NASA ARSET_ Monitoring Earthquakes_ Volcanoes_ and Landslides with NISAR_s InSAR Capability

0:01 Speaker 1: Hello, and welcome to this webinar series harnessing NYSSAR next Generation Radar observations for Earth applications. I'm Erica Potis. I'm a scientist that NASA's Jet Propulsion Laboratory, and I'm also an instructor with the RSET program. Today's session will focus on the use of nisar's InSAR capability that's interferometric synthetic aperture radar to monitor earthquakes, volcanoes, and lass lights, which will be delivered by invited expert doctor Eric Fielding from NASA's Jet Propulsion Laboratory.

0:41 Speaker 1: This is a training outline. Today is the third and last session of this three part webinar series. There is a homework associated with this training and it can be accessed through the training web page. The do date for the homework is August six, and a certificate of completion will be awarded to all of those participants who attended all three sessions live and complete the homework assignment by the do date. How to ask questions. To ensure we see your question, please write your question in the Q and A box which you can locate it by doing a click on the three dots in the bottom right of the window in the platform, and there you'll see a slider, a Q and A option, and you can also see a standalone slideer, tap or app write your questions.

1:47 Speaker 1: Are all your questions there, and we will answer them during the Q and A session at the end, and we will try to answer all the questions during the Q and A session. The remaining questions will be answered in a Q and A document which we will post on the training website about a week after the training. Today's guest instructor is doctor Eric Fielding, who's a research scientist that NASA's Jet Proportion Laboratory. He's a geophysicist and an expert in the use of radar interferometry or InSAR for studying tectonic movements, landslides, and other phenomena.

2:26 Speaker 1: Doctor Fielding is also part of the NISAR science team. He's a globally recognized expert in the use of InSAR and has supported many or all of the InSAR sessions that our set has offered. So I'm extremely grateful for his involvement today as a guest instructor in this training. Welcome back, Doctor Fielding.

2:58 Speaker 2: My name is Eric Fielding. I'm a geophysicists at the Jet Propulsion Laboratory operated by the California Institute of Technology in Pasadena, California. Today, I'm going to be talking about how NISAR can be used to monitor geologic hazards, including earthquakes, volcanoes, and landslides using the InSAR capability of NISAR.

3:25 Speaker 1: Here's the.

3:28 Speaker 2: Overview what we're going to cover. Interferometric SAR or InSAR measures the displacement of the surface. We're going to concentrate in this part on the geologic hazards. Earthquakes, volcanoes, and landslides are the main ones, and I'm going to be showing that. This image on the right here is just a eye candy image of the interferogram for the volcanic eruption of this volcano and Ethiopia called Hailey Goopy that erupted in November twenty twenty five, fortunately just after nine Star was ready to start a collecting science date.

4:15 Speaker 2: The prerequisites for this training are the fundamentals of remote sensing training. The prerequisites for this training, assuming if you have no background, that would be to follow these previous r set trainings on fundamentals of remote sensing. There was one on SAR processing and data analysis in twenty seventeen, and a more extensive introduction to SARN deffrometry in twenty seventeen, or some other equivalent experience in understanding SARN defrometry. I'm going to give a here, but not all the details, and then of course the two previous sessions in this series.

5:08 Speaker 2: Okay, so I'm going to be covering now the how SAR diffromatory works. By the end of this training you should be able to understand what InSAR products are in their characteristics, have some idea about the uses and limitations of INSER data from NISAR, and see how we can use NICAR insert to study large displacements of volcanoes and earthquakes, and also how to use that InSAR to study more gradual displacements of landslides and other more slow moving phenomena. So I'm just going to give a quick review of what was in that twenty seventeen R set training on the sarrow diffrometry theory.

6:10 Speaker 2: In sorrow defrometry, the key thing is the phase of the SAR signal. And many other applications for mapping vegetation or floods or biomass those use the radar amplitude, sorry amplitude, but we're going to be using the phase of the star signal. The SAAR phase is a measure base of the range between the satellite antenna and the ground, plus something about the complexity of the objects within a given radar pixel, because these objects are essentially randomly located within the pixel. The way that we separate those random effects from the actual measurement, which is the range from the satellite to the ground that we want to measure by using sorrow deffrometry or differencing of the radar phase.

7:25 Speaker 2: So it's just a simplified version of the SAR phase. We have a phase observation five one, which is equal to four high over the radar wavelength Lambda times the range row one plus other constants which are those distribution of scatterers within the pixel, and then plus in one which is noise system noise. And then for the second image, which is five two, again we have the same constant at the beginning for bio verlambda the radar wavelength times the range of the second acquisition between the satellite in the ground and other constants which we are going to assume are the same.

8:18 Speaker 2: And then there's some noise, which we hope is small. Because these other constants can't be directly measured, what we do is use sarn defrometry to basically we subtract phase one from phase two or phase two from phase one to cancel out those other constants. This is a simplified view of what we're looking at. If for what we call differential interfrometry, where we're trying to measure the displacement of the surface, if we're lucky to have our two satellite passes exactly the same location, then then T one and T two at T one and T two, the two antennas are in the same place, and then the range at T one and range at T two can directly measure the change in range, which is what we want to measure.

9:15 Speaker 2: Again, we get this, We want to look at this change in phase, which is then proportional to four pi over lamb. At times the change in range grow. Another key aspect of Sarron defrometry is that we can only measure the radar phase modulo two pi. Basically, the phase is going around in a circle and it's in the complex numbers, and we can only measure the phase modulo two pie. So This top graph here shows the actual phase that we want to measure on them. It's the distribution of phase across a line across the program.

10:04 Speaker 2: And after we do the measurement, then we end up with this wrap phase where all the values are between minus pi and plus pi. And you can see then we have these discontinuities where it suddenly goes from minus pi to plus pie. And there's several of these discontinuities here, and that's what we call the wrapped phase. And then what we do in the processing is use a computer program that specialized for doing what we call phase unwrapping. It tries to go through the whole image and figure out where these discontinuities are and how to reconnect them and get back to the original phase.

10:46 Speaker 2: Although the original phase, I mean the unwrapped phase will be different from the original phase by some arbitrary constant that we will have to deal with later. These other equations here we're not going to go into in detail, but there is some bird pairs of satellites where there's some difference between the two antenna locations. There's also some phases to the topography, and we will subtract that out by knowing the topography. Another key aspect of sorrow defrometry is that it's there's a coherence and that's basically a measure of that noise term that we had in the in the first equations, where there's some amount of noise in the phase one in phase two measurements, and if that noise is high, then the coherence is low, and if the noise is small, then the coherence is high.

11:53 Speaker 2: In the past, many people have used the word correlation because this coherence is is measured by the correlational phase between adjacent pixels. If you look at papers from a decade ago or more, you almost always see the word correlation. But in the last ten years people have started using the word coherence for this measurement. It's really the same thing, but they are all the recent papers and the InSAR products from NISAR use the word coherence. So the key things that cause this noise are thermal and process or noise.

12:37 Speaker 2: This is generally very small different differential, geometric and volumetric scattering. This is where the radar waves are bouncing off of a volume of scattering objects. Typically this is a forest where the radar may be bouncing off different layers in the forest, and that's heavily dependent on the radar wavelength. Another thing that can cause noise is rotation of the viewing geometry. With modern satellites, that's not a concern. And then there's the issue of random motions over time, which are maybe a tree fell down, or a rock rolled down the hill, or there's a landslide that completely disrupted the hill.

13:28 Speaker 2: So these our examples of the types of things that cause a loss of coherence or will be also cause decorrelation, and we can relate this coherence or correlation to the standard deviation of the phase or an it's basically an estimate of the standard deviation of the phase, so that if affects how accurately we can estimate height from InSAR. There are ways of using InSAR to measure elevations, although not with the nicar satellite or displacements which we're going to be talking about here. And the loss of coherence.

14:10 Speaker 2: If the coherence is low, then that also affects our ability to unwrap phase because it's more difficult to know which phase jumps need to be connected. So now I'm going to be talking about then insur data products and their characteristics, So this is a simplified processing diagram for the NiCr system. The NIC processing system. Level zero A is the data that came down from the satellite that gets reformatted and with additional information becomes a level zero B. These are the raw data products not usually used by regular users, although some power users do their own processing with the levels zero data.

15:07 Speaker 2: That's then processed into the Level one which is in radar coordinates. The R here means radar coordinates or range Doppler coordinates. The r SLC is the range Doppler single look complex image. This is then the key product that's used to make all the level two products. The Level two products are all in geographic coordinates, so they're geocoded. The first one is a geocoded single look complex image. This is a new data type that is as far as I know, not been produced by any other satellite mission, although the Opera project that JPL has been reprocessing Sentinel one data to a similar product that they call the CO registered SLC Courridge single complex as a post processed from the original ISA data products.

16:10 Speaker 2: That's an r as a Range Doppler SLC. There's a geocoded covariance matrix, the g COVE that's the amplitude images with radiometric terrain correction that's used by ecosystems and other other studies of the radar amplitude. And then there's these two types of interferometric products. The geocoded offsets which are used for cryosphere or the very large motions of ice, and the geocoded unwrapped into programs. The gui W with this red circle here, which is what we're going to be talking about today, that's the useful for solid earth and the parts of the criosphere that are not moving so fast.

16:58 Speaker 2: And these other are these other boxes here show the intermediate products that that they're used in the processing where the UH to make the final GEW coded products. There's a range Doppler offset field range Doppler interfer wrapped inter program. Then that's unwrapped to a Arrange Doppler unwrapped INTERFERGRAM and that's finally jew coded to make did GEW coded unwrapped into the g u n W.

17:29 Speaker 2: So the nice R Level two g u n W file is it's an HDF five file like all the other nice R products and has multiple layers inside of it. So these are some of the key layers. There's an unwrapped phase layer that's the actual unwrapped phase. It's at eighty meter resolution. Looking down to the below that we have the interfergram coherence. That's also for the wrap unwrapped phase that's also eighty meters. Then there's a wrapt phase. They actually include the wrapped into parrogram within the gu NW file at a much higher resolution.

18:15 Speaker 2: This is a twenty meters pixel spacing, and the wrapped into paragram then has these individual color contours that are the all the phases between minus pie and pie that repeats across the scene. And another key layer is the anospheric phase. In this particular slide we've taken the the anospheric phase is normally unwrapped. It comes as an unwrapped phase at eighty meters, but it's been re wrapped here just for a better comparison to the wrapt phase, which shows that almost all of these fringes in the in the wrapped phase interferogram are due to the honosphere.

19:03 Speaker 2: And that's one of the key things about SAR in defrometry. With the L band radar instrument. L band as a twenty four centimeter wavelength and it's much more sensitive to the anospheric effects than shorter radar wavelengths, So the anospheric phase is a key thing that we're going to need to subtract for most of the types of analysis that we do. The other two panels here show solid earth tide, which is the people are probably familiar, I'm sure familiar with the tide of the oceans, But it turns out that the gravitational attraction of the Moon actually also causes the solid earth to flex somewhat and that can have a small but a measurable effect on the interferogram.

20:09 Speaker 2: So we include a calculation of that in case you want to remove this small but easily determined effect. And then another layer is the tropospheric phase. So the homosphere is a part of the atmosphere that's very high above the earth over basically over one hundred kilometers, and the tropospheric phase is the lower fifteen kilometers where we all live and where most of the water vapor is, and that water vapor actually causes an effect on the propagation of the radar waves. So we also include an estimate of what the tropospheric phase is from a weather model provided by the European Center for Medium Range Weather Forecast.

21:00 Speaker 2: So as written down here, the unwrapped phase and coherence and the anospheric phase are stored at eighty by eighty meter pixels. There's there's a water and subswath mask also at eighty meters. There's the twenty by twenty meter pixels of the wrap phase, and then there's a there's also a wrapped phase coherence layer that also has that twenty meter pixel size, much much higher resolution than the wrapped into unwrapped coherence, and then there's these the tropospheric delay and solidar died are actually stored as three DQ's which is a little bit complicated to explain and more advanced topic.

21:47 Speaker 2: And the range Doppler into program and range Doppler unwrapped into program are only saved and stored in the in the archive for over criosphere some areas of Greenland, Antarctica, and other major ice sheets. So this is sort of a review from the session two from the ASF folks about how NYS data are stored in the NASA Earth Data System. The Earth Data System is hosted in the Amazon Web Services cloud. Earth data can be searched directly from the Earth data search tool at this URL find that not as easy to use as the more specialized search tool at the Alaska Satellite Facility ASF that's specifically designed for searching for SAR data including NYSAR, So I'll be showing how to use that.

22:50 Speaker 2: Heidi also showed and that her presentation, but I'll be showing it just for specifically for GNW data. So you can download either from either one of these user interfaces, and it's the same data, it's just the user interface is different. The ASF guide to nice Data download usage and Tutorials is excellent and I highly recommend going to the nicsartdocs dot ASF dot Alaska dot edu page to find out more information about how to do the downloads and various usage, and they have some tutorials there.

23:32 Speaker 2: At the time that we're that I created this presentation in June twenty twenty six, we only have a limited set of uncalibrated data that were released publicly in February, and that's what I'm going to be showing today because that's what's available now. The planned release of the calibrated data is in July, and this part of the af nice art docs page, the updated release timeline, which they update frequently, so you can see what most recent timeline is for the big calibrated data release in July.

24:12 Speaker 2: So I'm going to show here how to search for the GEOMW. Do you go to unwrapped into programs through the ASF search tool. Usually if you go to the search tool, it comes up with the data sets set to NISAR automatic as default, but if it's not, then you have to choose NISAR. Draw a box around your area of interest. This is an area in Ethiopia where the volcanic eruption happened in November. I just drew a box around the two volcanoes, and there's a work Ali volcano and the adjacent Highly Gooby. The eruption actually came out of Yily Gooby, but there was additional depth information underneath Workdale because the magma drained out of Urdali and went into the Hailey Gooby volcano.

25:09 Speaker 2: So once you've selected your box, then you can add a search filter with the filter button at the top of the page and come down here choose science product G and W two only get the j code unwrapped into programs. You can also choose a date range if you only want to see certain dates here, and this is the result of that search for the GNW files over the Hailey, Goobie and Erdale area in Ethiopia. One of the key things, of course, we're talking about interferograms. So the interferograms have two dates of the before date and the after date.

25:54 Speaker 2: In the list here, the interms are always listed by the first date of the pair. And because we're trying to see this eruption that happened on November twenty third, we choose this interfergram that has the first date on November twenty second, and then which then has a second date twelve days later on the December fourth, and so that's on the left here you choose the specific scene, and then in the middle here it shows details about that scene. Is a very complicated naming convention here, but the ASF interface shows the key piece of information down here extracted out of that name, the start time of the first date, stop time of the second date, the track number, the frame number.

26:48 Speaker 2: It's an ass sending track. This frame is a full frame. The interferogram polarization is HH that's horizontal transmitting, horizontal receive and the range bandwidth which determines the resolution is twenty mega hurts. And then there's this CRID that's the composite release ID that tells you the specific version of the software that was used to create that product. This is from the preliminary release of February and it's CRID x five zero ten. The release in July is going to be P zero five zero twenty two, and that's that's how you know the difference between what processing version.

27:48 Speaker 2: So then Oliver on the right side, you get the list of the actual products for that scene. There's actually a list of products. Probably height he covered this. The top one is the level two q n w HDF five file. That's a large two gigabyte file, but there's also other files here. You press this little download button to just download. If you have not logged into earth Data in this session, then you may have to log into earth Data again. But there's also some other items here that are useful to download.

28:26 Speaker 2: There's a browse image that's the image that actually shows in this lower part of this center panel. If you scroll down to overview of the phase of the whole scene, there's a QA report that has a number of images of all the different layers and those two can be used to make sure that the scene covers the area you want to cover without having to download that big two gigabyte file. And then it's not visible here because I haven't scrolled down in the slide. There's another file called the footprint KML.

29:04 Speaker 2: By combining that footprint KML file with the PNG, then you can visualize it in Google Earth. This is just an image taken from one of the pages of the QA or Quality Assurance Report. These reports are available for all nice level one to level three products. First page it's basic metadata data about the acquisitions, and then the other pages have these types of maps and or graphs of histograms. On the left here is the full unwrapped phase. You can see it goes from minus ten to plus thirty radians.

29:44 Speaker 2: That's much more than the minus pie to pie. On the right. Here, what they've done is taken that unwrapped phase and rewrapped it to the range between zero and seven pie. This is what the data and the February release drawing uh the new The new data that's being going to be processed and available in July will have a wrap at the rewrapping it zero to two pie, which is a little bit different and there's there's similar pages for all the other layers. So this is what the that browse product looks like.

30:25 Speaker 2: If you take that PNG image and and the KML and put them in the same directory and double click on the k mL file, it should be able to load into Google Earth if you have Google Earth on your computer, or you can also load it into Google Maps. It's a very convenient way to see the coverage of the scene. The Hayley Gooby area is this corner near the edge of the scene. As I mentioned before, this is the seven pile wrapped phase and the final products will have a two pie rewrap of the face which will make it look considerably different with more fringes.

31:09 Speaker 2: So now I'm going to talk about some of the uses of the NISAR INSUR data using some examples from previous satellites, since we don't yet have a lot of good examples from NISAR. This is a map that was made more than twenty years ago of volcanoes in the central Andes of South America, and Matt Pritchard, who was then a student but is now professor at Cornell, made this map and found that there were significant InSAR signals over many of these volcanoes in the central Andes. Some of these signals were over volcanoes that had known recent eruptions, but other of these volcanoes did not have any known eruptions and weren't even known to be active.

32:04 Speaker 2: But the InSAR here then showed that these were having some amount of displacement and deformation at depth and had to be considered active. This is a more extreme event that happened also an Ethiopia and in a part called the Assault rift. There was a huge injection of magma into a shallow fissure that opened up at the two sides of the fissure spread apart by more than five meters. These are people along the edge of the fissure, but for scale we can see that this wide fissure is basically the line here the middle of this interfer man and all these fringes show that show the displacement of the ground surface away from this dike that wasn't intruded into the crust here in East Africa, and this happened in two thousand and five.

33:08 Speaker 2: This is an example of an earthquake that we that I studied with some other people that happened in twenty fifteen.

33:18 Speaker 1: It was.

33:21 Speaker 2: Mage Tu seven point eight that happened has started at the location of this star here in an area near a town called Gorka in Nepal, but it all then propagated far to the east, past Katmandu and into more almost too eastern Nepal, a distance of over two hundred kilometers. And so this map here is made from ALOS two,

34:00 Speaker 2: the Japanese as SAR satellite called ALOS two, which is also an Alban system. It was launched in twenty fourteen and it's now being being moved to retirement because it's been replaced by a new satellite of Jackson's called ALOS four. But in twenty fifteen it had just started operations and it got this excellent image of the Gorka earthquake. We can show here the line of sight direction. As I mentioned earlier, what we measure with InSAR is the displacement in the range direction. That's also normally called the line of sight between the satellite and the ground.

34:55 Speaker 2: So a lot of times they'll see this LOS or line of sight arrow showing the horizontal projection of that line of sight. But that line of sight also has a vertical dimension, which is because the satellite's looking down at the ground,

35:16 Speaker 2: and the ins are only measures displacements that are parallel to that line of sight direction. And for this specific earthquake we have we have these GPS stations that were being operated in Nepal. The triangle here shows the location of the station and the arrow shows what direction that GPS measured the ground displacement, and you can see that that ground displacement is almost perfectly perpendicular to the line of sight, and that means that we're basically not able to measure that horizontal motion because it's it's perpendicular to the line of sight and the horizontal component of the line of sight, and what we do can measure is the vertical component uh this earthquake.

36:07 Speaker 2: It was a thrust earthquake, and that means that the area of the Humalias was thrust to the south and part of that resulted in uplift in the in this area of red, including the city of Katmandu, and the area further north in the high Himalias was actually dropped down as much as as one meter h There was a it was one little extra weird thing here that that's kind of an extra complication,

36:47 Speaker 2: but we're only seeing the vertical component of this this particular earthquake. In other cases, the radar line of sight here. Usually if you see a symbol like this, the longer area is the direction of the satellite was moving, and the smaller arrow is the radiar line of sight direction. In this case, the line of sight is west and down. But we're and we're looking at displacement along this central part of the San Andreas Fault in California, and this particular part of the San Andreas Fault moves continuously at almost the full rate, so we can see after a few years this large area of discontinuous motion across the fault, showing that the fault is creeping or or moving a seismically,

37:51 Speaker 2: and that this particular section between Parkfield and San Juan Bautista, the central San Andreas, is this creeping or a seismic slip that we believe means that while it may have some small earthquakes, may be up to magnitude four, it's not going to participate in large earthquakes. And it acts to separate the northern part of the San Andres to the north of west, which is the section that ruptured in nineteen oh six from the southern part of the San Andreas fault from parksfell to the south that ruptured in last ruptured in eighteen fifty seven.

38:35 Speaker 2: Another thing that we can use the InSAR for is looking at surface changes. As I mentioned earlier, coherence or correlation measures then change the noise in the radar between the two dates, and that can be affected by both vegetation or major changes in the surface. This is a paper that I wrote twenty five years ago, so of course at that time I used the word interferometric correlation, which is the same as what we now call coherence. And in this particular case, there was this horrible, devastating earthquake in the city of Bam in Iran at six point five in December two thousand and three.

39:36 Speaker 2: This was just after the ENSAT satellite started operations, so we were very we were able to get this Envy Sat into pherogram and one of the things we noticed when we first made the interferogram is that the correlation or coherence had this line extending south from the city. The geology there in Iran. We're trying to find where the fault rupture might have been. We sent them this map and they were able to go to the field and this was basically a desert area that nobody ever went to and find the surface ructures marked by these circles at the locations where there was this decrease in the coherence in the interferogram.

40:28 Speaker 2: And then we took this a step further and looked at the change in coherence if I go back this particular the city of Bam, it was mostly it is the main population center and largely houses, with some vegetation between some of the houses. But then there was an adjacent town over here called Barravont, and this is almost entirely date palm, so it's vegetation and this low coherence over Barravot is due to vegetation, whereas a lot of this low coherence over Bam is due to actual destruction of the buildings.

41:14 Speaker 2: And so by taking the interferogram that includes the earthquake the co seismic correlation or coherence, and subtracting the coherence of another interfrogram that we made for a pair before the earthquake, then that subtracts out the effects of the of the vegetation. So Barravat is now approximately zero, and these red areas in the city of BAM show the extreme destruction of large parts of the city and the fault ruptures going south are very clear in this red color. And we also found that there was an additional fault rupture north of the city.

42:04 Speaker 2: So this is what we call coherent coherent change detection or correlation change, and that's another use of interfometry is to actually use the coherence as a measurement of service change and possibly damage to buildings. The fault rupture goes down here to the south and it ruptured towards the city, which really focused the energy into the city and cause of this horrible devastation. I think over a quarter of the population was killed in this unfortunate earthquake. So landslide motion is another application of SARI indefrometry.

42:48 Speaker 2: This particular interferred interferometric SORROW map was made with a airborne radar system that now so owned called it's called UAVSR, but it actually flies on pilot as Goldstream airplanes, and that radar was built at JPL and it flies on the piloted airplanes out of NASA aims that also uses L band like NISAR, so this is sort of a preview of what NISAR is showing but using the airborne system. The airborne system, because it's closer to the ground, can have much higher resolution, so it's not quite the same, but it gives an idea of how we can measure these very high displacements.

43:47 Speaker 2: This particular landslide as a central section here that moves over two centimeters every day, and I'll be showing another example a little bit later in Los Angeles moving almost this fast.

44:08 Speaker 2: So one of the things to think about is what are the limitations on in solar measurements. So we observe a phase here the FI observed, and it's actually the sum of all these other effects. There's the line of sight displacement, which is what we want to measure. As I mentioned earlier, there's some amount of phase that comes from the troposphere, there's some amount of phase that comes from the on sphere. There can be some phase that's due to inaccuracies in the DM that you're using for the processing.

44:51 Speaker 2: And then there's this noise term which is basically that decorrelation. And so the these phases are added together. I mean, the noise is random, so we can't really we can't subtract it out the homosphere. We have an homospheric correction layer in NISAR and atropospheric correction layer in NSAR, so we can subtract out both of these effects and get back the line of sight displacements that we want to measure. The coherence effects are multiplicative, They're not like the phase effects that are additive, and that means that

45:37 Speaker 2: any one of these effects, if any one of these different effects that are low, then the resulting multiplier multiplied coherence is also low. It also means that we can't just subtract out one effect because they're multiplied together and not added. And in this particular separation, I've separated out the volumetric coherence due to trees. There's also some amount of effect of loss of coherence due to steep slopes.

46:22 Speaker 2: There's a a coherence change that's gradual like plants growing or that's that's one of the typical ones ah. And then there's a correlation loss due to sudden changes like the landslide disruption or the disruption of an earthquake. So these are the the key components of the coherence. So now I'm going to be talking about two case studies for how applying some nice are insert data to volcanoes because I want to have this accessible to everybody right away. I'm going to be using data that was released in that February release that's already at the ASF and Earth Data Data repositories.

47:17 Speaker 2: During the time of the that's included in that February release, there were no earthquakes with significant InSAR signals, so I'm gonna not gonna be able to show you an earthquake example. So we're going to be looking at this volcanic eruption in Ethiopia that's part of the East African rift system. As I mentioned earlier, there's these two volcanoes next to each other. The Ali volcano has been very active for decades and has had a lot of eruptions, and then adjacent volcano called Haley Goofy had not been active and in historic time, and there are suggestions that it hadn't been active for as much as five thousand years, although obviously we don't have five thousand years of recorded history.

48:13 Speaker 2: It's part of the East African rift system where the eastern part of the African Plate is being as pulling apart. That gives this extensional stress to the to the rocks there and this Hailey Goobi volcano erupted suddenly on the twenty third of November without people really expecting it. What we can see in the INDO program is that in fact, the magma drained out of the art Ali volcano and the Orte Ali volcano actually had a dyke that contracted, and that magma then moved into Haley Gubi and erupted into in a quite explosive eruption that spread ash over a large area and it even sent to ash into the stratosphere.

49:08 Speaker 2: So there were ash clouds that went they blew far to the east after this eruption. So first I'm going to show you some slides with wrapped into programs that were generated by Matt Pritchard who's at Cornell. As I mentioned, he's still working at volcanoes since he's since he's been a grad student, and he made these slides. These maps are just for a small subset of the full nice our frame. So Matt went back and looked at there was one nice or Into perogram for the twelve days before the eruption, and in fact we can see this is a wrapped into program.

49:57 Speaker 2: So the the faces are mine pie to pie, which we can approximately convert to surface displacements between zero and in twelve centimeters. There was a small amount of displacement here in the haley Goobi volcano. Probably the magma was starting to move into the volcano before November twenty second, and there was some indication that this dyke underneath the earth Ali volcano had started to close or contract, so a little bit before the eruption, there was this precursor, but nobody had seen this because they hadn't bought to look at it.

50:48 Speaker 2: And then the next interferogram is the interferogram that includes the eruption. As I mentioned the haley Goobi eruption, he sent a huge amount of volcanic ash out and it h ash and other larger rocks are are called tephra by the by the volcanologists, and that tefra has covered up the land surface and and made it completely different from the two radar images. And that's why we see this area of low coherence or noise in the wrapped into program. And there's a slight hint of some possible deformation just to the west of the main Heligov volcanic peak that is still being studied.

51:44 Speaker 2: And then this is the signal over there or to Ali volcano. And what we see here this is actually uh sort of the opposite of that dike intrusion that I showed from the the one in the assault rift. In this case, the two sides of the of the dyke moved together because the magma actually drained out of the dyke during this and moved into Haley goofy.

52:20 Speaker 1: And uh.

52:24 Speaker 2: So this is the wrapped into program I'm showing these uh here because the demonstration I'm going to show is with the q g I S program and the q g I S program cannot show the wrapped into programs. And then this is the the radar track angle. It's an assending track with the radar looking to the left for the west to the left. So you've probably heard in the in the previous session about how nice are are stored in this HDF five format the HDF five files or a version of what's called net CDF. The r g I S package can read the nice r HDF five files directly, but we're I'm going to be using g DOLL and q g I S. The QGIS is an open source and free and open g I S program.

53:30 Speaker 2: It's available for a wide variety of platforms, whereas RGS only runs under Windows and the present version of g doll needs to be told that this this HDF five file should be read with the net CDF driver in g DOLL, and because QGS uses g DOLL, we also need to do that for opening it in QGIS. So there's two options. We can actually renamed the file, take the dot h five file that we download from a s F or data and change it to b dot NC for net CDF. That's the preferred option. You can also add this net CDF colon to the front of the file name, and that allows you to open it when you're doing opening the file and q g I S, but that actually doesn't work very well in QGIS for doing additional processing steps, so I recommend if you're can open in q g A S to use this renaming option.

54:44 Speaker 2: But this net CDF colon can be used for doing other g doll operations that you can do from the command line or other ways. And as I mentioned earlier, the QGIS cannot display the wrapped into program layers in the gun W file because those are complex numbers and QGS doesn't know what to do with complex numbers. I'm going to stop here and.

55:14 Speaker 1: Go to q g I S.

55:20 Speaker 2: Okay, so I've already I pressed the load raster data set button. The load data set, I've selected raster here. I'm going to go to the directory. You can see here is my g O n W file. I've renamed it as dot NC, so now I can open it directly in QGIS. When I press AD here, it shows me all the possible layers that are inside this g on W file. I don't want to load all those files, all those layer I'm just going to load the coherence magnitude ononospheric phase screen ionospheric phase screen on for certainty, the unwrapped phase the mask, and the wrapped into pherogram coherence magnitude that has that twenty resolution.

56:21 Speaker 2: So this is the it opens this. It actually loads it into QGI S as a as a group of layers, all those layers from the same interferogram. I'll just turn off all the layers, but the coherence first so we can look at the coherence. Okay, So I've turned off all the layers except for the the wrapped in, the unwrapped into pherroogram coherence, and now I'm going to zoom to that layer. This is the coherence for the whole scene. You can see this is a desert area, so that coherence is very high. Almost everywhere coherence is a normalized number.

57:10 Speaker 2: It goes basically from zero to one. And there's large areas here which are very close to one. In this dry area with very little vegetation, there's this basically zero coherence in the top right corner. That's the ocean of the Red Sea adjacent to the land.

57:37 Speaker 1: Here.

57:40 Speaker 2: Water always has basically zero coherence because the surfaces is completely different twelve days apart or even an hour apart. And there's another water body here I'm not sure what the name of that lake is. And then there's some areas here that have sort of intermediate coherence values, and those are areas that probably have some loose sand and

58:17 Speaker 2: and and sand is one of the things that can cause there's a large amount of sand at the surface can cause low coherence just because it absorbs a fair amount of the radar and reduces the amount of signal sent back. But also in a lot of desert areas, the sand can move around and there will also cause lower coherence. And then we can see this sort of spot up here, and that's the area of the Hailey Gooby. This big blob here is the low coherence of the Hailey Gooby eruption and ash flows. And then there's an area of the Earth ali that where the displacements are so large that the coherence it gets lost.

59:12 Speaker 2: One of the rules of InSAR is if the displacement within the radar pixel size is more than half of the radar wavelength or with with nicear that's twelve centimeters. So if you have twelve centimeters within the eight meter pixel of our inter program, then you'll lose coherence and not be able to make a measurement. This is the unwrapped phase.

59:52 Speaker 2: One of the things we offer will I almost I always do when I load the unwrapped phase is go into the band rendering and change it from a gray scale to a pseudocolor I usually use this color ramp from red through white through blue, and that gives this.

1:00:17 Speaker 2: Now we can see clearly that the two sides of the dyke in order Holly moved in opposite directions. We can also see that the area of Hailey Gooby is basically pure noise, and we can go it back into that and change the color scale. This is still in radiance. It's unwrapped phase minus twenty two plus twenty radians. The displacements are larger on the east side of the dike than they are on the west side of the dike, and that's because we're measuring these displacements in the radar line of sight.

1:01:10 Speaker 2: The line of sight of this track is down into the west from the satellite to the ground. The blue means it moved to the away from the radar, and red means moved towards the radar, and that means that the two sides moved towards each other. And that's because this dike is is contracting. Because the radar moved up, the land moved out, and because so on the on the east side here the ground has moved away from the radar both horizontally towards the dike, but it also moved down as part.

1:01:49 Speaker 1: Of the.

1:01:51 Speaker 2: The displacement due to that withdrawal of the magma. And because the westward and the down motions are both in the same direction relative to this radar line of sight, those two are adding together and we get a larger signal on the on the east side here, and on the west side, the ground motion is to the east, but the ground all moved down, and that means that the horlental motion is towards the satellite, but the vertical motion is away from the satellite, and they tended to partially cancel out.

1:02:33 Speaker 2: And that's why the displacements on the west side are quite a bit smaller than the east side.

1:02:43 Speaker 2: So that's the interpretation of this dike. A dyke collapse or or contraction. Earthquakes have a very similar type of pattern you'll have. You're always measuring this combination of the horizontal and vertical displacements, so you'll see a similar You have to do this similar type of analysis of what displacements mean in the radar line of sight, and by getting a second radar line of sight, say this is the ascending track where the satellite was moving north. There was also a descending track over this area where the radar satellite was moving south.

1:03:36 Speaker 2: So that would then have the satellite looking to the east, and that would show an opposite pattern here, which would tell us that this has a large horizontal component. That descending track was not included in the February release, so I'm not going to show it to you here, but now I'm gonna zoom out to the whole frame. This is the whole frame,

1:04:11 Speaker 2: and so the displacements there or to ali are quite large.

1:04:19 Speaker 2: But if we go back and change this color scale to be more residsentative of the whole scene, then we can see this is from minus two to plus four radians. We can see that there's this large kind of gradient of the displacements across the scene, and that's largely due to the honosphere. Nice are assending track scenes are acquired at approximately six am local time, and that is when the atmosphere is less active. So generally the ascending tracks have less ionospheric effects, and the descending tracks are required at six pm local time, and that is a time when the answer is much more active.

1:05:21 Speaker 2: So you'll see in many places that there is a descending track into program that has many more ionospheric effects or fringes than the ascending track for even the same location. And it also depends on latitude. The atmosphere is much stronger near the equator. This is quite near the equator in Ethiopia, and they're less at middle latitudes, and then they're very strong again near the pole, so looking at Alaska or other high latitude areas, you'll see that the homosphere is even more extreme. So we're gonna just take a quick look here at the honospheric phase screen.

1:06:14 Speaker 1: This is the.

1:06:16 Speaker 2: And one of the tricks I can use in is well, I'll use the same uh pseudo color. We're gonna sent this to two percent count and we can see now that the amospheric estimates are generally our our are generally uh close to zero for this scene, but if it was a descending track scene, we'd see a much stronger So this particular scene is not a good example to show for the ionospheric phase screen calculation. This was processed with the older version of the nice R software and it has these blue areas are places that had that low coherence, and the atmospheric estimate is with that old software was it ends up with bad values in those areas of low coherence.

1:07:35 Speaker 2: That's been fixed in later versions of the software, and we'll see that the amospheric estimates for more recent that are going to be released next month or in July, we'll have much higher quality and not have these kind of weird patterns where there's low coherence. One other thing I just wanted to give a quick look at is the honest spirit. I mean the wrapped into program coherence. A turn off these phase layers. So this we go back to Haley goofy. Here the m.

1:08:27 Speaker 1: This is the.

1:08:29 Speaker 2: This is the coherence layer from the wrapped into paragram at eighty meter pixels. And if we turn that off, we can see the the coherence layer from the the wrapped into pharogram. And this has twenty meter pixels and shows you much greater detail. There must be some kind of of a tongue here of maybe a valley where the ash has moved down the valley here, so we get more much more detail in the coherence of the wrapped into program layer. That so the QJAS can display the coherence from the wrapped inter program, but it can't display the phase from the wrapped into program.

1:09:16 Speaker 2: So that's why I showed you the slides. So that's our example of a volcanic eruption and with similar earthquakes. And now I'm gonna go back and show you landslide example. So landslides are a major hazard in many places. We have a lot of landslides in California and They're also important process in any mountainous areas, but they can occur even in relatively mild hilly areas in a wide variety of locations.

1:10:03 Speaker 2: The amount of impact of the landslides depends on their specific location, and we're going to be getting this nice our data to be able to map landslides worldwide. I'm going to be talking today about a specific landslide in an area called Palas Verdes. This is a city, a city in Palos Verdes called Rancho Palos Verdes, where a landslide accelerated in twenty twenty three and went right through the edge of the landslide went right through this neighborhood, including this house right here. This is the vertical view.

1:10:48 Speaker 2: You see this landslide edge here that went right through the middle of this house and completely tore it apart in twenty twenty three. This house has now been was red tagged. That's the US equivalent of condemning the house is saying it's no longer useful for human occupation, and it's now been torn down. But these other houses nearby, a lot of them had only what we call yellow tag, which means that the the owners could actually still go into them. Some of them they've been trying to put some like supports underneath the underneath the house to try to keep it from from sliding down the hill more uh, and that's a that's an ongoing situation here.

1:11:45 Speaker 1: Ah.

1:11:47 Speaker 2: These landslides are as I mentioned in Palas Verdes. Pallas Verdes is this this circle is in the wrong place, it gets shifted a little bit. Pallas Verda is this funny little peninsula that sticks out from the coast just south of Los Angeles. Downtown Los Angeles is up here. These two black lines here are the low coherence of the lax runways. These views here are a combination of the wrapped phase for these pairs and the and the coherence. That's one of the tricks that I often use is actually combining the coherence with the phase.

1:12:35 Speaker 2: I'll show you that in a minute. Palos Verius Peninsula is southwest of the main part of Los Angeles, and the southern west coast of the Palas Verdes has this large landslide complex. Some parts of this landslide complex have been moving for about sixty years since actually the nineteen fifties, almost sixty five years, and that part was known to be has been moving for a long time, but we had very heavy rainfall in twenty twenty three and again in twenty twenty four, and that caused both those older parts of the landslide and other parts of the landslide to accelerate drastically and caused the damage that you saw on that previous slide.

1:13:34 Speaker 2: At one point it was moving at the rate of around five centimeters per day, even faster than that Slumgullion landslide that I showed earlier. It's based in Colorado,

1:13:51 Speaker 2: so these are wrapped into program results from the on the left here the descending track and on the right the ascending track for an interval in between November and twelve days November twenty third December fifth. Turns out that just one day later the ascending track November twenty fourth and December sixth, and you can see these two maps look very different, and that's again because we're measuring the surface displacements in this radar line of sight direction. The descending track, the line of sight is to the east slightly south, and on the ascending track gets to the west and slightly south.

1:14:50 Speaker 2: So the horizontal motions in these two look directions are going to have opposite signs, and the vertical motions have the same sign If we look at the said this area and the upper left, it's blue in the assending track, descending track and in the assending track, and that's because that area is moving downward. This is sort of the top of the landslide where the land is moving down and then out and other areas here the two the two measurements are in opposite directions, and that's because it's mostly we're measuring the horizontal motion.

1:15:40 Speaker 2: But again we also have to remember that the big motion of this landslide is actually almost due south, and that's in the direction that's roughly perpendicular to the right ero line of sight. So we're not at we're not seeing the big motion of the landslide southward, but we're seeing this vertical motion along the edges, and some parts of the landslide are moving west and some parts are moving to the east. And this is just the same.

1:16:17 Speaker 2: You'll notice on this that it's actually hard to tell which is positive and which is negative, and that's one of the difficulties of interpreting wrapped phase. And this is the same to inter proagrams with the unwrapped phase. Of course it's eighty meter pixels instead of twenty meters, so we have a course of resolution, but we can see which parts are positive and which parts are negative. Again, we have this upper part in the upper left corner which is red or positive away from the satellite in both ascending and descending, and that's because it's moving down.

1:17:00 Speaker 2: And then there's this section in the southwest where it's positive on the ascending track because it's moving towards the satellite to the west, and it's negative on the descending track because it's moving to the west, which is away from the satellite on the descending truck. Due to the way these particular tracks were required, they ended up excluding these particular scenes from the February release. So I'm only going to show you the ascending track data which is in the February release. And now we're going to go back to QGIS and open the Los Angeles into program.

1:17:58 Speaker 2: I already I did the same process of downloading the g n W from the ascending track for the area of Los Angeles. UH. And this is this is after we'reasking So this is the unwrapped phase.

1:18:19 Speaker 1: UH.

1:18:20 Speaker 2: Again, as we the ocean next to the p peninsula is is water, so that the unwrapped phase is a whole noise. So one of the one of the ways to to hide that unwrapped that the noise in the ocean is to go to this coherence map. This is the coherence map from the unwrapped phase, and make this particular coherence map set to multiply instead of normal. And now if we put the coherence map on with I only turn off this Google satellite with the unwrapped phase. Now we get this this view where the the low coherence of the ocean is is black and therefore easier to focus on the on the land on the land park where the measurements are good.

1:19:30 Speaker 2: So this is the the assending track. So again this has that area to the at the north part of the landslide that's uh moving away down and to the west slightly, and this lower part where it's also moving to the west. One of the other things I can show you with this inter program is the whole scene. If we look at the whole scene and change the stretch of the interferogram

1:20:13 Speaker 2: on this pair, we can see that there's a much stronger ramp from north to south and that's due to the anospheric effect on this pair. And if we go to the anospheric face screen layer, this is the face screen layer we turn offic coherence. We'll set back to be a pseudocolor, and let's turn this coherence back on. So now we can see basically the same pattern of gradient across the scene in the amospheric phase screen. So by subtracting this ionospheric phase screen from the unwrapped phase, then we get a corrected into fairground that we can use to look at a larger area of the scene.

1:21:20 Speaker 2: So one of the other things we can see in this overview of the whole scene are these little patches of low coherence. These are in the mountains, the San Gabriel Mountains that are just north of Los Angeles, and these are actually areas of snow. This was a pair that's in the winter, and these higher elevations and the San Gabriel Mountains are covered with snow and that's what's causing this low coherence in the high parts of the land. And further to the north is the actually the central Valley of California, and there's little patches of of This is a major agricultural region.

1:22:03 Speaker 2: And there's areas of low coherence here which are in some cases places where the farmers have plowed their fields. If a farmer plows his fields, he completely changes the surface and that causes low coherence, and that can also give you a loss of coherence. And this big sort of triangular area is the Mohave Desert, which has a lot of sand, so that also has generally a lower coherence than the more rock areas around it. There's this funny blob of a weird phase here that's actually caused by radio frequency interference that it's present.

1:22:50 Speaker 2: There's actually a Federal Aviation Administration FAA radar system there and it causes interference with the nice our data that's present in these older scenes. But we've now implemented a filter that removes this interference and you won't see that in the release data that's being released in July. And that's an overview of how we can work with the actual data of the GMW files in a GIS. And I think that's a good place to end up here.

1:23:31 Speaker 1: Thank you, thank you very much doctor Fielding for that great presentation and demonstration. Now I will do a summary of today's session as well. As the previous two sessions,

1:23:50 Speaker 1: so this is a summary of today's session. InSAR measures distance from the satellite to the ground with high precision by using the base of the reflected radar signals. The coherence of the insurface is a measure of surface or surface cover stability at the radar wavelength scale and phase cycles. In a repeat passenger paragram show change in distance to the ground by half the radar wavelength that's twelve centimeters for NISAR. The NISAR InSAR products enable user analysis of interferograms with few additional steps, and the nicer insert measurements of surface motion are useful for a variety of geological processes, some hydrological processes, dynamics of glaciers, and other effects that displace the surface or large structures.

1:24:52 Speaker 1: The nicer gun W interferometric product is geo geocoded and is the best suited product for users that are new to InSAR. And here's the summary for the second session. NICs OUR data are freely and openly available. The L band data can be accessed through the Alaska Satellite Facility, while the S band data along with coincident L band data can or are available through isroe's bounity platform. L BAND provides near global coverage every twelve days, whereas s band observations are acquired primarily over India and selected science and calibration validation sites outside India.

1:25:41 Speaker 1: NICs OUR data are distributed in HDF five formats. The level two NISAR products are geo coded, allowing them to be readily overlaid with other geospatial data sets Jupiter notebooks. These are open source notebooks that contain algorithms for the missions different science disciplines, and these notebooks are publicly available. The only Global three product generated by NISAR is sol moisture, which has a spatial resolution of two hundred meters. The GCOV product is the geocoded geometrically terrain corrected backscattered data in gamma knot and it's the most suited for those new to radar and wanting to do ecosystem type studies, so for example deforestation, biomass, agriculture, flooding, etc. The GUNW product is a geocoded wrapped and unwrapped interferogram and is it is the best suited data set for those new to instar and wanting to do surface deformation or movement type studies, and finally, NYSAR data can either be downloaded for local analysis or accessed and processed in the cloud, depending on your workflow and this is the summary from session one.

1:27:12 Speaker 1: NYSAR is designed to monitor continuous Earth surface changes from the odor of centimeter to meter scale globally day and night and through almost all weather cloud conditions. The key focus processes include tectonic deformation, I sheet motion, ecosystem change, and agricultural monitoring. NISAR has an L band sensor at twenty four centimeter wavelength and an S band sensor operating at nine point four centimeter wavelength. The satellite has an exact twelve day repeat cycle and an imaging swath of about two hundred and forty kilometers.

1:27:54 Speaker 1: The spatial resolutions vary between three and thirty meters depending on the radar acquisition mode. Polarization acquisitions vary depending on the acquisition mode. The core science disciplines cover the cryosphere, ecosystems, and solid earth, and NYSAR can also address different applications and support disaster and hazard management. Data processing levels span from level zero, which is the raw data, to level three, which is a geocoded derived product. With soil moisture as mentioned already being the only Level three global product being generated.

1:28:35 Speaker 1: Level one and Level two products contain information about amplitude, phase, interferometric products, coherence and pixel offsets. The Level one are in radar coordinates, range Doppler and Level two products are geocoded and analysis ready. In July twenty twenty six, the fully calibrated forward processed data will be released with the data availability latency of thirty six to seventy two hours.

1:29:06 Speaker 1: As a reminder, there is a homework assignment. You can access the homework through the training web page. As of today, the answers must be submitted via the Google forms and the do date is August six. A certificate of completion will be provided to those that attend all three or attended all three live webinars. The attendance is recorded automatically and complete the homework assignment by the deadline. The certificate of completion will be emailed approximately two months after today.

1:29:48 Speaker 1: I like to thank all of our guest instructors for this training series, doctor Franz Meyer from the University of Alaska, Fairbanks and the Alaska se Facility. Hidi Christiansen from the Alaska Satellite Facility as well, and Upama Sharma from the National Remote Sensing Center from Israel, the Indian Space Research Organization, and finally doctor Eric Fielding from JPL. And of course, if you have any questions about the material that was presented today, please don't hesitate to contact doctor Eric Fielding through the email that you see here, so please share your thoughts.

1:30:37 Speaker 1: Before I close, I just wanted to remind you that within a day or two you'll be receiving an imitation to complete a short online survey and this feedback will help us improve the RSET program. The participation is optional and all responses are confidential, but the survey data is really critical to help us understand how to better meet your needs. So we want to help you use your observation data more effectively. But we need your feedback, so please share your thoughts. And with that, we've reached the end of today's session, session number three and the end of this training series.

1:31:24 Speaker 1: So I'd like to thank doctor Eric Fielding for the great presentation today, and we will now begin our question and answer session. Great and yes, thank you for all of the great questions that have been coming in. We've been compiling those and we'll be sharing that document here on screen. But just to remind everyone that please there are three sessions as part of this training, and please feel free to contact either me or any of our guest speakers whose emails we posted for each session and as a heads up.

1:32:09 Speaker 1: So this is an introduction to nice our training, and we are planning several trainings that use nice our data throughout the next year, so please keep tuned into what's coming next. All right, So we do have doctor Fielding is here. He has been answering the questions that have been coming in and we'll try to get through as many questions as we have time. If we don't get through all of the questions, we will be posting this document on the training web page with all of the questions answered. Okay, so let's get started.

1:32:50 Speaker 1: Question number one, I need to know if it's possible to measure the height of a crop, for example, corn or soybeans using radar technology and what teristics should the radar have to achieve a result with an acceptable margin of error? Go ahead, Doctor Phey.

1:33:10 Speaker 2: Nice InSAR is not particularly sensitive to crop height because it's l band and especially for you know, relatively small crops. Like corn or soybeans. The radar is going to be going through and mostly bouncing off the ground. If you want to measure crop height, you probably want to use radar with a shorter radar wavelength like expand, where the

1:33:48 Speaker 2: crop height would be more have a bigger effect on the interfraometric phase. In addition, to really get a good measurement of height of crops, you would want to have different a larger baseline. Niceare is designed to fly with the satellite keeping the orbit small and all the baselines small, so the effects of height vegetation height are going to be less for trying to with the in differometric effects.

1:34:34 Speaker 1: Great, thank you. Question number two, I've been trying to download a gcov nice our file from ASF Vertex, but each time the download complete, it starts restarts instead of finishing. So what could this be? And the answer to that is, we have forwarded this question to the Alaska Satellite Facility team and we will post a response to this here. It could be the server you're using, but we will be responding to that. Sorry, the browser that you're using. All right, So the next question number three, can ins our time series be used to detect glacier surface deformation or glacier velocity changes, and are there any planned applications for monitoring glacier instability or glacier lake outburst flood hazards.

1:35:35 Speaker 1: Go ahead, doctor Feeling.

1:35:36 Speaker 2: Yes, the ins OUR time series and it can be used to measure a glacier surface velocities. That is one of the main applications of NSAAR. I'm on the solid Earth part of the science team, but we have several people in the Chrisphere part of the science team that are using or INSUR data to measure glacier velocities both in mountain glaciers and in the large ice sheets the

1:36:19 Speaker 2: the the Chriosphere people also use pixel offset tracking to measure the glacier velocities where the velocities are larger, and one of the examples of is this it's live project that's that's highlighted here that uses pixel offset tracking on the radar amplitude images both from NICE with from NICs are Sentinel one landst Sentinel two to get a complete picture of what the velocities are for for the faster moving glaciers, but the InSAR is good for the parts of the glaciers that are moving more slowly, so it's really depending on how fast the glaciers are moving which technique you need to use.

1:37:15 Speaker 2: InSAR can be used to see if the glacier is speeding up or slowing down with time, and that can be possibly helpful for looking at glacier lake outburst flood hazards.

1:37:37 Speaker 1: Great, thank you, Let's go on to then question number four. Has NASA evaluated integrating instar nic insert deformation measurements with machine learning models to predict landslides or glacier related hazards and high mountain environments And what are the main challenges?

1:38:00 Speaker 2: Well, we haven't. We don't yet have enough n INSER data to do machine learning analysis, but scientists have done machine learning analysis with INSER data from other satellites. One of the main challenges of applying machine learning models is that many existing machine learning models use either the time series of each pixel or the spatial distribution of pixel values, which are displacements for InSAR from a single image. The

1:38:38 Speaker 2: machine learning needs to consider both the spatial distribution of displacements and the time history at the same time, and that may require more advanced machine learning models to be able to handle that type of really three dimensional information of both two dimensions of spatial distribution and one dimension of time history.

1:39:07 Speaker 1: Great question number five, how do we map urban flooding? And I'll take that one because for that you need the gek of the amplitude data. And in fact, yeah, INSAAR is not useful for mapping surface water extent, whether it's open water or inundated vegetation. So the plan in terms of our set trainings is for actually the next training to be focused on mapping floods. So do keep tuned. The next question number six, what are some other phase on wrapping algorithms other than SNAFFU.

1:39:50 Speaker 2: So the standard algorithm that's used by ICE the for the nicear processing is SNAFU, but there are the ICE three software package that nice Mission uses for processing does have other unwrapping algorithms included in the software. So it is possible if you want to do your own processing with ICE three to use those other options. And there's also commercial packages such as Gamma that can process nice our data with their own phase unwrapping methods. There are a few examples here. There is called Spurt and Whirlwind.

1:40:44 Speaker 2: The nicear processing system may be moving to one of these other packages in the future, but at the moment, the present software is using SNAFU as a standard for the standard product. The Opera project uses a different phase on wrapping, and they will be producing displacement products from NYSAR for North America within a year or so.

1:41:26 Speaker 1: Great, thank you. Let's move on then to the next question number seven. How does SORROW and instart differ relating information about forest there canopy size and chlorophyll content? And how does one try to correlate water resources in a nearby environment with inundated or non inundated areas. Okay, so this is a this question has multiple parts. I'll answer part of it in that chlorophyll content. For that you need to use optical right, because radar is not going to be sensitive to the chemical properties of vegetation.

1:42:08 Speaker 1: However, as posted here, there have been attempts to estimate ndv I from SAR with deep learning techniques, and there's a link to that paper here. But uh, go ahead, doctor Fielding, you want to respond to the rest of us.

1:42:30 Speaker 2: Yeah, the in sorrow is not going to tell you anything about the chlorophyll content. It's only seeing the physical structure of the forests, so it can have some ah wait, some some effect due to the volumetric scattering in forests, so that's somehow related to the biomass.

1:42:59 Speaker 1: But the.

1:43:02 Speaker 2: The nice R biomass estimates are made from the optical the amplitude images the g coves, so the standard nice R biomass products are not using insert.

1:43:23 Speaker 1: Great. Then let's move on to the next question, which is question number eight. How effectively does nicer L band retain coherence over Korea's forested and mountainous terrain.

1:43:39 Speaker 2: I think one of the a SF people has did this search and pasted in the search url here so you can go and look at the There are a few scenes from the February release already online that you can look at the coherence layer and see what the

1:44:02 Speaker 2: what the coherence is for those areas. And in fact, we have looked at a number of forested areas around the world and the coherence is much much better than with any seed band radar, especially because of the twelve day and repeat time and high spatial resolution of NISAR.

1:44:31 Speaker 1: Wonderful. Okay, question number nine, will NISAR products be available in Indonesia?

1:44:39 Speaker 2: Yes, the global forward processing that's going to be released later this month or expected to be released later this month,

1:44:49 Speaker 2: will include global coverage, so it will include Indonesia and the whole rest of the land area of the Earth.

1:45:01 Speaker 1: Great question number ten. My intention is to learn how NYSAR can be applied for landslide early warning in extreme weather events.

1:45:11 Speaker 2: So NISAR is not going to be useful for early warning of landslides caused by extreme weather events. The InSAR coherence and amplitude images can be used to map the extensive landslides after they have occurred, but NISAR does not have any way to provide real time precipitation estimates that would be necessary for early warning of landslides during an event. NISAR does have this soil moisture estimates that we haven't really talked about so much in these trainings, But there's the two hundred meter so moisture estimates and that can measure what the swarm oister is before the event, and that could be confined with some receptation estimates to better predict where landslides might occur during the extreme weather.

1:46:12 Speaker 1: Good. Okay, the next question number and by the way, before I move on, Yes, in terms of future trainings, we will be doing one on the soul moisture product as well, sometime within the next year and a half. All right, The next question is number eleven. What are the recommended best practices or threshold values for using the coherent band to mask out severe phase noise.

1:46:47 Speaker 2: The exact relationship between the coherence measurement and the numbers and phase noise depends on the number of looks used in the in the processing and the way that coherence is estimated with nice our data. Typically, areas with coherence less than zero point two is completely decorrelated, so that would be like water surfaces, And in many cases coherence less than zero point four is getting to be too noisy for accurate phase unwrapping and displacement measurements, So that's typically the threshold that I would use for masking out unreliable phase and displacement measurements.

1:47:43 Speaker 1: Good along those lines, on landslides, number twelve, how effective is ins are for rapidly moving landslides where coherence is quickly lost and all there are there alternative processing strategies for such e buns.

1:47:58 Speaker 2: So the L band radar wavelength of NISAR is able to keep coherence for landslides that are moving up to one centimeter per day. Roughly, the Palas Verdes landslide example that I showed you is presently moving around to centimeter per day and still keeps coherence. If it's moving faster than that, then you need to use the pixel offset measurements. The standard NISAR processing to the geo coded offset products is only done in chrisphare areas, so you would have to do your own pixel offset measurements with the nice ART data to measure the faster land slides moving faster than one centimeter per day, but one sentiment per day is really unusual for landslides.

1:48:57 Speaker 1: Okay, great, The next one, what are the opportunities Question number thirteen the opportunities and limitations of using nice our InSAR for long term wetland hydrology monitoring.

1:49:13 Speaker 2: So nice ARE wetland extent is typically is measured with the nice are amplitude images the g codes, not with the InSAR, but InSAR data has been used to measure the surface in and flooded vegetation because with the double bounce of the radar you can often get coherent reflection returns, and the phase of that return will depend on the height of the water. In flooded wetlands so that's a technique that's been used in a few places.

1:50:07 Speaker 3: In most cases, if the if there's no if it's not full of vegetation, then the coherence will be zero and you can't use InSAR phase.

1:50:24 Speaker 1: Okay, the next question number fourteen, Can we use nicsor for infrastructure monitoring like airways, I guess airports, airports?

1:50:35 Speaker 2: Yes. Yes. InSAR and NICE are in general that can be used for monitoring dissurfaced displacements of a wide variety of infrastructure things levees, dams, roads, bridges, pipelines, airport run airports, and other other features. And that's so it's very useful for from monitoring infrastructure. The lband radar wavelength is going to give you better coherence in places where there's vegetation.

1:51:15 Speaker 2: The lband radar wavelength actually does not work so well for the extremely smooth surfaces in most airports. And you may have seen in that in my example of the Los Angeles area that the areas of the of the runways of LAX were very dark and incoherent. And that's because the smooth surface of the radar of the runways and the adjacent to closely cut grass does not reflect the radar back, so there's very low signal and we can't really get a good InSAR measurement. So you probably for for monitoring airports, you might want to use a shorter radar wave length.

1:52:08 Speaker 1: Great. The next question the question number fifteen. With nice or dual pool, how should the effects of the ionosphere be taken into account? For instance, how should we address errors arising from channel mixing caused by the by Faraday rotation? Yeah.

1:52:31 Speaker 2: Since most of the nice OUR data is only collected with two polarizations, it's very difficult to correct the Faraday rotation.

1:52:42 Speaker 2: You wouldn't be able to do that directly from the data. But because nice Our acquisition times are at six am and six pm local time, the Faraday rotation is probably going to be small in most cases.

1:53:06 Speaker 1: Okay, question, Oh, go ahead? Was there anything else? No? Okay? The next question number sixteen. The nice are products have been discussed in previous sessions. Is there a document available for S band or dual wavelength L and S band products?

1:53:29 Speaker 2: Yeah, so there's a link here to the ROW Data product specification for the S band data products. But the S band data products have the same format and layers as the L band products including GNW. So the presentations we have here applying to the S band data as well. It's just that you have to get the S band data FROMRO and not from the NASA repositories.

1:54:06 Speaker 1: Great the next question, let's see question number seventeen. How does nisar's L band improve temporal of decorrelation when compared to sentinels one C band? What are the limitations of L band?

1:54:23 Speaker 2: So, the big advantage of L band is that it's less sensitive to vegetation. Basically, for moderate and low levels of vegetation, the radar goes all the way to the ground. For denser forests, the radar is going to be bouncing off the trunks of the trees and maybe some of the larger branches, and not the smaller leaves and small branches that move around. So the coherence is much better with L band, both over time and even just for a single short interval. In fact, in a few places where they did the special Sentinel one one day repeat last month where they had the Sentinel one A and Sentinel one C collecting dat only one day apart, the nice R L band coherence is generally better over with a twelve day nice r L band into program than with the one day seed bend inter program from Sentinel one, and of course that that was only a temporary experiment that they did during June of twenty twenty five, So

1:55:55 Speaker 2: the regular seed bend time interval operationally is six days when they have both the use satellites working and acquiring data like now with something on one C and something on one D.

1:56:13 Speaker 1: So No.

1:56:13 Speaker 2: One A has now been decommissioned after that short experiment in June. The main limitation of L band is the amospheric effects are stronger, so you need to do the anospheric phase correction. But Naysaur has the extra L band frequency band. We didn't really talk about that in detail, but it gives us a much more accurate estimate of the amospheric phase screen that you can use to correct the amospheric effects.

1:56:52 Speaker 1: Okay, great, the next question number eighteen, what is the strategy for resembling eighty meters unwrapped phase from wrapped phase at twenty meters.

1:57:05 Speaker 2: The in the production process of the twenty meter interferrogram is then multilook to about eighty meter pixels before doing the phase unwrapping with the snap food program.

1:57:23 Speaker 1: Yeah, Okay, so we have a couple of minutes left, so let's just I'm gonna select a couple of questions down the list. So let's go to question number twenty one. How does the selection of a stable reference point differ between earthquake earthquake related crustal deformation analysis and landslide displacement monitoring? Considering their distinct spatial scales, deformation mechanisms, and temporal characteristics. Choosing a stable reference point is very important for deformation.

1:58:04 Speaker 2: Yes, of course, the reference point is very important. For landslides. We typically choose a reference point very close to the landslide because the landslide only affects a small area, so there's usually an area nearby. It's stable that we can choose the reference point for earthquakes or deformation. For volcanic eruptions, the displacements extend over a much larger area, so then the reference point needs to be chosen far away enough that you can assume that the deformation is near zero.

1:58:48 Speaker 1: Great, then let's go to question number twenty three. In the last nice AR data release in February, there was a warning about ionospheric contributions not being completely removed. Are there have there been any changes in the ionospheric and tropospheric models to remove the contribution of these errors and where can one find more details?

1:59:15 Speaker 2: The ICE three software is on public GitHub. You can go to the ICE three GitHub and see lists of what was changed in the different versions of the software. The version that we're using for the release this month is ICE three version five point zero two point three, which is then converted to this composite release id P zero five two three that's going to be used for the release it's going to be later this month. One of the big there's been a number of improvement improvements to the honospheric estimate.

2:00:10 Speaker 2: As I mentioned the one of the improvements was to mask out areas of low coherence and interpolate to get a better estimate in areas of low low coherence within a scene. Another big improvement is that the the new version has filtering of radio frequency interference that it caused errors in the amospheric estimates with software used in the February release. And there's there's a number of other improvements that have been made to the anospheric estimates. I think the tropospheric correction layers are approximately the same in the new version as in the February release, but the anospheric corrections are considerably improved.

2:01:04 Speaker 2: So when you get your data, you want to look at this the cr ID or composite release ID and make sure that you're getting the new data whether it's P zero five zero two three as opposed to the old February release which is crd X zero five zero one zero, which is much which is the old software.

2:01:34 Speaker 1: Great, Let's go then to question number twenty seven. Are the available GU and ws always only between the twelve day return pairs? Well, different temporal baseline interferograms need to be processed separately if we want them so, for example, when working with sentinel data, I've used pairs with a longer temporal baseline.

2:01:57 Speaker 2: Yes, the nice R g ONW use are only produced from the nearest neighbor in time pairs. If you want to get interfer grounds with different time intervals, you would need to do your own interferoground processing from either the g coded slcs or the range Doppler slcs to make your own interfer grounds for different temporal baselines.

2:02:27 Speaker 1: Great, and then one more question, Let's go to question number forty. Can you expand a little more in coherence. What is the difference between temporal and spatial coherence? You mentioned that coherence has multiplicative components. Are these for temporal or spatial and how do you evaluate them?

2:02:51 Speaker 2: Temporal coherence is UH is derived from time series analysis.

2:03:00 Speaker 2: It's different from the spatial coherence that we've used for many years, and the coherence layer in g O n W files is a spatial coherence. The multiplicative nature of coherence applies to both temporal and spatial coherence, so it's

2:03:29 Speaker 2: basically anything that makes the coherence low will make the temporal coherence low and the spatial coherence low.

2:03:40 Speaker 1: So the.

2:03:45 Speaker 2: Estimation of temporal coherence is a more advanced topic that we didn't cover here, but there's going to be other trainings in the future that talk about that.

2:03:58 Speaker 1: Great, and how about we do one more I always liked the Q and A sessions and if we had time, we would go through all of them. I know that we're at over our time, but let's just do one more, one more question. So how about we go to question number number forty four? So I would like to know the typical baseline separation associated with the twelve day repeat past nice are acquisitions and it's expected impact on coherence over forested areas.

2:04:42 Speaker 2: The nice baselines for particular baseline separation are almost always under one hundred and fifty meters and typically only on the order of fifty to one hundred meters. So this means that the the spatial baseline effects on correlation our coherence are quite a bit smaller than the temporal decorrelation effects to random motion of the forest branches, and even at lband there's some amount of temporal decorrelation.

2:05:29 Speaker 1: Great. Okay, So as mentioned, we have a lot of questions here. We have over fifty questions. We will be answering all of your questions in this Google doc and then we will be posting it on the training web page. So don't forget the homework. You can access it as of today. And also please if you can make sure you complete the survey. It helps us a lot for future trainings. Before I close, I'd like to thank the our set team for their incredible effort in putting this together. Block Blevins, Natasha Johnson, Griffin, Slwyn Hodson, o'doy, Jonathan O'Brien, Sue Mon team.

2:06:12 Speaker 1: Uh. There's a whole team, uh behind all of this Sherry Morris, and I'd like to thank our our incredible guest instructors, doctor Franz Meyer, Heidi Christensen, and Upama Sharma and today UH, doctor Eric Fielding. Before I close, i'd like to provide the opportunity for our guest instructors to give some final words. Doctor Eric Fielding, you like to say some closing words.

2:06:45 Speaker 2: Yes, I also thank the our set team for all the work they did to get this to make this webinar function smoothly and helped to keep the whole training system convenient for me to to do the webinar and be effective for the users.

2:07:18 Speaker 1: Thank you very much, doctor Fielding. It's been such a pleasure and an honor to have you here with us again teaching us about InSAR and specifically now in SAR related to NYSAR. So, before I close, I'd like to thank the help from the Alaska Satellite Facility that we had today. Two members from the a SF joined us today to support UH with questions. So I'd like to thank Alex and Zachary So Alex Lewandov's and Zachary Hopinen And yes, so this is the end of this session, but hopefully it's at this training but hopefully the beginning for you to start working with nice our data.

2:08:15 Speaker 1: Please, we want you to work with the data. We want to hear from the community in terms of what are the challenges, so please reach out to us, to any of the instructors if you have questions, and more will be coming. As mentioned, more trainings will be coming in the near future. So thank you to all of you, to all of the participants really for your interest and your enthusiasm for this amazing mission. And until next time, have a great day everyone. Bye bye,