5:01:08
The third fifth, monitor over the DOD where there is a relative.
5:01:15
There can be some changes of direction for the fruit off by the change, we interact, chance of actually had to go to a sheath and both of the ...
5:01:26
the mother of the capture the very rare both.
5:01:35
Basically, this is the example, how we can use high throughput smoke simulation, and.
5:01:45
Make sure next page, how to get that information.
5:01:51
Again, I, I empathize with Diego the rotational location and type 10 and a mouth and we handle it.
5:02:06
Different type of data for the fire, OK? Well, we have many data set that can be used for the fire detector.
5:02:15
So, the from the Fahrenheit, that perfect life from the muddy the water from the Earth and others TO 13 seminars, and you can utilize everyone. Well, so I mentioned the H M of Theta ... from the Nest, if that one needs to also spend a good product for the fire detection. and they actually have the information of the transport of love. The outdoors, mostly for a satellite state merges the information for the fire, the tip jar, and deflate month over. the fire them off. So that's another very strong that can be used to Florida.
5:02:59
Fire simulation, uh, but also their data on another fire, emission data base of death, ....
5:03:11
Defects are in F and then close it.
5:03:16
They have a very well defined by an image, an amount, to endorse it, OK?
5:03:22
So, the thing can be used for the High Spirits Malaysia, if you can convert them into the death image tags format.
5:03:35
Oh.
5:03:37
But let me just introduce one way of getting those tents, information, and.
5:03:46
Yeah, from our Redis server, if you have, if you are a registered user, you can go into the simulation, both of them archive data, and of how those here.
5:04:06
You can see the left hand.
5:04:12
Ah.
5:04:15
There's a bunch of stuff the thumb, whatever, location.
5:04:19
Boom, math near the Optic one place, and just try to learn.
5:04:27
And let me just use those early over June.
5:04:35
Oh, maybe for the sick fellows.
5:04:47
Oh, yeah, It's dirty the simulation.
5:04:50
And what I want to say here is that if you can, if they attend simulated, test the thumb state so you can get some information for the basic computer configuration focused, edified.
5:05:09
Really, it's the same with the birth and cultural five.
5:05:15
And you can see where if is it just another typical control fire.
5:05:27
You can see it has a baddie O P.
5:05:32
For the relocation lift actually differ from the controller, will be discussed later, but there are further ... the number of the location should be matching with though.
5:05:48
So, also that there, anytime, though, just have a thumb, though the location for the 128 oh 9, that means, that the N times information used for the K theta 128, OK.
5:06:07
So, we have with them simulation. Sure.
5:06:13
If they haven't meaning, but the length of stay is from here, you can have this kind of awesome soon Asia.
5:06:24
So, oh, because when we specify this location, it is using this location to learn, produced a emission mother and get the emission are marked and you can check that that specific conditions. five, though, is that anytime fires that the ... is a very detailed, one for every three minutes and the one we used in previous simulation, tester for the 1 9, demoting everything together.
5:07:05
Oh, oh, Maybe you can download that for five years, the four basic simulations. So, it can be a very good starting point for you want to try.
5:07:21
By simulation, we are using type three.
5:07:28
The case of, oh, that's, you go back to life, and, oh.
5:07:36
We read about the Future Directions, and the I mentioned about the equation for plug might actually break the occasion either give females, but the water the time for the predominate rise From that.
5:07:56
In the tariff tariff, though, it wasn't that the early, the data for the fire emission minimize the.
5:08:07
Maybe we need more options for that, that goes for co creation.
5:08:12
And, actually, because the rest of the satellite provided that information about the fire power, there, were the us can develop, the five top, here, to use the parking numbers, to get the plum lights.
5:08:32
And, actually, this part was implemented by the ..., and Chris Ruffner into the faith period.
5:08:45
You can not find the DevOps down from the GUI, but you can manually change it.
5:08:53
Uh, you can add the another line in the step five.
5:08:59
Because it's just text where you can add another, right?
5:09:01
So, when you have, uh, the blue nice one, if it is just using the verification, but equally a step of truth, it is used.
5:09:14
Stop here for a nice Kim and it will use the left column, the heap karloff as a number or maybe you can have the thermo the case law. And we are also testing this loan and when it.
5:09:32
Finalize to review.
5:09:35
Perhaps for the ... officially, at the end, the future direction for the ....
5:09:44
Another one over the other, in the previous session, the chaplain about how to use that.
5:09:56
Oh.
5:09:58
Therefore, the Hopefully, you caught the function and but TCM approach proposed fiscal mess today but to starve the wow. If that's the reason for the emission estimates, yeah. So basically what we are doing here is when you have a fire detection location, right?
5:10:19
She picks the simulation from each detected by authority from the rockin location and also for the deep content to it.
5:10:31
And then we use them, one and then the, oh, fire smoke detection from the satellite. Usually a third of that can provide the error, the depth AOT number and the data that's on products. That 10 best pay the AODA out, though.
5:10:55
Go smoke, fire smoke.
5:10:59
So you'd think at the earth the bowtie simulation.
5:11:05
Well, we can, we can know that the ...
5:11:10
derivation factor and then we combine them and and then compare with the life of the beta five that competitor beneath it down by the by constructing the PCM, make faith.
5:11:28
And we will find that that the what will be the best coefficient for each dilution factor.
5:11:36
That is paying the satellite observation The most efficiently, though, uh, using the right data data to constrain this.
5:11:50
Wow.
5:11:52
This PCMA thrift and usually, what are the steps?
5:11:58
Do this in relation, that, we set, the, hook, the function, if the faith can eat them, difference between the mother and the ...
5:12:08
end, though, minimizing the cost, function means the, whatever, be, the solution that minimize the difference between the third line, then the mother, by piece.
5:12:21
By doing, if the approach, by minimizing the function, we can have the best, the Photoshop for the by, any chance that can be present to the detector in the best ticket, from the tip of the fluff.
5:12:38
So, in the left corner, in the gray color, sometimes our traditional mother, and those are, all right, blue competitor: you can see, the April month has been improved, and that is by them.
5:12:56
The result advised that if the, uh, the corresponding ... method. And, of course, the inverse modeling technique.
5:13:06
We call that this shift in the Heinz has been the image, an inverse map system for the way, for the light, the fire.
5:13:17
And also, we have different than for other different the KT. Because if we can apply this concept for many cases, it's nothing like, your creative person.
5:13:31
Go for ..., oh.
5:13:36
This is a case for the.
5:13:41
The fire, though.
5:13:43
That's just the sunrise again.
5:13:46
So, basically what I showed you, how to get there from fire information for the simulation and we can use that information flow depth. That from our traditional way of using the one, and then we have some of the actions to get them more.
5:14:04
They're getting the water tonight options, or, or, we will provide some way, maybe next year, how to use a crappy already done, but they're not part of the training. And I will definitely be using the inverse mapping technique and to estimate the buyer information via ancient information.
5:14:31
So, when we finish the order of death or maybe they can provide very good database via Amazon. And maybe that over any Type five that you can utilize immediately. But that's our show.
5:14:44
And if you have any comments Or some JSON for the direct and just let us know because we have just working on building the new shift and those are if you already have.
5:15:03
Bookcase. I think that's it for me. And I don't have to mark off.
5:15:10
OK, yeah, thank you.
5:15:12
Thank you very much, ..., really excellent, and super important topic, as we all know, as numbers of fires have been, I'm increasing so much in recent years.
5:15:25
It's such big impacts on public health, And so thanks very much, I'm going to take the control back, and we're going to go now into a section on dust storms.
5:15:44
And let's see, second, here, There we go, Alice, doctor ...
5:15:52
Crawford, will be teaching this section.
5:15:55
And Alice, the CBC your screen, which is good, and can we hear you? see?
5:16:03
We can't hear you yet.
5:16:08
Ah!
5:16:09
Nice.
5:16:15
Let's see, Let me unmute you Now, can we hear you? There we go, OK, All right. Good. I'll now leave, it lets you take it away. Thanks a lot.
5:16:25
OK, thanks, Mark.
5:16:27
All right now we'll be talking about dust storms, which have, you know, a bit in common with the previous topic of fires and that a lot of the transport, as as hands-on mentioned, you've already learned a lot about high split, works in terms of transporting and modeling the dispersion of substances. And then the problem with, like, dust storms and fire.
5:16:52
A lot of it comes down to, you know, determining what the emissions are which can be really tricky in dust storms in particular, you know, determining how much material is going to be emitted from the ground is complicated. So basically, the physics, kind of behind it, is that, you know, the material, the amount of material that's going to be recent, re suspended, has some dependent on the meteorological conditions, right? Like, if it's really windy, you're gonna get more material re suspended, but it's also dependent on the properties of the deposit itself.
5:17:28
So, and those properties that can change over time, sometimes a place that was emitting a lot of dust kind of forms a crust, or you know, stuff starts growing on it and then it's less likely to emit material.
5:17:43
Um, so it's basically a momentum transfer, from the atmosphere to the deposit material.
5:17:50
And what we what we use to characterize us as the friction velocity, as a quantity called fixing velocity that's available, from almost all of the meteorological models now. So it comes in like, you know, her will provide an estimate of what the friction velocity is or the GFS or wharf.
5:18:12
So that friction velocity kind of characterizes the amount of shear stress on the surface.
5:18:17
And so that's what you want one part that you use to kind of ..., how mature machines are going to be.
5:18:25
They usually want like a threshold friction velocity. So first, we're gonna look at the various kind of simplest algorithm that's an highest split.
5:18:34
And that has a hardwire threshold friction velocity of 28 centimeters per second in it.
5:18:40
And the locations that are even allowed to a net are determined from the land use files.
5:18:47
So if the land use file, which we haven't really talked about before, like I think they were mentioned, that they don't usually play a huge role in a lot of the other high split things.
5:18:59
But in this particular application, in this particular algorithm, they play a big role because they actually determine which locations you can emit dust from. So, basically, only areas that are coded as desert locations can emit test.
5:19:15
And then the vertical dust flux is calculated from this kind of little simple algorithm, which says, you know, it's point 0 1 times the three times the friction velocity to the fourth power times the area.
5:19:29
So, the area of the, that you're considering emissions over.
5:19:39
So to demonstrate this, we will be looking at an event that occurred in 2010 in Utah, like a dust storm.
5:19:47
And we have like a little bit of data for that site, a huge amount of data I think you can see there.
5:19:54
So we'll get the file later.
5:19:56
It's a, it's a, a set of data points.
5:20:00
And first, we're going to start as usual by retrieving A.
5:20:12
Control dust to zero.
5:20:14
OK, I can save that.
5:20:22
And then.
5:20:28
I don't have to do anything else to it, but if you look at it.
5:20:37
So, if you look at it, it only has three points in it to begin with.
5:20:42
And you'll remember, if you remember back to when we did.
5:20:47
we did that trajectory calculation that was similar to this. That there are three points. The first two points kind of defined that the corners of basically a square area, that you are going to consider emissions out of.
5:21:03
And then the third point, aye, to find what fee, um, resolution within that area wise.
5:21:12
So, you can see that this is like, zero point two five passed the first point, so that means it's gonna space emission areas at zero point twenty five degrees.
5:21:25
So there'll be an admission area every zero point twenty five degrees within the square that's that's defined by these two.
5:21:32
And there's a little preprocessing algorithm that we're going to run that will take this initial control file and then turn it into a control file with more Emerson points.
5:21:47
So we're also going to make sure that we have our, um, but what I think will do this by hand, they have us doing this. We're going to skip this step five first sense and then we'll go back to step four. Step five, the setup dot c.f.t.
5:22:05
basically it knows that you want to use the emission schemes for dust.
5:22:12
If you set this, I can equals three, which says done and menu number 10 here.
5:22:19
So, you can see, actually, I already have it set, because I was doing this earlier.
5:22:25
You can see that this is kind of a parameter we use for a lot of different kind of specialized runs.
5:22:32
So, we're going to save that.
5:22:35
Then we're also going to open Menu four.
5:22:38
And the dust emissions scheme is one of the one of the things, like sometimes you have actually more particles released precise press cycle, than that maximum number of particles. So, I'm gonna set that.
5:22:55
And then we're going to do the special runs, So special hands that dust storm.
5:23:04
And, again, what the special runs does?
5:23:08
Is it first, friends as preprocessing, utility, that will take your three points that you defined, and then, Oh, we forgot to do something, but it's OK that doesn't run long. We forgot to do the land use files. That's why I skipped over step four.
5:23:25
So basically, um, the default Landis' files have a one degree resolution, so we'll just do this now while that's running.
5:23:35
Ah!
5:23:36
And we actually updated the Land use Files recently. So even the one degrees, I think, to have some desert areas, so you won't get nothing.
5:23:46
But I think in our old Lindsey's files, it used to be there weren't even any dust areas like in North America, because it's just too big of a, an area cell.
5:23:56
Even though there might be some small desert areas in there, they weren't defined that way until she would try this algorithm. You just wouldn't get any admissions at all.
5:24:03
But if you did a finer land use algorithm, then you would get some.
5:24:10
So we're going to go to See, P V Y Files. So here, you'll see them, where they are.
5:24:19
So this is the default one, and you'll see it says one Degree Spacing here.
5:24:27
We also provide half a degree facing, we actually provide, uh, a 10th of a degree spacing one's as well, and we're just going to use the half degree spacing, so we're going to take this file.
5:24:46
Are going to put it in our Working Directory.
5:24:56
And yeah, I want to replace it.
5:24:58
So now if I look at this, I have the half a degree spacing and it's pointing to the folder with the half a decree files in it.
5:25:12
I close that out.
5:25:14
Slow running I don't really like to exit out, clearly seems to take longer to figure out I've stopped the run than it does to actually let it go through.
5:25:23
Um.
5:25:31
So.
5:25:35
I know I wanted to look at the control file.
5:25:39
So the control file, remember I rewrote it.
5:25:43
And now it has this one only has 16 so I guess the one degree lantus files currently have 16 sources within that region that we defined that could possibly emit S here just indicates that we're using this cell, this, this simple algorithm. So the will tell it that it's using the simple algorithm and that the default threshold velocity is 28 centimeters per second.
5:26:17
Then this one is the area, so it's just A, so it's the area of D, the mission area. So it should be about zero point twenty five degrees squared.
5:26:28
Um.
5:26:31
And you'll see U and V is the half a degree.
5:26:36
Land use files will actually get 24 sources.
5:26:42
So that will look like like this.
5:26:52
And the other reason it kind of slows down at the end.
5:26:54
You'll notice what, that's what the dust as it can be kind of uneven because what happens? What's happening in this simulation is actually at the beginning of the time period. There aren't very many emissions.
5:27:04
So let's look at that what the message file is showing us.
5:27:12
So, we're looking at the message file, and see, at the beginning, the first time period, there is nothing there is, because what's happening is, the friction velocity is never exceeding the threshold. So nothing's being emitted.
5:27:28
And so, that's why the model, at the beginning, it was like going really fast, But then, you see around here, now, we start getting mass in the model.
5:27:36
Um, now, this is another interesting thing about high split. So you'll see that even when there's zero mass here, it's emitting particles, right? It sounds like there's 176 particles.
5:27:47
But, no, there's no mass.
5:27:49
So, a high spot will actually create the particles with zero mass, but then on the next time step, it texts for particles with sara mass, and then remove them from the simulation. So those protocols are being created, and then immediately removed. And so when the message file, you'll see that. And you'll see that.
5:28:10
Know, like here that the particles were not, but then they went back down, and then when they saw the emissions, when the real emissions started, we're actually having like 848 particles.
5:28:21
So this ends up kind of with us, you know, a weird particle number because of the way that high split estimates how many particles it needs to emit an intermission cycle and the fact that it does a sign some particles with zero mass when there's no emerson's.
5:28:38
So, you can see now.
5:28:40
It's running along, and.
5:28:45
Are getting, like, the masses increasing and around here. It looks like.
5:28:49
The emissions are, like, you see, just see them as being pretty constant, Um.
5:28:56
And, I can't remember.
5:29:00
If we define that that position, OK, so that completed.
5:29:03
So, now that we have the right, um, we can just look at it right away to span, and we won't play a step.
5:29:16
So, if we wanted to look at what the Yeah.
5:29:25
Fire looks like.
5:29:30
Just to get a better plot, ... did this.
5:29:34
So, to get a better plot on the second, kind of zoom into the right area, there's a bunch of things you can set, which I was like there.
5:29:45
The distance around the area and the center of the plot, the user set current contour levels, and we also want to look at it in micro grams, and the emissions were N N grams.
5:30:01
So, we're gonna add a concentration multiplier there.
5:30:06
So this is, this is what we would get.
5:30:08
And I didn't reset so, actually, I had done this earlier, and, and these are the measurement points that we'll be looking at.
5:30:18
So let's go ahead, I'm gonna quit out of that and we're going to just run the model again.
5:30:25
Now that we have different Oh, but I have to do.
5:30:30
So I do need to um, I need to retrieve the file again because I want to start with those three file data points and did a special runs.
5:30:45
Hmm, Special Runs Dust Storm.
5:30:58
OK.
5:31:00
So now let's look and see what our Control file looks like.
5:31:05
OK, so now I have 24 sources to find out that I'm using the higher resolution land use file.
5:31:12
And a few other things to note about this friend, which is a little different from maybe what we've done before is we have one concentration grant defined, which has a 24 hour averaging time, and 30 and it starts six hours after the release.
5:31:28
And the reason for that is because those measurements that we're going to compare to, where 24 hour samples that started six hours after, um, the really start time.
5:31:43
Now, we also defined a second concentration grad.
5:31:49
And this concentration grad is pretty much the same. It starts six hours after the release time. But we're just doing a one hour average, and SSL.
5:32:00
We'll look at, like, we just want to be able to see, you know, what's happening from hour to hour at this dust storm.
5:32:08
It's more for curiosity sake, we don't have any our averaged measurements for this event.
5:32:18
Oh, I also wanted to luck.
5:32:20
OK, so, ASO.
5:32:22
We do not have any like gravitational settling.
5:32:25
So, these are what's usually front desk's he would want to have, Um, PM 10 would have some kind of density And you did get some gravitational settling and deposition.
5:32:38
Yeah.
5:32:42
OK, that's all running.
5:32:49
Yeah, that's what, it's almost done.
5:32:52
OK.
5:32:53
Ah, OK, well, it's finishing running, I'll just say. So, there was an optional exercise, which was to use a 10th of a degree land use files.
5:33:03
And we can just, we're not going to do that, that we can see, um, the release locations will get slightly different.
5:33:13
And I think there's ends up being 29 locations.
5:33:20
And instead of 24, ah.
5:33:27
And then there's another exercise, where I just increased the spatial resolution of the Amison points.
5:33:36
So basically, if you looked at this control file by the system, and that goes to the preprocessing, this used to be 38.2, 5, and I changed it to 38.1.
5:33:49
And I changed this to 114.9, so now the spacing is, you know, zero point one degree, and, um.
5:34:00
The effect that has is it has it creates more emission points that each one has a smaller area, right?
5:34:07
So each one is representative of a smaller area.
5:34:12
And but it does have a little bit of an effect on what the plume looks like. And that's kind of interesting. You see.
5:34:19
And these are the measurement points that we're going to be looking at, right? And if you look at the concentration gradient here, like the contour lines are pretty closely spaced.
5:34:29
I mean, granted, these aren't.
5:34:32
Maybe we should go back and plot them.
5:34:34
So they're not, like, this is a factor of 10, and then this is a factor of two.
5:34:39
Factor of two, OK. So they're all about a factor of two, except for that first one.
5:34:43
So you can see that the concentration gradient here, as indicated by the close spacing of those lines, could mean that no, just a small shift in the location of the plume could mean a fairly big impact on, um, what you're seeing and these.
5:35:04
What what your modeled values would be at those measurement points? So, it could be that you'll see later that we actually have a fairly large under forecasting bias for the model, which could be due to the emissions being too low. But it could maybe also possibly be due to a displacement of the applicant.
5:35:23
Slightly in space, It's hard to tell, because that's why it's always good to have like more measurements.
5:35:33
Um.
5:35:36
Still running.
5:35:41
And then there's an advanced topic here, which says that there is a way to adjust the threshold friction velocity. So, if you wanted to use this simple algorithm, but you really don't like that 28 centimeter per second threshold friction velocity, there is.
5:35:58
There is a parameter that you can set in the setup that c.f.t. a file called P 10 F.
5:36:05
So if you set it to zero point five, then that estimation would occur at half the normal threshold velocities.
5:36:19
All right, 90% done.
5:36:30
Right.
5:36:33
And.
5:36:39
So, we're gonna go through it, and maybe I'll just talk about the next one, since we're pretty good on time, but they might not get through the whole chapter.
5:36:48
Um.
5:36:51
And the next one, we're going to talk about a little bit more complicated emission algorithm, and then.
5:37:00
we're also going to talk about adjusting the emissions a little bit, And, you know, one thing to keep in mind as, and if you don't like the emission algorithms provided with high splitter, if you have your own, or if you want to develop your own, um, then you can do that by.
5:37:18
Again, using that admit times file to create a complex, in the sense. I will be talking more about them at tennis filed tomorrow in Section 16. And Hanshaw referenced that to you, saw how it was being used.
5:37:32
For the smoke and fire smoke, that's generally how it's done is you have no sun model to create the emissions, and then you put you, in fact, the emissions and the highest split using the at times file.
5:37:45
So, you can use a similar approach with dust, whereas rather than using one of the algorithms, provided, with high split, to estimate estimations, you can estimate estimations on your own, create an at times file, and then put that into split.
5:38:03
All right, So now we're gonna go back to this section.
5:38:11
I'm just gonna, we're going to plot that.
5:38:20
Remember, we have If you haven't done this yet, Tschida sat out all these.
5:38:27
Yeah.
5:38:30
He is here, Then we'll look at.
5:38:34
So here's the results, and we might actually want to put that. Oh, I see what happened.
5:38:44
My tent turned into his hero.
5:38:49
There, that's better, um.
5:38:55
And then, we're gonna look at the datum. So we're going to create the theta M file.
5:39:04
So the measurement file as and this is under tutorial under the Dust folder and it's called Air Now Dust.
5:39:15
Um, and we're gonna get measurements are also in micrograms. So we're gonna leave that there.
5:39:25
And we're going to create the data file and we're going to compute the statistics.
5:39:32
And so our correlation coefficient is pretty high, but what you'll see is, um?
5:39:42
Hit the there's a huge under forecasting Dyas so basically, you know, you see that there's a really large bias here.
5:39:52
You can see that the ratio of the calculate into the measure, it is like zero point zero two.
5:40:00
And so, with that, if you look at your rank, it's pretty low because yeah, because of that, that bias.
5:40:07
And you can also look at the scatter plot, which tells you the same thing, like all that the points are down here.
5:40:14
I mean, it's very tempting, right? Like, I can just go in here. I was like, wow, that's a lot of factors.
5:40:20
42, So, we just put that there.
5:40:26
Computer statistics, And like, you know, everything looks better.
5:40:32
Um.
5:40:36
Yeah, And, like I said, there's only 1, 2, 3, 4, 5, 6 measurements. So, there's often a lot of games you can play it to match, to make things like better.
5:40:47
Especially, when the correlation is good, right? So, we just need it, too.
5:40:51
Basically, adjust emissions.
5:40:55
Um.
5:40:58
All right, Rick quit out of that. Oh, the other, the one last thing to Tijuana says just to take a look at the other.
5:41:07
The other file that we created, remember, this was the one that had a 24 hour averaging period, and this was the one that just has a one hour averaging period.
5:41:17
So, we're gonna look at this one, and we can leave everything else the same.
5:41:23
Just execute the display, and That doesn't appear.
5:41:32
Yeah, OK, So this is that, and you will see that? Remember, we looked at the message file, and we saw, there's like no emissions at the beginning, nothing.
5:41:42
We move forward in time.
5:41:44
This is the time, 8 to 9.
5:41:47
And OK, so finally, at nine o'clock, we start getting emissions, but only from you know, one point, it looks like now, now, we get emissions from two points, then those emissions actually kinda stopped, and they, they kind of go off, and they sort of, don't do much. Oh, now, we got emissions from a lot of points down here, and a few here.
5:42:11
So, about 1700 are getting a bunch of stuff.
5:42:14
And this is where it's kinda really picking up.
5:42:17
Um.
5:42:20
And That is the evolution, the temporal evolution of yore of your dust storm there.
5:42:35
So you can see it's nice to be able to define, you know, multiple concentration grids, um.
5:42:47
All right, sorry.
5:42:49
Let's go on to the next section.
5:42:54
So the revise algorithm is a little bit more flexible and ascents.
5:43:00
And this one, uh, you have the emissions defined by, yes K value, which is called the soil dusts density.
5:43:11
That's basically, It's almost like a little bit of a fudge factor, but I quite emissions area.
5:43:17
And then, um, the friction velocity minus, the threshold friction velocity.
5:43:23
And all of these values.
5:43:30
We're kind of developed from sort of a climatology based on satellite data.
5:43:35
And so if you wanna and I won't go into how that was done because it's there's a nice, um, publication on that.
5:43:46
So, here, I mean, and that, and so, this kind of describes that.
5:43:50
And, ah, there is, we provide, like, you can get the data from that here.
5:43:59
For this example, we just took the data for the whole continental US and just took the 21 locations that are in the region that we're interested in here to create this test control file.
5:44:15
And that's control file, you can see.
5:44:18
this is the threshold friction velocity right now in this column.
5:44:22
You can see up there most of them are larger than the 28 centimeters per second.
5:44:26
And then this is now that it's not just the area, it's the area multiplied by that K value.
5:44:35
So that's why the values are a little bit smaller than they were before.
5:44:41
So we're going to go ahead and retrieve, go back into our setup run, going to retrieve that control file, which I think has controlled dust too.
5:44:54
Yes.
5:44:58
Um.
5:45:01
Then we're, I think we don't, I'm gonna do this anyway, even though I configuration concentration are going to retrieve.
5:45:11
Yeah.
5:45:15
two, set up that test as well.
5:45:22
OK, and it also tells us.
5:45:26
To restore the one degree land use files.
5:45:30
Although it's fairly not critical, but well, do it.
5:45:35
So, we go back to where our land use files are.
5:45:39
We take the default, one de gray one, and we put it back in our Working Directory.
5:45:55
All right.
5:45:58
Ah.
5:45:59
And now he can run the model.
5:46:01
At this time, we're not going to do the special run because we don't need to run that little pre processor that will turn the three data points into more.
5:46:12
We already have them them in there.
5:46:16
Um.
5:46:22
And.
5:46:26
I guess. So.
5:46:30
Yes, I think this has probably been noted before, and probably will be again. There's a ... Guide online as, as well as help us with the tutorial and help, which is really useful. And something that's new in a, relatively new, I think, within the last year, is we, we put a whole listing of all the utilities that are available.
5:46:52
And so, um, that led all dust, utility, that was called.
5:47:04
So we can, I forgot what it was called, but it's, This is the script's.
5:47:08
So if you guys, I think what this has been talked about before, but we'll emphasize that these two little buttons will show you scripts that you can run to basically reproduce what was done in the GUI.
5:47:21
Um, and so here it is. So this is that little preprocessing utility, it's called dust beedi, why?
5:47:29
So, it should be in here somewhere.
5:47:33
Um.
5:47:40
Yeah, yeah, High split, configuration Utilities, So, here it is, and it gives you a little description of it, generates multiple latitude longitude ... points from a control file.
5:47:53
Um.
5:47:56
So if you're kind of, you know, if you're trying to write your own scripts, and you want to use the utilities that are provided with highest split. And there is now a nice, little handy guide to them.
5:48:14
Ah, Go back to.
5:48:21
Here.
5:48:30
Yeah, I feel like it just kept for this plan.
5:48:33
We can see again, it's like writing the plan TS file, plotting the times that says writing the control file, writing the setup file, running to high.
5:48:47
So this has a twice E S Underscore and standards is basically what it's high split standard.
5:48:53
Say you're running the highest split standard executable here.
5:48:58
Um, then can apply it as the planning program and stop mean as the program that computes those statistics. And then C to datum as the one that creates that data file from your C dump file.
5:49:16
Because you'll notice, you know, a lot of CLCS for going through the GUI, and we do make mistakes. And, and that's why.
5:49:25
QE is really great for exploration, and sort of, I guess, mapping out what you want to do.
5:49:32
But then I think when you're trying to do something really complicated, at least fi opinion is that the scripts are really nice because you have it all written out there.
5:49:41
All the steps are written out, and, and you can really reproduce it, figure out where you went wrong.
5:49:50
Or, sometimes, I think, your way, If things start to go wrong, you not sure if, you know, as you say, we have a race that, and sometimes we just shut it down and reset it, So.
5:50:03
And.
5:50:07
All right.
5:50:13
That's, We're waiting for that to finish, I'll also mention Bhat.
5:50:26
And it's my.
5:50:30
Here we are working on, um, here we go.
5:50:42
Researchers that Arielle developed, um, and in our air quality modeling cryptic. This ... Shah Emissions module for dust was developed. And we do have some work to couple that to highest, but to the highest that could use utilize the ...
5:51:01
algorithm lettuce.
5:51:05
Pretty sophisticated and very good and utilizes information like soil moisture.
5:51:11
Um, and not quite sure when that will be Dan. And certainly, if you, if, you, know, you could do it again, like I said, to for, without waiting for it to be automatically coupled, you could.
5:51:27
if you had access to this model, you could run it, created a net times file, and then feed that into high split.
5:51:36
Ah, All right, oh, So, I'll kinda go through what we'll see maybe when this is done.
5:51:50
So, we're gonna, once this is done, we'll well look at the datum file that the results are basically going to be actually not quite as good as the last one.
5:52:00
So, the correlation coefficient is similar, but a little lower, the under forecasting bias is even larger, and it's like almost a factor of a thousand.
5:52:13
So.
5:52:17
This is the, I guess, the plot.
5:52:21
And, so, what we do in the next section, and we probably won't get to the section, we'll skip it, but we basically, we look at the emissions files for the continental US.
5:52:33
And, say, ha, the maximum value is about two effects.
5:52:39
Um.
5:52:41
And so, we go under the control file, and this is the original one, And then we change all these values to the maximum amount of the emissions to E six.
5:52:54
Um, we keep the threshold, friction velocity is the same.
5:52:59
Which you can also play around with attending those, but, and actually doing that, just changing it to the, the maximum, um, improved the results a lot. So, the scatter diagram now looks kind of like this, so, it's still a little bit of an under forecasting bias, this would be the 1 to 1 line.
5:53:18
But it's better.
5:53:19
And the statistical results are, um, know, a little bit better as well.
5:53:30
Um.
5:53:32
No?
5:53:36
Yeah.
5:53:42
So one thing with both I think fire and dust is that there can be a large amount of uncertainty in the emissions estimates.
5:53:52
Um and it would be reasonable to go to an ensemble approach for those two, right? To two times we call it a source term ensemble or you're, you don't know what your source term is. There's a huge amount of uncertainty and what your emissions are.
5:54:07
And so you actually ran a bunch of different runs with different emerson's and treat that as a as an ensemble and provide uncertainty estimates, you know, based on those.
5:54:20
And you could use some of the same techniques that we talked about in the ensemble section on looking at the, you know, creating applied percentile agreement and percentile levels an agreement and threshold level plots for a source term ensemble.
5:54:45
Hmm, hmm, Well.
5:54:53
I know when we, when we do these runs, they do seem a little bit slow.
5:54:58
But, on the other hand, ah, they only, it seems slow, and, and I think, in relation to our eyes, like, Oh, it's not finished yet, but if you think about how long it takes, actually, to run like a numerical weather prediction model or how much computational power.
5:55:15
And we're just running this, you know, basically on our little by little laptop, it's actually pretty amazing that we can, um, we can even get that's number of simulations done within the workshop time period. Since he got a little bit spoiled.
5:55:32
You know you really, it's fairly easy to condition yourself to want everything like really, really, really, really fast.
5:55:39
And an emergency response situation, so, you know, we do try to have enough the computational power that these simulations would run faster if you, if you needed results in response to a serious incident, needed them within 10 minutes or so.
5:55:59
Um.
5:56:12
Well, we have four minutes until risks fester start. We're at 90%, let's check our message file.
5:56:25
Where are we?
5:56:30
So.
5:56:33
Ah, got about 37,000 particles there now and.
5:56:46
Oh, you know what?
5:56:46
I forgot to the person I'm summit doing the particle dump files, um.
5:56:54
So, test output, more than 10.
5:56:58
Sorry, 31.
5:57:01
Ah!
5:57:13
So 28, OK, so we're doing 24 hour, So it shouldn't be about 30 hours, right?
5:57:18
And we're on the 28th power.
5:58:14
Alice, this is Mark, yeah.
5:58:16
To your point about being spoiled about computational time, you could talk a little bit about the difference between Lagrangian Models and Larry and Model and how this model, you know, at least, it's just following the plume.
5:58:31
As opposed to trying to calculate everything everywhere in every single grid square.
5:58:36
So, it was designed to run fast.
5:58:38
So, if we say this is slow, try running these things, know, with the big alert, are there a model?
5:58:48
Not a contest, but. Well, OK, yeah, I guess I can talk a little bit about the difference.
5:58:57
Right, so, Anna, and now I, Larry, and model, I wish I had shared yet a nice slide for this, I think, for one of the presentation, Spark and the eulerian model.
5:59:09
You're defining these grid squares, and basically you have to do the computation for every single, um, square in your domain.
5:59:19
um, for every time step.
5:59:21
So even if there's no mass, even if your plume isn't in that square, you have to do the calculation.
5:59:29
And you also have to define your model domain.
5:59:33
Um, beforehand, so, if you don't know how big your plame's gonna get, it's possible, you can define your own larian. Domain too small, and you're playing with kind of like run off of it.
5:59:47
What the Lagrangian model, you know, are following these fluid parcel.
5:59:52
L C is computational particles that are kind of representing a fluid parcel, and the computational power that you need is just directly related to the number of particles that you have, and so you're doing a calculation for every computational particle in your domain, um.
6:00:12
So, especially for, kind of, any, and so it doesn't really cost you anything, when I defined my concentration grid, um.
6:00:24
No, most of the time You'll notice the Center Lat Lon.
6:00:31
Oh, that was Sarah Sarah, actually here we did a finite kind of a small one, but like when I am doing a volcanic eruption, which could be like, a global event. And it could have pretty long range transport. I actually will define the span to be, like, global, Like, I'll have a concentration grid that's going all the way around the world.
6:00:51
I don't have any computational costs for doing that, because the program is, oh, it's just computing stuff for each particle, no matter where the particle is on the concentration grid.
6:01:04
It's not doing a computation for each, you know, content concentration grid square.
6:01:11
Um, so, that's a little bit of a difference.
6:01:16
Now, sometimes when you get really long range transport, and you get things like spreading out, like all over the place, and you do have computational particles, you know, everywhere, then the computational benefits can be a little bit less clear.
6:01:31
Um?
6:01:36
And, like I said, I was talking about this earlier.
6:01:40
The way we sort of turn the position of the computational particles into a mass distribution is probably not like the most computationally efficient way to do it.
6:01:50
There are those kernel density estimators. Are we working on something?
6:01:54
You're saying Gaussian mixture models to do it and you can actually get away with using less particles.
6:02:02
The reason I haven't really caught on that much, I think.
6:02:04
As just because, it's a little bit more complicated, and it's just, even though you save on the computational power, the tradeoff just hasn't seemed worth it to people because these Lagrangian molecules are pretty computationally efficient, anyways.
6:02:19
So saving even more computational power hasn't, you know, then that attractive. When the trade off, as, you know, implementing something, that's a little more complicated to do.
6:02:30
I guess that could change. You know, if you really want to, especially now that these are really getting into, people are really exploring how far they can push the resolution of the models. You know, we get more and more people asking, How can I do? Can I do?
6:02:48
No 50 meter vertical resolution can aid to, know really high resolution and then you do need a lot more particles and so some of these techniques like the kernel density estimators and the Gaussian mixture models, they might start looking, you know, worth the extra effort to implement.
6:03:09
OK, so I completed the egg.
6:03:13
I probably tots longer than that, ticket to Complete, um.
6:03:20
Gonna quit out of these things.
6:03:22
And we'll just kind of actually showed you what the we're waiting for that to complete, But we already know what the answer is. But let's just look at it anyways.
6:03:32
So we wanted to convert that to datum.
6:03:36
And we're going to create the data file and we're going to compute the statistics. And again, we see and, you know, the correlation coefficient is, all, right?
6:03:45
But the actually has, the biases is really, really, really large.
6:03:53
And feel free, if we look at the scatter plot, we'll see that like all the measurement points are down here.
6:04:00
But then, like I said, if I change this to like guess change and let's have a factor of one thousand.
6:04:10
Um.
6:04:13
No, then you're scatter plot looks like this which has much nicer.
6:04:20
And then.
6:04:23
We'll skip the next section, which is basically just increasing the emissions quite a bit by changing that line and that, that last, this last entry in the location box, increasing it.
6:04:40
And now also, now it's increasing the emissions increases at that every day says here, under forecasting diez.
6:04:52
All right, I will pass it back to Mark for the end of the day.
6:04:56
All right, thank you very much, Alice.
6:05:01
Excellent. Excellent job, very interesting. And dust storms are becoming bigger and bigger.
6:05:07
Um, issue as well as fires, And I didn't even realize this, but in some areas of the United States, and of course other places around the world does storms can be a big issue.
6:05:19
I mean, not only for health, of course, but also things like if it happens near a big road and people suddenly can't see.
6:05:28
We see reports every year of big traffic accidents and things that are caused by, by dust storms.
6:05:34
So, it's a, it's a big issue, and we're doing our best to try to make high split, be able to do a better job of estimating the emissions, and we hope it doesn't become more and more useful in doing so.
6:05:49
I'll take back the E, control, and let me quickly move this file over and show.
6:06:03
Ah.
6:06:07
Yes.
6:06:10
OK, and so, um, this is the, just take a few minutes here to wrap up that this third day of of the workshop, and I want to just show a few slides about some simulations regarding deposition.
6:06:30
Um, here what I'm showing is a a series of a month long simulation and um, one line is looking at sulfur dioxide emissions and then the deposition will actually concentration downwind.
6:06:49
Um, the orange line is looking at particulate matter one micron, green of five microns, red, 10 microns and purple of 25 microns.
6:07:02
And what this is showing you is it's the maximum concentration over the entire month.
6:07:08
As a function of distance down, we end and the point that I'm trying to show here is that for small particles, like one micron particles, or even even five micron particles.
6:07:20
No, the deposition is not that much different than for, for a gas. And it's not that dramatic.
6:07:29
In fact, I think I show here ah, yeah, here is A Maybe even a better one to show with. The blue would be no deposition at all.
6:07:42
Then the orange and green are for S O 2 and 1 micron particles.
6:07:49
If you can see this, you can't really see, really any difference at all hardly between the no deposition.
6:07:55
And when you turn on that on the deposition, both dry and wet deposition. You do start to see something with 10 micron particles because there are Saddling out.
6:08:05
Um, but, but even then, you know, it's, it's, it's dramatic, but it's also, you know, um, even if you didn't include it, it's only, like a factor of two or something like that.
6:08:19
Um, and I'm showing actually here on two different grids.
6:08:23
This is from 0 to 50 kilometers out, and this is out to 250, uh, kilometers out.
6:08:30
I think the point really here is that you absolutely want to try to include no deposition and things in your modeling.
6:08:40
We recognize that it's uncertain, but it also doesn't make a huge difference in the results.
6:08:45
Sometimes, especially if you're looking at sort of close in first 50 kilometers, 100 kilometers, which is often some of the biggest public health impacts, and because concentration impacts, especially for, um.
6:09:04
Emergency response applications, have a couple of other here, once you and I'll share this presentation with you later. I also think haven't the report on this that goes into more detail about this. So, let me, let me stop that.
6:09:17
And, let me just say, I will share these slides.
6:09:21
With you just shows you a bunch of examples of what happens when you do include deposition or, or transformations and things, and how much it really impacts the the results.
6:09:32
Um, so, um, like we've said many times, you know, we're trying to put stuff like these presentations and things on the workshop webpage.
6:09:43
And we'll try to post things that were shown today on that page.
6:09:46
So you'll should be available for you for you tomorrow.
6:09:51
And, a quick recap of logistics, which I've seen, you've seen this slide a number of times.
6:09:56
We are asking you to, if you can, ask the questions in the forum and we're getting quite a lot of questions in the forum, which is great.
6:10:04
We're able to answer them much more deep detail and also other people can really benefit from from seeing the answers.
6:10:10
And I think that's it.
6:10:12
Here's a screenshot from earlier about the workshop webpage, so tomorrow, um, we're going to be looking at radioactive pollutants and dose search for accidents from nuclear power plants.
6:10:29
And we'll be looking also at volcanic modeling volcanic eruptions for volcanic ash, which can have really big impacts for aircraft.
6:10:38
Um, will end up with a section on a lot of different custom simulations that get into some kind of interesting aspects of how you can run, run high split.
6:10:50
And then we'll have a section at the very end where you'll be able to ask your questions kind of live.
6:10:58
And we'll have our panel of instructors try to answer them. We'll do the best job we can.
6:11:05
A lot of times, the questions you're asking are some of the same questions that we have, and so R might be, yes, that's a great question.
6:11:15
Please try to find the answer to that question and let us know what you find.
6:11:18
I mean, we're all a lot of us are doing research, and a lot of you are doing research, and a lot of us have some of the same questions about what's the best way to do something, or what are the uncertainties? And how do we reduce the uncertainties?
6:11:33
So, oh. Yes. Can I just I just wanted to put in a little bit of a plug for the Section 16.
6:11:41
It's not just about volcanoes um.
6:11:44
I think we've had this issue in the past where we say that's about volcanoes.
6:11:49
And then, there are people who are very specifically interested in volcanoes, but not everybody is. And they just kinda skip that section.
6:11:55
It also has a lot about the M at times file, as well as defining like a particle size distribution, which are relevant to other applications, thank you.
6:12:08
Thanks for mentioning that. That's exactly right.
6:12:10
In fact, I should've renamed the, this section I think I think you actually did rename the section.
6:12:16
And I didn't catch up with that, that the emit times file, which you saw in the, in the wildfire smoke and in the dust and then Alice will be explaining it in much more detail.
6:12:28
I think in the volcanic eruption section.
6:12:31
It's a really powerful way to specify the emissions for high split, and no simple menu that you saw where you're just sort of looking at one or a couple of pollutants and you're specifying an emission amount.
6:12:46
And then in a duration, um, too simple for most cases.
6:12:52
So, with the times file, it lets you specify really, you know, any number of locations with the starting and stopping of emissions at any given time, for each location, and you can restart the emissions, then the emissions can be you know varying in each record.
6:13:11
So, if it emits, you know, a certain amount for a couple of hours, and then another amount for another couple of hours, you can put that in that file. So, absolutely.
6:13:20
It's a very powerful method, and when you start getting into more complicated simulations, you definitely need to use use that.
6:13:28
OK, and I think that is the last, so here's what we did today, and here's what we'll be doing tomorrow.
6:13:36
And with that, I think we'll say goodbye for today. We hope we see you tomorrow.
6:13:43
And, like I said, we will post as many of these presentations as we can on the books on that page, OK, so with that, say Goodbye, and we'll see you tomorrow.