NASA ARSET Overview and Applications of OPERA-DISP to Monitor Groundwater Changes Part 3

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

Speaker 1: changes for Water Resources Management. This is Part three and

Speaker 1: the topic today is overview and applications of OPERA disp

Speaker 1: to monitor groundwater changes. OPERA is observational products for end

Speaker 1: users from remote sensing analysis and this is a splacement product,

Speaker 1: so we will be talking about that, and our guest

Speaker 1: picker for that today is doctor Eric Fielding from NASA

Speaker 1: Jet Proposion Laboratory. This is the final part of this

Speaker 1: training and there's one homework that's posted today. It opens

Speaker 1: today and it's available from the training website and the

Speaker 1: homework will be due on fifteenth of May. A certificate

Speaker 1: of completion will be awarded to those who attend all

Speaker 1: live sessions and complete the homework assignment before the two date.

Speaker 1: His brief review of what we saw in the last set.

Speaker 1: In Part two, we talked about Global Land Data Assimilation

Speaker 1: System version two point two. Land data assimilation systems, or altas.

Speaker 1: They integrate data from various sources using algorithms that represent

Speaker 1: knowledge of physical processes to fuel special and temporal gaps.

Speaker 1: Gilda's two point two integrates grace and grace followen observations

Speaker 1: we talked about these in part one of the Straining

Speaker 1: and so Gildas assimilates terrestrial water storage with other information

Speaker 1: using data assimilation techniques enabling spatial and temporal downscaling and

Speaker 1: vertical desegregation. Child's two point two soil moisture and groundwater

Speaker 1: output has been evaluated extensively using ground based observations. There

Speaker 1: are a couple of limitations to this data. The model

Speaker 1: does not simulate water management, that is, groundwater pumping, so

Speaker 1: it cannot effectively downscale grace or grace follow on observations,

Speaker 1: but direct human impacts are significant, and the model does

Speaker 1: not simulate confined a prefer storage changes, which may be

Speaker 1: an important components of total water storage changes observed by

Speaker 1: Grace and greasewallowing in certain regions. Gilda's two point two

Speaker 1: output are updated monthly on GEES disc site, and associated

Speaker 1: wetness and drought indicator maps and monthly forecasts are updated

Speaker 1: weekly on the National Drought Mitigation Center website. This brings

Speaker 1: us to today's session and so objectives for today are

Speaker 1: that by the end of part three you will be

Speaker 1: able to identify characteristics of opera surface displacement, product access

Speaker 1: and visualize opera dis data to infragroundwater change is at

Speaker 1: regional scale for applications. As in the previous two sessions,

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

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

Speaker 1: free to enter your questions as we go. We will

Speaker 1: try to get to all of the questions during the

Speaker 1: Q and A session after the webinar, and the remainder

Speaker 1: of the questions will be answered in the Q and

Speaker 1: A document which will be posted on the training website

Speaker 1: about a week after the training. With that, I want

Speaker 1: to introduce our guest speaker for today, doctor Eric Fielding.

Speaker 1: Doctor Fielding is a senior research scientist at NASA's Jet

Speaker 1: Proportional Laboratory and California Institute of Technology. He received his

Speaker 1: undergraduate degree in Earth sciences from Dartmouth College and his

Speaker 1: PhD in geological sciences from Cornell University. His expertise in

Speaker 1: processing synthetic aperture radar data and especially SAR interferometry. His

Speaker 1: research interests include mechanics of landslides, including transition from stable

Speaker 1: to unstable sliding, and effects of climate variations active tectonics,

Speaker 1: especially in the alpine Nimalayan Tibetan system on a variety

Speaker 1: of time scales, then fault interaction via the transfer of

Speaker 1: stresses through the lithosphere, and the possible implications for earthquake

Speaker 1: risk assessment. Doctor Fielding has received several awards including agu geotes,

Speaker 1: Ivan Muller Award for Service and Leadership, several NASA Exceptional

Speaker 1: Scientific Achievement Medals, and Edward Stone Award for Outstanding Research Publications.

Speaker 1: Doctor Fielding has more than one hundred peer defeat publications

Speaker 1: and so we invite doctor Fielding to talk about Opera

Speaker 1: disk Eric.

Speaker 2: Thanks Amitia for the introduction. I'm going to be covering

Speaker 2: the use of the Opera displacement products to monitor groundwater changes,

Speaker 2: and I'm using some slides from Grace Ponto, who is

Speaker 2: also a JPL and part of the Opera Displacement Validation team.

Speaker 2: So I'm going to be doing an overview of the

Speaker 2: Opera displacement products and show some specific applications to understand

Speaker 2: how this particular type of product can be used for

Speaker 2: groundwater mount during. These are the objectives. I'm going to

Speaker 2: do a very quick overview of sorrow defrometry just to

Speaker 2: get the basic concepts we have previous rset trainings that

Speaker 2: cover this topic in much more detail, so I will

Speaker 2: to those as a good play to get the more

Speaker 2: details about the background. A little bit about what the

Speaker 2: interferometric phase, which is the basic measurement of soorro intofrometry

Speaker 2: tells us about the land surface and surface subsidence. And

Speaker 2: then understand what the opera which OPERA stands for the

Speaker 2: observational products for end users from Remote Sensing Analysis program

Speaker 2: displacement product measurements mean. And then we'll in the demonstration

Speaker 2: show some examples of what the OPERA displacement time series

Speaker 2: reveals about land subsidence in one of the larger subsidence

Speaker 2: areas in the United States, the California Central Valley. These

Speaker 2: are the previous are set, so these are good places

Speaker 2: to start if you want more detailed information on the

Speaker 2: process InSAR and SORROW analysis. So here's a review of

Speaker 2: the synthetic APPETU radar or sorrow interfrometry. The twenty seventeen

Speaker 2: R set training and twenty nineteen are set training on

Speaker 2: sorrow defrometry. I have much more details. I've reduced that

Speaker 2: it here to make it fit into the time that

Speaker 2: we have available. So the key thing there is that

Speaker 2: what we measure with sorrow defrometry is the phase of

Speaker 2: the radar signal. You may have seen in other sor

Speaker 2: analysis training or papers that there's the radaritude, but for

Speaker 2: interferometry we don't really pay attention to the amplitude when

Speaker 2: you look at the phase of the SAR signal, and

Speaker 2: the SAR phase is a measurement of how many oscillations

Speaker 2: of the radar wave there are between the antenna and

Speaker 2: the satellite and the ground. And one of the things

Speaker 2: to keep in mind is that we have the radar

Speaker 2: resolution and all the elements are objects within that resolution

Speaker 2: element which is roughly ten by ten or twenty by

Speaker 2: twenty meters, are added together into one phase measurement, and

Speaker 2: we also have this very large number of cycles of

Speaker 2: the radar wave between the satellite in the ground. But

Speaker 2: by taking the difference or interferogram, we basically subtract the

Speaker 2: phase from one observation from them the phase of another observation,

Speaker 2: and by that cancels out almost all of these effects,

Speaker 2: and we then get a measurement that's just the surface displacement.

Speaker 2: So basically we can simplify the phase as just being

Speaker 2: due to the distance, which is the letter row. Here

Speaker 2: at some other consonants that are due to the the

Speaker 2: radar of propagation and distance from the satellite and noise.

Speaker 2: And by taking the same measurement of a second image

Speaker 2: uh these other constants, we assume that these other constants

Speaker 2: are are the same. And when we then subtract phase

Speaker 2: one from phase to these other constants, then cancel out

Speaker 2: and therefore get rid of those and the remaining signal

Speaker 2: is the difference between the distances between of the two

Speaker 2: satellite images plus some amount of noise. So this is

Speaker 2: just a graphical representation of the same thing. At we

Speaker 2: take two different images at time one and time two,

Speaker 2: and the distance to the ground can change up by

Speaker 2: a small amount, and the phase tells us how much

Speaker 2: that distance changed between time one and time two. One

Speaker 2: of the key things about the sarfrometry. I didn't have

Speaker 2: time to get into the fact that we can also

Speaker 2: measure topography with SAR in defrometry. But here we're using

Speaker 2: SARN difframetry to measure displacement. That's this equation here, This

Speaker 2: this equation the second equation here, and you can see

Speaker 2: that the sensitivity to displacement is depending on the wavelength

Speaker 2: of the radar lambda four pi over lambda. That's the

Speaker 2: sensitivity to displacement, and we were using here today. We're

Speaker 2: going to be talking about data from the Sentinel one

Speaker 2: satellite that the radar wavelength is six centimeters. Other satellites,

Speaker 2: including the the nice Are NASA is SAR satellite that

Speaker 2: was just launched last year and now being calibrated, has

Speaker 2: a radar wave length of twenty four centimeters. So the

Speaker 2: Sentinel one satellite has a greater sensitivity because the radar

Speaker 2: wavelength is shorter. There is a sensitivity to topography, but

Speaker 2: we're not going to talk about that here. So the

Speaker 2: Opera Observational Products for end Users from Remote Sensing Analysis

Speaker 2: Project does systematic measurement of displacements from the Sentinel one project,

Speaker 2: but they also do several other types of analysis. It's

Speaker 2: a large scale project that's supported by the NASA Satellite

Speaker 2: Needs Working Group, which gets money directly from the Office

Speaker 2: of Management and Budget to support these products that are

Speaker 2: requested by other US government agencies. And all of the

Speaker 2: data is free and open, and we've been working hard

Speaker 2: to make this data what we call analysis ready, which

Speaker 2: means that the data is processed to a level where

Speaker 2: you can start using the data without having to do

Speaker 2: your own sophisticated processing. So there's these different products and

Speaker 2: I'll go into more in the next slide and Offergos

Speaker 2: uses data from a wide variety of missions, including a

Speaker 2: lance AT eight, lance AT nine, Sentinel one the SAR

Speaker 2: from the Copernicus European constellation, and Sentinel two is the

Speaker 2: optical imaging from the Copernicus uh constellation, and then NISAR,

Speaker 2: which is the new NASA satellite that is obviously a

Speaker 2: SAR mission. So these are the products that are that

Speaker 2: Opera is producing. H there's level two products. These are

Speaker 2: basically intermediate products that we don't that we are producing

Speaker 2: for the as part of the processing, but could be

Speaker 2: useful for other people. They're not so easy to use directly. There.

Speaker 2: Level three products are what most people are going to

Speaker 2: be using. There's the dynamics surface water extent that's a

Speaker 2: map of the surface water over time. It's dynamic because

Speaker 2: it's over time, and there's two different versions of the

Speaker 2: surface water extent. One is made from the the harmonize HLS,

Speaker 2: the harmonized Lancet Sentinel two data sets. So that's optical

Speaker 2: imaging and then a second version that's been It's made

Speaker 2: from the Sentinel one SAR imagery and in the future

Speaker 2: will be made from ANSAR. Those products are near global

Speaker 2: wherever the data is acquired, and they're updated as often

Speaker 2: as we get new images from Sentinel one and the

Speaker 2: harmonized Lancet Sentinel two. Similarly, the disturbance product. This is

Speaker 2: a measurement primarily of a disturbance to vegetation and other

Speaker 2: other things that happen to the surface, could be due

Speaker 2: to fires or people cutting down tree or whatever. These

Speaker 2: are very similar in there both there's two different versions,

Speaker 2: one made from the harmonized LANSATS Sentinel two and one

Speaker 2: made from the Sentinel one, and again these are near global,

Speaker 2: available on land almost everywhere. Today I'm going to be

Speaker 2: talking about the surface displacement product. This is mapping surface

Speaker 2: displacement from the Sentinel one and soon from Sentinel from NYSAUR.

Speaker 2: The coverage of this data set is in North America,

Speaker 2: not global, because of the large amount of cost of

Speaker 2: doing the more advanced processing for necessary for the displacements.

Speaker 2: In the future, the Opera Project is going to be

Speaker 2: making a vertical land motion product. The present placement product

Speaker 2: is in the radar line of sight, not UH separated

Speaker 2: into vertical or horizontal motion, and it's not tied to

Speaker 2: any any datum, whereas the vertical land motion product will

Speaker 2: be separated into vertical and horizontal displacements and it will

Speaker 2: be tied to GPS or other absolute datum. To make

Speaker 2: a product that's directly usable for measuring vertical land motion.

Speaker 2: UH from North America, similarly to the displacement product, and

Speaker 2: just recently just UH released is a troposphere e zenith delay.

Speaker 2: One of the effects that I didn't mention on radar

Speaker 2: you know from tree is that as the radar beam

Speaker 2: goes through the atmosphere, it can be slowed down slightly

Speaker 2: by the presence of water vapor in the atmosphere. And

Speaker 2: this tropospheric zeenus delay product can be used to correct

Speaker 2: for those atmospheric delays, and it is useful for applying

Speaker 2: to the opera displacement product. But it's actually a generic

Speaker 2: global estimate for any place in the world and can

Speaker 2: be used for any sore mission. I won't have time

Speaker 2: to go into that today. So this is the map

Speaker 2: of the displacement products for North America. This is the

Speaker 2: the the descending track. Now this is the assending track map.

Speaker 2: The assending is where the satellite is going north relative

Speaker 2: to the Earth's surface. Due to the way that the

Speaker 2: Copernicus system operates the Sentinel one satellite, they have good

Speaker 2: coverage of the eastern part of North America. Only with

Speaker 2: the assnding tracks. They have limited capacity to do two

Speaker 2: different look directions there, so for the eastern part of

Speaker 2: the US, only assending data is available. You'll notice that

Speaker 2: the the displacement product is only available for the part

Speaker 2: of Canada that's within two hundred kilometers of the border

Speaker 2: with the United States. Due to the wide spread a

Speaker 2: snow cover in northern that we decided to not include

Speaker 2: that in the processing. But the processing also goes down

Speaker 2: through Central America to the southern border of Panama. The

Speaker 2: coverage is North America with an asterisk because it doesn't

Speaker 2: cover the northern part of Canada. The Sentinel one data

Speaker 2: coverage varies with time. In some time intervals time periods,

Speaker 2: the passes were acquired with six day separation, other times

Speaker 2: twelve days, or another times that's twenty four days, and

Speaker 2: that depends on also on the location. For the western

Speaker 2: part of the US, especially the west coast because of

Speaker 2: the larger amount of deformation due to groundwater and earthquakes

Speaker 2: and other processes. They tend to they acquire bold assending

Speaker 2: and descending data and generally have more frequent coverage. And

Speaker 2: the opera processing starts in July twenty sixteen. The satellite

Speaker 2: Sentinel one A, the first Sentinel one satellite, was operating

Speaker 2: before that, but the coverage was less, so we just

Speaker 2: decided to start with July twenty sixteen. Nice Are data

Speaker 2: processing is waiting on the final calibration of the nice

Speaker 2: ART data, which should be completed by June this year,

Speaker 2: and the Opera team is working hard on getting that ready.

Speaker 2: And this is the first insur product from anywhere that's

Speaker 2: been certified by the Committee on Earth Observation Systems COOS

Speaker 2: as application ready data. So OPERA products have a number

Speaker 2: of layers. I'm going to be talking about some of

Speaker 2: these in detail. The frame number here is part of

Speaker 2: the final name because the OPERA products are processed on

Speaker 2: a frame on a frame basis using a frame definition

Speaker 2: that was created by the ASM. So I'm going to

Speaker 2: be specifically talking about the main primary data later the

Speaker 2: displacement and meters, but there's a second version of the displacement,

Speaker 2: which is what we call the short wavelength displacement. The

Speaker 2: short wavelength displacement layer has the the larger scale variations

Speaker 2: in displacement filtered out. Everything over about twenty five kilometers

Speaker 2: in spatial scale has been filtered out, and that short

Speaker 2: wavelength displacement is what they use for display in the

Speaker 2: ASF displacement portal, which I'll be talking about later. And

Speaker 2: then there's a whole set of these different quality metrics.

Speaker 2: I'm not going to go into this in great detail

Speaker 2: because it's kind of an advanced topic, but these years

Speaker 2: can be used to determine which which pixels of the

Speaker 2: of the product have good quality measurements and which may

Speaker 2: have a more noisy measurement. Okay, so that's the we're

Speaker 2: ready to go to the demonstration. I'm going to be

Speaker 2: giving a demonstration of that displacement portal and then doing

Speaker 2: some analysis. The displacement portal, as I mentioned, uses these

Speaker 2: short wavelength displacement filter versions. Then I'm going to show

Speaker 2: a Jupiter notebook for more into this analysis of the

Speaker 2: displacement products. And when you go to the displacement portal

Speaker 2: which is here, the displacement dot A SF DOT Alaska

Speaker 2: dot E d u uh. It comes up with this.

Speaker 2: This a disclaimer, which is just a reminder that what

Speaker 2: they're showing is the filtered displacement, which can work. It

Speaker 2: works well to highlight small scale deformation features, but large

Speaker 2: features like the Central Valley that I'm going to be

Speaker 2: talking about later are not very well image by this

Speaker 2: short wavelength. Okay, So now I'm going to go to

Speaker 2: the demo. Okay, So now I'm in the displacement portal.

Speaker 2: Right now, I'm looking at the descending track upward displacement

Speaker 2: products from Sentinel one. As I mentioned earlier, there's kind

Speaker 2: of a gap here in the eastern part of the

Speaker 2: United States where they have not been acquiring the data

Speaker 2: with the descending tracks. That's going to be one of

Speaker 2: the big advantages of nice Are is the nice Are

Speaker 2: has global coverage with both ascending and descending tracks. The

Speaker 2: nice Are engineers spent a lot of time developing a

Speaker 2: whole hardware and data system and downlink it did downlink

Speaker 2: to be able to cover the whole world with assending

Speaker 2: and descending data. There is coverage in the very northeast

Speaker 2: part of the United States and also around Houston, but

Speaker 2: most of the eastern part of the US is not

Speaker 2: available in the descending tracks, So I'm going to zoom

Speaker 2: in here. You could just zoom in with a with

Speaker 2: a scroll bar or or the plus button over here.

Speaker 2: The scroll bar works for a nicely if you have

Speaker 2: a scroll a wheel on your mouse. And here we

Speaker 2: can see that these are groundwater subsidence areas in Arizona,

Speaker 2: in New Mexico and Texas, and also I think northern Mexico.

Speaker 2: Because these are relatively small areas of subsidence, the short

Speaker 2: wavelength filter is not changing the values too much. You

Speaker 2: can see that there's kind of a halo of what

Speaker 2: looks like apparent uplift around the edges. That's a filtering artifact.

Speaker 2: But we can click on a point here and the

Speaker 2: displacement portal goes out and gets the opera displacement products

Speaker 2: for that point and will show us a time series

Speaker 2: for the whole data set at that point. Takes a

Speaker 2: little bit of time because they have to go through

Speaker 2: a lot of files. And here we can see the

Speaker 2: time series from July twenty sixteen or August twenty sixteen

Speaker 2: through the last point here is September twenty twenty two.

Speaker 2: They are working on continuously updating this time series, so

Speaker 2: they're trying to get up to the very most recent

Speaker 2: data soon. And one of the things to notice here

Speaker 2: is that this is subsidence due to groundwater extraction, and

Speaker 2: it's highly seasonal. During the winter, when it's raining and

Speaker 2: they're not trying to grow so many crops, the surface

Speaker 2: is relatively stable. And then in the summer, when they're

Speaker 2: using lots of groundwater to irrigate their crops and extracting

Speaker 2: it out of the subsurface, then the ground surface subsides.

Speaker 2: So that's why we can see this very systematic level

Speaker 2: and then subsidence pattern over the whole almost almost ten

Speaker 2: years here, and we can see the same thing on

Speaker 2: other groundwater basins. It's a very common feature. That first

Speaker 2: point had a total subsidence of zero point four meters

Speaker 2: that's forty centimeters in nine years, and this new point.

Speaker 2: So the portal has a way of showing points that

Speaker 2: are not well determined as these kind of open or

Speaker 2: dash circles. Because I clicked on a point that has

Speaker 2: probably some agricultural fields in it during the time that

Speaker 2: they're plowing those fields. The opera displacement product cannot be measured,

Speaker 2: so then it ends up being these dashed circles. But

Speaker 2: the way that the opera displacement product is processed is

Speaker 2: able to get an overall time series even when there's

Speaker 2: these gaps where the measurements are not good quality? Can

Speaker 2: we erase those points? Uh? Oh before I go away

Speaker 2: from this area, I'm gonna just switch. You could use

Speaker 2: this little uh flight direction button here to change the

Speaker 2: flight direction. So these are This is now the assnding

Speaker 2: track data. Yeah, we can see that the groundwater line

Speaker 2: of sight displacement in the radar line of sight is

Speaker 2: very similar for the ascending and descending tracks. In this

Speaker 2: particular area, it seems that there's more gaps in the

Speaker 2: ascending track than they are in the descending track.

Speaker 1: Uh.

Speaker 2: Where you don't see a dot, that means that they

Speaker 2: didn't actually turn on the satellite on that date. So

Speaker 2: I'm gonna zoom out again. This is the ascending track.

Speaker 2: As I showed earlier in the slide. It has much

Speaker 2: more complete coverage, including all of the eastern US. Most

Speaker 2: of the eastern US doesn't have so much groundwater extraction,

Speaker 2: but there are some hotspots. Houston Texas in Galveston is

Speaker 2: one of the areas where there's a lot of groundwater extraction. Yeah,

Speaker 2: we can look at a point here and you can

Speaker 2: actually zoom in here to see the full resolution thirty

Speaker 2: meter pixels if you want to get the pixel that

Speaker 2: includes your house or whatever or some other feature. So

Speaker 2: here we can see that the ground the groundwater extraction

Speaker 2: seems to be more constant without that strong seasonal variation,

Speaker 2: and the total subsidence here is a little bit more

Speaker 2: than it's around twelve zero point one one two or

Speaker 2: meters or twelve centimeters. Now I'm going to take a

Speaker 2: quick look at the California Central Valley and here you'll

Speaker 2: notice that it looks very well. If you know what

Speaker 2: the subsidence in the Central Valley looks like, this is

Speaker 2: going to look very strange. And that's because the overall

Speaker 2: subsidence over a long spatial scale has been filtered out,

Speaker 2: and what we're seeing is just local variations. So it

Speaker 2: looks like there's little areas that are going up and down.

Speaker 2: But that's because the overall subsidence has been subtracted out.

Speaker 2: And that's why I'm going to show you next how

Speaker 2: to look at the full the full unfiltered data with

Speaker 2: a notebook Jupiter notebook. This other feature over here is

Speaker 2: the signal from the twenty nineteen Ridgecrest earthquake. It was

Speaker 2: a nineteen seven point one earthquake that was actually preceded

Speaker 2: by a six point four earthquake. This line here was

Speaker 2: the six point four and this was the rupture of

Speaker 2: the seven point one. This is also a little bit

Speaker 2: affected by the filtering, but it shows the earthquake signal. Okay,

Speaker 2: so I'm going to switch to a different tab now. Okay,

Speaker 2: So this is a Jupiter notebook that was Yes, this

Speaker 2: is a Jupiter notebook that we're going to use to

Speaker 2: download a subset of offer our displacement products for a

Speaker 2: part of the Central Valley and then look at the

Speaker 2: full unfiltered displacement products. I'm not going to go into

Speaker 2: a lot of detail on this first set up part.

Speaker 2: It's possible to run. You can run this by setting

Speaker 2: up a condo environment and then running these commands here here.

Speaker 2: So this next, this is the cell where we actually

Speaker 2: specify what area we want to look at. You'll remember

Speaker 2: earlier I said that the opera products are processed on frames.

Speaker 2: I was about to mention that the in the displacement portal,

Speaker 2: when you click on a point, it shows you what

Speaker 2: frame I D that point is part of. So that's

Speaker 2: the kind of the easiest way to get the frame

Speaker 2: I D. But you can also get it by searching

Speaker 2: for yeah upper products in the regular Earth data or

Speaker 2: or asf vertex search. Then I specify a bounding that's

Speaker 2: just a pulling on with the four corners of the

Speaker 2: rectangle that I want to analyze this particular notebook. I'm

Speaker 2: using just the data from August twenty twenty two to

Speaker 2: August twenty twenty four. This is a time interval that

Speaker 2: was very I had had a lot of rainfall. The

Speaker 2: twenty twenty two twenty twenty three winter in California was

Speaker 2: extremely wet, one of the wettest in the last century,

Speaker 2: and the twenty twenty three twenty twenty four winter was

Speaker 2: also well above normal. So this time interval is a

Speaker 2: very wet period for California rainfall. So once you've specified

Speaker 2: your area, then the system will prompt you to enter

Speaker 2: your NASA Earth Data log in. You can't downlo load

Speaker 2: than there uper displacement products without having a NASA Earth

Speaker 2: Data log In account. This is free, but it requires

Speaker 2: registration and you have to enter your username and password.

Speaker 2: Here you can set up a file called net rc

Speaker 2: in your home directory with that Earth Data log in information,

Speaker 2: which is what I do to avoid having to enter

Speaker 2: this every time. It just reads from that file in

Speaker 2: your home directory. So this sell checks to make sure

Speaker 2: that I have the credentials for Earth Data, and then

Speaker 2: it's going to go out and download the data. Because

Speaker 2: I've already downloaded this data. It's going to very quickly

Speaker 2: realize that I've already downloaded it. So then it finishes.

Speaker 2: It can take quite a while, depending on your internet speed.

Speaker 2: It could take twenty minutes or or longer too, and

Speaker 2: depending on how big of an area you're trying to download.

Speaker 2: So it's now put those files in this subfolder called

Speaker 2: subset NCS. That's net CDFs and frame thirty eight five

Speaker 2: h two. So these are all the files. It's fifty

Speaker 2: nine files for that two year time interval. This this

Speaker 2: just checks to make sure there's nothing wrong with the files.

Speaker 2: So the next step looks at the files and gives

Speaker 2: basic information about each file. I didn't have a chance

Speaker 2: to I didn't make a slide earlier but the opera

Speaker 2: data is process and what we call mini stacks, about

Speaker 2: fifteen dates in a row are combined into a mini stack.

Speaker 2: This is a key part of the processing efficiency because

Speaker 2: we can't process the whole time series together. But in

Speaker 2: this case, over those two years, there's four different mini stacks.

Speaker 2: Each one has fifteen scenes except for this one that

Speaker 2: seems to be missing one scene, and within each mini stack,

Speaker 2: the reference date is the first date in the mini snack,

Speaker 2: but the last date of a mini snack is the

Speaker 2: same as the first date to the next mini stack,

Speaker 2: So this notebook is able to then reconstruct the whole

Speaker 2: time series without any any real additional calculation. So we

Speaker 2: can see here the first date is August tenth, twenty

Speaker 2: twenty two, and the last date is remember fifty eight,

Speaker 2: it's between. So the first day here is the reference

Speaker 2: date of that mini stack. This mini stack goes from

Speaker 2: February to July, and the last date is July thirtieth,

Speaker 2: twenty twenty four. So this then quickly does the referencing

Speaker 2: of those mini snacks. It sees all the files and

Speaker 2: then we need to choose a reference point.

Speaker 3: As MH the opera products and as with most in

Speaker 3: is our products.

Speaker 2: They're not referenced to any absolute reference system. They have

Speaker 2: their own internal it's just basically relative to the satellite orbit.

Speaker 2: So we need to choose a point that we think

Speaker 2: is not moving. In this case, I chose a point.

Speaker 2: It makes this plot here. This is the whole time series.

Speaker 2: I made a point that over here on the side

Speaker 2: of the of the central valley where it's stable ground

Speaker 2: and not not affected by the subsidence, as the reference point.

Speaker 2: But you can pick one here and it'll tell you

Speaker 2: what lat long. What the lat long of the point

Speaker 2: that you picked is here, and then you can enter

Speaker 2: this value into this next cell. I already entered in

Speaker 2: the reference point here that I had chosen earlier. But

Speaker 2: you can also use this tool to choose another point

Speaker 2: that's in an area of subsidence and get that latitude

Speaker 2: and longitude. And then this next cell. If you enter

Speaker 2: in the point for the plumbing the time series, then

Speaker 2: this next cell shows the full time series for that point. First,

Speaker 2: it shows the map it's from. Now re reference the map. No,

Speaker 2: that's the that's the w's this one. We're going to

Speaker 2: use that reference point.

Speaker 3: Uh.

Speaker 2: The opera products are actually in UTM coordinates. But the

Speaker 2: notebook is is converting the long point that I gave

Speaker 2: it into UTM coordinates to uh use with the Opera

Speaker 2: data product. Yeah, it's now re referenced the time series

Speaker 2: of that reference point that I gave it. It's going

Speaker 2: to make a new map here. Yeah, ok yeah, Okay,

Speaker 2: So this is now the accumulative displacement over the over

Speaker 2: the four years relative to the reference point. And the

Speaker 2: reference point is this point up here with the yellow dot.

Speaker 2: You can see it's at the edge of the Central Valley.

Speaker 2: This is all the big Central Valley of California, near

Speaker 2: the city of Fresno, and there's a small city here

Speaker 2: called Corcoran. This point here is near the city of Corcoram,

Speaker 2: and we can see the time series for that point

Speaker 2: in this next plot. In this case, I chose a

Speaker 2: point which in this time in a whole actually had

Speaker 2: very little net displacement. There was a little bit of

Speaker 2: displacement here in twenty twenty two before the heavy rains

Speaker 2: of the winter of twenty twenty two twenty twenty three started,

Speaker 2: and then it started moving back up later. So this

Speaker 2: is one of the things that is very which is

Speaker 2: a little bit unusual about this time interval, you can

Speaker 2: see this is the fully This is not the filtered

Speaker 2: time series. This is the full time series. And you

Speaker 2: can see that the eastern part of the central Valley

Speaker 2: here has continued to subside, but this western part over

Speaker 2: here has actually rebounded, and that's because there was enough

Speaker 2: rainfall that it started recharging the shallow aquifers enough that

Speaker 2: the ground surface actually started moving back up during this

Speaker 2: time at aval we have seen this before. There was

Speaker 2: a previous time between twenty fifteen and twenty nineteen where

Speaker 2: we also got a lot of rainfall, and there was

Speaker 2: a previous time when this western part of the Great

Speaker 2: Valley also moved upwards. So this is one of the

Speaker 2: This is one of the key pieces of information that

Speaker 2: the InSAR can provide where Grace is going to be

Speaker 2: averaging over this whole area and showing that maybe there's

Speaker 2: a reduction in the water groundwater extraction, but it won't

Speaker 2: show you the spatial distribution of the surface effects of

Speaker 2: what the groundwater is doing. And with the ins are

Speaker 2: we here we can see that the western part of

Speaker 2: the Grade Valley is actually moved up in these two

Speaker 2: years due to significant groundwater recharge, and this eastern part

Speaker 2: is still moving down, but as I'll show you shortly,

Speaker 2: the rate of subsidences much less than in a previous

Speaker 2: time at all.

Speaker 1: Yeah.

Speaker 2: So then there's a bunch of more notebooks that show

Speaker 2: more information about the details of the equality metrics. This

Speaker 2: first one is the there's a number of different quality metrics.

Speaker 2: I don't really have time to get into all of them.

Speaker 2: Here there is a layer that you could look at.

Speaker 2: It's the short wavelength displacement. I'm not gonna we already

Speaker 2: saw that in the displacement portal, ah, one of.

Speaker 4: The yeah, yeah, yeah, the main it's.

Speaker 2: So I'm gonna show the here.

Speaker 1: Yeah.

Speaker 2: So this is showing the whole set of time of layers.

Speaker 2: This is the overall displacement, with the subsidence in the

Speaker 2: eastern part of the valley and in the a slight

Speaker 2: uplift in the western part. This is the short wavelength displacement, which,

Speaker 2: as I mentioned in the in the portal, it has

Speaker 2: this very strange fact where there's very large scale long

Speaker 2: wavelength displacement has been removed. One of the layers is

Speaker 2: called the recommended mask, and this is a starting mask

Speaker 2: to mask out the less accurately measured locations, and you

Speaker 2: can see that there are many of these fields that

Speaker 2: were not well measured int This area here was actually

Speaker 2: flooded for almost a year due to the extremely high

Speaker 2: precipitation in twenty twenty three. It formed a lake there

Speaker 2: they called Lake Tulari, which was a historical lake but

Speaker 2: then was drained for agriculture, but came back because there

Speaker 2: was enough rainfall. These are some of the other quality metrics.

Speaker 2: Temporal coherence is showing how coherent the measurements are with time.

Speaker 2: Corcorant is this in cities show up as very highly

Speaker 2: coherent because the buildings reflect the radar very well. Estimated.

Speaker 2: Phase quality is similar to the temporal coherence, and you

Speaker 2: can see this area here with the low phase quality

Speaker 2: is the area of that lake that formed in twenty

Speaker 2: twenty three, Yeah, causing the measurements to be inaccurate during

Speaker 2: that time in the hole. And then there's there is

Speaker 2: a water mask where permanent water is massed out, but

Speaker 2: there's very few permanent lakes here. This Lake Tulare is

Speaker 2: a temporary lake. There are some other types of measurements

Speaker 2: that I'm not going to go into here for the

Speaker 2: sake of time. So Finally, this cell here does a

Speaker 2: velocity estimation, and it's just going to take a linear

Speaker 2: fit to all the points in each pixel to estimate

Speaker 2: the average volney over the two year interval. And this

Speaker 2: particular step can actually take a lot of time to

Speaker 2: run and and quite a bit of memory. So I

Speaker 2: wouldn't run this unless your computer has at least thirty

Speaker 2: two gigabytes of round for for this very large area.

Speaker 2: For smaller areas, it doesn't use so much memory. It's

Speaker 2: something that I want to improve later. And then down here,

Speaker 2: this is the estimated UH average velocity for the for

Speaker 2: the two year time in a hole. And here we

Speaker 2: can see UH this western part where the recharge occurred

Speaker 2: with velocity on the order of six to six centimeters

Speaker 2: a year, or a little bit more in some places.

Speaker 2: And over here in the in the east part of

Speaker 2: the central valley it was subsiding around six centimeters per

Speaker 2: year during these two years. So this is a this

Speaker 2: is a significant a product that you can extract from

Speaker 2: the opera time series using this notebook. Then this later

Speaker 2: part is actually taking these products that we generated, the

Speaker 2: velocity and other other products and exploiting them to file

Speaker 2: formats that we can load into a time series program

Speaker 2: called mint pie or q GIS. And I'm not going

Speaker 2: to go through all these details the QGIS products.

Speaker 1: It.

Speaker 2: Actually there's different versions of the recommended mask. This is

Speaker 2: the most strict. It removes the most area, especially in

Speaker 2: this area of Lake Chuwari, whereas this one over here

Speaker 2: is a more lenient mask that allows it to try

Speaker 2: to estimate the velocity in that area from the points

Speaker 2: before and after the formation of the lake. And so

Speaker 2: it actually exports three these different files and it writes

Speaker 2: out the geotfs here. They can be loaded into h GIS.

Speaker 2: It also makes images for all the dates, so those

Speaker 2: export files go into this export directory. UH. These are

Speaker 2: the different files.

Speaker 1: UH.

Speaker 2: These are the as As I mentioned, there's three different

Speaker 2: versions of the recommended mask.

Speaker 1: UH.

Speaker 2: And then there's a velocity UH file here. It's based

Speaker 2: on the recommended mask converted to sentimeters. So now I'm

Speaker 2: going to bake a quick jump to the UH pretty

Speaker 2: popular q g I s g I S program. UH.

Speaker 2: This is available for Mac, PC Linux and so it's

Speaker 2: a very multi purpise international, free and open source GIS package. UH.

Speaker 2: It's can also be the same thing kind of things

Speaker 2: can be done with r g I S for those

Speaker 2: who have the commercial r GIS package or or other

Speaker 2: GIS packages. Okay, okay, So this is the the average

Speaker 2: velocity map loaded into q g i S. It's the

Speaker 2: notebook actually reprojected it from the UTM coordinates to allow

Speaker 2: long coordinates to make it easier to look at. But

Speaker 2: i've I'm actually using QGIS to to re display it

Speaker 2: in the in UTM coordinates, so it's not as a

Speaker 2: more accurate map coordinate system. And here I have also

Speaker 2: added in QGIS a Google satellite image, which is one

Speaker 2: of the layers that q I GIS offers as a background.

Speaker 2: It's a I can wait a minute. This is This

Speaker 2: is the twenty twenty four velocity map with the rebound

Speaker 2: in the western part of the Central Valley. And if

Speaker 2: we turn this off you can see the all the

Speaker 2: agriculture and you can see that they Google Earth actually

Speaker 2: captured this when there was so a significant part of

Speaker 2: the Tolare Lake here. So this is the twenty twenty

Speaker 2: twenty four average velocity. And I did the similar analysis

Speaker 2: with the time interval from August twenty twenty nineteen to

Speaker 2: twenty twenty two. This was a three year time interval

Speaker 2: of precipitation, and here you can see that the average

Speaker 2: subsidence rate is much much higher. This is the the

Speaker 2: color scheme here is that the the notebook converted the

Speaker 2: subsidence rate into centimeters per month, and these areas of

Speaker 2: deep blue here are subsiding at one and a half

Speaker 2: centimeters per month, or over thirty centimeters per year. That's

Speaker 2: that's a big amount. Whereas if we go back to

Speaker 2: the twenty twenty two twenty twenty four we can see

Speaker 2: that these subsidence rates are are much much lower in

Speaker 2: the area that's still subsiding and the western part is rebounding.

Speaker 2: So that's just a very quick introduction to how we

Speaker 2: could load these products into QGIS or other GIS programs

Speaker 2: to have a better view of what the results look like.

Speaker 2: And these particular versions of the are using the recommended

Speaker 2: mask with a pretty with a moderately strict coverage. So

Speaker 2: the Tlare lake is largely massed out here in the

Speaker 2: twenty twenty two twenty twenty four image, but more complete

Speaker 2: in the twenty nineteen to twenty twenty two image. There's

Speaker 2: also a lake down here that actually stayed even longer.

Speaker 2: I forget the name of that lake. So now I

Speaker 2: think that's a very quick demonstration, but I hope that's informative.

Speaker 2: And I'm just going to go back to these slides

Speaker 2: to show some resources of web pages that you can

Speaker 2: use to go back and look at the background for

Speaker 2: Opera and the displacement products. Okay, so these are some

Speaker 2: web page resources about the NASA Earth Data System, that

Speaker 2: is where all the Opera of Data products are stored,

Speaker 2: Satellite Needs Working Group, which supports the Opera project and

Speaker 2: other projects. In fact, the Harmonized landsat Sentinel to data

Speaker 2: products are another Satellite Needs Working Group product that we

Speaker 2: in the Opera project build on their product for some

Speaker 2: of those displacement I mean disturbance and water extent products.

Speaker 2: This is the displacement portal that I demonstrated. There's some

Speaker 2: product information about OPERA in general. There's a link to

Speaker 2: the GitHub page where the Opera Jupiter notebooks are shown,

Speaker 2: and in fact, you can also look at the Opera

Speaker 2: products in NASA Worldview. It's not as effective for the

Speaker 2: displacement products, but is very effective for the other OPERA

Speaker 2: products such as the water extent and disturbance products and

Speaker 2: I'd be happy to answer questions and I'll hand it

Speaker 2: back to a reader.

Speaker 1: Thank you so much, Eric for your presentation and information

Speaker 1: about opera disp especially for your demonstration about ASF disc

Speaker 1: portal and sharing your Jupiter notebook code with everyone. Just

Speaker 1: to summarize, we had a review of SAR interferometry, especially

Speaker 1: that phase of SAR signal is very important when you

Speaker 1: consider interferometry. So when two observations are made from the

Speaker 1: same location in space but at different times, the interferometric

Speaker 1: phase is proportional to any change in the range or

Speaker 1: distance of a surface feature, and that is used in

Speaker 1: producing opera DISP. It is derived from sentinel ones are

Speaker 1: imagery using interferometry. Opera disk products provide information or movements

Speaker 1: of the Earth's surface such as subsidence due to groundwater

Speaker 1: or oil and guest extraction, uplift due to water injection,

Speaker 1: and ground motion from tectonic faults, landslides and volcanoes. We

Speaker 1: also had an overview of opera displacement visualization portal available

Speaker 1: from Alaska sur Facility and after filling demonstrated Jupiter notebook

Speaker 1: code for in depth analysis of opera disc products. We

Speaker 1: saw how subsidence in Central Valley, California, looked at from

Speaker 1: disp indicated groundwater changes. This brings us to the end

Speaker 1: of this training. And in this training we focused on

Speaker 1: three different ways of looking at groundwater. So we first

Speaker 1: looked at Grace and Grace follow On, then Gilda's model

Speaker 1: and then Opera data products and we saw their characteristics.

Speaker 1: So Grace and Grace follow On they provide measurements of

Speaker 1: variations in mass or gravity changes over Earth surface and

Speaker 1: these gravity variations are interpreted in terms of changes in

Speaker 1: equivalent water thickness or terrestrial water storage t WS. Gled's

Speaker 1: version two point two integrates Grace and Grace follow on

Speaker 1: observations of t WS with other information using data assimilation

Speaker 1: enabling spatial and temporal downscaling of soil, moisture and groundwater.

Speaker 1: And finally Opera which uses central ones are as displacement

Speaker 1: product or disks that is also useful for monitoring groundwater

Speaker 1: depletion and recovery. As we saw, these are the characteristics

Speaker 1: of these data sets. Grace and Grace follow on missions

Speaker 1: they provide terrestrial water storage change data. These are global

Speaker 1: data sets. Resolution original is approximately three degree by three degree.

Speaker 1: Let you to longitude, but they're reprocessed at one degree

Speaker 1: by one degree and also at half a degree resolution.

Speaker 1: Temporal coverage is February two thousand and three to present,

Speaker 1: and there is a gap between July twenty seventeen to

Speaker 1: May twenty eighteen. That's the gap between Grace and Grace

Speaker 1: Fallen data are available on monthly timescale. Gilda's groundwater storage

Speaker 1: data that's also global and it's available at higher resolution

Speaker 1: quarter degree by quarter degree latitude longitude the same time

Speaker 1: period as Grace and Grace follow on as they're assimilated

Speaker 1: in here, and these data are available on daily timescale. Finally, Operatis,

Speaker 1: which is based on Centinel one data as a higher

Speaker 1: resolution thirty meters. It's right now away label only for

Speaker 1: North America and the duration of the data is July

Speaker 1: twenty sixteen to twenty twenty five and revisit time is

Speaker 1: about six to twenty four days depending on the location.

Speaker 1: We also had demonstrations and you had exercises for data access,

Speaker 1: analysis and visualization for Grace and Grace fall on terrestrial

Speaker 1: water storage changes. We looked at Grace data analysis tool.

Speaker 1: I want to bring another visualization tool which we did

Speaker 1: not have time to go through. But this mask on

Speaker 1: visualization tool is similar, but it also includes a GSFC

Speaker 1: mask on data at half a degree so you can

Speaker 1: explore that. But we looked at maps and time series

Speaker 1: based on this original Grace data analysis tool. Then we

Speaker 1: used GEO one n QGIS to look at Gilda's groundwater data.

Speaker 1: We had maps and time series and also looked at

Speaker 1: difference in groundwater so Groundwotter story change over different periods. Finally,

Speaker 1: today we had a demonstration on opera disc using ASF

Speaker 1: displacement portal and we will have some exercise based on that.

Speaker 1: There's one homework that's posted today on the training web

Speaker 1: page and answers must be submitted via Google forms. The

Speaker 1: homework is due on fifteenth of May. The certificate of

Speaker 1: completion will be awarded to those who attend all three

Speaker 1: live webinars, complete the homework assignment by the deadline, and

Speaker 1: then you will receive a certificate email approximately two months

Speaker 1: after completion of the course. Just to note here that

Speaker 1: for each train session we had exercise and some questions

Speaker 1: are based on those exercises, so please go through the

Speaker 1: exercises before you go to the homework. Once again, we

Speaker 1: want to thank our guest speakers, doctor Matthew Rottel and

Speaker 1: doctor Eric Fielding for their presentations and information about different

Speaker 1: data sets. You can see the trainer information here Doctor

Speaker 1: Rodell and doctor Fielding. You can contact them if you

Speaker 1: have any questions. This is the ourset team, my colleague

Speaker 1: Sean McCartney and edicup bodies. Along with that, you can

Speaker 1: contact any of us if you have any questions. This

Speaker 1: is our set website, our set YouTube link and for questions, comments,

Speaker 1: or to share how you have applied our training to

Speaker 1: your work or studies, please email at NASA dot R,

Speaker 1: set at gmail dot com and join our quarterly newsletter

Speaker 1: to stay up to date on our latest trainings. And

Speaker 1: for that you can send an email with no subject

Speaker 1: line to this email address and follow the instruction sending

Speaker 1: the response. Here are some useful resources used in this training.

Speaker 1: Thank you all, and we'll start with our question and

Speaker 1: answer session. We have a doctor Fielding here and so

Speaker 1: we can start with the question answer session. Welcome America

Speaker 1: and thank you for being here. So we'll start with

Speaker 1: question one. Is downscaling important for looking at groundwater drought

Speaker 1: index in a particular region? And Eric, you can unmute

Speaker 1: and answer the question.

Speaker 2: Please hello, Yes, this is a The drought index that

Speaker 2: the National Oceanic the National Weather Service in the United

Speaker 2: States makes is a estimate of the availability of surface

Speaker 2: water and shallow saw moisture. It does not directly measure

Speaker 2: ground water, so uh downscaling of that would would be possible,

Speaker 2: but they they estimate that based on the weather including

Speaker 2: precipitation and temperature, which are not which they use only

Speaker 2: a large scale measurements. The conversion of the drought index

Speaker 2: to the drought index is not directly measured groundwater, but

Speaker 2: areas of drought are often places where cities or farms,

Speaker 2: especially farms, will extract groundwater to continue their farming operations.

Speaker 2: So it's not the same thing. It's a different type

Speaker 2: of measurement, but they are related to.

Speaker 1: Great Thank you. Question two is what happens during a

Speaker 1: storm if opera has difficulty due to distance to capture data.

Speaker 2: So the opera surface water and disturbance products that use

Speaker 2: the optical harmonized landset Sentinel two inputs will be blocked

Speaker 2: by cloud cover and they will not have any measurement

Speaker 2: for the dates and pixels. Where there's cloud cover, the

Speaker 2: OPERA products based on SORROW measures, including the Sentinel one

Speaker 2: and soon Nice hours are images are not affected by clouds,

Speaker 2: and they have measurements for every past.

Speaker 1: Great question three. Can we apply OPERA DISP anywhere? For

Speaker 1: example the Middle East?

Speaker 2: So NASA only has funding to process the displacement products

Speaker 2: for North America, but the software that Opera uses is

Speaker 2: fully open so people could use the same software to

Speaker 2: process other areas.

Speaker 1: Thank you. Question for is I want to process Sentinel

Speaker 1: one data into displacement for small area of interest in

Speaker 1: different regions outside North America. Since Opera DISP is limited geographically,

Speaker 1: how difficult would it be to build a similar workflow?

Speaker 2: As I mentioned, the Opera software is available. Well, it's

Speaker 2: open source and on a GitHub. Processing a small area

Speaker 2: should be relatively easy. The main time series program is

Speaker 2: a program called Dolphin, and it can be used with

Speaker 2: Sentinel one or other SOUR data to analyze the time

Speaker 2: series and using the same algorithm that is used by Opera.

Speaker 2: The OPERA system does have a sophisticated workflow to enable

Speaker 2: continental scale processing, but you don't need that for processing

Speaker 2: a small area.

Speaker 1: Thank you. Question five. If we generate s pas insur

Speaker 1: time cities for a region using tools like by GMTSR,

Speaker 1: MINTBI or list s POS. How can we distinguish groundwater

Speaker 1: driven defamation from other processes like oil and gas extraction, tectonics,

Speaker 1: earthquakes or snow loading. How should we handle intracomparison and validation,

Speaker 1: and how to operate up products ensure consistency between seabands

Speaker 1: andinel one and ALB and NYSAR for long term monitoring.

Speaker 2: So the full separation of different geophysical and groundwater effects

Speaker 2: is beyond the scope of this training, but the time

Speaker 2: series data that's provided by the just Opera displacement product

Speaker 2: is very helpful for distinguishing different effects within the earth.

Speaker 2: The groundwater extraction in most places is typically seasonal, more

Speaker 2: in the dry seasons and less in the wet seasons,

Speaker 2: So if you see a seasonal variation in the displacements,

Speaker 2: that's a strong clue that the displacement is due to

Speaker 2: groundwater and not tectonic effects. Cut The second part of

Speaker 2: the question is about the how to opera products ensure consistency.

Speaker 2: The opera displacement products are processed separately for each Sentinel

Speaker 2: one track or and soon each nice our track. There

Speaker 2: is no additional processing to ensure consistency between the tracks,

Speaker 2: but the future Opera vertical land motion products, so that

Speaker 2: will have an additional processing to make to combine different

Speaker 2: sentinel one displacement tracks ascending and descending and with calibration

Speaker 2: with the local GNSS or g GPS data to make

Speaker 2: a consistent map of the vertical land vertical and horizontal

Speaker 2: land mush.

Speaker 1: Question six is for the Python demonstration, what is the

Speaker 1: velocity accuracy of this result? See five millimeter per year.

Speaker 2: Yes, the Opera products have a requirement to have an

Speaker 2: accuracy of five millimeters per years and that has been

Speaker 2: validated for places where the data has high quality. There

Speaker 2: are some areas where the InSAR coherence is low due

Speaker 2: to farm plowing fields or flooding or dense vegetation that

Speaker 2: is incoherent in central one seed bend data, and those

Speaker 2: are markets unreliable in the recommended matter that's included with

Speaker 2: the products, and there's no useful velocity measurements in those areas.

Speaker 1: Question seven for the Python demonstration are subsidence and delish

Speaker 1: and equal.

Speaker 2: There's a quite complex relationship between the volume of groundwater

Speaker 2: extracted and the amount of service. Subsidence This is determined

Speaker 2: by the characteristics of the groundwater reservoir, which varies greatly

Speaker 2: in different places, even within the same sedimentary basin, and

Speaker 2: the situation could be especially complex in places where there's

Speaker 2: groundwater reservoirs at different depths. They are not equal, but

Speaker 2: they're closely related it and require some advanced hydrological modeling

Speaker 2: of the ground border reservoir to better understand what the

Speaker 2: the reservoir characteristics and response are.

Speaker 1: Thank you. Question eight is I'm seeing some missing monthly

Speaker 1: data while working with the ASF data. What is the

Speaker 1: reason for this?

Speaker 2: The opera products are only available when the source data

Speaker 2: was acquired. The various satellites, especially Sentinel one, sometimes miss

Speaker 2: a few acquisitions or or change their change their acquisition plans.

Speaker 2: The Sentinel one constellation has had some changes over time.

Speaker 2: Initially they had just one satellite, they Sentinel one A.

Speaker 2: Then they added Sentinel one B, but then the Sentinel

Speaker 2: one B satellite stopped working in December twenty twenty one,

Speaker 2: so then they were back to having only one satellite

Speaker 2: until twenty twenty five when they got Sentinel one C

Speaker 2: in operation and they just I think earlier this month,

Speaker 2: declared Sentinel one d to be operational. So they're now

Speaker 2: back to full operational strength with two Sentinel one satellites.

Speaker 2: But during the especially during the time in the world

Speaker 2: where there was only one Sentinel one satellite, they had

Speaker 2: to reduce their data coverage because if they had less

Speaker 2: half the acquisition capacity, and if they didn't acquire the data,

Speaker 2: then Opera couldn't has no data point.

Speaker 1: Great, thank you. Question nine is when we downloaded that

Speaker 1: disk files, what is the best way to assign the

Speaker 1: reference based on a point or border of the mounting box?

Speaker 1: How is a reference point selected for measurement of displacement?

Speaker 1: Is there any specific point we should select?

Speaker 2: So you want to choose a reference point in a

Speaker 2: place that you know or have good reason to believe

Speaker 2: that the ground surface is not moving or stable. For

Speaker 2: groundwater studies, you typically want to have a point that's

Speaker 2: outside of the sedimentary basin nearby but not in the

Speaker 2: not likely to be affected by the subsidence. And the

Speaker 2: reference point also needs to be coherent in all the

Speaker 2: dates of the ins our time series. So for the

Speaker 2: Sentinel one display some products, you can't choose a point

Speaker 2: in an area where there's dense vegetation that will not

Speaker 2: have a good measurement and being coherent.

Speaker 1: The next question is question ten. What's the difference between

Speaker 1: a line of sight velocity deformation along slope horizontal and

Speaker 1: vertical deformation and how are they estimated using SAR?

Speaker 2: So the fundamental measurement that we get from the InSAR

Speaker 2: is the line of sight velocity. The single a single

Speaker 2: line of sight velocity from one track of data is

Speaker 2: not sufficient to be able to separate horizontal and vertical displacements.

Speaker 2: There are ways to combine two or more line of

Speaker 2: sight displacements to estimate what component of the displacement for

Speaker 2: each point is vertical and what part is horizontal. This

Speaker 2: is what the opera vertical land motion processing we'll be

Speaker 2: doing in the future. The if you only have two

Speaker 2: line of sight directions like you get from the ascending

Speaker 2: and descending sentinel one, then you only the only horizontal

Speaker 2: component that can be estimated is the east west direction.

Speaker 2: You can't estimate the north south component.

Speaker 1: Question eleven is is there any atmospheric correction applied to

Speaker 1: the level three or raw results when analyzing the displacement

Speaker 1: time series? Should we apply corrections like ionosphere delay, solid

Speaker 1: or tides et cetera.

Speaker 2: So the operator displacement products, the raw products that you

Speaker 2: get with from the time series files or do not

Speaker 2: have the atmospheric directions applied, but they do have extra

Speaker 2: layers in the in the file that include the ionospheric

Speaker 2: delay and solid earth tide estimation.

Speaker 1: The UH.

Speaker 2: These effects are really quite small for the c bend

Speaker 2: Sentinel one data, so for most purposes this is doing

Speaker 2: that those two corrections are not necessary. As I mentioned briefly,

Speaker 2: the opera product is providing a separate tropospher delay estimation

Speaker 2: that can be used to estimate and apply a troposfer

Speaker 2: delay correction. And these are much much larger than the

Speaker 2: amosphere or solid earth tide corrections, but requires some a

Speaker 2: little bit of extra work to add to your analysis

Speaker 2: group not covered here, just to.

Speaker 1: Make a point that the simple file that you have

Speaker 1: in your exercise will show a different layers that includes

Speaker 1: what doctor Filling is mentioned and question twelve in the

Speaker 1: displacement portal, there are many pixels over a water body.

Speaker 1: Why why were the pixels not removed by the water mask.

Speaker 2: The AESF displacement portal does include a water mask and

Speaker 2: a mask for areas of low InSAR coherence. There should

Speaker 2: not be pixels over a water body unless unless that

Speaker 2: water body was dried up for some reason. The Tu

Speaker 2: Laari Lake in the Central Valley of California is typically

Speaker 2: dry most of the time, but it did get water

Speaker 2: in it during the very heavy rainfall of the winter

Speaker 2: of twenty twenty three, so in that area we get

Speaker 2: a we had we could see that there was a

Speaker 2: measurements in the time in twenty three, and then during

Speaker 2: twenty twenty three when there was water there there's no measurements.

Speaker 1: Interesting question thirteen. In the time series that you shoot,

Speaker 1: does each point represent deformation with respect to the reference

Speaker 1: data or to the data immediately before. If we have

Speaker 1: deformation in time T times T one, T two, T three,

Speaker 1: the shown deformation in T three is deformation with reference

Speaker 1: to T two or T one. Assuming T one is

Speaker 1: a reference date.

Speaker 2: The time series is with reference to the reference date.

Speaker 2: So when I the opper products are process in these

Speaker 2: mini snacks, so there is a separate reference date for

Speaker 2: each mini stack. But the the Jupiter notebook that I

Speaker 2: had ties together the reference states for each mini stack,

Speaker 2: so that the whole time series is relative to the

Speaker 2: reference date of the first mini snack.

Speaker 1: That's great, And if you again look at the supplemental

Speaker 1: document in today's exercise, you will see that the file

Speaker 1: LIMB convention is explained in there. And then question fourteen,

Speaker 1: when analyzing displacement time series data, should we use unfiltered

Speaker 1: displacement or shortwave displacement?

Speaker 2: For almost every analysis you want to use the unfiltered displacement.

Speaker 2: The short wavelength displacement is just a visualization tool that

Speaker 2: makes it easy to show the displacement over a large

Speaker 2: area without worrying about local reference points, and it's not

Speaker 2: really usable for detailed analysis.

Speaker 1: Next question fifteen, is the displacement time series comparable with

Speaker 1: s PASS derived displacement time series?

Speaker 2: Yes, it's it should be comparable. The OPERA time series

Speaker 2: analysis uses a more sophisticated what we call persistent scatterer

Speaker 2: and distributed scatter analysis to get higher quality measurements, but

Speaker 2: the fundamental result should be similar to s PASS for most.

Speaker 1: So the next question also is like what algorithms? What

Speaker 1: algorithm uses opera to create the time series PSI or

Speaker 1: s BUS.

Speaker 2: The Opera algorithm is what we call it's a phase linking.

Speaker 2: It's called phase linking. It actually uses both persistent scatterer

Speaker 2: PSI InSAR and distributed scatterer in SAR, so it's kind

Speaker 2: of a combination of PSI and s PASS and the

Speaker 2: two measurements are combined into each thirty meter by thirty

Speaker 2: meter pixel.

Speaker 1: Next question is I received this message time series service

Speaker 1: earn no upera ice on burst frame IDs were found

Speaker 1: over the given AOI and the point is given here,

Speaker 1: what should I do? This is twenty nine longitude, which was.

Speaker 2: Yeah, this is outside of North America. There's not going

Speaker 2: to be any opera data outside of North America.

Speaker 1: I think it's not in North America. So question eighteen

Speaker 1: are the velocity maps and average or total for years?

Speaker 2: The calculation that I showed in the notebook is the

Speaker 2: average over the time and of ALE. So I showed

Speaker 2: two different calculations. One was the average velocity from twenty

Speaker 2: nineteen to twenty twenty two, the dry years, and the

Speaker 2: other was the average velocity from twenty twenty two to

Speaker 2: twenty twenty four, which was the wet years. So those

Speaker 2: were the average velocities over those two time intervals. You

Speaker 2: can't you can do other types of analysis. That's what

Speaker 2: the full time series is available for other types of estimation.

Speaker 2: I also showed some maps that were the total or

Speaker 2: cumulative displacement.

Speaker 1: Question nineteen is the velocity map still in line of

Speaker 1: site direction or does the script download both directions and

Speaker 1: decompose the vertical element.

Speaker 2: The script the analysis that I showed in the notebook

Speaker 2: is still in the line of site direction. I only

Speaker 2: have it only used one track of the opera displacement products,

Speaker 2: and that's just one line of sight. There's no decomposition.

Speaker 1: Question twenty is what should be the minimum time being

Speaker 1: able to measure surface displacement for groundwater applications?

Speaker 2: Well, that's going to the minimum time interval is going

Speaker 2: to depend on how fast the groundwater displacements are happening

Speaker 2: in a given area. In the Central Valley, the displacements

Speaker 2: can be up to thirty centimeters per year, so you

Speaker 2: can get a good measurement in just a few months.

Speaker 2: Other places, the displacements are much slower, maybe only a

Speaker 2: centimeter a year, and you might need to have two

Speaker 2: or three years of data to get a good estimate

Speaker 2: of what the rates are if the rates are quite slow.

Speaker 1: Question twenty one, why are these why are there positive

Speaker 1: subsidence points?

Speaker 2: Yes, that's quite a surprising effect there in the twenty

Speaker 2: twenty two to twenty twenty four time interval, because this

Speaker 2: was a time interval where the California received a huge

Speaker 2: amount of rainfall. The groundwater actually was recharged and the

Speaker 2: surface rebounded due to the very large amount of rainfall

Speaker 2: during those two years.

Speaker 1: Thank you. Question twenty two. While excluding the ASF portal,

Speaker 1: in my region, I'm seeing localized pockets of subsidence a

Speaker 1: few hundred meters per yards across. Is it reasonable to

Speaker 1: interpret this as subsidence? What sort of factor should be

Speaker 1: wary of for misinterpreting At fine scales.

Speaker 2: The displacement products are indicating that there's some type of

Speaker 2: ground motion. It's probably a subsidence, but it could be

Speaker 2: some other effect like such as landslides moving or in California,

Speaker 2: we have earthquakes also in Nevada and other places. But

Speaker 2: most of the measurements should be useful. Uh, it's it's

Speaker 2: just a measurement of the surface displacement. It doesn't tell

Speaker 2: you necessarily the cause of that surface displacement. So you

Speaker 2: do have to do some interpretation as to what would

Speaker 2: I actually is causing the displacement. In other places, there's

Speaker 2: some displacement where landfills or former landfills that the material

Speaker 2: keeps compacting with time, and you can see that on

Speaker 2: the on the opera products in urban areas where there's

Speaker 2: large landfills that are been they've been covered over with

Speaker 2: because they stopped using them, but the material keeps compacting

Speaker 2: with time. Mhm H.

Speaker 1: It's interesting. Question twenty three. Land displacement can be caused

Speaker 1: by multiple factors such as erosion to be separate them

Speaker 1: from groundwater subsidence in the dispretas.

Speaker 2: Sold land surface modification by erosion or other. You know,

Speaker 2: construction that that changes the land surface in a major

Speaker 2: way will become incoherent in the InSAR measurements and therefore

Speaker 2: it'll it won't have a valid measurement. So the InSAR

Speaker 2: measurements in the opera displacement products are only in places

Speaker 2: where the surface is not being changed significantly over time

Speaker 2: and major erosion. You can't measure erosion directly with insert.

Speaker 1: Thank you. Question twenty four. Could you provide key references

Speaker 1: documentations that could guide us to produce opera land displacement

Speaker 1: or vertical land motion, et cetera output data sets for

Speaker 1: study area located in Africa, for example.

Speaker 2: Yet the main uh opera algorithm is UH. All the

Speaker 2: software is online at gitab. The main there is a

Speaker 2: paper that's coming out shortly on the whole opera algorithm

Speaker 2: and methodology. It's not yet published, so I can't give

Speaker 2: you that link yet.

Speaker 1: Okay. Question twenty five is there a way to enter

Speaker 1: the codinates of my point of interest directly into the

Speaker 1: opera disc search bar?

Speaker 2: The displacement portal doesn't have that capability at this time.

Speaker 1: In NASA Earth Data, you can do that. You can

Speaker 1: just click on a point and then modify the let

Speaker 1: alone exactly the way you like.

Speaker 2: Ah, yes you can. You can. You can search for

Speaker 2: the opera displacement products, but not in the displacement portal.

Speaker 1: Not in the portal, Yes, thank you. Question twenty six,

Speaker 1: how can operate this time series be quantitatively integrated with

Speaker 1: Grace and Chielda's data to improve groundwor the storage estimation

Speaker 1: considering the different special and physical sensitivities.

Speaker 2: This is important question, but it's really beyond the scope

Speaker 2: of this training that you have to consider the whole

Speaker 2: response of the reservoirs in the areas, and the spatial

Speaker 2: resolution is a major effect that requires some computer modeling

Speaker 2: or data assimilation techniques beyond the scope of this training.

Speaker 2: But people have done this work and there's a number

Speaker 2: of papers on combining INSERT and GRACE and grace follow

Speaker 2: on data with insert to get a better estimation of

Speaker 2: what the local variations in the ground order extraction.

Speaker 1: Question twenty seven is, since OPERA disc uses insert radar

Speaker 1: from satellites, does the radar beam actually penetrate underground and

Speaker 1: see the water table directly?

Speaker 2: Now, the radar beam does not see the water table.

Speaker 2: It penetrates into the ground surface at most a few centimeters.

Speaker 1: In twenty eight, when OPERA data is processed, are there

Speaker 1: different reference points for the North America region or is

Speaker 1: there a unique point? Can we know the location of

Speaker 1: such point?

Speaker 2: The Opera displacement products are are not processed to any

Speaker 2: global reference point for all of North America, and that's

Speaker 2: what the for the future vertical land motion product will

Speaker 2: you be So each each track is basically each frame

Speaker 2: has its own local reference system and you need to

Speaker 2: provide your own reference points and when you do your analysis,

Speaker 2: there's no absolute reference point for the Opera displacement products.

Speaker 1: Question twenty nine for insult, what are some best practices

Speaker 1: if studying groundwater related subsidence in a seismically active area,

Speaker 1: in addition to analyzing both ascending and descending data sets.

Speaker 1: Do you have further recommendations.

Speaker 2: Yeah, Well, earthquakes are our fault motions. Gradual fault motions

Speaker 2: are a concern in most cases that the faults are

Speaker 2: are not in the middle of the groundwater basins, but

Speaker 2: in other places there are. There's an additional effect where

Speaker 2: many times the groundwater reservoirs are actually have an edge

Speaker 2: right at the location of a fall, and you may

Speaker 2: see that there's a bigger ground water variation on one side

Speaker 2: of the fault on the other side of the fault.

Speaker 2: It doesn't mean that the fault itself is moving. It

Speaker 2: just means that the aquifer layers in the groundwater are

Speaker 2: of different thickness on the two sides of the fault.

Speaker 2: We've seen. We saw that very clearly in the area

Speaker 2: of what's called the Santa Clara Valley that include San

Speaker 2: Jose in California. There's a few papers on that.

Speaker 1: Question. Thirty is InSAR measures displacement relative to a reference point,

Speaker 1: but no point is truly stable due to tectonic strain.

Speaker 1: And other factors in a basin with both regional tectonics

Speaker 1: and groundwater pumping. What the minimum temporal baseline is needed

Speaker 1: to separate these signals.

Speaker 2: Yes, the the InSAR measurements are are always relative to

Speaker 2: a reference point. In some cases, you may have a

Speaker 2: GPS station or or g n SS station that you

Speaker 2: can use to get the full absolute three dimensional UH

Speaker 2: displacement time series for that point and use that as

Speaker 2: your stable reference for the InSAR. That's a that's an

Speaker 2: effective way to try to separate the two signals. UH

Speaker 2: different signals, but you would have to do some your

Speaker 2: own analysis of the tectonic signal to see how it

Speaker 2: might be affecting the groundwater. Typically they're they're on different

Speaker 2: spatial scales, So just by looking at the spatial resolute,

Speaker 2: the spatial area a special pattern, you can separate different

Speaker 2: effects by using what when you have a local reference point,

Speaker 2: it's basically saying that that local reference point has zero displacement.

Speaker 2: So if there's some kind of you know, large scale

Speaker 2: plate motion, that that plate motion has already been subtracted

Speaker 2: out because you set that you have that local reference point.

Speaker 1: So next question, let's see if it's relevant. What if

Speaker 1: we want to do groundwater temporal changes as one of

Speaker 1: the environmental factor to demonstrate a casual relationship coursal relationship

Speaker 1: to the lake geo hazard, which data set is the

Speaker 1: best available option place as its own limitations.

Speaker 2: I'm not sure what they mean by the lake gudo hazard,

Speaker 2: but typically the in Sorrow or other just InSAR is

Speaker 2: going to give you the higher spatial resolution that you

Speaker 2: can more easily tie to or local effects. The grace

Speaker 2: data is only going to give you a regional scale

Speaker 2: look at what's happening in the given area.

Speaker 1: Question thirty two is if the area of selection is

Speaker 1: bigger than one frame, does that this Jupiter script music

Speaker 1: them and fix the phrase jump in between?

Speaker 2: No, this Jupiter script only works with one frame. In fact,

Speaker 2: you have to make sure that your area of interest

Speaker 2: fits within frame boundary or it'll it'll stop running.

Speaker 1: Thirty three is pont Opera process Canada even by removing

Speaker 1: wintertime data? Is there a limitation in processing six months

Speaker 1: of data every year?

Speaker 2: It was a decision that they made some years ago.

Speaker 2: I don't know the exact reason, but they decided not

Speaker 2: to process the part of Canada out north of more

Speaker 2: than two hundred meters two hundred kilometers from the border.

Speaker 2: That's just what they decided to do, mostly because of

Speaker 2: the snow cover issues. But so there is a snowcover

Speaker 2: issue in Alaska, and when you look at the displacement

Speaker 2: data in Alaska, you'll see that there's large gaps in

Speaker 2: the time series during the winters when there's snow cover.

Speaker 2: It is possible to do that kind of processing, but NASA,

Speaker 2: the NASA funding wasn't sufficient to do that for the

Speaker 2: large area of northern Canada.

Speaker 1: Thank you. Question thirty four. How can OPERA disc data

Speaker 1: be used at the local scale to assess groundwater level

Speaker 1: variations linked to vegetation eveport transpiration, particularly in comparing water

Speaker 1: use between fast growing and slow growing forestry species.

Speaker 2: So the OPERA displacement products are measured with the whatever

Speaker 2: temper resolution is available from the Centinel one data and

Speaker 2: in the future NISER data Sentinel one. As I mentioned earlier,

Speaker 2: some of the time they've had two satellites, in which

Speaker 2: case they in some places they actually acquired data every

Speaker 2: six days, especially over the West coast of the United States,

Speaker 2: but in other places they acquire data only every twelve days.

Speaker 2: And NISAR will also be twelve days. So the temporal

Speaker 2: resolution is determined by the input data, so you're not

Speaker 2: going to see variations at time intervals less than the

Speaker 2: data acquisition.

Speaker 1: Great, we just have time for a couple of more

Speaker 1: questions and then we'll stop. We're almost at the end

Speaker 1: of the time. What's the cause of the positive effect

Speaker 1: artifacting around bodies of water in the ASF.

Speaker 2: Yes, So this is due to the way that they

Speaker 2: calculate the short wave length filtered version of the placements.

Speaker 2: To calculate a short wavelength displacement, you take a window

Speaker 2: it's about twenty five kilometers across, take the average of

Speaker 2: that window and subtract it from the data, and that

Speaker 2: means that the center of the subsidence ball gets still

Speaker 2: has the subsidence, but then the areas next nearby end

Speaker 2: up having an apparent uplift. That's just due to the

Speaker 2: way the filter was calculated.

Speaker 1: Next question, when calculating the linear recurrence velocity of pixels,

Speaker 1: what is the improvement to accuracy of using a polynomial

Speaker 1: fitting instead, At least over smaller area where computation time

Speaker 1: is not a major concern.

Speaker 2: You can use higher order polynomial fitting, but then you

Speaker 2: may be fitting a noise in the data. There is

Speaker 2: a significant noise in each measurement. This can be reduced

Speaker 2: if you do the tropospheric corrections that I mentioned earlier.

Speaker 2: So you don't want to overfit and stay fitting noise

Speaker 2: in the data. That's why we typically use averages over

Speaker 2: at least a few months.

Speaker 1: Yeah. The last question is it is commonly known that

Speaker 1: Mexico City is sinking due to the compression of the

Speaker 1: clay beneath it. In our exercise, this negative lend displacement

Speaker 1: was surrounded by a ring of positive displacement. Why is this?

Speaker 2: Yeah, this is the same thing that I just described.

Speaker 2: It's due to the way the filter was applied. When

Speaker 2: you do the short wave length filter, it shows up

Speaker 2: as positive around the edges the subsidence in the middle.

Speaker 2: If you look at the unfiltered displacements from Mexico City,

Speaker 2: you will see that there is no uplift around the edges.

Speaker 2: It's only subsidence inside the basin.

Speaker 1: Thank you so much, Eric. We thank Doc right Fielding

Speaker 1: for his presentation and answering all the questions today and

Speaker 1: he's sharing his expertise on this subject. So thank you

Speaker 1: all for attending this webinar series. This is the last

Speaker 1: part of this series and we hope to see you

Speaker 1: in our future training, so please stay updated with our

Speaker 1: SAT listener you'll get information about our upcoming trainings. Also,

Speaker 1: please take a few minutes to complete the survey that

Speaker 1: you receive so that we get feedback from you. And

Speaker 1: we want to thank our set team for their help.

Speaker 1: Our coordinator Natasha Johnson Griffin, our editor Maria Marapito and

Speaker 1: other coordinators Roc Plevins, Salvin Hadsondoy, Shry Morris, and our

Speaker 1: editors Jonathan O'Brien and Sarah Corschel. Also want to thank

Speaker 1: our institutional designer Susan Monthy and my colleagues Sean McCartney

Speaker 1: and Erica Potters for everybody's help. And once again, Doctor Fielding,

Speaker 1: we thank you for your help and time with this

Speaker 1: training and it looks like it generated so many questions

Speaker 1: and people are very interested in this topic. So hopefully

Speaker 1: through the exercise and through the jupitter notebook you can

Speaker 1: explore this data and then, as as we told you earlier,

Speaker 1: the homework is you on fifteenth of May, and we

Speaker 1: hope to see you in our future training.

Speaker 2: Thank you all, Thank you, and you're welcome

The Story