NASA ARSET Overview and Applications of GLDAS Groundwater Data Products at Regional Scale Part 2

0:02 Speaker 1: Hello everyone, Welcome back to this training on monitoring groundwater

0:06 Speaker 1: changes for water resources management. This is part two of

0:10 Speaker 1: the training and the topic today is overview and applications

0:14 Speaker 1: of Global Lend Data Assimilation System groundwater data products at

0:18 Speaker 1: regional scale. This is a meta meta again and we're

0:22 Speaker 1: fortunate to have doctor Matheurodeal once again as our guest

0:25 Speaker 1: speaker today. Last week we had an overview analysis of

0:32 Speaker 1: NASA terrestrial water storage data from Grace and Grace follow

0:36 Speaker 1: on missions. Today we will focus on Gilda's groundwater data

0:41 Speaker 1: and next session will be on opera disc to monitor

0:45 Speaker 1: groundwater changes. As I mentioned last week, there will be

0:49 Speaker 1: one homework that will be posted on the training website

0:52 Speaker 1: on thirtieth of April and it will be due on

0:54 Speaker 1: fifteenth of May. A certificate of completion will be awarded

0:58 Speaker 1: to those who attend all live sessions and complete the

1:02 Speaker 1: homeburg assignment before the two date. Just a brief review

1:08 Speaker 1: of what we saw last week. We had an overview

1:12 Speaker 1: of GRACE missions and data products by doctor Rottel and

1:16 Speaker 1: he also talked about data applications to flood and rout

1:20 Speaker 1: We had a demonstration of Grace and GRACE fall on

1:23 Speaker 1: data Analysis tool or GRACE Interactive browser. Using that, we

1:28 Speaker 1: examine in terrestrial water storage for river basin Colorado River Basin,

1:33 Speaker 1: and we also looked at area of Ogallala aquifer and

1:38 Speaker 1: time series for that region and we saw that groundwater

1:45 Speaker 1: has strong multi year variability and there is a decreasing

1:51 Speaker 1: trend as well in this region. We went through an

1:55 Speaker 1: exercise that you must have worked on also. Now will

2:00 Speaker 1: start with today's session its overview and applications of Gilda's

2:05 Speaker 1: two point two its version two point two groundwater data

2:08 Speaker 1: products at regional scale. Our objectives for today are that

2:14 Speaker 1: by the end of this part, participants will be able

2:17 Speaker 1: to identify characteristics of groundwater data product from GRACE assimilated

2:22 Speaker 1: gildas system monitor inter annual to ticital changes in groundwater

2:28 Speaker 1: at regional scales using Gilda's data product. Our outline is

2:34 Speaker 1: that doctor Rottel is going to provide a review of

2:38 Speaker 1: key aspects of GRACE and GRACE. Follow on then overview

2:42 Speaker 1: of land data assimilation system about GRACE data assimilation in

2:47 Speaker 1: Elda's and groundwater data and then we will have a

2:50 Speaker 1: demonstration of how to analyze Gilda's based groundwater data, how

2:56 Speaker 1: to access it analyze it, visualize it using Giovanni. It's

3:01 Speaker 1: a web twol and we will also use jis to

3:04 Speaker 1: do some post analysis. Just a note about asking questions.

3:10 Speaker 1: Please put your questions in the questions box and we

3:13 Speaker 1: will address them at the end of the webinar. Feel

3:17 Speaker 1: free to enter your questions as we go and we

3:19 Speaker 1: will try to get to all the questions during the

3:22 Speaker 1: question and answer session. After the webinar, the remainder of

3:26 Speaker 1: the questions will be answered in the Q and a

3:28 Speaker 1: document which will be posted on the training website about

3:31 Speaker 1: a week after the training. With that, I want to

3:36 Speaker 1: introduce our speaker for today, doctor Matthew Roddel. Doctor Matthew

3:41 Speaker 1: Roddell is the Deputy Director of Earth Sciences for Hydrosphere,

3:45 Speaker 1: Biosphere and Geophysics, or HBG at NASA Gorder Space Blight Center.

3:51 Speaker 1: HBG comprises more than three hundred and fifty scientists and

3:55 Speaker 1: engineers within five laboratories involved in remote sensing and numerical

4:00 Speaker 1: modeling of land and ocean processes and related applications. Doctor

4:05 Speaker 1: Rodel has also served as Associate Deputy Director for HBG

4:10 Speaker 1: and as Chief of the Hydrological Sciences Lab. He is

4:14 Speaker 1: a member of the science teams for NASA's Grace follow

4:18 Speaker 1: On Mission and future Grace Continuity or Gray Sea Mission.

4:23 Speaker 1: He leads the Global Land Data Assimilation System and projects

4:26 Speaker 1: focused on monitoring groundwater storage changes, mapping and forecasting drought wetness,

4:33 Speaker 1: and detecting climate related variations in the watcycle. Doctor rodel

4:38 Speaker 1: is a past associate editor for the Journal of Hydrology

4:43 Speaker 1: and a current editor for the Journal of Hydrometology. He

4:51 Speaker 1: has also chaired the Hydrology program for the American Geophysical

4:55 Speaker 1: Union or AGU Fall meeting from twenty nine to twenty eleven,

5:00 Speaker 1: and led various national and international scientific working groups. He

5:05 Speaker 1: received a Presidential Early Career Award for Scientists and Engineers

5:10 Speaker 1: in two thousand and six, a NASAJSFC Earth Science Achievement

5:14 Speaker 1: Award in two thousand and seven, a Robert H. Goddard

5:18 Speaker 1: Award for Exceptional Achievement in Science in twenty eleven, and

5:23 Speaker 1: an Arthur S. Fleming Award for Outstanding Federal Service in

5:27 Speaker 1: the area of Basic Science in twenty fifteen. He was

5:32 Speaker 1: elected to the rank of AGU Fellow in twenty twenty two.

5:36 Speaker 1: Dotor Rottel has more than one hundred and sixty peer

5:39 Speaker 1: reviewed publications and appears on clariveet Web of Sciences twenty

5:44 Speaker 1: eighteen to twenty three lists of highly cited researchers. He

5:48 Speaker 1: holds a BS in Environmental science from the College of

5:51 Speaker 1: William and Mary and a PhD in Geological sciences from

5:56 Speaker 1: the University of Texas and Boston. With that wind night

5:59 Speaker 1: talk mate.

6:05 Speaker 2: Okay, thank you, Amita. So I'm Matt Rodell, and I'm

6:08 Speaker 2: going to be presenting on the GLEDS two point two

6:13 Speaker 2: groundwater product in which we assimilate data from the GRACE

6:16 Speaker 2: and GRACE follow on satellite missions. So in a nutshell,

6:21 Speaker 2: GLS two point two is one of the versions of

6:24 Speaker 2: the Global Land Data Simulation System. In this case, it's

6:29 Speaker 2: the land Information system running the Catchment Land Service Model.

6:33 Speaker 2: It's a global simulation excluding Antarctica and Greenland where there's

6:38 Speaker 2: permanent ice cover. The resolution is a quarter degree by

6:42 Speaker 2: quarter degree and it's about twenty five by twenty five kilometers.

6:46 Speaker 2: The time period of GAILEDS two point two is two

6:49 Speaker 2: thousand and three to present. For the input meteorological forcing,

6:54 Speaker 2: it uses the European Center for Media Weather Forecasts analysis data,

7:00 Speaker 2: which is actually proprietary. We get that through A through

7:03 Speaker 2: a special agreement with them, and in this case, JAILS

7:09 Speaker 2: two point two assimilates data from the Grace and Grace

7:12 Speaker 2: follow On satellite missions, which are trust your water storage anomalies.

7:18 Speaker 2: So I'm just going to go back, and you've heard

7:20 Speaker 2: from a meta about Grace and Grace follow On, but

7:22 Speaker 2: I just want to remind you of a few key

7:24 Speaker 2: aspects before we get into Jailed's two point two. So

7:29 Speaker 2: the big thing is that it's Grace and Grace follow

7:33 Speaker 2: On are unlike any other Earth observing satellite observations that

7:37 Speaker 2: we have because they're not looking downward at the Earth.

7:40 Speaker 2: It's actually two satellites, one following the other about two

7:44 Speaker 2: hundred kilometers apart, about five hundred kilometers initial altitude, and

7:49 Speaker 2: the distance between the satellites is continuously measured by a

7:53 Speaker 2: k Ban microwave ranging system and also a laser system

7:56 Speaker 2: on Grace follow On. It's measured down to a very

8:00 Speaker 2: high precision about the size of a red blood cell.

8:04 Speaker 2: This and we're talking about two and a chlometers, so

8:05 Speaker 2: it's a huge distance and a very fine measurement. And

8:10 Speaker 2: these measurements made every five seconds or so as the

8:14 Speaker 2: satellites orbit the Earth, and over the course of the month,

8:17 Speaker 2: there's enough information in this inner satellite range, along with

8:22 Speaker 2: the precise positioning of the satellites, for it to be

8:26 Speaker 2: put into a super computer model and come up with

8:30 Speaker 2: a new map of Earth's gravity field based on the

8:33 Speaker 2: information on the satellite orbits. And basically what's happening is

8:39 Speaker 2: if there's a mass anomaly at the land surface or

8:42 Speaker 2: the ocean. Over the ocean, that mass anomily will cause

8:47 Speaker 2: a difference in the gravity field which will affect the

8:49 Speaker 2: satellites as they pass over. So the example here shows

8:53 Speaker 2: a mountain range. There's more mass in the mountain range.

8:56 Speaker 2: That means as the two satellites approach, the first satellite

8:59 Speaker 2: will be pulled for a little bit, and then distance

9:01 Speaker 2: between the two satellites will increase as they sort of

9:05 Speaker 2: pass over. The second satellite will catch up and even

9:07 Speaker 2: get a little closer as the first satellite is kind

9:09 Speaker 2: of held back, and then things will even out afterward.

9:14 Speaker 2: So that's how they would detect a mountain range. But

9:18 Speaker 2: because the measurements are so precise, they can actually detect

9:21 Speaker 2: changes in the gravity field from month to month that

9:23 Speaker 2: are caused by changes in terrestrial water storage. What is

9:28 Speaker 2: trust water storage? You ask, well, treastor water storage is

9:31 Speaker 2: the sum of all the groundwater, so moisture, snow, surface water,

9:36 Speaker 2: vegetation of water storage, et cetera. And to visualize this,

9:41 Speaker 2: we have time series from the state of Illinois where

9:44 Speaker 2: they actually have measurements of all these quantities on the ground.

9:48 Speaker 2: So they will to measure groundwater and the somisture measurements

9:52 Speaker 2: and snow measurements, and we created a time series. Show

9:56 Speaker 2: to the top here where the blue is the groundwater,

10:00 Speaker 2: and then superposed on top of that is saw moisture

10:03 Speaker 2: and red, and then the white you can just barely

10:05 Speaker 2: see on top is the or snowater equivalent water storage.

10:10 Speaker 2: In fact, surface water in Illinois is small enough you

10:12 Speaker 2: wouldn't even see it on this time series on a

10:16 Speaker 2: relative based relative to the other components. So Grace is

10:20 Speaker 2: measuring the total treasher water storage. So that's sort of

10:22 Speaker 2: the top contour here. It does not measure the individual components.

10:26 Speaker 2: It can't break it down into these components the way

10:29 Speaker 2: we have with the time series that's shown here. And

10:33 Speaker 2: I also want to point out that you can see

10:35 Speaker 2: there's a seasonal cycle each year treasure water storage goes

10:42 Speaker 2: gets higher and lower just based on how much precipitation

10:45 Speaker 2: is falling and evaporation and runoff. But you can also

10:47 Speaker 2: see there's interannual variability. And the two yellow oval show

10:51 Speaker 2: the first one shows a major drought in the Midwest

10:54 Speaker 2: during nineteen eighty eight eighty nine, and the second one

10:58 Speaker 2: shows nineteen ninety three when there was ascid flooding in

11:01 Speaker 2: the upper Mississippi River basin. And the Grace observations look

11:09 Speaker 2: like the animation at the lower rate. And so there

11:13 Speaker 2: are these sort of large blobs of red where it's

11:15 Speaker 2: there's less water than than normal and blue where there's

11:19 Speaker 2: more water than normal. And if you sort of follow

11:22 Speaker 2: along with the months there, you can see that you

11:24 Speaker 2: can sort of see the seasonal cycle there where for example,

11:27 Speaker 2: in South America it's it's wetter in the boreal summer

11:32 Speaker 2: and and and drier around November, December, January. So these

11:38 Speaker 2: are the the monthly anomalies, but there's also this sort

11:42 Speaker 2: of inter annual variability on top of that. And I

11:48 Speaker 2: want to reiterate that Grace and Grace follow and measure

11:51 Speaker 2: all the trust water storage components. And we've we've taken

11:54 Speaker 2: a look at you know, where those different components have

11:57 Speaker 2: a greater or lesser impact. It turns out that on

12:01 Speaker 2: a monthly basis, so I'm moisture, it tends to dominate

12:04 Speaker 2: in temperate regions. Snow not surprisingly dominates in the high

12:12 Speaker 2: latitude and in alpied regions. Surface water is actually dominant

12:17 Speaker 2: in certain areas like the Amazon, and then groundwater storage

12:21 Speaker 2: is relatively important everywhere, but becomes even more important on

12:26 Speaker 2: longer time skills, particularly annual and multi annual time skills.

12:31 Speaker 2: But GRACE cannot separate these components, and that's one of

12:34 Speaker 2: the reasons why we have gel US two point two,

12:36 Speaker 2: which let me get into that. First, I want to

12:39 Speaker 2: talk about land data simulation systems in general. So obviously

12:44 Speaker 2: geld AS is a global land data simulation system. Land

12:49 Speaker 2: data simulation systems have land surface models as there as

12:54 Speaker 2: their main foundation. And these models are are complex computer

13:00 Speaker 2: models that simulate the processes at and below the land surface.

13:06 Speaker 2: Or they take as input the meteorological what we call forcings,

13:10 Speaker 2: so things like precipitation, downward, short wave, belong wave radiation,

13:16 Speaker 2: wind speed and that sort of thing, humidity. Those are

13:19 Speaker 2: the inputs to the model, and then they The models

13:24 Speaker 2: basically simulate what happens to that water and energy effort

13:28 Speaker 2: hits the land surface using systems of equations that describe

13:32 Speaker 2: our understanding of these these physical processes. They break down

13:37 Speaker 2: the Earth into into a grid, and then within each

13:40 Speaker 2: grid they often have subgrid heterogeneity where they model separately

13:45 Speaker 2: the different vegetation types. And then they also are have

13:51 Speaker 2: a vertical stratification, so they might model the surface so

13:56 Speaker 2: moisture and roots on saw moisture and the case. So

14:00 Speaker 2: the cash mad Landservice model, which we're gonna be talking

14:02 Speaker 2: about here the groundwater storage. The reason we have these

14:09 Speaker 2: data assimilation systems is largely for integration of data. So

14:15 Speaker 2: if you think about it, NASA has you know, something

14:18 Speaker 2: twenty something earth observing satellites in orbit making various types

14:23 Speaker 2: of measurements, but it's not always the measurement that we want.

14:27 Speaker 2: For example, if we were interested in snowater equivalent, which

14:30 Speaker 2: is basically the amount of water stored in the snow

14:34 Speaker 2: on the ground, we don't currently have a satellite observation

14:39 Speaker 2: for that. We can we can observe snow cover, but

14:43 Speaker 2: not but not the more important snowater equivalent. So what

14:47 Speaker 2: a landed is assimilation system would do. Is it uses

14:51 Speaker 2: inputs like precipitation, and we might do some intercomparison optical

14:56 Speaker 2: merging of the various precipitation products that are available. That's

15:00 Speaker 2: one of the inputs to the ld as. They are

15:04 Speaker 2: also I didn't mention it here, but there are also

15:07 Speaker 2: some static inputs things like soil types and vegetation types

15:11 Speaker 2: and that sort of thing. And then there are other

15:15 Speaker 2: forcings like short radiation. And then we can do data assimilation,

15:21 Speaker 2: which is where you take the model estimate of the

15:26 Speaker 2: the quantity you know, in this case snowater equivalent, and

15:29 Speaker 2: then you made have an observation of things like the

15:32 Speaker 2: snow cover and you use that observation to constrain the model.

15:37 Speaker 2: And neither of them are perfect. The model has its

15:40 Speaker 2: own imperfections because you know their errors and the forcing

15:43 Speaker 2: and errors and the way we simplify complex processes in

15:46 Speaker 2: the model. But you know, and the observation is not

15:51 Speaker 2: perfect either. But when you combine the two, taking into

15:54 Speaker 2: account their their error characteristics and their advantages this is vantages,

16:01 Speaker 2: you can come up with a best estimate and that's

16:03 Speaker 2: the point of data similation. And then finally we might

16:06 Speaker 2: have on the ground estimates we can use to evaluate

16:10 Speaker 2: the output of the ld ASS and use that to

16:15 Speaker 2: improve it and and tweak it to get better results.

16:21 Speaker 2: So let's get onto grace and GRACE follow on data

16:24 Speaker 2: a similation. And this is really what makes GILDS two

16:28 Speaker 2: point two different from the other geled as products. So again,

16:36 Speaker 2: the JAILS two point two uses the land information system

16:40 Speaker 2: with the cash Bent landsurface model and it runs a

16:44 Speaker 2: three dimensional ensemble and ensemble common smoother. And I'm not

16:49 Speaker 2: going to get into all the mathematics of this, but

16:52 Speaker 2: just to be simple about it, it's a way that

16:57 Speaker 2: you run an ensemble of of model runs over the

17:02 Speaker 2: course of a month where you perturb the forcing we need,

17:06 Speaker 2: adjusted the forcing a little bit to account for errors,

17:09 Speaker 2: and you come up with twenty different runs and then

17:12 Speaker 2: those are used to estimate the uncertainty on the model side.

17:17 Speaker 2: We also have an estimate of the uncertainty in our

17:19 Speaker 2: observation and we use those too uncertainty piece of uncertain

17:24 Speaker 2: information to sort of weight the model versus the observed estimate.

17:30 Speaker 2: And the smoother allows us to take a monthly Grace

17:34 Speaker 2: trust water storage anomaly observation and use it to produce

17:40 Speaker 2: results on a daily or sub daily basis. So if

17:44 Speaker 2: you're interested in learning more about data simulation, I encourage

17:48 Speaker 2: you to look at their reference here, and there are

17:51 Speaker 2: various other resources on data similation. So here are some

17:57 Speaker 2: results comparing in the up or left the trust to

18:01 Speaker 2: order Storge anomalies from the what we call the open loop,

18:04 Speaker 2: which is the model running without data similation. The upper

18:08 Speaker 2: right shows the Grace observations and you can see the

18:11 Speaker 2: difference between the two. There's a lot more of fine

18:14 Speaker 2: scale variability on the left and the model output, whereas

18:19 Speaker 2: the grace observations are pretty coarse. There they look very

18:22 Speaker 2: smooth out. The lower right shows the results when we

18:26 Speaker 2: assimilate the grace data into the model, and what we

18:30 Speaker 2: like to see here is sort of the best of

18:31 Speaker 2: both worlds. It looks a lot like the grace observations

18:36 Speaker 2: at the larger scale, but we have the finer scale

18:39 Speaker 2: variability from the model. And the reason the model is

18:42 Speaker 2: able to have that is because it has inputs like

18:45 Speaker 2: precipitation that are available with much finer scale. In fact,

18:50 Speaker 2: all the inputs are much finer scale than the grace observations,

18:54 Speaker 2: so we hope that the bottom right, which is what

18:56 Speaker 2: you would get with the Gail does two point two,

18:58 Speaker 2: is again sort of the best of both worlds. So

19:04 Speaker 2: we've done quite a bit of evaluation of the results.

19:08 Speaker 2: I'm going to show you a few different examples, this

19:10 Speaker 2: one from Kumar at All twenty sixteen, showing results if

19:15 Speaker 2: you look on the right of comparisons of the surface

19:18 Speaker 2: so moisture and roots and sol moisture output from the

19:23 Speaker 2: Grace data assimilating model with on the ground observations, and

19:29 Speaker 2: in this case we're looking at the correlation with the

19:32 Speaker 2: ground observations. And when you see warmer colors, that means

19:37 Speaker 2: the data similation results are better than at least in

19:42 Speaker 2: terms of correlation, they're better than the open loop or

19:44 Speaker 2: no data similation results from the model. So in most

19:48 Speaker 2: cases we're getting either grays which means no difference, or

19:51 Speaker 2: warmer colors. So that's that's what we want to see.

19:57 Speaker 2: Looking at the groundwater output in this paper by biling

20:02 Speaker 2: Lee at All in twenty nineteen, they actually looked at

20:06 Speaker 2: about four thousand wells and twenty two regions and basins

20:10 Speaker 2: around the world. You can see in the map here

20:12 Speaker 2: that they're not evenly distributed around the world. There are

20:15 Speaker 2: some areas where like southern India, where there are a

20:18 Speaker 2: lot of observations available. Similarly, the US is pretty well covered,

20:23 Speaker 2: but then large areas of the world where there's really

20:26 Speaker 2: no groundwater data available for comparison in these In these regions,

20:34 Speaker 2: you know, with the with these data would be very

20:36 Speaker 2: careful about which observations we used. We had some selection

20:40 Speaker 2: criteria that were very important in order for a proper evaluation.

20:45 Speaker 2: One is that we only used wells that were installed

20:49 Speaker 2: in unconfined okfurs, meaning in a water table open to

20:53 Speaker 2: the basically open to atmospheric pressure, as opposed to a

20:57 Speaker 2: confined oc forur, which has a confine layer, and it

21:00 Speaker 2: has much different sort of reasons for water storage changes

21:07 Speaker 2: observed in the well we want to have. We want

21:13 Speaker 2: to have wells that were not directly influenced by pumping

21:18 Speaker 2: or injections. And then we also looked for places where

21:22 Speaker 2: there was at least a five year data record and

21:24 Speaker 2: typically wanted something that was monthly or bi monthly in

21:28 Speaker 2: order to get a recentably good time series for comparison.

21:31 Speaker 2: So that actually cut out quite quite a number of

21:34 Speaker 2: wells that we would have otherwise been able to use. Further.

21:37 Speaker 2: In order to convert a depth to water and a

21:40 Speaker 2: well to a change in water storage that we can

21:44 Speaker 2: compare with the model output or with grace, we need

21:47 Speaker 2: to know the specific yield, and so we carefully went

21:50 Speaker 2: through and looked at the aqua types and an estimated

21:53 Speaker 2: specific yield for each of those to do that conversion.

21:58 Speaker 2: So here are some of the results for a few

22:01 Speaker 2: different regions. And so the upper left is the Toilkentans

22:07 Speaker 2: River basin, upper right is the Upper Mississippi River basin,

22:10 Speaker 2: and then the Uganda and the lower left. In each

22:14 Speaker 2: of these plots, they're red is the catchment landsforst model

22:18 Speaker 2: open loop output meaning no data similation. The blue line

22:23 Speaker 2: is the is the data similation output, and then the

22:27 Speaker 2: black dots are the mean observed groundwater storage anomalies. And

22:34 Speaker 2: in these three cases at least, you can see that

22:37 Speaker 2: that when we do the data simulation, the blue is

22:39 Speaker 2: much closer to the black dots, and that means that

22:42 Speaker 2: the data sublation is doing what it's supposed to be doing.

22:46 Speaker 2: On average around the world, the Great Data Simlation reduced

22:50 Speaker 2: the root means square error by thirty six on a

22:55 Speaker 2: regional and basin, river basins and regions around the world

23:00 Speaker 2: and increased the correlation by sixteen percent on average. However,

23:04 Speaker 2: I will point out that as you've seen all three

23:07 Speaker 2: cases here, the amplitude of the open loop is higher,

23:13 Speaker 2: so it was overestimating the amplitude of the groundwater variability.

23:19 Speaker 2: And and this is these are the types of cases

23:22 Speaker 2: where the grace data simlation performed very well. It didn't

23:25 Speaker 2: do as well in the opposite direction, where you know,

23:29 Speaker 2: for example, if the if the red line had been

23:31 Speaker 2: had a low too low amplitude, the grace data simlation

23:35 Speaker 2: didn't necessarily make it larger to meet the observations. Looking

23:43 Speaker 2: at this a different way, here are twenty two regions

23:46 Speaker 2: around the world shown in the in the table in

23:49 Speaker 2: the middle. Here for each of these regions, and the

23:52 Speaker 2: top we're showing the data simulation Rootming square error in

23:57 Speaker 2: blue and the open loop Rootming square error in orange,

24:01 Speaker 2: and then the same for the correlation and the bottom

24:05 Speaker 2: plot here and you can see the average statistics again

24:10 Speaker 2: on the upper right. So looking at this a little

24:15 Speaker 2: bit differently, we can look at these regional scale means

24:18 Speaker 2: and we can see that the comparing the DA and

24:22 Speaker 2: the open loop Routman square error and correlation changes, where

24:27 Speaker 2: the rootmy square error decreases on average by thirty six percent,

24:31 Speaker 2: the correlation increases by sixteen percent. If we look at

24:35 Speaker 2: point scale means, meaning all the groundwater wells individually, the

24:41 Speaker 2: rootmate square or decreases by ten percent correlation increases by

24:45 Speaker 2: twenty two percent. So either way we can see that

24:48 Speaker 2: around the world on average, the data simulation is doing

24:51 Speaker 2: what it's supposed to be doing and making the simulated

24:54 Speaker 2: output groundwater storage better match the observations. This is from

25:03 Speaker 2: another another publication by riger at All in twenty fifteen,

25:09 Speaker 2: looking at the GILDS two point two output prior to

25:15 Speaker 2: a major flood event in the Missouri River basin in

25:18 Speaker 2: twenty eleven, and what we see is again the open

25:22 Speaker 2: loop from the cash bt lansfers model. It underperforms the

25:27 Speaker 2: data similation results. If you look at the right the

25:31 Speaker 2: right most column in the table at the bottom. In

25:34 Speaker 2: comparison with the USGS groundwater anomalies for the Missouri Roof basin,

25:39 Speaker 2: the open loop had a correlation coficient at point five eight,

25:43 Speaker 2: whereas the data similation output at two point eighty six,

25:46 Speaker 2: so pretty significant increase in correlation and likewise a decrease

25:52 Speaker 2: Inman square error switching gears a little bit. I just

25:59 Speaker 2: wanted to talk about sort of how this this GLS

26:04 Speaker 2: two point two got started. Originally, we were looking at

26:07 Speaker 2: just the continental the United States to do the Great

26:10 Speaker 2: Data Similation, and we had a project where one of

26:13 Speaker 2: the goals was to create drought and wetness indicator maps

26:17 Speaker 2: that could then be used by the authors of the

26:19 Speaker 2: US Drought Monitor. So it was a similar sort of

26:23 Speaker 2: a configuration of the lands Information System running the catchment

26:29 Speaker 2: landsurface model. In this case, we used NLS two meteorological

26:34 Speaker 2: forcing and and we produced these maps that are basically

26:40 Speaker 2: the percentiles of wetness for each location. So what that means,

26:47 Speaker 2: and I'm looking at the on the lower right of

26:50 Speaker 2: the surface so moisture, rugelans, soysure and groundwater maps. What

26:54 Speaker 2: these percentiles mean is is, for example, if you're the

26:57 Speaker 2: fifth percentile, that means it's only been drier at that

27:01 Speaker 2: location and the same time of year five percent of

27:05 Speaker 2: the time in the long term record. Similarly, if it's

27:09 Speaker 2: at the ninety eighth percentile, it's only been wetter at

27:12 Speaker 2: that location in time of year two percent of the time.

27:16 Speaker 2: So we produce these maps in the same way we

27:19 Speaker 2: have in this example, the lower right is showing the

27:22 Speaker 2: grace dress or water storage anomalies in May twenty fourteen.

27:26 Speaker 2: It's a pretty coarse, pretty coarse scale. It's actually even

27:30 Speaker 2: coarser than you get the feeling for with this gridded product.

27:35 Speaker 2: We assimilate that into the model, and then the model

27:37 Speaker 2: is able to give us the nice looking maps on

27:40 Speaker 2: the lower right, and then compare that with the twentieth

27:46 Speaker 2: of May twenty fourteen US Drought Monitor in the upper right.

27:50 Speaker 2: You can see the comparison's pretty good in terms of

27:52 Speaker 2: the overall patterns. There's more information in the grace data

27:58 Speaker 2: simlation maps because they have different levels, whereas the drought

28:03 Speaker 2: monitors really supposed to represent sort of all levels uh

28:07 Speaker 2: and and also the drought monitor does not indicate, you know,

28:13 Speaker 2: wet periods. So since since about twenty eleven, we've been

28:19 Speaker 2: running running this and it's actually used as one of

28:22 Speaker 2: the inputs to the US Drought Monitor. The Drought monitor

28:25 Speaker 2: authors look at look at our products as well as

28:28 Speaker 2: many other things when they're drawing their weekly maps. And

28:33 Speaker 2: if you go to the National Drought Mitigation Center website

28:37 Speaker 2: if you have the If you have this this presentation,

28:42 Speaker 2: you can hover over Drought Mitigation Center in the bottom

28:44 Speaker 2: and you'll get actually great NASA GRACE dot U n

28:48 Speaker 2: L dot E d U. If you go there, you

28:50 Speaker 2: can you can choose some different options that the radio

28:54 Speaker 2: buttons to the top and so the left is the

28:57 Speaker 2: the US Shallow groundwater a drought indicator, and you can

29:02 Speaker 2: flip between that roottone so moisture and surface solar moisture.

29:06 Speaker 2: The middle is basically drought indicators Global drought indicators based

29:10 Speaker 2: on Gildas two point two. And the right is is

29:14 Speaker 2: an experimental forecast product for one, two and three months

29:18 Speaker 2: into the future. If we compare the US Drought monitor

29:26 Speaker 2: with with our basically GRACE d A drought and wetness indicators,

29:34 Speaker 2: we can look at the different drought categories and if

29:37 Speaker 2: you look at the lower right there you can see

29:40 Speaker 2: wetness percentile ranges for each drought category. And once plotted

29:45 Speaker 2: here is is in red the US drought monitor and

29:51 Speaker 2: in blue the the GRACE uh DA based drought indicators,

29:58 Speaker 2: and in the black dash line shows you where you

30:02 Speaker 2: want the average to be. This is This is the

30:05 Speaker 2: accumulation over the entire U S So, so it basically

30:08 Speaker 2: shows the percentage of area in the US versus time

30:13 Speaker 2: in each of these plots. And so where where you

30:18 Speaker 2: see a higher number, that means there's more drought you

30:20 Speaker 2: know around the US, and lower number means there's less

30:23 Speaker 2: drout And you can see that in general, especially if

30:27 Speaker 2: you look at the lower right all drought categories. You

30:30 Speaker 2: can see that the the grace based drought indicators are

30:35 Speaker 2: pretty well correlated with the US Drought Monitor. The other

30:39 Speaker 2: thing I'll notice is the US drought monitor numbers or

30:42 Speaker 2: or percentage areas tend to be a lot higher. That's

30:46 Speaker 2: actually a sort of a flaw in the US Drought Monitor.

30:49 Speaker 2: The dash black line is supposed to show the long

30:53 Speaker 2: term mean for each of these. So if you looked

30:56 Speaker 2: at again at the lower right the percentile ranges, for

30:59 Speaker 2: example Z zero percent, that implies that about ten percent

31:04 Speaker 2: of the US should be in drought at any given time.

31:07 Speaker 2: But look at the top left the top left plot.

31:11 Speaker 2: There you can see that the d zero drought category

31:15 Speaker 2: for the red line, which is the US drought monitors,

31:18 Speaker 2: is almost always above ten percent. So this is just

31:21 Speaker 2: sort of shows the tendency of the US Drought Monitor

31:23 Speaker 2: authors to sort of overestimate drought and they may have

31:28 Speaker 2: a little bit of I won't call it bias, but

31:32 Speaker 2: they've may able to feel a little bit of pressure

31:34 Speaker 2: to put different, you know, counties in drought because it

31:38 Speaker 2: affects the affects the distribution of disaster funding. So but overall,

31:47 Speaker 2: the you know, the important message here is that that

31:50 Speaker 2: our drought maps tend to line up pretty well with

31:53 Speaker 2: the the US Drought Monitor. This is one of my

31:59 Speaker 2: last slides and it's an animation of the US drought

32:04 Speaker 2: or or sorry, of the the Grace based drought indicators

32:07 Speaker 2: over the US. The left is the Rootston so moisture

32:11 Speaker 2: wetness percentile. On the right is the shallow brownwater wetness percentile.

32:17 Speaker 2: And uh, what's cool here, I think is is you

32:19 Speaker 2: can sort of see the the evolution of of of

32:22 Speaker 2: drought and pluvial or wet events and the routland so

32:27 Speaker 2: moisure which is in you know, basically direct communication with

32:32 Speaker 2: the atmosphere changes much more quickly. So if you have

32:35 Speaker 2: a you know, you might be in drought, but any

32:37 Speaker 2: a bit have big rainstorm come through, boom you go

32:39 Speaker 2: straight from red to blue. But on the right there

32:42 Speaker 2: the shallow groundwater. You can see that that because it's

32:45 Speaker 2: not in direct communication, because groundwater is a bigger store

32:48 Speaker 2: of water and it and it reacts and and evolves

32:51 Speaker 2: more and more slowly. You can see that these these

32:54 Speaker 2: droughts sort of take a while to evolve and and

32:57 Speaker 2: uh and take you longer to to dissipate than they

33:02 Speaker 2: do with their roots and so and much. Sure, so

33:03 Speaker 2: I think this is this is sort of fun to watch.

33:06 Speaker 2: And we've done this for the global scale as well,

33:09 Speaker 2: but it's kind of nice to see the defined scale

33:12 Speaker 2: granularity for the US. So let me summarize what I've

33:19 Speaker 2: talked to you about with the GRACE Data Similation and

33:24 Speaker 2: the GLEDUS two point two product. Grace and GRACE follow

33:28 Speaker 2: on are unique in that they observe changes in trust

33:32 Speaker 2: your water storage, including all forms of trust or water storage,

33:36 Speaker 2: which is not something that any other satellite system can do.

33:40 Speaker 2: We use land data similation systems to integrate data from

33:44 Speaker 2: various sources, which allows us to fill spatial and temporal

33:48 Speaker 2: gaps using algorithms that represent our knowledge of the physical

33:53 Speaker 2: processes in jailed US two point two. We assimilate the

33:57 Speaker 2: GRACE and GRACE follow on observation of trust water storage,

34:02 Speaker 2: and it allows us to basically downscale those trust water

34:07 Speaker 2: storage observations in space and in time, and also perform

34:11 Speaker 2: vertical disaggregation so we can separate the groundwater from the

34:15 Speaker 2: sow moisture and snow, et cetera. We've used the gilled

34:20 Speaker 2: ass two point two sow moisture and groundwater output and

34:25 Speaker 2: we evaluate those extensively and use them for producing drought products.

34:29 Speaker 2: But they do have some key limitations versus that the

34:33 Speaker 2: model does not simulate water management, so especially in areas

34:37 Speaker 2: where there might be intensive irrigate irrigation that's it's fed

34:41 Speaker 2: by groundwater pumping and you might have a long term

34:44 Speaker 2: decline in groundwater storage, or maybe there's a seasonal aspect

34:49 Speaker 2: to it. You know, when that's the growing season, you

34:52 Speaker 2: might have a big decline in groundwater storage because there's

34:55 Speaker 2: pumping in an agricultural region that's not simulated by the

34:59 Speaker 2: cash maint LANDSERCE model and and therefore well it is

35:05 Speaker 2: detected by the GRACE observations. If it's something that's happening

35:09 Speaker 2: with some fine scale variability, we're not able to effectively

35:14 Speaker 2: downscale the Grace and GRACE follow On observations over those

35:18 Speaker 2: sorts of those sorts of regions and that sort of case.

35:22 Speaker 2: The second issue is that the models not stimulate confined

35:25 Speaker 2: AWKWAD storage changes and those those could be an important

35:29 Speaker 2: component of the trust or orders trade sort of changes

35:32 Speaker 2: observed by Grace and Grace follow On in certain regions.

35:38 Speaker 2: And I'll just end by saying that these output are

35:40 Speaker 2: available on the Goddared Earth Sciences Data Information Services Center,

35:46 Speaker 2: and we can also get the drought and wedness indicator

35:49 Speaker 2: maps and forecasts from the National Drought Medication Center website.

35:55 Speaker 2: Thank you so much for listening, and I'm going to

35:57 Speaker 2: pass the mic back over to a meet and I'll

36:01 Speaker 2: be here for the question answer session.

36:05 Speaker 1: Thank you so much, Matt for your presentation and information

36:09 Speaker 1: about GRACE assimilated giled as groundwater storage data. Next, we

36:15 Speaker 1: are going to have a brief demonstration of access, analysis

36:19 Speaker 1: and visualization of giled as groundwater storage data using a

36:24 Speaker 1: web tool geomanny. We will also do some processing using qgis. First,

36:31 Speaker 1: we'll learn to create maps and time series of groundwater

36:34 Speaker 1: storage data.

36:35 Speaker 3: Using geovanny.

36:38 Speaker 1: We will look at groundwater storage differences over inter annual

36:43 Speaker 1: to interticketal time scale. The case study I've chosen is

36:48 Speaker 1: a small river basin, Plemat River Basin, and I'm going

36:52 Speaker 1: to share my screen with you to talk about the river.

36:59 Speaker 1: So Clemat River basin. It straddles California and Oregons or

37:06 Speaker 1: northern California southern Oregon, and it's a small river basin.

37:11 Speaker 1: If you recall last week we saw increase interactive browser.

37:17 Speaker 1: Major river basins are covered in there. If you're interested

37:21 Speaker 1: in looking at region of your own interests, you can

37:25 Speaker 1: use childas to do that, and that's what we are

37:28 Speaker 1: trying to see here. So this river there is groundwater issues,

37:32 Speaker 1: but also there are multiple dams on this river. It's

37:36 Speaker 1: used for water, is used for agriculture, flood controlled hydropower, recreation,

37:41 Speaker 1: fish and wildlife habitat, and also has tribal treaty rights.

37:47 Speaker 1: Just to see the basin, this is the basin that

37:50 Speaker 1: we will be focusing on, and for that, I'm going

37:55 Speaker 1: to start with Giovanni. To access Gilda's groundwater storage data.

38:00 Speaker 1: This is the Giovanni web page and you have this

38:03 Speaker 1: link in the presentation slide that you can access. One

38:07 Speaker 1: note is that you will have to register to NASA

38:10 Speaker 1: Earth Data to be able to download images or data

38:13 Speaker 1: using Giovanni. Just go through this page quickly. If you

38:19 Speaker 1: know what data you're interested in, which parameter, then you

38:23 Speaker 1: can just enter here and search by keyword. If you

38:27 Speaker 1: are just searching in general, you can go to observations,

38:32 Speaker 1: different disciplines, measurements platforms, so there are multiple options that

38:36 Speaker 1: you can explore. If you are just looking for different

38:40 Speaker 1: data sets. If you already know, you can just enter

38:43 Speaker 1: here and I'm going to do that.

38:46 Speaker 4: Gilled As sorry, and you can search.

39:04 Speaker 1: You will see two options. We want this data assimilated.

39:08 Speaker 1: Grease Assimilated goes from two thousand and three to January

39:14 Speaker 1: twenty twenty six. The daily data units are in millimeter

39:19 Speaker 1: and resolution is a quarter degree. Once you choose the parameter,

39:24 Speaker 1: you can pick analysis and visualization option, so select plot here.

39:31 Speaker 1: You have multiple options as well. You can make maps.

39:34 Speaker 1: You can make time series, difference of map and difference

39:38 Speaker 1: of time series. There are other you can compare and

39:42 Speaker 1: make correlation maps, scatter plots. There are hermular diagrams. You

39:47 Speaker 1: can make instagram zono mean and vertical cross sections from

39:53 Speaker 1: different data sets. If there are three D data sets.

39:57 Speaker 1: We are going to stick with time averaged map for now,

40:00 Speaker 1: and here's the time selection window from and two. So

40:08 Speaker 1: let's do one thing. We're going to first look at

40:11 Speaker 1: two different decades and average groundwater storage data for that.

40:16 Speaker 1: So I'm going to start with February first, two thousand

40:22 Speaker 1: and three when the data starts, and go to two

40:30 Speaker 1: thousand and thirteen December thirty first, so we look at

40:38 Speaker 1: first period, and then we'll go from twenty fourteen to

40:41 Speaker 1: twenty twenty five to see the difference between groundwater storage

40:46 Speaker 1: over Klamath River basin. This is the window that allows

40:52 Speaker 1: you to select region of your interest, and there are

40:54 Speaker 1: multiple ways to do that. You can enter longitude and

41:00 Speaker 1: altitude if you know exact box that you are interested in,

41:04 Speaker 1: or you can go to the map by clicking here

41:07 Speaker 1: and make a box here or any region that you're

41:11 Speaker 1: interested in and those code in its will appear here.

41:17 Speaker 1: Another way to do is to click here. And this

41:21 Speaker 1: has different shape files already available in Giovanni, so it's

41:27 Speaker 1: different countries, lakes and reservoirs, land only see only tribal lands,

41:35 Speaker 1: are there, US states, are there, World regions and so

41:39 Speaker 1: ezy world regions, and most importantly what we are going

41:42 Speaker 1: to look at is watersheds. Different river basins are given here.

41:48 Speaker 1: For Klemath River, which is a smaller river and not

41:52 Speaker 1: included in here, I have already looked at code in

41:57 Speaker 1: its and then I'm going to just enter those here.

42:01 Speaker 1: So these are longitudes. So it's west one twenty five

42:07 Speaker 1: to one twenty box and this is forty two forty

42:12 Speaker 1: three point five north. And if you click here, this

42:15 Speaker 1: will show you where the box is. So here between

42:22 Speaker 1: Oregon and California. Here's where Clemat River basin is. So

42:26 Speaker 1: that's what we're picking. And once you have picked all

42:30 Speaker 1: the options, all you have to do is just say

42:33 Speaker 1: plot data and it launches a workflow. So here I

42:40 Speaker 1: have already done that to save time. It takes depending

42:44 Speaker 1: on the data resolution, it may take a few seconds

42:48 Speaker 1: to maybe a minute or two. But I have already

42:52 Speaker 1: done that. And so this is the map over the

42:56 Speaker 1: region that we chose. And this particular map is from

43:01 Speaker 1: two thousand and three to twenty thirteen December, and you

43:07 Speaker 1: can see this is the color bar here there is

43:12 Speaker 1: red is lower groundwater storage and in the upper part

43:17 Speaker 1: of the basin there is more groundwater storage.

43:21 Speaker 3: As you can see.

43:23 Speaker 1: Once you have this map, what you can do is

43:26 Speaker 1: you can go back to data selection and I'm going

43:30 Speaker 1: to go to this window. It's the same map we

43:33 Speaker 1: got here. You can go back to data selection and

43:38 Speaker 1: keep the same. But now look at the next decade.

43:43 Speaker 1: And here I'm going to pick two thy fourteen January first,

43:51 Speaker 1: sorry first, two thousand and twenty five December thirty first.

44:01 Speaker 1: So this is next ticket from what we saw previously.

44:05 Speaker 1: And again you can say plot data and you will

44:10 Speaker 1: be you will see the map again. This map is

44:16 Speaker 1: given here. It starts in twenty fourteen, and you can

44:20 Speaker 1: see that. Of course, the pattern's pretty much the same magnitude.

44:25 Speaker 1: We will compare in QGIS in a couple of minutes. Next,

44:30 Speaker 1: what we want to do is look at time series

44:34 Speaker 1: of groundwater storage data. And for that, I'm going to

44:38 Speaker 1: pick the entire time period that data is available, so

44:44 Speaker 1: February first, and let's just go to.

44:47 Speaker 3: Twenty twenty five December.

44:50 Speaker 1: So here I'm going to go to select plot options

44:55 Speaker 1: and say time.

44:57 Speaker 3: Series area averaged.

44:59 Speaker 1: So this going to average groundwater storage data over that

45:03 Speaker 1: entire box, and it's going to give a time series

45:07 Speaker 1: and you can plot it and the workflow launches again

45:11 Speaker 1: to save time.

45:12 Speaker 3: I've already.

45:15 Speaker 1: Turn the time series analysis and then if you see

45:18 Speaker 1: this is the time series here, it's in millimeter groundwater

45:24 Speaker 1: storage from two thousand and three all the way to

45:28 Speaker 1: two thousand and twenty five December. Clearly you do see

45:32 Speaker 1: annual variations. These are daily data as we saw, so

45:36 Speaker 1: you can see annual variation, but you can also see

45:40 Speaker 1: that there is inter annual variation in groundwater storage. And

45:47 Speaker 1: what you can do is if you're interested in any

45:50 Speaker 1: particular year, you can go back to data selection and

45:57 Speaker 1: just picks it two thousand and five to two thousand

46:00 Speaker 1: and six and make a map or make a time

46:05 Speaker 1: series for that particular year, so you can do it

46:08 Speaker 1: year by year, and then you can have different maps

46:12 Speaker 1: for different years. So can go to results, and this

46:18 Speaker 1: is a quick way to see the time series. Next

46:23 Speaker 1: thing that we are going to see is how to

46:25 Speaker 1: download this. You can just go to download. You can

46:28 Speaker 1: save this file as net CDF for PNG image, or

46:31 Speaker 1: you can have a CSV file so you can then

46:35 Speaker 1: either go to Excel and plot it the way you

46:38 Speaker 1: like it, or you can then parse it by time

46:42 Speaker 1: different temporal domain. You can do all that with CSV

46:45 Speaker 1: as well for map. Also you can download data in

46:51 Speaker 1: various formats, either net, CDF data file or KMZ for

46:56 Speaker 1: Google Earth PNG just image as we saw on screen,

47:00 Speaker 1: and geotif something that you can work with either r

47:05 Speaker 1: GIS or QGIS. And I have already saved these files

47:11 Speaker 1: in geot format so we can look at them in QGIS,

47:15 Speaker 1: and have also downloaded CSV file for time series. So

47:22 Speaker 1: you will be following these steps for another river basin.

47:27 Speaker 3: So just keep note.

47:28 Speaker 1: That you will be downloading some of these data. And

47:31 Speaker 1: so now what I'm going to do is let's go

47:35 Speaker 1: back to data selection. And just we looked at map,

47:40 Speaker 1: we looked at time series. We know how to change

47:42 Speaker 1: time and how to change special domain. So now I'm

47:46 Speaker 1: going to close Giovanni and now go to qgis to

47:50 Speaker 1: see how groundwater storage changed between those two periods that

47:54 Speaker 1: we looked at. So let me share QGIS project with you.

48:02 Speaker 1: So this is the QGIS project. You need to install

48:07 Speaker 1: QGIS on your computer. There is a link in your

48:11 Speaker 1: presentation for that as well. You can follow steps and

48:14 Speaker 1: have QGIS on your computer.

48:17 Speaker 4: Once you have.

48:18 Speaker 1: Qgs you can click open and this window will open.

48:23 Speaker 1: You can go to project and open a new project.

48:26 Speaker 1: That is what I've done, and this is the new project.

48:29 Speaker 1: Now you can go to web here first and add

48:35 Speaker 1: quick map services.

48:37 Speaker 3: I'm going to use Google Roadmap.

48:43 Speaker 1: You can choose other options as well. And now you

48:49 Speaker 1: can follow the instruction in your exercise. If you do

48:55 Speaker 1: not have this plug in quick Map services, and you

49:00 Speaker 1: can go to plug in and install that particular plug in,

49:05 Speaker 1: and there's an instruction given to you for that, but

49:09 Speaker 1: you can follow that and you will be able to

49:12 Speaker 1: have a background map or base map on qgis like this.

49:18 Speaker 1: Now we can start adding the data layer we saved

49:21 Speaker 1: from geoanny. So for that, go to layer and first

49:26 Speaker 1: we're going to add those map or raster layers. So

49:30 Speaker 1: add layer and add raster layer. And here I have

49:36 Speaker 1: saved them in a Folderoda's geomanny and if you look

49:42 Speaker 1: at the names, this is twenty thirteen, so two thousand

49:49 Speaker 1: and three to twenty thirteen.

49:52 Speaker 3: And then this is the other ticket.

49:54 Speaker 1: You can highlight and shift and both and add both

50:00 Speaker 1: at the same time. Open and add, and now you

50:05 Speaker 1: can close this window and you can you can just

50:09 Speaker 1: use your computer mouse to zoom in, or you can

50:15 Speaker 1: use this to zoom in here and use this hand

50:21 Speaker 1: symbol to move your map. So, now these rasters are added.

50:31 Speaker 1: So once the rasters are added, we can change color

50:36 Speaker 1: symbology here.

50:37 Speaker 3: So go to.

50:40 Speaker 1: Control click and go to properties and in the properties

50:44 Speaker 1: select symbology. Here we're going to pick single band pseudo color.

50:52 Speaker 1: And these are the mean and max values for the

50:55 Speaker 1: groundwork storage for this particular raster. We can make it

51:00 Speaker 1: round number.

51:03 Speaker 3: If you like.

51:04 Speaker 1: Sorry about that. And then this is the color ramp.

51:12 Speaker 1: You can change different colors we are going to keep

51:15 Speaker 1: the same and here, instead of continuous, let's have equal

51:23 Speaker 1: interval and you can decide how many intervals or classes

51:28 Speaker 1: you want. I'm going to pick twenty one. One more

51:33 Speaker 1: thing is that you can select label precision. I'm going

51:37 Speaker 1: to say two, so that number of digits after the

51:42 Speaker 1: decimal point. And these are the groundwater storage intervals for

51:48 Speaker 1: different colors that we see. Once you do that, you

51:51 Speaker 1: can say okay, okay. One more thing is you can

51:54 Speaker 1: do legend setting here and instead of continuous, just turn

52:00 Speaker 1: that off so that we can see different intervals here.

52:04 Speaker 1: And now you can say okay, So now you can

52:08 Speaker 1: see that this raster has colors showing different groundwater storage

52:15 Speaker 1: range from four fifty to twelve fifty. Next, what you

52:22 Speaker 1: can do is click again and go to styles and

52:26 Speaker 1: copy style. And now go to the next raster layer,

52:32 Speaker 1: go to styles and paste the same style. So you

52:37 Speaker 1: have the same groundwater storage scale for both both the

52:42 Speaker 1: maps for you to compare. And now you can click

52:45 Speaker 1: and compare. We're also going to see the differences soon,

52:49 Speaker 1: but this is the way to do symbology and at

52:54 Speaker 1: the colors that you like. One more thing is that

52:58 Speaker 1: I have downloaded vector shape file for Klemath River basin

53:06 Speaker 1: and I'm going to display that. So this is the

53:09 Speaker 1: shape file. I'm going to pick through a layer and

53:14 Speaker 1: add vector layer and then open and add and so

53:20 Speaker 1: you can move the layers around. This is the Clemat

53:23 Speaker 1: River basin. You can again use properties and symbology to

53:28 Speaker 1: make it transparent transparent fiel and you can change.

53:38 Speaker 3: Line colors if you like.

53:39 Speaker 1: I'm just going to keep it the same for now.

53:42 Speaker 1: And here now you can see the boundary of the river.

53:45 Speaker 1: So here in the upper part of the basin you

53:48 Speaker 1: see more groundwater storage compared to the lower part, which

53:52 Speaker 1: is in California. You can also clip your rast or

53:58 Speaker 1: crop your raster so that it is within the river basin.

54:02 Speaker 1: And for that you can go to raster extraction and

54:08 Speaker 1: then say clip raster by mask layer. You can choose

54:13 Speaker 1: input layer. This is the twenty two thousand and three

54:18 Speaker 1: to thirteen and this is the mask layer, which is

54:22 Speaker 1: the shapefile of Clamath River basin. And you can keep

54:31 Speaker 1: everything else as default and then run and you will

54:37 Speaker 1: see that there is a raster that is cropped here.

54:43 Speaker 1: This is just within this of course, that is because

54:47 Speaker 1: of the resolution. There are some areas which are not there.

54:50 Speaker 1: You can interpot it and feel that up.

54:52 Speaker 3: But for now, this.

54:54 Speaker 1: Is good approximation for the river basin. You can you

55:00 Speaker 1: name this clipped us. You can control click and go

55:04 Speaker 1: to rename layer and click here again to add the.

55:08 Speaker 3: Name you like.

55:09 Speaker 1: I'm going to pick gildas groundwater Storage two thousand and

55:16 Speaker 1: three to thirteen. And you can use the same color

55:22 Speaker 1: style for this layer. Suppose a copy style and then

55:32 Speaker 1: it's the same style. So now you can see that.

55:37 Speaker 1: Let's turn these two layers off so that we can

55:42 Speaker 1: see the groundwater storage within this river basin. I have

55:48 Speaker 1: done the same processing with raster layer from twenty fourteen

55:55 Speaker 1: to twenty twenty five. So just to extract and here

56:03 Speaker 1: I picked twenty fourteen to twenty five raster river basin

56:08 Speaker 1: as mask. One more thing I've done is I set

56:13 Speaker 1: the file name here in clicked mask and then run

56:18 Speaker 1: and I have this raster here. Also I have added

56:22 Speaker 1: symbology to it the same using styles for copy and paste.

56:28 Speaker 1: So we have these two layers. Now you can see

56:33 Speaker 1: how groundwater is changing. And to do this calculation. To

56:38 Speaker 1: see the differences between them, go to raster and raster calculator.

56:44 Speaker 1: Here window will open with raster bands and these are

56:48 Speaker 1: the operators for calculations. I'm going to add a temporary

56:54 Speaker 1: name for difference file as g astiff and then now

57:02 Speaker 1: you can go here. When you click on the raster

57:06 Speaker 1: you're interested in. So this is twenty fourteen to twenty five.

57:10 Speaker 1: Double click and it appears here minus two thousand and

57:14 Speaker 1: three to thirteen. And now you can say, okay, now

57:20 Speaker 1: this raster has been calculated, and we want to make

57:23 Speaker 1: sure that it has the same CRS or cod reference system,

57:29 Speaker 1: so that is EPSG four three two six, And now

57:33 Speaker 1: you can see the raster here as you can see

57:38 Speaker 1: everywhere groundwater storage is negative.

57:42 Speaker 3: That means groundwater has decreased.

57:44 Speaker 1: Between two thousand and three to thirteen decades and then

57:48 Speaker 1: two thousand and fourteen to twenty five. You can symbolize

57:53 Speaker 1: it just as we did before. To see numbers better,

57:57 Speaker 1: you can go to symbology single band pseudo color. You

58:02 Speaker 1: can keep the same range and same here we can

58:07 Speaker 1: go to equal interval and maybe.

58:10 Speaker 3: Say tan for now and.

58:16 Speaker 1: Have proper numbers displayed in the color table and say okay,

58:22 Speaker 1: and now you can see that the differences are shown here.

58:28 Speaker 1: As you can see, there are some regions where it

58:31 Speaker 1: is yellow and red. These are the much larger differences.

58:38 Speaker 1: So in this area, in this area, and this so

58:43 Speaker 1: you can look at how ground or changed in different

58:47 Speaker 1: parts of the basin. So this concludes our demonstration of

58:53 Speaker 1: how to get data from geovanny and analyze the data

58:58 Speaker 1: in QGIS.

59:00 Speaker 3: You will be doing.

59:01 Speaker 1: Similar exercise with another river basin. I'm going to go

59:05 Speaker 1: back to the presentation window.

59:08 Speaker 3: Thank you.

59:11 Speaker 1: This brings us to the end of our demonstration of

59:15 Speaker 1: how to access and visualize child dust groundwater. You also

59:20 Speaker 1: have an exercise on the training website that you can download.

59:24 Speaker 1: You will be working on this exercise for your homework,

59:27 Speaker 1: so you can download and follow the steps. Next session

59:34 Speaker 1: will be on overview of observational products for end users

59:38 Speaker 1: from remote sensing, analysis or opera surface to displacement products

59:43 Speaker 1: or dispase. The product we'll be focusing on. The homework

59:48 Speaker 1: will be posted on the training web page on thirtieth

59:51 Speaker 1: of April and it will be due on fifteenth of May.

59:54 Speaker 1: Homework answers must be submitted via Google forms. A certificate

59:59 Speaker 1: of com will be awarded to those who attend all

1:00:02 Speaker 1: three live webinars and complete the homework assignment by the deadline.

1:00:07 Speaker 3: You will receive a.

1:00:08 Speaker 1: Certificate where email approximately two months after completion of the course.

1:00:15 Speaker 1: We thank doctor Rodel for his presentation on Gilda's Groundwater

1:00:19 Speaker 1: and also for his presentation on Grace missions last week.

1:00:25 Speaker 1: Yes the contact information that you have any questions, please

1:00:29 Speaker 1: contact us. Our set, website and YouTube links are given

1:00:33 Speaker 1: here for questions, comments, or to share how you have

1:00:37 Speaker 1: applied our trainings to your work or studies. Email NASA

1:00:41 Speaker 1: Arset at gmail dot com and join our quarterly newsletter.

1:00:46 Speaker 1: To stay up to date on our latest trainings. You

1:00:49 Speaker 1: can join our list serve at this email address and

1:00:52 Speaker 1: follow the steps, and thank you very much for attending

1:00:57 Speaker 1: this session. You have time to start with the exercise

1:01:01 Speaker 1: and then we will have question and answer session.

1:01:04 Speaker 3: Thank you WELLO welcome doctor Rutell.

1:01:20 Speaker 1: We are here in the question and answer session. I

1:01:24 Speaker 1: hope you all had some time to work on the

1:01:27 Speaker 1: exercise and have made some progress working with Giovanni downloading

1:01:32 Speaker 1: Gilda's data and you can process that in QGIS and

1:01:37 Speaker 1: save your results so that Homer questions can be answered

1:01:41 Speaker 1: based on that. And now we will start with question

1:01:46 Speaker 1: and answer session. We'll go through a few questions till

1:01:51 Speaker 1: the end of this will be in our time. So

1:01:54 Speaker 1: we're starting from question one, how does the assimilation work

1:01:58 Speaker 1: daily when Grace and Ways follow on data monthly?

1:02:04 Speaker 2: So first of all, hello everyone, can everyone hear me?

1:02:07 Speaker 2: I just want to make sure yes, yes, okay, great.

1:02:11 Speaker 2: So this is actually not an easy question to answer.

1:02:15 Speaker 2: We use something called an ensemble common smoother and and

1:02:21 Speaker 2: you can read about it in some of the some

1:02:25 Speaker 2: of the publications that we've that we've done on on

1:02:28 Speaker 2: Grace and Grace follow on data similation. But essentially it

1:02:32 Speaker 2: sort of mimics the observation type and and does a

1:02:37 Speaker 2: correction to the model, typically three times per month, and uh,

1:02:45 Speaker 2: you know, sort of based on the trajectory of the changes.

1:02:50 Speaker 2: So it looks at the difference between a the similation

1:02:55 Speaker 2: system looks a difference between the monthly average from the

1:02:58 Speaker 2: model and the monthly average from race really the month

1:03:03 Speaker 2: and month changes of each and then it makes a

1:03:07 Speaker 2: makes an adjustment to the model based on based on

1:03:11 Speaker 2: the difference and applying information on the uncertainty in both

1:03:19 Speaker 2: the model and the observation to do sort of a

1:03:21 Speaker 2: waiting between the two. So that's that's in a nutshell,

1:03:26 Speaker 2: but it's it's really quite a bit more complicated.

1:03:28 Speaker 3: Than that, Thanks matt Our.

1:03:32 Speaker 1: Next question is how well does roots and soil marsha

1:03:35 Speaker 1: from gild as represent seasonal soil motion dynamics in agricultural regions,

1:03:40 Speaker 1: considering differences between summer and winter crops and the impact

1:03:44 Speaker 1: of irrigation.

1:03:47 Speaker 2: So let me take the last part first. So, first

1:03:49 Speaker 2: of all, we do have an irrigation simulation routine in

1:03:55 Speaker 2: the Land Information System you know as the software behind

1:03:58 Speaker 2: giald ASK, but it hasn't been earned on in any

1:04:00 Speaker 2: of the the current jailed as versions, so JAILS two

1:04:03 Speaker 2: point two does not stimulate irrigation and therefore in irrigated regions, Yeah,

1:04:10 Speaker 2: I would definitely say that that there are going to

1:04:12 Speaker 2: be significant differences between you know, on the ground soil

1:04:17 Speaker 2: moisture where there's irrigation and what the model is is simulating.

1:04:24 Speaker 1: And uh.

1:04:27 Speaker 2: Yeah, summer in winter crops, so so yeah, there's multiple

1:04:31 Speaker 2: crop types, then then that's not going to be simulated

1:04:35 Speaker 2: in jailed as either. So that could have an effect,

1:04:39 Speaker 2: you know, but overall we are making use of precipitation data.

1:04:43 Speaker 2: They're basically observation based and and so the precipitation is

1:04:49 Speaker 2: going to have a large impact on the roots, on

1:04:51 Speaker 2: sol moisture, and so where there's where there's not irrigation,

1:04:55 Speaker 2: I think, you know, we do a region reasonable job

1:04:57 Speaker 2: of simulating the soil moisture, but but where there's irrigation

1:05:01 Speaker 2: and that that's it's going to be less than perfect

1:05:06 Speaker 2: by a lot.

1:05:10 Speaker 4: Yeah.

1:05:11 Speaker 1: Question four, is there a reference to look up the

1:05:15 Speaker 1: formula to convert well data using specific yields to compare

1:05:19 Speaker 1: with gield us.

1:05:23 Speaker 2: Well, the formula is very simple, but finding this specific

1:05:26 Speaker 2: yield data is that is the difficult part. So specific

1:05:29 Speaker 2: yield is going to be a number that ranges between

1:05:32 Speaker 2: about point oh two and point five, and it depends

1:05:38 Speaker 2: on on the type of the type of box for

1:05:42 Speaker 2: you know, basically the poorest media that the water is in.

1:05:44 Speaker 2: So something like a sand or gravel is going to

1:05:47 Speaker 2: have a high specific yield. It could be point you know,

1:05:49 Speaker 2: from put to five to point four. You know, it's

1:05:53 Speaker 2: a If it's a limestone ocher with huge you know,

1:05:56 Speaker 2: caverns in it, it could be closer to point five

1:05:59 Speaker 2: or even more. But uh, you know, if it's something

1:06:03 Speaker 2: like a clay or a sandstone and it's it's you know,

1:06:10 Speaker 2: not fractured or anything like that, then serpeific yield could

1:06:12 Speaker 2: be very low. So the formula that you ask about

1:06:16 Speaker 2: is is simply the the well, the water level, and

1:06:20 Speaker 2: the well multiplied by this specific yield, it converts to

1:06:25 Speaker 2: to to a storage value. So if you had a

1:06:29 Speaker 2: change of of ten centimeters of water and you're well,

1:06:35 Speaker 2: and you would a specific yield of point two five,

1:06:38 Speaker 2: then you multiply two point point two five by ten

1:06:41 Speaker 2: and you would have a two point five centimeter change

1:06:44 Speaker 2: in water storage. You directly compare with with the grace observation.

1:06:51 Speaker 2: But again spif gield is going to vary from place

1:06:54 Speaker 2: to place, so finding those stiff field values is difficult.

1:06:58 Speaker 2: You basically need to know something about the opera and

1:07:00 Speaker 2: make an assumption about this skild or you know there's

1:07:05 Speaker 2: actually been a well log and they've removed you know,

1:07:10 Speaker 2: when they tugg the well. You might be able to

1:07:12 Speaker 2: get more accurate number, but those are again hard to

1:07:14 Speaker 2: come by.

1:07:18 Speaker 3: Thanks Matt.

1:07:19 Speaker 1: Question five is does terrestrial water storage data also include

1:07:23 Speaker 1: information and water pollutants and does it vary based on

1:07:27 Speaker 1: differences in geographic location?

1:07:31 Speaker 2: Has no information on pollutants.

1:07:36 Speaker 1: Question six is when groundwater elevation drops, eventually stream based

1:07:41 Speaker 1: flow drops, then disappears and the stream runs dry. This

1:07:46 Speaker 1: has been observed in western and central Kansas. Has anyone

1:07:50 Speaker 1: looked at time series of stream flow data looking for

1:07:54 Speaker 1: gauges transitioning from perennial to intermittent flow regimes with base

1:07:59 Speaker 1: flow as an indicator of drought, climate change, or human impacts.

1:08:07 Speaker 2: I don't know the answer, and I don't mean to

1:08:08 Speaker 2: sound glib, but I would you know, to find the answer,

1:08:10 Speaker 2: I'd probably go to Google, scholar and and and look

1:08:14 Speaker 2: and see if if anyone's done anything like that, I

1:08:18 Speaker 2: don't know of it off the top of my head,

1:08:19 Speaker 2: but certainly people could have done that type of research.

1:08:22 Speaker 2: I'm not sure, thank you.

1:08:26 Speaker 1: When we calculate animal is in Jaileda's two point two,

1:08:30 Speaker 1: what years do we used for normalization?

1:08:37 Speaker 4: So?

1:08:37 Speaker 2: Uh, I think we've we've tried to align with the

1:08:42 Speaker 2: with the Grace years. Uh, you know the Grace baseline. However,

1:08:49 Speaker 2: so what we do is JELLS two point two is

1:08:52 Speaker 2: when we assimilate the Grace data, we have to basically

1:08:59 Speaker 2: align the model with the Grace and make an adjustment.

1:09:03 Speaker 2: So so for the we might take an averaging period

1:09:07 Speaker 2: of two thousand and two to twenty twenty. I don't

1:09:10 Speaker 2: remember the exact averaging period that we use exactly, to

1:09:13 Speaker 2: be honest. And we take the average and look at

1:09:17 Speaker 2: the anomalies relative to that baseline and then do the

1:09:20 Speaker 2: same thing with Grace. Use it, you know it could

1:09:24 Speaker 2: it would be the same baseline two to twenty twenty

1:09:26 Speaker 2: if that's what we chose. And then when we do

1:09:29 Speaker 2: the data simulation, we make an adjustment if the anomaly

1:09:33 Speaker 2: is off by one centimeter. Let's say that you know,

1:09:35 Speaker 2: we need to increase water storage by one centimeter or

1:09:39 Speaker 2: the model to better align with the Grace observation. We

1:09:43 Speaker 2: then add that one centimeter to whatever water storage value

1:09:48 Speaker 2: what was in the model, so the model doesn't have

1:09:52 Speaker 2: an anomaly. The model has an actual amount of water

1:09:57 Speaker 2: store in each soil layer and in its representation of

1:10:01 Speaker 2: the OKFUR and those numbers are also sort of arbitrary.

1:10:08 Speaker 2: I mean, they're not going to align with with the

1:10:11 Speaker 2: real world. But but that's how we make the adjustment

1:10:14 Speaker 2: is with the data simulation, as we look at the

1:10:16 Speaker 2: difference between the anomalies Grace and and and and the model,

1:10:20 Speaker 2: and then we add whatever adjustment we're making. We actually

1:10:24 Speaker 2: call that an increment. We add that increment to to

1:10:27 Speaker 2: the model. States. Yeah, so I'm not I didn't. I

1:10:33 Speaker 2: sort of danced around your question there. What years use

1:10:38 Speaker 2: for normalization? You could? You could basically use whatever years

1:10:40 Speaker 2: you wanted to if you wanted to, you know, average

1:10:44 Speaker 2: over a certain time period and and calculate the baseline

1:10:48 Speaker 2: you know that mean as a baseline and then subtract

1:10:50 Speaker 2: it to get to get anomalies.

1:10:55 Speaker 4: Mhmm, thanks Matt.

1:10:58 Speaker 1: Next question eight is if we more monthly estimates of

1:11:01 Speaker 1: soil moisture changes from the for the conus that include

1:11:06 Speaker 1: conditioning on remote sense soil moisture should be used gelled

1:11:11 Speaker 1: as two point two at a version of nal das.

1:11:17 Speaker 2: So I'm trying to read the question and it keeps

1:11:20 Speaker 2: moving around here.

1:11:21 Speaker 1: Question eight.

1:11:22 Speaker 2: Yeah, uh, well, you know, if you're if you're interested

1:11:32 Speaker 2: in the conus and and remotely sense so moisture. Uh yeah,

1:11:41 Speaker 2: this is diff difficult to answer. It depends on what

1:11:44 Speaker 2: remote sensing you're talking about. If you're you know, you're

1:11:46 Speaker 2: talking about the Grace remote sensing, then you'd probably want

1:11:48 Speaker 2: to use the Gilds two point two. But if you're

1:11:51 Speaker 2: talking about you know, uh snap so moisture, you know,

1:12:00 Speaker 2: the n old ASS has a higher resolution and somewhat

1:12:05 Speaker 2: better input data. So general, and people are interested in

1:12:11 Speaker 2: in the conus and they're not interested in the groundwater

1:12:16 Speaker 2: more in the so much or I'd probably point them

1:12:18 Speaker 2: to analtass. But they both have their own advantages and disadvantages.

1:12:26 Speaker 1: Great, thank you. Question nine. I'm looking to analyze drout

1:12:32 Speaker 1: patterns across southern Alberta and would like to compare grace

1:12:36 Speaker 1: based drout indicators with existing roundwater well records in the province.

1:12:42 Speaker 1: Are there any guidelines you can suggest for the best

1:12:45 Speaker 1: types of wells to use for such an exercise? Do

1:12:48 Speaker 1: I need to select between active and inactive wells for

1:12:52 Speaker 1: such analysis?

1:12:55 Speaker 2: Yeah, the the things that we look for in wells

1:12:59 Speaker 2: to here with GRACE data or number one, it should

1:13:03 Speaker 2: be a well that's in an unconfined oct FORER. So

1:13:08 Speaker 2: a lot of wells go deep and they're they're installed

1:13:11 Speaker 2: in confined octfers and those are not appropriate for comparison

1:13:16 Speaker 2: with GRACE observations. Secondly, you want to have a long

1:13:20 Speaker 2: you know, the longer the data record, the better you know.

1:13:22 Speaker 2: If it's you mentioned inactive wells, I guess you mean

1:13:27 Speaker 2: a well that's not observations aren't being made anymore, and

1:13:31 Speaker 2: if you know, if there were, you know, if there's

1:13:34 Speaker 2: a long time series of observations that happens to end

1:13:37 Speaker 2: in twenty nineteen or something like that, you could still

1:13:39 Speaker 2: compare with the older GRACE data that would be that

1:13:42 Speaker 2: would be fine, But you know, you'd really like to have,

1:13:48 Speaker 2: you know, if you're interested in in current conditions, and

1:13:51 Speaker 2: clearly you want to have a well that's still actively monitored.

1:13:56 Speaker 2: And you also, you know, depending on whether you're interested in, uh,

1:13:59 Speaker 2: you know, the seasonal cycle and short term variability versus

1:14:03 Speaker 2: you know, just long term trends. Then that sort of

1:14:07 Speaker 2: helps to determine whether we need you know, is one

1:14:10 Speaker 2: observation per year enough or do you want to have

1:14:12 Speaker 2: you know, at least seasonal observations and that sort of thing.

1:14:15 Speaker 2: But and then the final thing I'll mention is you

1:14:20 Speaker 2: don't want to use measurements for well that's actively pumped

1:14:24 Speaker 2: or right next to you know, a municipal water supply

1:14:27 Speaker 2: where they're doing a lot of pumping, or agricultural area

1:14:30 Speaker 2: where they're pumping a lot of water for irrigation, that's

1:14:33 Speaker 2: going to have you know, variability that's determined you know,

1:14:38 Speaker 2: by the pumping and not so much represent the larger

1:14:42 Speaker 2: scale ambient ambient you know, groundwater conditions.

1:14:51 Speaker 3: Thank you.

1:14:52 Speaker 1: On the next question is it's about exercise, and I

1:14:56 Speaker 1: think we've gone through that. So I'm going to question

1:14:59 Speaker 1: to I am unsure if I understood the reach of

1:15:03 Speaker 1: gildas Is it a model that incorporates different data sets?

1:15:07 Speaker 1: Are the outputs only for the US territory? If that

1:15:12 Speaker 1: is the case, can we modify code? So no, I

1:15:15 Speaker 1: think we answered that question Matt, if you want to

1:15:18 Speaker 1: look at it, that GILL does is a global landed

1:15:21 Speaker 1: assimilation system, incorporate different data sets and so data are

1:15:26 Speaker 1: available globally. You want to add anything, we can.

1:15:31 Speaker 2: No, that's you're correct, it was already. Yeah, GILS runs

1:15:34 Speaker 2: globally already so that you don't need to modify the

1:15:36 Speaker 2: code for other areas. And it does incorporate multiple different.

1:15:45 Speaker 1: Question eleven will skip that's about Gianni and q jis.

1:15:50 Speaker 1: The next one is question twelve. As leakes don't really flow,

1:15:55 Speaker 1: is it easy to capture its decrease in volume?

1:16:01 Speaker 2: Well, uh, if you're referring to based on a Grace observation,

1:16:08 Speaker 2: you know, first of all, most lakes, unless you're talking

1:16:10 Speaker 2: about the Great Lakes, are are too small to have

1:16:15 Speaker 2: a discernible effect on on the Grace observations. If it's

1:16:20 Speaker 2: a group of lakes, you know, like in parts of

1:16:22 Speaker 2: Africa where there you know, you know, several lakes close

1:16:26 Speaker 2: to each other in southeastern Africa, those those effects can

1:16:30 Speaker 2: be seen in the GRACE data for GLS two point two.

1:16:34 Speaker 2: We don't actually model lakes explicitly. Uh. And you know,

1:16:41 Speaker 2: as regards to your question, lakes don't really flow, they don't,

1:16:45 Speaker 2: you know, somebody doesn't have to flow for it to

1:16:47 Speaker 2: have a change in water storage. So if the lake

1:16:49 Speaker 2: level is going up and down, you know, like a

1:16:52 Speaker 2: like a reservoir level going up and down, that you know,

1:16:55 Speaker 2: that translate to a change in mass that is detectable

1:17:00 Speaker 2: grace theoretically, if it's a large enough change in mass, great,

1:17:06 Speaker 2: thank you.

1:17:06 Speaker 1: We're going to question fourteen. What is the latency of

1:17:12 Speaker 1: chilled US two point two can it be used for

1:17:14 Speaker 1: early morning.

1:17:16 Speaker 2: Jailled US two point two is typically made available within

1:17:21 Speaker 2: about a month of real time. I think although no,

1:17:26 Speaker 2: we have we have an early product. I should have

1:17:29 Speaker 2: answered this. It's either a month or it's either within

1:17:32 Speaker 2: a week or a month of real time that it's

1:17:34 Speaker 2: it's updated. So it might actually be weekly updates at

1:17:39 Speaker 2: this point. So it's it is more it's updated more

1:17:43 Speaker 2: frequently than than the data available from from grace because

1:17:46 Speaker 2: grace is has a latency of typically two to four months. Uh,

1:17:53 Speaker 2: and we can run up to neural time with chielled

1:17:57 Speaker 2: us two point two, but then it it takes a

1:17:59 Speaker 2: little longer for us to if we only run once

1:18:02 Speaker 2: a week, and then we have to upload it to

1:18:04 Speaker 2: the guest discs, so it might be a little longer

1:18:07 Speaker 2: than that. So for some things that could be useful

1:18:12 Speaker 2: for early warning for some applications, but you know, others

1:18:16 Speaker 2: might need you know, data within twenty four hours of

1:18:19 Speaker 2: real time for to be useful.

1:18:23 Speaker 1: Thanks Matt. I just wanted to clarify that in geomaney

1:18:26 Speaker 1: or gs disc there may be one or two months

1:18:31 Speaker 1: of latency for data, but if you go to NASA

1:18:34 Speaker 1: Earth Data Search, you find weekly data there. Question sixteen

1:18:45 Speaker 1: do we need know that that's also for exercise? We

1:18:47 Speaker 1: answered that similarly. Question okay, so question eighteen for the demonstration,

1:18:54 Speaker 1: I'm a little with that also answered. I'm sorry, she

1:18:58 Speaker 1: questioned nineteen or also has been answered. Question twenty. Are

1:19:03 Speaker 1: there correction for developments such as urban built out, mining,

1:19:07 Speaker 1: et cetera?

1:19:10 Speaker 2: This question twenty, Yeah, so we when we've looked at

1:19:16 Speaker 2: things like uh, you know, oil extraction or mining, even

1:19:23 Speaker 2: though those might be you know, large amounts of mass

1:19:26 Speaker 2: at a small scale, because it's such a small scale,

1:19:28 Speaker 2: it does not have an effect on on the grace observations.

1:19:34 Speaker 2: So so no correction is really necessary. I mean even

1:19:36 Speaker 2: in a place like Saudi Arabia where they pump huge

1:19:38 Speaker 2: amounts of oil that the resulting change in mass is

1:19:43 Speaker 2: less than than you would see from a water storage change,

1:19:47 Speaker 2: So it's below the aeror level urban built out, you know,

1:19:54 Speaker 2: I don't. I don't know if any cities have built

1:19:57 Speaker 2: you know, they've brought in construction material fast enough to

1:20:01 Speaker 2: have an effect on you know, to be detected by grace.

1:20:06 Speaker 2: I tend to doubt it.

1:20:11 Speaker 1: Thank you. We're on question twenty two. We've answered several

1:20:16 Speaker 1: questions for exercise, I'm just keeping some questions. Question twenty

1:20:20 Speaker 1: two is can we use the groundwater data to calculate

1:20:23 Speaker 1: the velocity or rate of change of groundwater of any area?

1:20:32 Speaker 2: Yeah? I mean you could, you know, if you you

1:20:35 Speaker 2: could plot the time series and then look at, you know,

1:20:38 Speaker 2: how the the change, you know, how fast the change

1:20:41 Speaker 2: is happening. You know, it's probably more accurate on an

1:20:47 Speaker 2: inter annual basis than it would be, you know, on

1:20:50 Speaker 2: a sub monthly basis or something like that. But but yeah,

1:20:53 Speaker 2: theoretically you can. You can take the changes in groundwater

1:20:57 Speaker 2: versus time and look at the rate of change.

1:21:04 Speaker 3: Thank you.

1:21:06 Speaker 1: The next question is about availability of Gilda's two point

1:21:10 Speaker 1: two in Google Earth Engine. Is it available in there.

1:21:15 Speaker 1: I believe some Gelda's data are available, but I'm not

1:21:19 Speaker 1: sure about Gilda's two point two met if you know,

1:21:22 Speaker 1: if it's there in Google Earth Engine.

1:21:27 Speaker 2: I don't think GLS two point two is on Google

1:21:29 Speaker 2: Earth Engine. But then again, you know Google tends to

1:21:31 Speaker 2: take things they find online and put them up without

1:21:35 Speaker 2: asking us, So it's possible that I don't know that

1:21:38 Speaker 2: it's there.

1:21:39 Speaker 1: Yeah, I was trying to check, and I don't see

1:21:43 Speaker 1: exactly two point two. But I'll come from that question

1:21:47 Speaker 1: twenty four. Is it better to correlate change in ground

1:21:52 Speaker 1: with sentinel one inside of the same location to determine

1:21:56 Speaker 1: the reason for uplift or subsidence? I'm asked king this

1:22:00 Speaker 1: while considering injection, welse.

1:22:09 Speaker 2: I I I think that. So, first of all, we're

1:22:13 Speaker 2: not you know, gailed US wouldn't model an injection well,

1:22:18 Speaker 2: so it's not going to be very good at downscaling

1:22:21 Speaker 2: the GRACE data in in areas where there's you know,

1:22:25 Speaker 2: a direct human impact, like like groundwater pumping for irrigation

1:22:29 Speaker 2: and direct injection. So I wouldn't necessarily trust, say, you know,

1:22:37 Speaker 2: one degree pixel from grace in a region where where

1:22:41 Speaker 2: there's an anthrogetic effect like that, that's that's significant, and

1:22:46 Speaker 2: comparing with within sur data to look for upflip, uplip

1:22:50 Speaker 2: or subsidence. Yeah, I think I think it would be

1:22:59 Speaker 2: you we look and maybe I'm wrong, may maybe find

1:23:02 Speaker 2: something interesting there, but I wouldn't necessarily expect to see

1:23:06 Speaker 2: a good correlation between the two.

1:23:13 Speaker 3: Thank you.

1:23:13 Speaker 1: The next question is question twenty seven, referring to the

1:23:19 Speaker 1: key limitations on slight thirty one. It is noted that

1:23:23 Speaker 1: Hilda's two point two does not simulate groundwater pumping in

1:23:28 Speaker 1: the context of Indonesia, particularly in the regions dominated by

1:23:33 Speaker 1: massive localized pumping for large scale mining operations. How severely

1:23:39 Speaker 1: does this limitations kew the GRACE fall on downscale data?

1:23:44 Speaker 1: Does the model misinterpret this localized anthropogenic depletion as broader

1:23:50 Speaker 1: climate driven animally.

1:23:53 Speaker 2: Yeah, there are a few a few issues here, so yes,

1:23:57 Speaker 2: First of all, it is a significant limitation. If there's

1:24:00 Speaker 2: localized pumping, you know, that's not going to be captured

1:24:05 Speaker 2: by the model, and it's so it's not going to

1:24:06 Speaker 2: be effectively downscaled from GRACE. So theoretically GRACE would detect

1:24:13 Speaker 2: the effects of pumping, but typically only if it's over

1:24:16 Speaker 2: a significantly large region. So if there's a you know,

1:24:21 Speaker 2: for example, the high planes off for this Southern high

1:24:23 Speaker 2: and Central high planes offer, there's skipping an amount of pumping.

1:24:26 Speaker 2: There's been a long term decline in groundwater storage there,

1:24:30 Speaker 2: and that's that's detected by GRACE. But if you're talking

1:24:34 Speaker 2: about a smaller scale operation, then that that's going to

1:24:38 Speaker 2: be sort of smoothed out in in the GRACE data,

1:24:40 Speaker 2: just because the Great Grace has a course resolution. But

1:24:44 Speaker 2: the other issue here is that, you know, because Indonesia's islands,

1:24:49 Speaker 2: GRACE does does not do well resolving Thrusser water store

1:24:54 Speaker 2: changes over islands because the much larger oceanic mass changes

1:25:00 Speaker 2: are not you know, are not perfectly modeled and removed

1:25:04 Speaker 2: from the GRACE data. So there's a lot of you

1:25:07 Speaker 2: know a lot of errors of our islands like Indonesian,

1:25:09 Speaker 2: So we're really lying relying on in Indonesia. In the

1:25:13 Speaker 2: case of GLS two point two is the model itself,

1:25:16 Speaker 2: and again the model doesn't simulate the groundwater pumping, so

1:25:18 Speaker 2: it's it's not really going to be any good at Similarly,

1:25:24 Speaker 2: the sort of anthrogenic effects.

1:25:29 Speaker 1: Thank you for your own question thirty one now. So

1:25:38 Speaker 1: this is again based on exercise, so I'll skip that

1:25:43 Speaker 1: and maybe come back later to thirty one.

1:25:46 Speaker 2: Yeah, I mean, if I can interrupt for a second,

1:25:49 Speaker 2: I found the answer on the latency question for GLS

1:25:51 Speaker 2: two point two, and it's it's actually available right now.

1:25:55 Speaker 2: The g LS two point two EP, which is the

1:25:58 Speaker 2: extended product where runs there's no more GRACE data to assimilate,

1:26:03 Speaker 2: but the model continues to run using other inputs. That's

1:26:06 Speaker 2: available right now up through April twenty second, which is

1:26:11 Speaker 2: six days ago. So I think it typically is it's

1:26:15 Speaker 2: on the order of seven to fourteen days latency, but

1:26:19 Speaker 2: you apparently it's only six days today. I think the

1:26:22 Speaker 2: data might have just been posted.

1:26:24 Speaker 3: That's great, that's great to know. Thank you.

1:26:31 Speaker 1: Yeah, so question thirty three, we're going to go to

1:26:34 Speaker 1: question thirty three. When you say groundwater storage in the model,

1:26:38 Speaker 1: are we actually measuring round water or we are just

1:26:41 Speaker 1: calculating whatever is left over after subprect soil snow plants.

1:26:49 Speaker 2: Yeah, this is I would encourage you to look at

1:26:52 Speaker 2: the publications that we have on graces. In ways, it's

1:26:57 Speaker 2: even more complicated than what you said. The cashment landsforce

1:27:01 Speaker 2: model simulates groundwater as a basically what's left after you

1:27:08 Speaker 2: you know, if you have a certain amount of water storage,

1:27:10 Speaker 2: it actually stimulates this uh this this water uh sort

1:27:16 Speaker 2: of loss from from a fully saturated state and uh

1:27:20 Speaker 2: and then you know a sokhomofore about Hey, how GRACE

1:27:23 Speaker 2: data simulation works. Are basically comparing the anomaly in the

1:27:25 Speaker 2: model to the anomaly and Grace and then making adjustment

1:27:29 Speaker 2: to the model's anomaly. So, uh so when you you know,

1:27:38 Speaker 2: we're not measuring groundwater using a well observation, We're we're

1:27:42 Speaker 2: really calculating it, uh you using a model simulation and

1:27:48 Speaker 2: then and then correcting it using an information from GRACE.

1:27:53 Speaker 2: I hope that's helpful, but I really would encourage you

1:27:55 Speaker 2: to look at the the publications if you're interested in

1:27:59 Speaker 2: in the de tells.

1:28:04 Speaker 1: I'll go to question forty and that's the last question,

1:28:09 Speaker 1: and then we can go back to a couple of

1:28:11 Speaker 1: more questions later on because we're almost at the end

1:28:14 Speaker 1: of the time. So question forty is I'm looking for

1:28:17 Speaker 1: a groundwater animal is on the east coast of James Bay. Unfortunately,

1:28:23 Speaker 1: there is no observation or any kind of Well, what

1:28:27 Speaker 1: is your suggestion for the validation and the comparison. What

1:28:31 Speaker 1: can I do for the downscaling.

1:28:36 Speaker 2: Okay, so I'm not familiar exactly where where you're looking

1:28:40 Speaker 2: for data. I can say that, you know, most of

1:28:43 Speaker 2: the world there is not there's not good groundwater well

1:28:47 Speaker 2: observations available. And what we've done is, uh, you know,

1:28:52 Speaker 2: we've done our evaluation of the GLS two point two

1:28:55 Speaker 2: model output using you know, areas in the world where

1:28:58 Speaker 2: there is you know, a substantial number of well observations

1:29:02 Speaker 2: to use for validation.

1:29:07 Speaker 4: Uh.

1:29:07 Speaker 2: And then you know, really I always like to say,

1:29:10 Speaker 2: you know, we've we've sort of proven that this is

1:29:12 Speaker 2: working reasonably well already. I don't know that there's how

1:29:16 Speaker 2: much science stuff you've done in terms of evaluation, and

1:29:19 Speaker 2: maybe just think about what's your real scientific question and

1:29:23 Speaker 2: then and then take a look and see if the

1:29:25 Speaker 2: data are useful for that.

1:29:29 Speaker 3: Thank you so much, Mett.

1:29:31 Speaker 1: So we are at the end of our webinar time,

1:29:35 Speaker 1: and thank you doctor Rudle for your presentation and your

1:29:40 Speaker 1: answering all the questions. We also want to thank all

1:29:44 Speaker 1: of you for attending today's session. There are some questions

1:29:48 Speaker 1: about Gioannie not responding. I think it's the same data

1:29:52 Speaker 1: being accessed by many people at the same time might

1:29:55 Speaker 1: do that. You can try it at a different time

1:29:58 Speaker 1: and it will work. So with that we thank you

1:30:01 Speaker 1: again and hope to see you on Thursday on thirtieth

1:30:04 Speaker 1: of April at the same time. And want to thank

1:30:07 Speaker 1: our set team here for their help, especially our coordinator

1:30:12 Speaker 1: Natasha Johnson Griffin and Maria Maravito, Cherry Morris brock Levin

1:30:19 Speaker 1: say our editors Mariam Arabito, Sarahkshell, Jonathan O'Brien and my

1:30:27 Speaker 1: colleagues for instructional design so monthly and Sean McCartney and

1:30:35 Speaker 1: Erica Bordas we all work together on these webinars, So

1:30:40 Speaker 1: thanks our set team and thank you all for atturning

1:30:43 Speaker 1: to this session once again. Matt, great to have you

1:30:48 Speaker 1: here with our set trainings and we've got a lot

1:30:51 Speaker 1: of information about both Grace last week and Jill does today,

1:30:56 Speaker 1: so we really thank you for your time and your

1:30:59 Speaker 1: help with this.

1:31:01 Speaker 2: Yeah, thanks to everyone who attended and helped with the

1:31:05 Speaker 2: with the training and much appreciated.

1:31:11 Speaker 4: H