Sacramento Delta Levees: Radar-based monitoring from 41,000 feet

NASA’s Dr. Cathleen Jones discusses an ongoing pilot study that uses radar technology to monitor Delta levees

Monitoring levees is currently done using ground-level observations and instrumentation, but with over a thousand miles of levees in the Delta, remote sensing could turn out to be a game-changing technology for determining the health and status of levees on a broad scale. Dr. Cathleen Jones with NASA’s Jet Propulsion Laboratories is the principal investigator working on a pilot study to determine the feasibility and to develop methods to detect and quantify levee deformation, seepage, and general subsidence rates using remote sensing. In July of 2015, she gave this presentation at a brown bag seminar jointly presented the Delta Science Program, the Ecosystem Restoration Program, and the Surface Water Ambient Monitoring Program.

Dr. Cathleen Jones began by saying that she has been working on a project to develop methods to use radar to detect threats to levees for about six years now, using the Sacramento-San Joaquin Delta as the prototype area for testing this technology. One of the advantages of using the Delta for this pilot study is that the Delta is close to Pasadena, where JPL and the plane is based, so it has been convenient for NASA.

“The Sacramento-San Joaquin Delta is a very critical water resource for the state, but it’s also a vital ecosystem, and balancing that weight between ecosystems and water resources is very delicate,” she said. “The area has over 60 islands, and they are contained by about 1100 miles of levees. That is an incredible amount of levees. This is the second largest levee system in the United States, exceeded only the Mississippi River levee system. People across the country don’t realize just how large the levee system is here, and how critical it is.”

The Delta has a lot of agriculture, oil and gas production, and tourism as well as a lot of people live there, she said. “However, because of the way this Delta has been managed, it has subsided substantially and most of the islands are below mean sea level,” she said. “Along the Mississippi River you have a case where the levees are only stressed during high water conditions, but in the Sacramento Delta, 24 hours a day, 7 days a week, year in year out, the levees are stressed so the maintenance of the levees in this area is extremely critical.”

It was around the 1880s when the levees were built so the islands could be used for agriculture. “Prior to the levees being built, you had a natural estuary – a freshwater tidal marsh, where you had water covered land and the vegetation would die, it would form into soil, you would get a natural balance between accretion and subsidence from anaerobic decay of the materials,” she said. “Once the islands were drained, however, you had a whole lot of new subsidence mechanisms. You had oxidation of the soils which was the main cause of subsidence, you have compaction from dewatering of the soil, and for a long time, you had windblown loss of soil from the area which not such a problem anymore because of the practices that have gone on, but this has been going on for over 100 years and now most of these islands are below mean sea level.”

Dr. Jones noted that on the map on the left of the slide, the dark blue islands around the edge are the ones that are above mean sea level. “The most critical islands are the ones in the western Delta, because if the levees fail there, then you get salt water intrusion coming in at a big scale from the San Francisco Bay,” she said.

There are also major faults close to the Delta, she said. “You have the monster, the San Andreas Fault 50 miles away, you have the Hayward Fault 30 miles away, and then you have a bunch of faults that are within the 10 to 20 miles of the Delta,” she said. “The shaking that could come from an earthquake on those faults could cause liquefaction and loss of multiple levees in the Delta.”

There are several advantages to using radar remote sensing for monitoring the levees. “In an area that’s relatively large like this with a large number of levees, you really can’t use a ground-based visual surveys on the time scale with the frequency and the coverage that you need in order to have a consistent, continuous picture of what’s going on,” she said. “Remote sensing surveys can give you very large swath images of the area and very rapidly, so you can make a rapid assessment of a snapshot of the conditions all across the Delta at one time.”

Equally important is that you can get consistent monitoring, Dr. Jones said. “You use the same technique to monitor every levee in the area, so you have a relative understanding of what’s going on at each island, and if you’re flying in an airplane, you don’t have to worry about getting into areas that are difficult to access.”

“The most important thing is that if you use radar remote sensing, then you can detect changes that occur on a very small scales and by very small amounts, so you can see something that nobody would detect with their eye driving past or walking past or surveying in that way,” she said.

The goal is to develop techniques that could be used to inform a targeted monitoring program. “We don’t envision getting rid of ground surveying, but instead targeting the ground surveying so that you go to the areas that look like they have problems,” she said. She noted that it also can be used to provide information for emergency response.

She then presented a slide of the aircraft, noting that the project has been going on since 2009. The project was initially funded by NASA Applied Science; since then it’s been funded by the Department of Homeland Security and now more recently by the California Department of Water Resources.

“We fly this aircraft over the Delta approximately every 7 weeks, so there’s been about 51 flights since July of 2009,” she said, noting that they image the area shown in the yellow box on the map. “That area corresponds to nine flight tracks, so we go east-west across the Delta, then we go north-south across the Delta, and what the means is that every area, every levee is covered with three lines and so we get a view from different directions. We get cross validation data, so we get a measurement in one line, we can expect to get the same measurement in the other line.”

Dr. Jones said that since she would be showing a lot of radar images, she would briefly discuss some basics about radar imaging. “Radar is not a photograph or an optical image, and the most obvious thing is the resolution is not what a photograph would be,” she said. “The resolution of this image is about 20 feet on a pixel, and you’re not actually seeing the real colors on the ground.”

There are numerous advantages to using radar, Dr. Jones said. “One advantage is that radar is an active instrument so you’re sending out a pulse towards the ground,” she said. “That pulse can go through clouds, so you can see through clouds, it images the surface day or night, and if there were a fire, it could see below the smoke and haze, so you can rapidly see the ground regardless of the day-night light conditions.”

It’s also very good at detecting standing water, so you can very quickly pick out the areas that have standing water, and that’s quite useful in a flood, for example, she said. Different types of radar can be used to identify the surface types so you can tell water from land and structures on the land. “You can detect these very small changes in movement of the ground, and by very small, I mean very small fraction of an inch, millimeter changes can be detected at the ground,” she said.

She then presented a picture of NASA’s UAVSAR, noting that UAV stands for Uninhabited Aerial Vehicle. “It’s actually deployed currently on a Gulf Stream 3 aircraft,” she said. “The radar is here in a pod that hangs below the wings of the aircraft. This aircraft has been operation as a science instrument for NASA since 2009.”

Dr. Jones then presented a slide of with pictures of the insides of the radar, noting that she has been involved with this project from the very start, even having helped design it. She pointed out some of the features of the radar, such as the inlets at the front of the pod that provide ambient air cooling, it’s independent of the platform that it’s mounted on, and the different antennas that can be used. “It operates at the same frequency as NASA’s next earth observing mission, so it’s a prototype for NASA’s upcoming NISAR mission, so we can use it for a kickstart for studies that could be done with this satellite once it comes online.”

“NASA has an open data policy so any NASA data is available for free and can be downloaded from the web,” Dr. Jones said. “That is another great advantage whenever you start looking at the costs of radar images from a commercial provider.”

There are two complete copies of the entire pod so they can test with one and fly with the other. “This is used for all kinds of experiments,” she said. “It flies 500 science hours a year, and it’s been flying 400-500 science hours a year for 6 or 7 seven years now. Many studies are ongoing, in particular like looking at volcanoes, earthquakes, fault lifts and those kinds of things. One of those studies has been the levee study.”

“The aircraft operates at 1.3 gigahertz, which means it has a 10” wavelength for the radiation,” she said. “The radiation wavelength tells you about the size of the structure that it’s builds a scatter from, so if you have a bunch of leaves on the top of a bush, it’s going to penetrate through that bush and scatter on the ground. So this is a radar that can actually see below vegetation, very useful whenever you want to study bare earth in an area like the Delta.”

The radar has very good resolution at less than 6 feet for a single pixel, she said. It operates from 41,000 feet; the swath width is 14 miles, each line about 200 miles long, so you can cover the area in about 20 minutes. “That’s why we can fly up to Sacramento, fly nine lines, and collect all images in about a 2 to 3 hour time frame.”

UAVSAR is a full polarization instrument, Dr. Jones said. “The polarization of this instrument is exactly analogous to polarization of light,” she said. “Surfaces reflect differently from different polarizations … you get different backscatter from different polarizations depending upon what’s on the surface.”

She then presented an example slide, explaining that the polarization is called horizontal and vertical, just as it is for optical frequencies. “In our notations, VV means you scattered, sent out a pulse vertical, and you measure the components that come back in that same orientation, and likewise HH is horizontal,” she said. “HV would be where you sent out one and you receive in another; something like a branch that is at an angle, would cause that kind of thing to happen – you’ll get more backscatter that changes the polarization when you have a structure that’s at an angle.”

“What you see here is that water is brighter in the VZ channel, soil looks like a combination of the two channels, and whenever you add more water to the soil, it begins to look more and more like higher and higher VZ so you can tell something about the water content in the soil,” she explained. “The green is false color for the cross polarization channel and it kind of picks up the vegetation.”

The most important aspect of this type of radar imaging is its ability to detect deformation and small-scale changes of the surface, Dr. Jones said. She explained that in order to do this, they use GPS which makes it possible to fly repeat tracks accurate to plus or minus 16 feet and at exactly 41,000 feet.

“So if I’m on this aircraft day 1, I come back I’m sitting the same location within about 16 feet,” she said. “It’s exactly the same distance from the ground, so once you fly that same distance from the ground, you track the distance to the ground very carefully and the ground moves, it deforms in some way, you can measure that deformation accurately by doing the difference between the two images that you select at the two different times. So this aircraft I show you, this is just two right here, this outline here would be the 16 feet, and usually it’s in about 3 feet, so plus or minus one and a half feet is the accuracy of the repeat track.”

Dr. Jones then explained how the technique of synthetic aperture radar interferometry or InSAR is used to measure surface deformation. “The first time the plane flies over, it sends out a pulse down to the ground, it scatters, and whenever it comes back, we measure the time to the ground and back, and then we measure exactly where it is along that fraction of a wavelength,” she said. “If you watch the surface, the next time we come through, if the surface has moved, then that path length changes, and exactly where it is in the wavelength changes, so you can measure small fractions of the wavelength very well.”

She noted that the picture in the lower right area of the slide is a color contour map of change around the Lost Hills oil field from oil extraction, showing about 7 to 10 inches of change over this time period. “This is used to measure surface deformation; If the surface is all jumbled up, you cannot necessarily tell how much the surface moved, but you can tell that it changed a whole lot,” she said. “Secondly, you’re measuring along the wave that the pulse went to the ground. That’s called the line of sight direction, so you’re really measuring not this full change in subsidence, but the fraction that’s along your line of sight direction.”

Dr. Jones then explained how this technique can be used for monitoring levees. The state has identified several threats to levees: cracks in levees, seepage, slope instability, and slumps as well as general subsidence across an area that causes the levees to drop down. “Many of those can be detected if you can detect ground movement,” she noted.

She then gave some examples of what they have discovered using this technique in the Delta since the project was started in July of 2009.

The first was an example when a ship rammed a levee on Bradford Island in 2009. “The aircraft was flying back from Alaska from a science deployment, and my collaborator at DWR called and said, ‘Can you image and see if you see anything that happened on this Bradford Island levee, and I’m not going to tell you where it was. You’re going to have to look at this and tell me where this has happened.’ So of course this is like throwing down the gauntlet, so we actually diverted the plane and refueled, and then we collected all these images. When I looked at the data, I could pick the point where the levee was impacted, so this is an example where the impact was so great that I couldn’t actually measure the amount of deformation, but I could tell that something had moved a whole lot along this stretch of levee.”

Afterwards, the state was interested in how the new levee was holding up. “One thing that we’ve been doing post repair is looking at the rate of subsidence of that stretch of levee,” she said. “I have two examples here: A point along the stretch of levee that was repaired, and an inland point, and the graphs are a plot of the displacement that occurred as a function of time. Each of these points on the graph is measurement from one of our flights, so what you see is that inland, you get a small amount of a seasonal trend with a small amount of subsidence. On the repaired levee section, what we see is an exponential decay following the repair; you’ve added material, you’ve weighed down the levee, you get some compaction, and then it settles out over time. So it behaved exactly like one would have predicted it would, based on what you expect adding soil to a particular area.”

She then presented an example of an area of the levee on the southwest side of Sherman Island. “There’s a hint that there’s something different in this stretch of levee from our signal – the blue colors are more subsidence and the yellows are less. If you drive along that stretch of levee, you actually find a dip in the road right there, so this is an area where we’ve confirmed that we are actually detecting a problem with the levee.”

On Sherman Island, one thing the radar is detecting is the great variability between the amounts of subsidence that occurs inland in different fields. “This is an example from the western side of Sherman called the Whale’s Mouth where inland at this point, you see large seasonal trend and negligible subsidence at that location. Just a short distance away, maybe a mile away, at this point, you see that same seasonal variation on top of a much larger subsidence rate.”

Dr. Jones pointed out that if she came through and measured very infrequently, she wouldn’t be able to differentiate between what is seasonal change and what is a long-term trend. “The seasonal change is substantial in this area, and the seasonal variation is larger in many areas than the subsidence, so you need to measure as frequently as I have been measuring in order to detect the subsidence rates and be able to tell where the problem spots are.”

Dr. Jones noted that the box in the upper left of the image is an area where the radar sees a very distinct subsidence trend, but a photograph of the area doesn’t show anything in particular going on. “This is an area where repairs were done relatively recently. Probably this is new fill subsiding, you pack the material on and it begins to subside. You would drive along this road and not know that there was that big of a change. The radar is actually quite good at detecting this kind of change.”

She then turned to the image in the box in the lower part of the screen, focusing in on it. “This is a scour pond. In the past there was a levee break at this location; it scooped out a bunch of sand and now after post-levee repair, you have a pond there in that area. All along the levee, you see very large subsidence. The levee in these locations are fairly stable … but inland you’re getting 25 centimeters of subsidence, 10 inches of subsidence over a five year period. Now you’re eyeball is not going to tell you 2” over one year, but for the radar, this is like the worst area in the Delta that we’ve seen. It’s huge for the radar.”

Dr. Jones said they have instrumented this site recently to use it to validate the data. “We have a ground measurement and we have radar measurements and we validate the two. I think that’s really critical and it’s something that’s expensive to do, so we’ve been building up to get enough funding to instrument this site. I think we plan to finish that collection of validation by the end of this year.”

She then showed an example from Jersey Island, which is located just southeast of Sherman Island. “In one area, you see subsidence kind of centered in the fields off of the levee and it does seem to be affecting the levee, but the waterside slope (the green line) is fairly stable in that area.”

“On the other hand, you go slightly east from there, and you see a similar thing going on the field, but you see a stretch where the waterside slope is being affected. We can actually differentiate something that’s happening on the waterside slope from something that’s happening on the landside slope with this particular radar.” She noted that the radar was also able to detect an area where the levee has been repaired within the last ten years, and there is some subsidence from that.

She next showed an example from Webb Tract, noting that subsidence across the whole island is generally low, but in the one area, there is subsidence of the levee and subsidence inland from the levee. “This is not an area that’s been recently repaired, so it’s an area that one would flag as needing watching, based on this kind of data.”

On Holland Tract, a large stretch of levee was repaired, but only some stretches are subsiding, post repair. “Probably that indicates that there’s some difference in the soil underneath – some relatively localized changes in the soil types and so you get different responses to the soil from the loading from the fill. The loading from the fill seems to be extending inland from the levee, so it has an effect past where the load was placed. That’s what appears to be happening in this particular location.”

The next example was on Mandeville Island in the central Delta. “This is an island that has experienced the most subsidence in the Delta. In general the levees look pretty good, but there is an area where, similar to Jersey Island, where inland you see some modest subsidence trend, but at this one stretch of levee, you see much larger trend of subsidence.”

“I think this technology is really game changing in a way as it really can tell you in a very rapid way what’s going on across a large area,” she said.

[Slide 28] She then turned to discuss seepage, noting that it isn’t something that the interferometric technique normally is used for. “We flew three imagings of the Delta in one day, and we chose the day in the summer when you have the highest tidal extremes, so we flew at low tide, we flew at mid tide, and we flew at high tide. The idea was to look at changes in the elevation based upon the tidal level. We didn’t particularly see much change from that, but what we did see a huge decorellation in this one area. High tide was 3 am, mid tide was 6 am, we just didn’t believe that farmers were out there doing a whole lot between 3 and 6 am, so that doesn’t leave a lot of things that could be … it could be seepage because the water is changing in the channel next to it, so we identified this stretch of levee and we went out and actually found seeps at that location. The next year, we repeated this same test, and we found an area that had a seep and was repaired. So it’s serendipitous that you could use this data to identify seeps.”

“I think that seeps are important, but deformation of the levees are probably more important,” she said. “These levees, they probably seep a lot. It’s good to know that, because you want to know for example, during an earthquake, those might be areas more subject to liquefaction because they have higher water content, so it’s very good to know, but it’s not, but the focus of our study has been to look for this movement, which is much harder to do.”

Her final example was from the spring 2011 high water eventon the Mississippi River. “There were a lot of threats to the levees and even a levee break in one area,” she said. “You can see that these fields were entirely flooded, but what I noticed actually in this area of the field adjacent, in addition to this VZ backscatter from the river and from the flooded area, you see an indication that you have more moisture in the soil. This is the levee and these are basically fingers of moisture coming in underneath the levee, so this indicates that there was seepage. We classified the area and were able to identify the areas with the most seepage in an automated way, so we wrote an algorithm that could process this data in about 20 minutes, and the implication of that is that you could have this on board your aircraft and in the future, this product could be sent to the ground from the aircraft in near real-time, and that’s a pretty cool thing for emergency response.”

Dr. Jones said what she’s been doing is a proof of concept. “You go out and take a lot of measurements, you try to figure out what will make them better and what makes them work. It’s a prototype study. To take this to an operational program, you really have to automate it and integrate it with your decision support system.”

In terms of emergency response, you need to know in advance where InSAR is going to work, Dr. Jones said. “Do you have line of sight or do you big trees blocking you, so that in an emergency, you’re not going to trust the data from that area. You have a map showing you where you can trust the areas, but you have to acquire a baseline so that you know the conditions before the emergency happens.”

“Then you have really develop and integrate all the products under test scenarios so that everyone believes the products and everyone is used to the products,” she said. “That’s the difference between what I’ve been doing and where I think this program would have to go in order to be an operational program.”

Dr. Jones then said in conclusion that this technology could be very useful for identifying change on earthen levees. “I think the difference between what’s been done before and what this is really doing is with earthen levees where you have low signals, low subsidence, and small backscatter, you need these long time scales in order to be confident of your results. And we are able to achieve the accuracy we did because we have acquired 51 flights over this area over a long time period.”

“In general, a simple assessment could tell you where InSAR is likely to work. It can tell you whether you have trees in the way, whether you have line of sight to that particular levee, and so you can actually in advance to a lot of work to make this useful during an emergency.”

During the discussion period, Dr. Jones was asked about the resolution if the radar is mounted on a satellite. “The difference between the aircraft and the satellite is that for the satellite, the image resolution wouldn’t be quite as good, so you would have slightly larger pixels. Instead of one pixel being about six feet, it would be about 18 feet, roughly a factor of three for your pixel resolution. The other thing is the satellite would have a higher noise level, so you wouldn’t have quite as good a measure in each measurement; however, it’s compensated by the fact that the satellite goes over every 24 days or every 12 days and so you get more images. You get a long time series, so all that noise is adjusted a little bit.”