An Introduction To Synthetic Aperture Radar (SAR) and Its Applications

0:01 Speaker 1: Hello and welcome to our Introduction to SAR and its Applications webinar series. I'm Dr. Erika Podest and I'm a scientist at NASA's Jet Propulsion Laboratory focusing on the study of terrestrial ecosystems using SAR. I'm also an instructor with the RSAT program. Although this webinar series is an introduction, it covers topics in greater deep depth than the first Introduction to SAR RSAT training from 2017. The open availability of operational SAR data from Sentinel-1 and some PALSAR scan SAR data is making the use of SAR more prevalent.

0:41 Speaker 1: And this will continue to grow with upcoming radar satellite missions such as NISAR and Biomass. And this is why we believe this introductory training is so timely. Before I begin talking about SAR, I would like to provide a bit of background about the RSAT program. NASA's Applied Remote Sensing Training Program, RSET, offers free trainings on the use of remote sensing and model data, along with methods and tools for the global science and application communities. Our trainings are tailored to different experience levels, ranging from introductory to intermediate to advanced.

1:25 Speaker 1: RSET trainings focus on six thematic areas, agriculture, climate and resilience, disasters, ecological conservation, health and air quality, and water resources. ARSA trainings are either online or in person. All trainings are free of charge and use open source data and software. Some of our trainings are delivered or presented in both English and Spanish, like this one. Yet, all of our trainings have a Spanish PowerPoint presentation available.

2:06 Speaker 1: And next, I will provide you with an overview of this webinar series. This webinar series consists of three sessions. Today, we will cover an introduction to SAR. The second session on November 13 will focus on an introduction to interferometric SAR in SAR with guest lecturer Dr. Eric Fielding from JPL. The third and final session will be on Wednesday, November 20th, and will provide an overview of SAR data sources and tools with guest lecturers, Dr. Franz Meyer from the Alaska Satellite Facility and the University of Alaska Fairbanks, as well as Heidi Christensen from the Alaska Satellite Facility.

2:50 Speaker 1: There will be a homework assignment that opens on November 20th, the last day of this training, and it will have a due date of December 4th. A certificate of completion will be awarded to participants who attend all the live sessions and complete the homework assignment by the due date. The prerequisite to this webinar series is the fundamentals of remote sensing.

3:20 Speaker 1: And these are the learning objectives for this webinar series. So by the end of this webinar series, participants will be able to recognize basic features and functionality of SAR, evaluate SAR sensor characteristics for addressing different science questions and application areas, interpret the information content in SAR images to distinguish different features detected by the sensor, evaluate the creation of an interferogram through interferometric SAR, interpret an interferogram to measure surface deformation and small movements, compare and contrast the capabilities of historic current and upcoming SAR data, and finally, to access and visualize SAR data for a given location and time.

4:12 Speaker 1: Now, this first session will provide an introduction to the basic principles of SAR. I am the instructor for this training. I'm a scientist at NASA's Jet Propulsion Laboratory, and my expertise is in the use of microwave remote sensing, primarily SAR, for the study of wetland inundation dynamics, vegetation growing season in the northern high latitudes, and mapping and monitoring land cover and land use change. I'm also the community engagement and training coordinator for an upcoming SAR satellite called NYSAR, which is a collaboration between NASA and the Indian Space Research Organization.

4:56 Speaker 1: NISAR is scheduled to launch in early 2025. So these are the objectives for today's session. By the end of the session, participants will be able to understand the radar remote sensing signal characteristics, recognize how the radar signal interacts with the surface, interpret a SAR image to distinguish different features detected by the sensor, and identify application areas where different SAR sensors are most applicable.

5:32 Speaker 1: How to ask questions. Please write your questions in the questions box, which is at the bottom right. You'll have, there'll be three points. Click on that and there'll be a pop-up menu. Select questions and write your questions in that box and we will address your questions at the end of this session. Feel free to enter your questions as we go along, and we'll try to get to all of your questions during the Q & A session. The remaining questions that we cannot answer, we will answer them in the Q & A document, which we will post to the training website in about a week.

6:15 Speaker 1: So now let's start with our introduction to synthetic aperture radar.

6:22 Speaker 1: So let's start with the electromagnetic spectrum, which is the range of electromagnetic energy that spans from very long wavelengths, such as radio waves, which can be the length of a football field, to very short wavelengths, such as gamma rays, which are the length of an atomic nucleus. Note that most of the energy on the spectrum is not visible to our eyes, and that visible light is just a sliver of energy from the total amount that surrounds us. Now, I want you to think of this energy as waves propagating, not much different than waves crossing an ocean, except that electromagnetic radiation travels at the speed of light.

7:06 Speaker 1: Remote sensing sensors are designed to operate at specific regions of the electromagnetic spectrum. according to their intended application. Microwave sensors operate within the range delineated in this figure, which is at a much lower frequency range than optical and infrared sensors. Now, to put things into context, the wavelength of light is about 390 to 700 nanometers, while for microwaves, it's on the order of 0.3 to 100 centimeters. because of this huge disparity in wavelengths, the features on the Earth's surface appear differently in the microwave range than in the optical range.

7:56 Speaker 1: In general, there are two main types of remote sensing observations, passive and active. So generally speaking, passive sensors, they measure energy that's emitted or reflected by the Earth's atmosphere system. And examples of passive sensors include optical and thermal sensors, as well as passive microwave sensors. Then there's, in general terms, active sensors, which provide their own source of illumination. And these instruments emit a burst of energy, let's call it a signal, and then they measure the portion of that signal that's reflected back.

8:32 Speaker 1: Examples of active sensors include lidars and radars. Now, the term microwave remote sensing refers to making observations in the microwave range of the electromagnetic spectrum. And these observations can be passive. Passive sensors, known as microwave radiometers, are analogous to optical sensors. However, they operate in the microwave part of the spectrum. All objects emit microwave energy in very small amounts. Passive microwave sensors detect this naturally emitted energy, which is related to the temperature of the emitting surface.

9:13 Speaker 1: The antenna on the microwave radiometer detects this emitted energy, and because this emitted energy is very small or has a low signal-to-noise ratio, the spatial resolution of microwave radiometers is typically low, on the order of 10 to 50 kilometers. A large field of view is therefore necessary to detect enough energy to record a signal, which is why most passive microwave sensors are characterized by having a low spatial resolution. All right, so now let's talk about active microwave sensors, which can be further categorized into imaging and non-imaging sensors.

9:55 Speaker 1: Examples of non-imaging sensors include radar altimeters, and scatterometers. Typically, these sensors take measurements in one linear dimension, as opposed to the two-dimensional representation of imaging sensors. Radar altimeters, they transmit short microwave pulses and measure the round-trip time delay to targets. in order to determine their distance from the sensor. Generally, altimeters look straight down at nadir below the platform, measuring height or elevation. Then there are scatterometers.

10:32 Speaker 1: These are also non-imaging radar sensors that provide quantitative measurements of the energy returned from targets. They usually have a large footprint and coarse spatial resolution on the order of, again, 10 to 50 kilometers. and are primarily used in ocean studies to measure wind speed and direction. Now, the main difference between synthetic aperture radar and scatterometers is that a synthetic aperture radar sensor has a high spatial resolution on the order of 10 to 50, 100 meters, but relatively low radiometric resolution, around 1 dB.

11:16 Speaker 1: while scatterometers have low spatial resolution but high radiometric resolution on the order of a tenth of a decibel. So finally, there's synthetic aperture radar. That's an imaging active microwave sensor. And SAR is analogous to an ultrasound. SAR remote sensing is commonly referred to as radar remote sensing. And in this webinar series, we will focus only on radar remote sensing, specifically SAR. This slide details the advantages and disadvantages of radar remote sensing compared to optical remote sensing.

12:00 Speaker 1: Radar systems can operate under almost any kind of weather condition, which is a significant advantage over optical sensors, which are hindered by clouds. And this capability is particularly valuable given that many regions, especially in the tropics, are frequently cloud covered. Radar can observe the Earth's surface at any time, regardless of whether it's day or night. And this is very useful during winter months where large areas experience long hours of darkness, as well as in, especially in polar regions where darkness can persist for several months.

12:38 Speaker 1: Radar signals can penetrate through various mediums, such as vegetation canopies, snow packs, and soils, although the extent of this penetration is dependent on the frequency. In contrast, optical sensors only detect information from the very surface of these mediums. Radar remote sensing requires minimal to no atmospheric corrections, whereas optical remote sensing is heavily reliant on atmospheric corrections for proper image interpretation. And then finally, radar is highly sensitive to the moisture and structural characteristics of the surface being observed, while optical remote sensing is sensitive to the spectral properties of the land surface.

13:25 Speaker 1: It's important to note that the information content in radar data differs from that in optical data. And this distinction is not necessarily an advantage or a disadvantage, but it is crucial for understanding the types of insights that each sensor provides. However, now there are disadvantages associated with radar. The information content in radar data differs from that in optical data, which can make interpretation challenging. In addition, radar images often exhibit speckle, which is a salt and pepper effect that complicates image interpretation and analysis.

14:08 Speaker 1: And then finally, the presence of topography can introduce distortions in radar images that must be accounted for during analysis. This slide highlights the advantage of using radar data, which can observe the Earth's surface regardless of most weather conditions. The figure displays an average of all satellite cloud observations from MODIS between 2002 and 2015. The color gradient ranges from dark blue, indicating no clouds, to light blue, indicating some clouds, and white, indicating frequent cloud coverage.

14:50 Speaker 1: As shown, certain regions of the world, such as tropical areas, experience significant cloud cover. In these locations, radar is an ideal sensor for obtaining reliable and consistent observations.

15:08 Speaker 1: This slide shows a comparison between an optical composite and a radar composite of Panama. The image on top is a Sentinel-2 RGB optical composite, while the image on the bottom is a Palsar-Skansar radar composite from the Japanese space agency JAXA, ALOS-2 satellites. Both composites cover the same timeframe from November 1st through the 30th, 2019. Cloud cover is common in Panama, resulting in gaps in the optical data even over an entire month. However, this is not an issue for radar sensors. So now let's discuss the fundamentals of radar image formation.

15:52 Speaker 1: Let's begin with some basic concepts of radar. Radar stands for radio detection and ranging, originated from investigations into the reflection of radio waves in the late 1800s and early 1900s. It was developed in earnest during World War II, primarily for navigation and target location. As previously mentioned, radar can be categorized into two types, imaging and non-imaging. This webinar series is focused on imaging radar, and there are two types of imaging radars. real aperture radar, and synthetic aperture radar.

16:31 Speaker 1: Now, as the acronym suggests, radar functions as a ranging or distance measuring device. Essentially, a radar emits a signal toward the Earth's surface and measures the time it takes for that signal to return. All imaging radars are side-looking. If a radar were to look straight down, as depicted in the figure on the left, it would not be able to differentiate between two points, A and B, on the surface. In this scenario, the signal would reach point A and B simultaneously and return to the sensor at the same time, which is why you wouldn't be able to differentiate them.

17:14 Speaker 1: However, when the radar is side-looking, as shown in the figure on the right, the time it takes for the signal to reach point A and B differs. allowing for points A and B to be differentiated. Both real aperture and synthetic aperture radar are side-looking systems with an illumination direction usually perpendicular to the flight line. A radar system fundamentally consists of a transmitter, a receiver, an antenna, and an electronic system for processing and recording data. The transmitter emits short bursts or pulses of microwave energy at regular intervals, which are focused by the antenna into a beam.

17:59 Speaker 1: This beam, a radar beam, illuminates the surface obliquely at a right angle to the motion of the platforms. the antenna receives a portion of the transmitted energy that is reflected or backscattered from various objects within that illuminated beam, as shown in this figure. By measuring the time delay between the transmission of a pulse and the reception of the backscattered echo from different targets, the radar can then determine their distance and location. As the sensor platform moves forward, the recording and processing of the backscattered signals build up a two-dimensional image of the surface.

18:43 Speaker 1: One unique aspect to radar is that the spatial resolution in the X and Y directions is independent of each other, and I will explain this concept further in the upcoming slides.

18:57 Speaker 1: So let me first explain some of the terminology in radar geometry. The instrument measures distances between the sensor and the point on the Earth's surface where the signal is backscattered. This distance is flat range. So look at the illustration here. It's the direct distance between the sensor and the target. The flight direction is also referred to as a long track or azimuth direction. And the direction perpendicular to the flight path is the across track or range direction. As the radar on the satellite or aircraft moves in the direction of flight, the ground is scanned in two dimensions.

19:40 Speaker 1: One dimension is the range dimension. Objects are placed along in this dimension according to their distance from the radar. The second dimension is the along track or azimuth dimension. It is parallel to the motion of the radar platform. In this dimension, the ground is scanned by the beam moving across the ground at a rate equal to the speed of the platform of the aircraft or the satellites. and objects are placed in this dimension according to the position of the aircraft along the track. An image is built from the reflected signals in both dimensions.

20:22 Speaker 1: Range resolution is the resolution perpendicular to the sensor's movement. So it's the resolution in the direction that the sensor is looking, okay, and it Range resolution in SAR refers to the ability of the radar system to distinguish between two targets that are located at different distances from the radar sensor. It is the minimum distance between two objects that allows them to be separated in the radar image. So if two targets are closer together than this resolution, they will appear as a single target.

21:00 Speaker 1: In order for two targets to be distinguished, their echoes must necessarily be received at different times. This image shows a pulse of length L, which is determined by the duration of the pulse transmission. This pulse is moving toward buildings A and B. The slat range distance or the direct sensor to target distance between the two buildings is D, as shown in the top image. For a side looking radar, to image separately two ground features that are close to each other in the range direction, it's necessary for the reflected signal from all parts of the two objects to be received separately by the antenna.

21:45 Speaker 1: Any time overlap between the signals from the two objects will cause their images to be blurred together. Note that the slant range distance between the buildings is less than the pulse length, or L over 2. So because of this, the pulse has had time to travel to B, and part of its return signal overlaps with the return signal from A. The signal overlap will cause objects A and B to be imaged as one large object, extending from A to B. If the slat range distance between A and B were greater than the pulse length over 2, then the two signals would be received separately and would be two different objects on the image.

22:33 Speaker 1: Shorter pulse lengths lead to better range resolution. If the pulse is shortened, however, it has a negative effect on the signal-to-noise ratio and decreased radiometric resolution. To address this limitation, modern SAR sensors use a technique to generate longer pulses with linear frequency modulation called chirping. So instead of sending short pulses with a constant frequency, long pulses are transmitted that sweep up or down in the frequency. And that's why it's called chirp. The extent of that frequency sweep is known as the pulse bandwidth.

23:11 Speaker 1: The frequency modulation must then be processed after reception to focus a pulse to a much shorter value. Now, for the user, the result is the same as if a very short pulse length had been used throughout the system. The pulse bandwidth is proportional to the pulse length. Shorter pulse length means larger, it's inversely proportional, so shorter pulse length means larger bandwidth and better range resolution. Pulse bandwidth is expressed in units of hertz rather than distance. Now, the range resolution or the slat range resolution is not the same as the spatial resolution of the final image.

23:56 Speaker 1: For analysis, we're interested in the ground range resolution, which is the actual distance between two points on the ground. So the figure on the bottom left shows that equidistant points on the ground, the ground range, result in differently distanced points on the slat range. The ground range describes the ability to discriminate objects that are situated on the ground and is calculated from using the local incidence angle, which is the angle between the line of sight of the radar and the vertical to the terrain.

24:29 Speaker 1: And I'll talk more about this later on in this presentation. Azimuth resolution describes the ability of an imaging radar to separate two closely spaced scatterers in the direction parallel to the motion vector of the sensor. In the case of a real aperture imaging radar, such as in the image, when two objects are in the radar beam simultaneously, For almost all pulses, they both cause reflections and their echoes will be received at the same time. So this figure is an example of azimuth resolution for a real aperture radar.

25:11 Speaker 1: Azimuth resolution is determined by the angular beam width of the antenna and the slant range. So note the projection of the antenna beam on the ground. As the antenna beam, it fans out with increasing distance from the aircraft, the azimuth resolution deteriorates. So the objects at point A and B in the figure would be resolved at the slant range of SR1, but not further away from the sensor at SR2. So that is at distance SR1, A and B result in separate return signals. And at distance SR2, A and B would be in the beam simultaneously and would not be resolved.

25:54 Speaker 1: For a real aperture radar, two targets in the azimuth or a long track resolution can be separated only if the distance between them is larger than the radar beam width. Now, the radar beam width is inversely proportional to the antenna length, which is also referred to as the aperture, which means that a longer antenna or aperture will produce a narrower beam and finer resolution. The actual length of the antenna is limited by what can be carried on an airborne or space-borne platform. For airborne radars, antennas are usually limited to 1 to 2 meters, and for satellites, they can be 10 to 15 meters in length.

26:37 Speaker 1: So to overcome this size limitation, the forward motion of the platform and signal processing based on the Doppler effect of the backscatter echoes are used to synthesize a very long antenna. and therefore increase the azimuth resolution.

27:00 Speaker 1: So the long track resolution depends on the length of the antenna and to produce high resolution images of the Earth's surface, the antenna needs to be hundreds of times greater in size than what can be carried on a satellite. So this is synthesized by using the movements of the satellites to simulate a very long antenna. Hence, it's called synthetic aperture radar. And typically, the larger the aperture, the higher the image resolution. SAR therefore allows us to create high resolution images with comparatively small physical antennas through the use of signal processing of the Doppler shift associated with the motion of the aircraft.

27:39 Speaker 1: This synthesizing process is possible because a scatterer or target, A in this point, on the ground remains within the real aperture radar beam for many radar pulses. Adding up the reflections from all these pulses appropriately allows us to synthesize the large antenna with a much narrower beam width, resulting in a better or higher spatial resolution in the azimuth direction. Radar pulses travel at the speed of light and can only measure the part of the signal that is reflected back toward the antenna.

28:20 Speaker 1: Most radars measure amplitude and phase. Amplitude is the strength of the reflected signal, and this amplitude is the backscatter coefficient, also known as sigma-nod, which is a measure of the radar signal that is scattered back to the sensor from the Earth's surface, normalized to a unit target area. Sigma naught is expressed in decibels by taking the log 10 of the energy ratio. Usually values vary from minus 25 dB, which are areas that have low backscatter or dark tones, or little energy reflected back to the radar, to values that are greater than 1 dB.

29:00 Speaker 1: So those are high backscatter areas or bright tones where a large amount of energy is reflected back to the radar. The phase is the position of a point in time on a waveform cycle. And phase is usually measured in angular units like degrees or radians. The phase difference between two images is a new kind of image called an interferogram, which is a pattern of fringes containing all of the information on relative geometry. And so the next session in this webinar series will be focused on interferometric SAR.

29:35 Speaker 1: Here, we will only discuss amplitude.

29:41 Speaker 1: Before moving on to the next session, the most important thing to remember is that with radar, the resolution in the x and y directions can be different because they are determined by different radar parameters. Also, since radar looks sideways, The distance between objects in the slant range is not the same as in the ground range. And usually, images need to be transformed from slant range to ground range. I will talk about this later. Finally, radar uses or measures signal amplitude and phase.

30:14 Speaker 1: Now, not all radar systems measure phase. Amplitude is expressed in decibels and phase in radians or degrees. In this next section, I will talk about the different ways that the radar signal can scatter back to the satellite when it interacts with different surfaces or targets. There are different types of backscattering mechanisms, and as a useful rule of thumb, the higher the backscattered intensity, the rougher the surface. First of all, surface roughness refers to the average height variations in a surface from a plane surface, and is measured on the order of centimeters.

30:57 Speaker 1: If a surface appears rough or smooth to the radar, this depends on the wavelength and the incidence angle. The wavelength will determine the interaction with the objects on the surface. And generally, if the roughness of an object or surface is comparable to the size of the waves, then there will be interaction. The surface will appear rough and some energy will be reflected back. A surface is considered smooth if the height variations are much smaller than the radar wavelength. The first type of backscatter mechanism is a smooth surface, also known as a specular reflector.

31:35 Speaker 1: A smooth surface acts like a mirror for that incident radar energy, and most of the incident radar energy is reflected away from the sensor, and those areas therefore appear very dark. So open water surfaces tend to be specular reflectors. When the surface height variations begin to approach the size of the wavelength, then the surface will appear rough. A rough surface will scatter energy approximately equally in all directions diffusely, and a portion of the energy will be backscattered to the radar.

32:11 Speaker 1: Then there's the opposite mechanism, which is called double-bounce. and illustrated here with a building. It occurs when two smooth surfaces form a right angle facing the radar beam and the beam bounces twice off the surfaces and most of the radar energy is reflected back to the radar sensor. Double balance is commonly seen in urban areas such as areas where there are roads and high rise buildings or also in flooded vegetation where there's upright vegetation and in standing water. The smooth surface, whether it's a road or standing water, will bounce the signal away from the radar, and the building or the tree trunk will then reflect the energy back to the radar, and these areas will appear very bright in the image.

32:59 Speaker 1: Finally, there is volume scattering, which is the scattering of radar energy within a volume or medium, and usually consists of multiple bounces and reflections from different components within the volume. You can also have volume scattering within a snow path, for example, or within a vegetation layer.

33:23 Speaker 1: And this is an example of volume scattering in a forest. Scattering may come from the leaf canopy at the tops of the trees, the leaves and the branches further below, and the tree trunks and soil at the ground level. Volume scattering may serve to decrease or increase image brightness, depending on how much of the energy is scattered out of the volume and back to the radar. This figure illustrates all of the possible scattering mechanisms within a forest. You've got direct scattering from the tree trunks, ground-crown scattering, crown-ground scattering, ground-trunk scattering, trunk-ground scattering, crown-volume scattering, and then direct scattering from the soil surface.

34:08 Speaker 1: So the following couple of charts are examples of the different scattering mechanisms discussed and how they look like in a radar image. So this example shows specular reflection, and this image is a mosaic of the Amazon Basin from the SMAP radar, which was an L-band radar. The yellow circle delineates an area of specular reflection, which is open water, and in this case is part of the Amazon River. So flat surfaces such as paved roads, runways, or calm water normally appear as dark areas in a radar image since most of the incident radar energy is reflected away, away from the radar sensor.

34:55 Speaker 1: This example shows rough surface reflection delineated by the yellow circle on this map radar image. And this area is an area of very low to no vegetation. an area that has been deforested. Agricultural fields with low vegetation, tilt fields, savannas, bare fields, or in general, areas that have low vegetation have rough surface scattering. And the brightness of the pixel is slightly higher than a specular reflector.

35:27 Speaker 1: This example shows volume scattering delineated by the yellow circle on the SMAP radar image. This is a forested area, and note the pixel brightness is higher than rough surface and specular scattering. The intensity of volume scattering depends on the physical properties of the volume, such as variations in moisture and structure, and of course the characteristics of the radar, wavelength, polarization, and incidence angle. Finally, this example shows double bounds delineated by the yellow circle on the SMAP radar image.

36:02 Speaker 1: This is an inundated wetland, meaning that the trees are on standing water. The reason that the signal is so strong is because it bounces off the water, which is a specular reflector, onto the tree trunk and other components of the tree, and then bounces back to the satellite. We cannot measure the amount of standing water, just that there is water above the surface. Double bounce also occurs in urban areas where streets are smooth surfaces reflecting the signal away from the radar and then the buildings reflecting the signal back toward the satellites and so urban areas also appear very bright.

36:42 Speaker 1: There are radar parameters and there are surface parameters that influence the characteristics of the signal. We will start with the radar parameters. There are three radar instrument parameters that influence the transmission characteristics of the signal. These are wavelength, polarization, and incidence angle. Wavelength is the length of the peak of one wave to the peak of the next wave, or really the position of any point in the wave to the same point in the next wave. In radar remote sensing, we often talk about wavelength rather than frequency.

37:22 Speaker 1: And this is because the length of the wave defines the interaction of the signal with the surface or medium, and it is therefore easier to associate the capability of the sensor. Wavelength is inversely related to frequency. It is the speed of light divided by frequency. The higher the frequency, the shorter the wavelength, and vice versa. The table on the right lists the common wavelength bands in radar. The letter codes KALP For example, we're originally selected arbitrarily to ensure military security during the early stages of radar development, which is why the denominations are not in alphabetical order.

38:05 Speaker 1: As you can see, each band has a wavelength and a frequency range. So, for example, when someone talks about a specific L-band radar, and I'll give you the example of one called PALSAR, which is an L-band SAR sensor on board the Japanese ALU satellite. Pulsar operates at a specific frequency within the L-band range, in this case at 1.2 gigahertz. To convert that to wavelength, we take the speed of light and divide it by 1.2 gigahertz. That will equal to 25 centimeters. Note that a C-band sensor has a wavelength of around 6 centimeters, while an L-band sensor has a wavelength of around 24, 25 centimeters, and P-band on the order of 68 centimeters.

38:51 Speaker 1: There is therefore a wide variation in wavelength between bands. There are two important things to remember about wavelength. The first is that the longer the wavelength, the greater the penetration of the signal through the medium. That means soil, snow, or vegetation. The figure on the lower left shows that a shorter wavelength, such as X-band, which is around 3 cm, will see the top part of the forest or probably the top third or fourth of the canopy, while C-band will penetrate halfway through a forest and L-band will penetrate all the way through.

39:30 Speaker 1: And this is limited by vegetation density, but that's the general idea. The same goes for soils. In the case of dry soils, X-band will see the top part of the soil, while C-band will penetrate further and L-band even further. It has been shown that in areas where the soils are very dry, such as in deserts, the L-band signal can penetrate on the order of two meters deep. The second important thing about wavelength is that the length of the wave will determine the interaction with surface objects.

40:02 Speaker 1: In general, if an object or surface roughness is comparable to the size of the wave, there will be interaction. The surface will appear rough and there will be energy scattered back. For example, Little energy is backscattered from a surface with a high fluctuation of the order of 5 centimeters of L-band, which has about 25 centimeters wavelength. If L-band SAR is used, then the surface, therefore, will appear dark. However, the same surface will appear bright due to increased backscatter in a C-band SAR image, which has a wavelength of about 5 centimeters.

40:41 Speaker 1: The table in this chart shows examples of applications most relevant to specific bands. If you're interested in forest-type studies where you'd want deep penetration through the canopy and information about forest structure, then you would go with L or P bands. If you're looking at agricultural-type studies where the biomass is not as large as that of a forest and you want penetration through the vegetation layer and information about crop structures such as leaves or stems, then you would go with C band or X band.

41:14 Speaker 1: For studies of ocean surface roughness, you would go with X or KU bands, or studies where you're interested in sensitivity to snow particle size. And here's an example of radar penetration into soils. This area is part of the Nile River near the fourth cataract of Sudan. Each image is about 50 by 19 kilometers. The top image is a photograph actually taken with color infrared film from the space shuttle Columbia in November of 1995. And the radar image at the bottom was acquired with Sir CXR in April of 1994.

41:56 Speaker 1: It is a false color image using a combination of L band and C band at different polarizations. The radar image shows a previously unknown paleo channel lying north of the Nile, that dark band. And this paleochannel cannot be seen on the photograph. These observations indicate that the course of the Nile recently shifted to the south due to relative uplift of adjacent regions of the Nubian soil. Here we have an example of penetration depth through vegetation with two different wavelengths in Kalimantan, Indonesia.

42:39 Speaker 1: So the top image is a C-band with a wavelength of about 5 centimeters, and the bottom one is P-band with a wavelength of around 68 centimeters. Note the difference

42:52 Speaker 1: between the two and the greater level of detail in the P-band image. The purple colored areas in the C-band image are likely areas of low vegetation. However, the P-band image shows a lot more complexity as well as More information in the areas south of those purple areas. Also note the purplish areas north of the river in the P-band image and the road structure in the forest area in the P-band image. And here's another example of signal penetration through vegetation to detect flooding. Here we have multi-frequency SAR images at CL and P-bands.

43:37 Speaker 1: obtained from JPL's airborne radar called AirSAR. The area imaged is a wetland in the Manu National Park in Peru back in 1993. Now, remember that C-band has a wavelength of around 5 centimeters, L-band around 24 centimeters, and P-band around 68. So, as you would expect, the longer wavelength penetrates deeper into the canopy to detect flooded vegetation, which are the bright areas, that's flooded vegetation. So note all of the white areas in the P band image that are either not present or less bright and extensive in the L band image or the C band image where there is hardly any flooded vegetation detected.

44:23 Speaker 1: The other radar parameter is polarization and it refers to the plane of propagation of the electric field of the signal. Irrespective of wavelength, radar signals can be transmitted and or received in different polarizations. Most radars are designed to transmit microwave radiation either horizontally polarized or vertically polarized. And this figure illustrates these planes of propagation. So if you had a long string tied to an object and you stand at the very end of the string and move it up and down in a vertical manner, such as the figure in the very top, then the signal is vertically polarized.

45:06 Speaker 1: If you move the string horizontally from side to side, then the signal is horizontally polarized. Similarly, the antenna can receive either the horizontally or vertically polarized backscatter energy by applying specific filters. And some antennas can receive both. So there can be four combinations of both transmit and receive polarizations as follows. HH, which means It's horizontally transmitted and horizontally received. VV, vertically transmitted and vertically received. HV, horizontally transmitted and vertically received.

45:44 Speaker 1: And VH, vertically transmitted and horizontally received. The first two polarization combinations are referred to as light polarized because the transmit and receive polarizations are the same. The last two combinations are referred to as cross polarized because the transmit and receive polarizations are opposite of one another. Systems that collect in all four combinations are called quad-pole, and they can more fully characterize structure on the surface. In summary, polarization provides information on the horizontal and vertical components of the target.

46:23 Speaker 1: This is an example of L-band SAR data from a NASA airborne sensor called UAB-SAR. In 2013, it acquired imagery over the Pacaya Samaria Natural Reserve in the Peruvian Amazon. This reserve is a vast wetland ecosystem. The images in black and white are individual HH, HV, and VV polarizations. Note the difference in intensity between the different images, especially with the cross-polarized HV image. False color images are an ideal way to visualize the contributions of the different polarizations to color combinations.

47:07 Speaker 1: And this example has HH in the red channel, HV in the green channel, and VV in the blue channel. You can see the information content unique to each polarization, which in this case is representative of different types of wetland vegetation or different types of vegetation. HV is in the green channel, and anything that is green means that the return is higher in HV than VV or HH. Anything that is magenta means that there is a combined contribution between HH and VV. An important point to highlight is that if the radar energy is reflected from multiple scatterers within the canopy structure, it often becomes depolarized, leading to a strong reflection in the cross-polarized channels.

47:56 Speaker 1: Essentially, the more times the signal interacts with different components, as happens in volume scattering, the higher the likelihood that the signal will depolarize. And this is why cross polarizations like HV or VH are good indicators of the presence of vegetation and its structural characteristics. Incidence angle is the angle between the direction of the incident wave and the Earth's surface plane. In radar, the incidence angle increases across the swath from the near to the far range. Large angles will be more sensitive to surface roughness and will penetrate less into the medium as opposed to small angles.

48:40 Speaker 1: Low incidence angles or small angles will result in high backscatter and greater penetration or higher backscatter. So generally, slopes facing towards the radar will have small local incidence angles causing relatively strong backscatter to the sensor, which results in bright-toned appearances in an image. This is an example of the effect of incidence angle variation. On the left is a Sentinel-1 BV image, and on the right is the incidence angle variation. for the same image. The radar is right-looking and notes that as we move across the swath from near to far range, the image becomes increasingly darker in tone, around a 3 to 5 dB difference in backscatter.

49:37 Speaker 1: If you measure, for example, the same forest at different incidence angles, then the measured backscatter will be different even if there has been no physical change in the forest. Every surface feature will have backscatter that is a function of incidence angle, so you need to be careful when comparing the backscatter of a feature when there is a large variation in incidence angle.

50:07 Speaker 1: Up to now, we've talked about the radar parameters. Now let's talk about the surface parameters that influence the signal.

50:17 Speaker 1: The radar backscatter contains information about the Earth's surface, which drives the reflection of the radar signal. And this reflection is driven by the characteristics of the signal, which is a function of the wavelength, polarization, and incidence angle, and then the characteristics of the surface. So there are two surface parameters that is unique to radar. The information content that is unique to radar images is related to structure and to moisture. There are three parameters related to structure in general.

51:00 Speaker 1: That's density, size relative to wavelength, and size and orientation. Let's start with size relative to wavelength. If an object or surface roughness is the approximate size of a wave, then there will be interaction. The surface will appear rough and there will be energy scattered back. If the length of the wave is much larger than an object, there will be no or minimal interaction between the wave and the object and the surface will appear smooth. And there will be minimal to no energy scattered back.

51:35 Speaker 1: Backscatter from vegetation is strongly dependent on the size of the scattering elements within the canopy. And there is greater response to scatters which are of similar size to the length of the wave. In this example, this is an example of an Austrian pine. The primary scatters in a tree canopy are the leaves, branches, and stems with a size on the order of the wavelength or larger. So we're looking at this pine tree with X-band will primarily be seeing the stems, the smaller stems and the needles.

52:15 Speaker 1: So on the order of elements that are around three centimeters. If we're looking at this pine tree with L-band on the order of 27 centimeter wavelength, then we'll be looking at primarily the longer branches and part of the trunk. And finally, if we're looking at this tree with P band, which has a wavelength on the order of 70 centimeters in this case, we're looking at the even longer branches and pretty much all of the trunk. Now let's look at size and orientation. This example shows two leaves and two trees that have different orientations.

53:01 Speaker 1: If we look at the leaves with a shorter wavelength, say X band on the left, we will get a slightly larger backscatter response in the vertical than in the horizontal polarization for the first leaf. And this is because the components of that first leaf are slightly more vertical in blue than horizontal in red. The second leaf is distributed equally in the horizontal and vertical directions. and the return in those polarizations will be about the same. For the trees which we are looking with a longer wavelength, say L or P band, the first one will have a higher return in the vertical polarization from both the trunk and the canopy.

53:45 Speaker 1: And the second one will also have an overall stronger return in the vertical polarization because the trunk component dominates. However, note that in the crown, The crown has slightly more horizontal than vertical components. This is an example of multipolarization images from Elos Pulsar over a part of the Pacaya-Samaritan Natural Reserve in Peru, which, as mentioned, is a vast wetland ecosystem. The images are at L-band and at HH, HV, and VV polarizations. Now, the very bright white areas that you see, these are inundated vegetation.

54:31 Speaker 1: The very dark areas are open water. And you can clearly see a river through the middle of the image, as well as some open water bodies north and south of the river. Note the difference in backscatter magnitude between the different polarizations. If you focus now on inundated vegetation or the very bright areas, HH polarization is the most useful for distinguishing flooded from non-flooded vegetation. VV shows flooded vegetation to a lesser extent and even lesser by HV polarization. In general, HH polarization penetrates deeper into the vegetation canopy than VV.

55:18 Speaker 1: Why is this? Because forests tend to have more vertical than horizontal structure. So the signal is more likely to reflect back from the vertical components before reaching the bottom of the canopy. The horizontally polarized wave has less structure to interact with and hence has a higher probability of penetrating all the way through the canopy. And then there's density and the denser the vegetation, the less the probability of the signal penetrating through. So there are saturation levels, and that's going to vary with the type and structure of forests.

56:06 Speaker 1: And so in this case, for a broadleaf, evergreen, and coniferous forest, this study indicated that C-band saturates at around 20 tons per hectare, L-band at around 40 tons, and P-band at around 100 tons. There have been many studies, though, in looking at this saturation issue, saturation of the signal, and there are many different ranges. There are other studies that have indicated that in tropical forests, the signal at L band will saturate between 80 and 150 tons per hectare, while at P band between 200 and 350 tons per hectare.

56:48 Speaker 1: And then the other surface parameter is dielectric, the dielectric constant, which is controlled by the amount of moisture. So radar backscatter is influenced by the amount of moisture in the vegetation and soil. the larger the water content in the vegetation or soil, the less the penetration. The magnitude of the radar backscatter is proportional to the dielectric constant of the surface. For dry, naturally occurring materials, this is in the range of 3 to 8. For liquid water, it's around 40 to 80.

57:25 Speaker 1: It is frequency dependent. And therefore, the amount of moisture in the soil or vegetation can greatly influence an increase in radar reflectivity. So generally, reflectivity or image brightness increases with increased moisture content, which means that there is less penetration through the medium. For example, surfaces such as soil and vegetation cover will appear brighter when they are wet than when they are dry. The lower the dielectric constant, the more incident energy is absorbed and the darker the object will appear on the image.

58:03 Speaker 1: And this is an example of the differences in image brightness for different landscape dielectric values. The dielectric of frozen water is very low on the order of 3.15 compared to liquid water, which is around 80. In the northern high latitudes when the land surface transitions from frozen to thawed conditions in the spring, there is a large change in the dielectric. the surface dielectric, and consequently a change in signal intensity from dark to bright. This example shows that. It is a multi-temporal time series of an area near Fairbanks, Alaska, from the JRS-1 satellite, which was a Japanese satellite that carried an L-band SAR and flew in the 1990s.

58:50 Speaker 1: You can see that The February image, when the land surface was frozen, is darker than the June image, when the land surface was thawed. And therefore, probably there was a certain amount of moisture in the surface from the frost. spring melt. So in freeze-thaw studies utilizing radar signals, the dielectric is therefore the controlling mechanism causing changes in backscatter behavior.

59:26 Speaker 1: There are a number of geometric and radiometric distortions that one has to account for in radar images, and in this next section, I will discuss these distortions. The first one is the side-looking nature of imaging radars results in certain geometric distortions, and one of them is the slant range. which is the distance of the signal between the sensor and the target, not the true horizontal distance along the Earth's surface. And this results in a varying image scale moving from the near to the far range and causing targets in the near range to appear compressed relative to the far range.

1:00:12 Speaker 1: Using trigonometry, ground range distance can be calculated from the slant range distance. and platform altitude to convert to the proper ground range format. This figure shows a radar image in slant range at the top, where the fields and the road in the near range on the left side of the image are compressed, and the same image is then converted to ground range in the bottom with the features in their proper geometric shape. For most SAR datasets, this correction has already been done. Similar to the distortions encountered when using cameras and scanners, radar images are also subject to geometric distortions due to relief displacement.

1:01:02 Speaker 1: Radar foreshortening and layover are two consequences of relief displacement. Layover on the left occurs when the radar beam reaches the top of a tall feature, in this case B, of a mountain, before it reaches the base of the mountain, A. the return signal from the top of the mountain will be received before the signal from the bottom. And as a result, the top of the mountain will be displaced towards the radar from its true position on the ground, and it lays over the base of the feature. So it looks like it's upside down, basically.

1:01:44 Speaker 1: And then you have foreshortening on the right. which occurs when the radar beam reaches the base of a tall feature that's tilted towards the radar, so a mountain, before it reaches the top. Now, because the radar measures distance in the slant range, that slope from A to B will appear compressed, and the length of the slope will be represented incorrectly. Depending on the angle of the mountain slope in relation to the incidence angle of the radar beam, the severity of foreshortening will vary. And maximum foreshortening will occur when the radar beam is perpendicular to the slope such that the slope, the base, and the top are imaged simultaneously.

1:02:31 Speaker 1: This is an example of an image over an area of complex topography where there is foreshortening. And notice the bright slopes on the middle and bottom of the image. They appear compressed compared to how they look after the image is corrected. Correction of the image is done using a digital elevation model. And many times SAR data that's available is already corrected for this. So radiometrically terrain corrected SAR data has been corrected for these sorts of things. And then the other issue to account for in areas of complex topography is radar shadow.

1:03:17 Speaker 1: which occurs when the radar beam is not able to illuminate the ground surface because you have big mountains that are in front of the signal, right? So shadows occur behind that vertical feature, behind the mountain or slopes with steep sides. And since the radar beam does not illuminate the surface, the shadow regions will appear dark on an image because there's no energy available to be backscattered.

1:03:48 Speaker 1: So this image shows radar shadow effects on the right side of the hill sites, which are being illuminated from the left. And there are ways to correct for shadows through interpolation. However, I personally prefer to leave the shadow areas as areas of data gaps rather than use interpolated values.

1:04:12 Speaker 1: Now let's talk about radiometric distortions. There's several radiometric distortions, including those related to terrain relief. And these often cannot be completely corrected. Radiometric correction involves removing the misleading influence of topography on backscatter values using a DEM. For example, the correction eliminates bright backscatter caused by radar reflection from steep slopes, leaving only the backscatter that reveals surface characteristics such as vegetation and soil moisture.

1:04:45 Speaker 1: Radiometrically terrain-corrected SAR images or RTC images, for example, have had radiometric and geometric corrections already applied. And the last topic I wanted to touch on was speckle. So let's talk about speckle. Speckle appears as a grainy, salt and pepper texture in an image. It is not noise. It is characteristic of SAR images. And this is caused by multiple scattering returns that occur within each resolution cell. So as an example, in this figure, we have a homogeneous target, such as a large grass-covered field.

1:05:28 Speaker 1: Without the effects of speckle, this field would generally result in light-toned pixel values, such as what you see in image A. However, reflections from the individual blades of grass within each resolution cell, so those, remember those individual blades of grass are in different orientations and positions. And so those reflections within each resolution cell are going to be different from cell to cell. And that is going to cause then some pixels being brighter and some being darker, such as what you see in image B. And that field is going to appear Speckled may make interpretation of radar images more difficult.

1:06:17 Speaker 1: So it's generally desirable to reduce speckle prior to interpretation and analysis. And speckle reduction can be achieved in three ways. Multi-look processing, spatial filtering, and temporal filtering. So multi-look processing refers to the division of the radar beam in this case, into several narrower subbeams, so one to five. The number of statistically independent images or subbeams being averaged is called the number of looks. And each of these subbeams will also be subject to speckle, but by averaging them together to form the final output image, the amount of speckle will be reduced at the expense of spatial resolution.

1:07:05 Speaker 1: So spatial filtering consists of moving a small window of a few pixels in dimension, say it's a window of three by three pixels or five by five, whatever you define it to be. And you move that window over each pixel in the image, applying a mathematical calculation using the pixel values within that window. There are many different mathematical calculations, many different speckle filters, but say here we're just calculating the average of the pixels within that window. And then we replace the center pixel with that new value, that new average.

1:07:40 Speaker 1: The window is then moved along in both the row and column dimensions one pixel at a time until the entire image has been covered. And by calculating the average of a small window around each pixel, a smoothing effect is achieved and the visual appearance of the speckle is reduced, as you can see in the two images in this example. Now, temporal filtering is similar, whereas instead of using a spatial filter, you average pixel within an image stack. The problem with temporal filtering is that if you're trying to detect change, it will likely be averaged out.

1:08:21 Speaker 1: The good thing about temporal filtering is that the spatial resolution is not degraded. So for multi-looking and for spatial filtering, you reduce speckle at the expense of spatial resolution. So the amount of speckle reduction desired must be balanced then with the particular application that you're wanting to do, right? So how much, what's the level of detail required? If a high level of detail is required, you need high spatial resolution. then you apply a lower number of looks or a small spatial filter.

1:09:08 Speaker 1: If you don't really need such a very fine level of detail, then you can apply a larger filter. number of looks or a larger spatial filter, which will reduce the speckle further, but it will also reduce your spatial resolution. So there's a balance there that you need to explore and define what is the best trade-off for your specific needs. Next, I'd like to touch on some additional considerations when working with radar data. So one of them is the look direction, the role of the look direction. And for those of you that have worked with radar data, you notice that there's ascending and descending passes.

1:10:00 Speaker 1: So first of all, the look direction of a SAR refers to the direction the radar antenna is pointed when emitting and receiving the radar beam. And a SAR look direction is determined with respect to the flight direction of the sensor. So typically, SAR sensors are configured to look either left or right and how an area is illuminated by a radar beam changes during ascending and descending overpasses of an area. So basically, the viewing geometry of that area is different. And this figure is an image subset from Sentinel-1 over Romania.

1:10:38 Speaker 1: And you can see the differences between ascending and descending, especially in the area's where you have mountains and complex topography. So it is advisable that if you're doing time series analysis to not combine data from ascending and descending passes. The other factor to consider is the role of moisture. So at the beginning, I said that the advantage of radar is that it has an almost all weather capability. Note the stress on the almost because it's not truly all weather. Sometimes when you have big rain events, especially the type of events that you see in the tropics, that will affect your signal.

1:11:29 Speaker 1: So here we have an example of the effect of an increase in surface moisture on the radar signal. This is a Sentinel-1 seaband image over an area in Ecuador for three different years. And 2017, which is the image in the middle, looks different than the rest of the images. It has some bright and dark anomalies. The dark areas are associated with active rain causing very strong signal attenuation. And the bright areas are the result from wet vegetation and soils from the rain events. So this is why it's always important to look at your data to identify if there are any sort of artifacts like this that would make your image non-usable for any sort of analysis.

1:12:21 Speaker 1: And now let's take a look at some applications of SAR data. This is an example of the use of radar to detect oil spills on water. It's an image from the UAV SAR sensor, which is an airborne SAR that operates at L-band. And this is the Gulf Coast of the United States. This image was acquired shortly after the Deepwater Horizon oil spill to determine the extent of the spill and the movement of the spill into the coastal areas. So this is a false color image of HH, HB, and BV. And the presence of oil smoothens out the sea surface and makes for an especially good specular reflector, much smoother than the ocean surface where there is no oil.

1:13:10 Speaker 1: And here we can clearly see the oil extent as areas that are very dark in contrast to the ocean water, which appears as blue. So ocean water has a little bit of roughness while the areas covered by oil are very, very smooth. This is an example of the use of radar for land cover classification. It's using SAR data from the Japanese JRS-1 satellite, which had an L-band SAR and operated in the 90s. The top image, the top left is a radar image of an area near Manaus in Brazil where there's a combination of forested and deforested land.

1:13:48 Speaker 1: The deforested areas have a darker tone than the forested areas. The image on the top right is the resulting classification delineating forest regrowth and non-forest. So remember, radar is sensitive to structure and moisture. And in this case, you can clearly identify even just visually the differences in tone due to structure, due to the vegetation density. Now, the bottom image is a part of a savanna region in the Empton area. which is characterized by short woody vegetation as well as non-woody vegetation.

1:14:28 Speaker 1: And the classification on the bottom right shows that through machine learning, a supervised classification approach, these two different types of savannas can be differentiated. SAR is very useful for wetland inundation studies because, as mentioned, ability to penetrate through the vegetation canopy and detect standing water underneath. So there's a very high backscatter return whenever there's inundated vegetation. Multi-temporal SAR data can be used to study wetland inundation dynamics, such as the study on the right, which used PALSAR data in the Amazon to look at inundation extent and duration which helps better understand seasonal and year-to-year hydrological changes in these ecosystems.

1:15:21 Speaker 1: And then there's also soil moisture monitoring. Since the radar signal is sensitive to the moisture content in the surface, it can detect the moisture in the soil. And this is an example of a global soil moisture product from the SMAP radar. The last example I wanted to show is crop classification. This is work by Dr. Heather McNairn from Agriculture and Agri-Food Canada, where she used Sentinel-1 images, multi-temporal Sentinel-1 SAR images to map different types of crops. So these examples that I've shown are not exhaustive at all.

1:16:05 Speaker 1: There are many other things that can be done with amplitude data, such as estimating biomass, assessing or detecting forest degradation, looking at open water bodies and how they're changing through time. So there are many different things that radar can support in terms of ecosystem type studies.

1:16:31 Speaker 1: And to close, I wanted to point you to some sources of openly available SAR data. The first source is the Alaska Satellite Facility, and they have a number of different SAR datasets, including data that has been acquired at C-band, at L-band, and data that goes back a couple of decades. So it's an excellent resource in addition to the data. They also have many, many different resources. tools and trainings that are extremely valuable. The third session of this webinar series will discuss all of the capabilities of the Alaska Satellite Facility.

1:17:21 Speaker 1: There is also the Copernicus Hub, and they are the repository for all of the Sentinel data, including Sentinel-1 data. And I noticed that they also have in their archive Sentinel-1 mosaics. So these are monthly mosaics, I believe. But it's just another way to access the data. And they also have a number of resources and tools that can be very valuable for working with the data. JAXA, the Japanese Space Agency, has a number of data sets that you can access also for free through their portal. They have global Pausar mosaics.

1:18:09 Speaker 1: These are yearly. They have some derived products on forest and non-forest and other different types of data sets on their website. For all of these sites that I've mentioned, you do need to register and have an account. However, there is no cost to do this and the data access is free. And finally, there is Google Earth Engine, which has a repository of different data sets, including Sentinel-1 and PALSAR, the global yearly PALSAR mosaics, as well as some PALSAR scansar data. And so the data on Google Earth Engine are analysis ready.

1:19:01 Speaker 1: So all you need to do is really just apply a speckle filter. And here we have a list of legacy current and future radar sensors in space. The ones that have a green box around them means that their data is freely accessible. In terms of future missions, there are two very exciting missions coming up. One is called NISAR, which is a collaboration between NASA and the Indian Space Agency. You'll have an L-band SAR as well as an S-band SAR. And it's scheduled to launch in 2025. And then there's Biomass, which is a P band SAR sensor from the European Space Agency.

1:19:49 Speaker 1: And that's scheduled also to launch in 2025. So next year is an exciting year in terms of radar.

1:20:00 Speaker 1: I wanted to briefly touch on NISAR because I'm part of the NISAR science team. And so are two of the guest lecturers in this webinar series. Dr. Eric Fielding in session two will be talking about InSAR and Dr. Franz Meyer in session three will be talking about the capabilities of the Alaska Satellite Facility. So NISAR is launching 2025 LNS band SARs. All the data will be freely available. It'll have amplitude and phase. And the data will address many science and application needs. And that brings us to the end of this session.

1:20:45 Speaker 1: Next, I will do a brief summary of the main concepts covered. In summary, in SAR, the long track resolution is different from the across track resolution. The three radar parameters, that you need to always keep in mind are wavelength, polarization, and incidence angle. So the longer the wavelength, the greater the penetration depth. The length of the wave will determine the interaction with the surface objects. Polarization provides information related to the structural characteristics or properties of the objects in the surface.

1:21:24 Speaker 1: Incidence angle will influence the signal penetration into the target. The two surface parameters that influence the radar signal are structure and moisture, and that's super important. So radar is sensitive to structure and to moisture. The main backscatter mechanisms in radar are specular scattering, rough surface scattering, volume scattering, and double bounce. Radar images have geometric distortions in areas of complex topography, and these can be corrected up to a certain extent using a digital elevation model.

1:22:03 Speaker 1: Speckle is that graininess inherent in SAR images, and it can be reduced through multi-looking or with a spatial or temporal filter. And then finally, radar can be used for different ecosystem studies such as mapping land cover, mapping crops, wetland inundation, and soil moisture, to name a few.

1:22:25 Speaker 1: The next session, number two, will focus on interferometric SAR. The guest lecturer, Dr. Fielding, will discuss the basic concepts of interferometric SAR or INSAR. He'll explain how to generate a SAR interferogram and interpret it in order to measure surface deformation. And he'll also discuss some of the applications that INSAR can address. And as a reminder, there is a homework associated with this webinar series. We will post the homework on the training webpage on the last day of this training, which is November 20th.

1:23:07 Speaker 1: And the homework is due by December 4th. A certificate of completion will be given to all those participants that attended the three live webinars. and completed the homework assignment by the deadline.

1:23:28 Speaker 1: If you have any questions about the material presented today, please don't hesitate to reach out to me via email.

1:23:38 Speaker 1: And with that, thank you very much to all participants. I know this was a lot of material, but hopefully it provided a good solid background to better understanding the information content in radar images. For those that want to dive deeper, I've included an appendix as part of this PowerPoint presentation, and there are many different references there. So with that, then let's begin our question and answer session.

1:24:15 Speaker 1: So thank you very much for everyone who's online for all of the questions. I'm so pleased to see so much participation and so many people from around the world. As you know, I'm passionate about SAR. I've been working with SAR for many years. And I'm super excited about, you know, this next year with the launch of NISAR and other satellites like biomass. So there's so much capability with SAR. that is either unique or complementary to optical. So I'm so happy and pleased to see such great interest.

1:24:52 Speaker 1: So let's start with the first question. And we have little time. I will try to go through as many questions as possible. I will stay over time. But however, any questions that I can't get to, I will answer on the Google Doc and I will put it online. All right, so I'll try to kind of breeze through these. Question number one, can we use electromagnetic waves to carry the particle from the surface to the satellite for measurements or analysis of the soils? The answer is that radar is able to measure the moisture in the soil.

1:25:26 Speaker 1: So radar is sensitive to humidity. And for example, NYSAR will have a global soil moisture product. However, if I'm reading this question correctly, you know, in terms of actually bringing a sample, that's unfortunately not possible. Question number two, when you say it can penetrate vegetation soil, can you say anything about depth of penetration and what parameters affect penetration? So this was a question early on before I... started touching on in more depth on some of these concepts. So the depth of signal penetration will depend on the wavelength.

1:26:08 Speaker 1: The longer the wavelength, the greater the penetration through the medium. That means a volume of vegetation, snowpack, or soil. Also, the amount of water in the medium will reduce penetration of the signal. So I showed that example of the signal at L-band penetrating through very dry soils up to two meters. However, if those soils were wet, you certainly would not be penetrating that depth. You would probably be penetrating the first couple of centimeters, five centimeters, eight centimeters, depending on the wetness.

1:26:41 Speaker 1: Question three. You've mentioned that the effect of topography is one of the major drawbacks of radar. I'm thinking of mapping wetland classes of the Hindu Kush region using Sentinel-1 imagery. Due to the topographic effects in the Hindu Kush Himalayas, Do you think it's better for me to use Sentinel-1, or maybe I can reduce, minimize the topographic effect with some image preprocessing procedures? So the answer is that even though topographic areas do present challenges with radar images, you can do time series as long as the images are radiometrically terrain corrected and you are using the images from the same path.

1:27:22 Speaker 1: So that means you either use images from ascending or descending. orbits, but do not mix them. In general, as a general rule of thumb, at least for myself, I try not to mix ascending and descending images. This is not such a big issue in areas that are flat, but in areas, obviously, where you have topography, it is a big issue because the viewing geometry is slightly different, so things do look different. Question number four, does the radar rotate in the cross-track direction to build a line of the image?

1:28:00 Speaker 1: And that's a great question. Some radars rotate like SMAP, but most do not. Please explain, question five, please explain if there's a difference between resolution of an optical and resolution of a SAR image, and how do they differ? Okay, great question. So optical, yeah, we're all used to these square pixels, 10 by 10s. With radar in the slant range, you're not going to have square. It's going to be like five by eight or something, something that's different. Now, when you reproject from the slant range to the ground range, then you grid things to a square pixel.

1:28:34 Speaker 1: So that's why like Sentinel-1 images in Google Earth Engine, for example, or on the Alaska Satellite Facility, the ones that are GRD, so that they're gridded already in the ground range, they have a square pixels. Now, one thing that's super important to remember is that with radar, do not take the spatial resolution at face value. So say Even on Google Earth Engine, Sentinel-1, yeah, that resolution of Sentinel-1 is 10 meters. Yes, it's 10 meters. Unlike optical data that you can straight out use a 10-meter or whatever 20-meter spatial resolution data and generate stuff with it, analysis, classifications, with radar, the 10 meters, remember, you need to apply a spatial filter, usually it's a spatial filter, to reduce the speckle.

1:29:28 Speaker 1: And what that does is it degrades the spatial resolution. So things might be still gridded in the same way, but your spatial resolution is going to reduce. So yeah, with radar, there's the posting of the spatial resolution. But in actuality, after applying a spatial filter, that resolution is going to degrade. So whatever you're generating is likely going to be at a lower spatial resolution. question six is there any possibility to share literature in the course web page where this information is taken from.

1:30:05 Speaker 1: And there is a list of tutorials and references in the appendix of this presentation and, and so I have I put a whole list of previous are set trainings that include demos for doing different types of things. Floods, looking at floods, earthquakes, landslides. I think there's also oil spill detection. And there are also other resources that I posted in the appendix. The Severe Handbook is a super great resource. I have a link to it further down. But there are many other great resources. If you're looking for something a little more challenging, just send me an email and I will Let me know specifically what you're looking for and I will give you some appropriate links.

1:30:52 Speaker 1: Question seven, could you tell us comprehensively how the Doppler effect is applied in SAR? So basically, Doppler refers to the change in frequency or wavelength due to the relative motion between the source and the object on the surface. So as the radar moves along its flight path and it's sending pulses, The pulses are reflected back to the antenna from the Earth's surface, but because the relative velocity between the radar and the ground surface causes a shift in the frequency of the return signal.

1:31:27 Speaker 1: And that frequency of the received signal will be slightly different from the transmitted frequency, depending on the velocity of the movement of the radar with respect to the object. Think of it as if you've ever heard an ambulance coming toward you. It tends to be higher pitched. And then when it moves away from you, it's lower pitched. That's Doppler right there. And I've put in a great video explaining that concept right here. Question eight. If I want to study the technicalities of mathematics involved in SAR data in greater depth, are there any resources that the panelists can recommend?

1:32:06 Speaker 1: Again, I put a series of resources at the very end. And there is a book that is kind of the the guideline, the basic book of microwave remote sensing. And that book is by Ulabi and Long. It's really a superb resource. I think there are, initially there were three volumes. I know it's been republished. So it's just a superb resource to have. Question nine, how can SAR systems be designed to optimize either slant range or ground range resolution for specific applications? So they are designed to optimize slant range, and that will depend totally on the need for your application, right?

1:32:53 Speaker 1: So do you need greater spatial resolution or higher or lower spatial resolution in the slant range, depending on your application? How do I identify which is backscatter? So that one, I didn't quite understand. If that person can please rewrite their question to clarify. Question 11, I'm interested in mangrove mapping using SAR imagery with the help of cloud computing. Can you guide me on how I can map mangroves and non-mangrove areas? This is a great question. There was an RSAT webinar and one of the sessions focused on monitoring mangroves.

1:33:32 Speaker 1: There's also the Severe SAR Handbook that has a chapter on mangrove mapping. I have a colleague at JPL, Mark Simard. who has done a lot of work on mangrove mapping using radar, and he's got a body of literature. I'll include that here. It's very difficult to use SAR to map mangroves because mangroves will look different when it's high tide and low tide. And also, because of the structure of mangroves, those aerial roots, they will cause problems. this signal to do some really weird things that are different from what you see in forests.

1:34:10 Speaker 1: So sometimes those aerial roots will cause, the signal will not penetrate all the way through or cause attenuation. So it's a bit challenging to use just SAR for mapping mangroves. Question 12, the double bound signal looks like specular reflection plus volume scattering. Why does it have higher intensity than the signal from volume scattering directly? And I think I know what this person is asking, and perhaps that drawing on volume scattering, that cartoon, can be modified to show that with volume scattering, I mean, you've got the signal penetrating within the volume, and you've got part of that signal being...

1:34:52 Speaker 1: reflected away from the radar, but you got a large portion of the signal also being reflected back toward the satellite. However, with double bounce, you have an even larger portion of the signal being reflected back to the satellite. It's basically two bounces with double bounce. So you've got the bounce from the smooth surface, which would be water, or you've got a street, for example. So it reflects it away from the radar. And then It bounces off of something geometric, say a tree trunk or a building, which redirects all of that energy that was being reflected away, redirects it back toward the satellite.

1:35:34 Speaker 1: And that's why double bounce is typically very, very bright. And you see double bounce in flooded vegetation and in urban areas. Question 13, is SAR imagery multispectral or hyperspectral based on wavelengths? Yeah, so in SAR, we don't use the term multi- or hyperspectral as you do in optical. So when you have multiple wavelengths or frequency, you call it multi-frequency SAR. There is no currently operational and openly available SAR data that is multi-frequency data. all of the openly available operational SAR data sets are single frequency.

1:36:17 Speaker 1: However, NYSAR will have multi-frequency. It will have L and S bands. Now, one thing to emphasize or to note is that SAR will be collecting L band, will be doing L band acquisitions globally and S band acquisitions only over certain areas. So that would be S band over India and S band over the CalVal site that we have around the world, which are just small areas where we're doing calibration validation of the signal. But S band collections will not be global, just the L band. Question 14. And by the way, all of the NISAR data will be openly available.

1:37:03 Speaker 1: Question 14. Based on atmospheric effects, low... I'm assuming that SAR is not frequently used in atmospheric composition studies, but I'm wondering what cases SAR data may be valuable for atmospheric studies. In your opinion, what SAR data may be of most valuable for atmospheric and air quality studies? I'm guessing land cover, land use in cloudy areas, but I'm not certain. So SAR will not really provide information about atmospheric composition or air quality. However, it is used and it's widely used for like weather prediction, right?

1:37:42 Speaker 1: So to look at atmospheric conditions related to precipitation, storms, rainfall intensity, tracking storms, tropical cyclones, hurricanes, typhoons, et cetera. And usually it's the shorter wavelengths because the shorter wavelengths are either getting attenuated by the amount of water in these rain cells or they're interacting with the particles. the rain droplets because the wavelengths can be on the order of the size of the rain droplets. So that's why atmospheric studies using radar use the shorter wavelengths and even shorter than these, KUKA, sometimes X-bands, and sometimes even shorter.

1:38:29 Speaker 1: Do we have free data? for L and P bands? Yes, we do. There is freely available data for L bands. There is a JRS1. So I put it down here, JRS1 which was a Japanese satellite that flew in the 1990s. You can obtain that data through the Alaska Satellite Facility. There is also PALSAR data from, again, the Japanese Space Agency. That's LBAND 2006 through 2011. Amazing data sets. And then there is PALSAR-2, which has limited release. So the PALSAR scan SAR data is available through Google Earth Engine and through the JAXA website.

1:39:17 Speaker 1: And in terms of PBAND, there is no satellite PBAND data available that I'm aware of. There might be some airborne, freely available airborne PBAND data over select regions. Question 16, in signal penetration into dry soils, can radar detect an improvised explosive device that has been put under the soil? This is a great question. I think it would be difficult. I think it would depend on the size of the IED relative to the spatial resolution of the radar and how different that IED is the characteristics, the structural characteristics of the IED.

1:40:08 Speaker 1: So I'm honestly, I don't know, but I think you would need very high spatial resolution. Can anyone tell me about the Early Adopters Program? So NYSAR has an Early Adopters Program, and I put the link here. You can join this program. So the idea behind the Early Adopters is to have the community start the potential, in this case, NISAR applications community indicate how they can use NISAR data and the NISAR program will provide some guidelines on how that data, what are the capabilities of the data and how the data can be used.

1:40:58 Speaker 1: So it's just a community of Again, the name says it's early adopters of NISAR data. Question 18, which satellites offer free L-band images? So I just answered that one. Question 19, you mentioned that in desert environments, the penetration depth can reach up to two meters. Given the water scarcity in our region, I'm curious about how pixels representing desert areas with underlying water would appear in SAR images. So this would totally depend on how deep is the water. You know, if there's some moisture that is not too deep, especially in desert areas, so if you have like a very dry upper layer and then wetness, again, say it's two meters, two meters and a half, three meters even, you might be able to sense a difference,

1:41:56 Speaker 1: but it would totally, it would depend on that. Question 20. It's not always true that higher wavelengths SAR images are better than lower wavelength counterparts. Is this true? Can you explain using examples? Because I remember coming across a point that for some applications, C-band is better than L or P-band. Is this correct? Is this correct? Yes, this is absolutely correct. The wavelength will totally depend on your application. So if you want to do agricultural studies, you want to use either C-band or X-bands.

1:42:34 Speaker 1: For forest-type studies, then you want to use L- or P-bands, for example. So totally dependent on your application. You don't want to use P-band, for example, on agricultural-type studies. That's not the right wavelength, or that's not the right band. Okay, question 21. Is it possible to penetrate concrete to detect water leaks In a pipeline network, can SAR be used to detect leakages and drainage? This is a great question. I think it's unlikely. However, if the concrete has some level of moisture that is different from the concrete surrounding it, then you might be able to detect it.

1:43:19 Speaker 1: So it could be that the moisture does seep up through the concrete. And so the dielectric or the moisture content in the area where there's a leak underneath might be slightly different than in the areas where it's dry. And you might be able to detect that. 22, in what instances would it be useful to explore polarization differences or ratios apart from just the bands? Actually, in all sorts of studies related to structural differences of your objects on the surface. So if you're looking at like differentiating agricultural crops, different types of vegetation, you do then want to use different polarization ratios or polarization differences.

1:44:14 Speaker 1: Question 23, which radar properties, like frequencies, bands, are used for soil moisture estimations? Can SAR be used to estimate soil moisture content through all the phenological stages of crop growth over vegetated areas to retrieve soil moisture content, to be precise? OK, great question. SAR can be used and is being used to estimate soil moisture.

1:44:45 Speaker 2: It works well in areas where you.

1:44:58 Speaker 1: Don't have dense vegetation. So, for example, SMAP, the SMAP radar, and SMAP in general, it works well in areas with low vegetation, like in areas... in agricultural areas, and it's estimated, at least for SMAP, that the soil moisture retrievals were within 4% volumetric, had a 4% volumetric accuracy in areas with a full-grown corn crop. So anything above that level of vegetation density, then your estimates of soil moisture would be lower or poorer. The reason for that is because with radar, it's not so straightforward to get to soil moisture.

1:45:50 Speaker 1: If it's a desert, yes, because it's a desert and you're observing, or if it's a barren area, because there's nothing else there, it's just the soil. But as soon as you have vegetation, then remember, remember that signal is interacting with different components, right? So basically... Part of the return signal has information about the moisture in the soil, but it also has information about whatever else is there, whatever your crop is or the vegetation. And so you need to tease out the part of the signal that belongs to the vegetation to get to the soil moisture.

1:46:38 Speaker 1: So yes, you can, and there are ways to estimate soil moisture in vegetated areas. And the more vegetation you have, the more complicated it gets and the lower the accuracy of your retrieval. Because once you have like dense vegetation, first of all, you're not even like in tropical areas. It's uncertain whether the signal is penetrating all the way through the vegetation canopy and actually detecting the moisture in the soil and also it's uncertain whether the yeah whether whether sorry whether it's detecting the moisture in the soil all the way through the canopy and and such a large part of the signal and vegetated area densely vegetated areas is part of the vegetation and a very small portion of it is a function of the moisture in the soil.

1:47:38 Speaker 1: Question 24, for ground motion, we use buildings and stable surfaces because these correlate less. However, as you explained, these surfaces are sending back less signal. Could you explain a little about what part of the signal is coming back?

1:48:05 Speaker 1: I think this is more of a question on INSAR. You know, when you're talking about decorrelation, yeah, this is more of looking at the phase and the de-correlation of the phase. And this is something that we will talk about in the next session, session number two. Question 25, to what extent can we use SAR to differentiate between open water and ice? In the case of less calm seas, will there be too much backscatter? There are many studies that are using SAR to identify ice. And the way you can do that is it's all about roughness, right?

1:48:50 Speaker 1: So I talked about structure, which refers to roughness. The ocean surface will be much smoother than an ice, than sea ice. And so the backscatter is going to be different between the ocean surface and the sea ice, and you can differentiate them that way. Question 26, what are the open sources where we can get the global coverage of each SAR band data? Yeah, so I touched on some at the end. And session three of this webinar series, We'll discuss all of the different SAR data sources available at the Alaska Satellite Facility.

1:49:42 Speaker 1: But I can tell you, yes, so you've got the JAXA site that has the PAL-SAR data, some PAL-SAR data available. There's Google Earth Engine, which has the Sentinel-1 analysis-ready SAR data. It's got some PAL-SAR scan SAR data, some PAL-SAR global mosaics. yearly global mosaics. There's a Copernicus hub by the European Space Agency. They have all the Sentinel-1 data. And then the Alaska Satellite Facility, they've got a bunch of different currently operational data and historical SAR data. Okay. How do you control polarization?

1:50:25 Speaker 1: Is it controlled by the antenna design? So, It's controlled by the system design, and so you have a filter where you emit the signal in a given polarization, and then whatever comes back, you're just receiving, so you filter just to receive in a given polarization. Okay. Question 28, which band is most preferable for flood monitoring, C-band or L-band? Also, what polarization is best to work with, VV, HH, or VV plus HH? Do the RGB bands always correspond to the waves for HHHV and VV, or will it be very, wow, this is a multi-tiered question.

1:51:19 Speaker 1: All right. The band that's more preferable for flood monitoring would be L-band, especially if you're looking at forested inundation because with L-band, you'll have a greater probability of penetrating through and you'll have a greater likelihood of seeing inundated vegetation. However, if it's crops, like for example, inundated rice fields, C-band is perfectly fine. HH is a better polarization because as mentioned, there's a higher probability that HH will penetrate all the way through than VV- or much less the cross polarized.

1:52:01 Speaker 1: Do the RGB bands always correspond to the waves? No, you can create an RGB band however you want. It's always better to put HV or VH in the green channel because the cross polarized will always be a better indicator of the presence of vegetation. So visually, for visual interpretation, HV, putting it in the green channel, whatever is vegetated will appear green. 20 questions. Let me see. Why is VH polarization not measured? It totally depends on the satellite. Some satellites measure VH, like Sentinel-1 has VV and VH.

1:52:56 Speaker 1: Does it share information with the HHVV or HV? So usually the cross-poll, like VH or HV, that's usually the same. It's the same information content or very similar. But each of your polarization channels has different information content. And that's why the more polarizations you have, if you have HH and HV, the better, because you have more information about the unique structural characteristics of your target of the surface. Question 29, is there any software open source hopefully you recommend to work with SAR data, mostly for visualization?

1:53:40 Speaker 1: I use software to estimate ground deformation, but I'm just exploring the amplitude characteristics of SAR. So the European Space Agency has a great open source software called SNAP, the SNAP Toolbox. And you might want to try that. Google Earth Engine, there you need to do a little bit of coding, but it's a great way to mine through entire time series data sets and do visualizations. Question 30, are there any studies applying SAR for coastal flooding? Could you provide some papers on this topic?

1:54:24 Speaker 1: Yes, there are some studies, and I'd be happy to provide some references when we post this document. Question 31. How large should the area be so that we need to be careful about incidence angle differentiation in the image? Also, can we correct for it somehow? Oh, this is a great question. Okay. Okay. So the incidence angle variation will vary. largely depend on the acquisition mode, right? So for example, the PALSAR scan SAR, they have a large incidence angle variation. And so there are ways to normalize for incidence angle.

1:55:11 Speaker 1: However, I prefer to just honestly just cut the edges and just work with kind of the central parts of the image and not deal with the variations in backscatter because of the large incidence angle and just work with the minimal variations that you'll find just within the middle of the image. Question 32. Can you explain for which scenarios which SAR polarizations are ideal? Like in optical remote sensing, we know which spectral bands are to be used to identify which surface characteristics based on spectral reflectance curves.

1:55:54 Speaker 1: In SAR remote sensing, how do we determine which polarization we need for a specific application? For example, identifying floodwaters or landslide debris. So I think part of this is exploring the different polarizations. And in the appendix, I have some links to previous RSET trainings that are looking at some of these things that you've identified here. But I can tell you right now, if you're looking at inundated vegetation, you want to use HH or VV because there's a higher probability that you'll detect inundated vegetation.

1:56:38 Speaker 1: If you're looking at open water, so that means like water surfaces, no standing vegetation, the better polarization is a cross-polarized. HV or VH. If you're looking at landslide debris, you'll have to see. Maybe a polarization ratio might be good for that, but that's something that you would have to explore. Question 33. Are there new good methods to reduce speckle using neural networks? Oh, that's a great question. I can tell you there are many different speckle filters and everyone has their favorites.

1:57:25 Speaker 1: I think it's always good for you to explore different speckle filters and see which one works best for you. I personally just like the simplest speckle filter, which is just a simple average because I don't like to change or alter the I like to alter as little as possible the original values of my image. Now, are there methods to reduce speckle using neural networks? I don't know. I can look into it, and if there are any literature references, I'll be sure to post them. Question 34. How can SAR be...

1:58:08 Speaker 1: used in glacier area mapping? Which polarization is best to choose? And how do backscatter values change depending on elevation and season? Aha. This is a great question. So NYSAR actually has a cryosphere component. And they are using INSAR. So interferometric SAR to look at glacier movement, so like velocity vectors of glaciers. And Dr. Eric Fielding will touch on that application tomorrow. Now, which polarization is best to use? I'm not sure because I don't do any sort of glacial studies, but I can...

1:58:57 Speaker 1: refine this answer and update it. How do backscatter values change depending on elevation and season? So backscatter will change depending on season, and it depends on what you're looking at. So if you're looking at areas where there's a phenology, forest phenology, where the trees have leaves and then they lose their leaves, your backscatter is going to be different. I also showed the example in the northern high latitudes where you have a seasonality of the land surface going from frozen to thawed.

1:59:33 Speaker 1: And you are going to have a difference in backscatter due to that state change from frozen conditions to thawed conditions. So there's going to be a higher backscatter as things go from frozen to thawed. So how do backscatter values change depending on elevation? If you're talking about the dynamic nature of elevation, like snow melt, you will see differences with the dynamic nature of things changing, right? So snow melt, for example, in a time series, you'll certainly see that. Can you please elaborate about geological applications of SAR and how to process data sets in order to detect geology, lithology of any particular area?

2:00:26 Speaker 1: So remember, radar is not sensitive to the spectral properties of the surface. However, if you're looking at faults and if you're looking at maybe different types of terrain based on roughness, you can map that with radar and I'll provide some links here. What is the capability of GE for SAR data analysis? You can do a lot with Google Earth Engine for SAR data analysis. As mentioned, there's several SAR data sets in there and you can not only visualize them, you can do time series analysis. You do have to do your own coding, and the links I shared with you in the appendix of this presentation, there are some demos in there, and there are some demos that have been done with Google Earth Engine, and the code is there.

2:01:24 Speaker 1: So you can run that and take a look at the types of things that you can do on Google Earth Engine for flood mapping, for example. And then you can take that code and modify it for your application of interest.

2:01:42 Speaker 1: Where does phase information come in handy apart from INSAR application, for instance, for land cover classification? So phase information, right, I mean, it's INSAR, but say for land cover classification, there are nifty ways of looking at phase to look at, for example, vegetation height, and there was a RSAT training back in 2020 on using INSAR to look at decorrelation of the signal to look at vegetation height, I believe. I can post more here in the answer. Question 38, is the dual polarization mode like polarization or cross polarization always?

2:02:35 Speaker 1: It depends. Dual polarization mode usually is like is VV or VH or HH and HV, for example. Question 39. Are there SAR-based indices available for assessing vegetation health, hydric stress, and crop yields similar to those used with optical? Can one use SAR to detect the degradation of agriculture similar to NDVI with optical data? We have vegetation indices for optical imagery. Is there a similar thing in the case of SAR? Yeah, SAR has some vegetation indices, but they're not related to the health as you would see it with NDVI, which is more related to the spectral properties.

2:03:26 Speaker 1: It's more of the change in structural properties of your vegetation that might be related to the health of the vegetation. Let me see. Vegetation health, hydric stress, and crop yield. We've done many, I think three or four now, RSET trainings on the use of SAR for different agricultural applications. Some of them are looking at identifying different types of crops.

2:04:04 Speaker 1: Hydric stress, so that's a difficult one because crops tend to grow very quickly, so you've got a change in the signal that's related to the growth in the crop, and it could be also a change in the signal that's related to the vegetation water content, for example, or the soil moisture conditions. So I'll have to look into this further, and I can post some references. Question 40, does the dielectric signal depend on the type of sensor, or is it unique per element? Are the different corrections unique per satellite, per type of satellite image?

2:04:47 Speaker 1: For example, can I correct an image of radar site using the same techniques and tools used to correct images of Sentinel? So the influence of dielectric is dependent on the wavelength or the frequency. It's not dependent on the type of sensor. Are the different corrections unique per type of satellite images?

2:05:15 Speaker 1: They might be slightly different. I mean, overall, you need to correct the images for radiometric and geometric distortions. And the way you apply that correction or the information you need to apply that correction is definitely different. sensor dependent. Now, the good thing is that a lot of the data that's already available has been corrected already for radiometric and terrain and geometric distortions. Okay, the next question. Is the ground range correction calculated from a smooth earth with curvature?

2:05:59 Speaker 1: Are there more complex corrections that use an elevation model? Yes, absolutely. So there are Ideally, you want to do a correction using an elevation model. If you assume a smoother earth than anywhere where you have topography, you're definitely going to have issues. Question 42. I thought the wavelength of SAR bands is long enough to penetrate water, and that's why they're not impacted by clouds. Why do wet ground surfaces have higher backscattering? Does SAR somehow not penetrate the water at the ground surface?

2:06:35 Speaker 1: So the SAR wavelength does not penetrate water. The reason it penetrates clouds is because the wavelength is longer than the droplets or the particles that you find in clouds. So it's not interacting with them. It goes through. Why do wet ground surfaces have higher backscattering? So the reason that wet... that a soil that has a large amount of moisture in it has a higher backscatter is because there's more reflectivity. Think of it as a signal comes in, and there's a large level of moisture, so it's just going to reflect back rather than penetrate.

2:07:23 Speaker 1: And so that's why that backscatter is higher. Does SAR somehow not penetrate the water at the ground surface? No, it does not. Question 43, referring to slide 58, do the layover and foreshortening appear on all types of bands, like C-band, P-band? Yes, it does. Question 43, I'm using INSAR data for structural health monitoring. I use the ESA Sentinel-1 band. How could the data processing and interpretation be easier and faster? Ooh, that's a great question. That's for session two. Dr. Eric Fueling can answer that.

2:08:02 Speaker 1: And he's the expert on INSAR. Question 44, how is radiometric correction applied on coastal regions where elevation is very low? I think, so the radiometric correction is applied evenly over all regions. And usually, the radiometric correction tends to be, at least for topography, right? It's a big issue in areas where there's topography. In coastal regions, it's not such a big issue because coastal regions tend to be flat. Referring to slide 63, image A and image B represent the phase and intensity images in pixel form.

2:08:53 Speaker 1: I'll have to answer that. I'll have to go back and take a look at slide 63. Question 46, for temporal filtering, this would use two images. How are they correlated? Let's see. So you know that the pixel being filtered matches in both images. You have to stack the images. Basically, you have to create a pixel stack where that same pixel in different dates that, yeah, that same, you have the same pixel. Say you have 10 different dates. As you look down through the different dates, you know, for a given X and Y position, it's the same pixel.

2:09:33 Speaker 1: Question 47, is it necessary to always apply radiometric correction, speckle filtering before using a SAR image? So radiometric correction is different than speckle filtering. So yeah, usually you want your image to be geometrically and radiometrically terrain corrected. Now, if your image is not radiometrically corrected in, say, an area that's flat, I don't think it would be a huge deal. But is it always necessary to apply a speckle filter? It depends. You can try it, but you know what's going to happen.

2:10:16 Speaker 1: Say you run a classification, your classification is going to have a lot of defects of speckle. So it's going to have a lot of noise that you somehow need to clean up. So if you apply a speckle filter to your image and then you run a classification on it, your classification will be cleaner. So you can try and you can try using different sizes of speckle filters and see how that translates to the product that you're trying to generate. Question 48, referring to slide 65, which spatial filter is recommended for speckle reduction?

2:10:53 Speaker 1: Can you also tell by applying the speckle reduction filter? If we trade off information, these details present in the image, is this true? So as mentioned, there are many different speckle filters. I can put here some recommendations of the most commonly used speckle filters. But anytime you apply a spatial speckle filter, you will degrade the spatial resolution. Question 49, in which instances can you combine ascending and descending? If you want to create a composite image for each season, can I combine these?

2:11:30 Speaker 1: Do you need to correct the rain cells in an image before creating a composite image? If yes, how can you apply this correction? Okay, so first of all, You can combine ascending and descending. I personally do not. Like you can do what I would rather do is a separate analysis with the ascending image, like time series images and descending time series images, and then combine the results, for example. But you could combine them. And you can do that if your area is flat. So if you're looking at a wetland or a flat area, that should not be an issue.

2:12:13 Speaker 1: Now, if I want to create a composite image for each season, can I combine these? Yes, as long as, again, they are, as long as they, it's a flat area. do you need to correct the rain cells in each image before creating a composite? You cannot correct for the rain cells. So that example I showed, that image is not usable. You can't correct. I mean, you've got some very bright areas and you've got some very dark areas. You throw that image away. So if you use an image like that as part of your composite, your composite is going to look odd.

2:12:55 Speaker 1: So you want to throw that image away, and that's why it's important to look at the images in your stack. Question 51, is it possible to derive a high-resolution digital surface model out of SAR images? It is possible in SAR. So it is possible, yes, there are some ways to do this.

2:13:25 Speaker 1: Question 52, besides the interpolation, would it be useful to consider slope and aspect to correct shadows on the star image? Yeah, basically,

2:13:41 Speaker 1: you can use aspect to identify, yeah, you can use slope and aspect to correct shadows, and there are ways to correct it, like through interpolation. Again, as I mentioned, I rather treat those areas with shadow as areas that have no data. Usually I find interpolation. It just introduces some weird artifacts. But there are ways to correct it, and it just depends what you're comfortable with. I can put some resources in here. All right. One more question. My next question is about comparing star images.

2:14:18 Speaker 1: Could we subtract one image from another to see the difference? Yes, of course. Would that be helpful? Of course. If we do that, should we compare the same polarization like HH between images from different years? Do we need to make sure things like wavelength and angle are the same in both images? Yes. Okay, so if you want to do like an image difference, you want to use the same polarization because the information content between different polarizations is different, right? So if you want to look at the difference between time A and time B, use the same polarization.

2:14:57 Speaker 1: And you want to use the same wavelengths. And angle are the same in both images. Ideally, you're doing the difference with images from either ascending passes or descending. Again, if it's a flat area, it shouldn't be such a big thing, but I would stick to either ascending or descending. All right, one more. I'm just so excited reading your questions. Let's see. One more. Can you please list methods to separate water pixels from others during flood analysis other than histogram thresholding? Can we go ahead with some machine learning for the same?

2:15:44 Speaker 1: If so, what are the key factors to consider? So, yeah, there are other ways. And in the list of tutorials, the references I put in the appendix, there is actually either one or two RSET trainings on looking at flooding that I did where I did a classification. So it was beyond just thresholding. It was a classification to map flooded areas. So you actually train your classifier with in native education so double bonds high backscatter with open water so that's really dark and then with everything else that's in your image and then you run the classification okay okay two more can you tell us the paper reference that looked at woody and non woody vegetation with supervised machine learning yes I will put that reference here Referencing to slide 70, is this image a composite image of all the polarizations as mentioned HH, HV, and VV?

2:16:57 Speaker 1: I have to go back and look at slide 70. I think that was just an RGB. I have to go look back. Question 57, what freely available platforms provide access to SAR imagery and could you outline their spatial and temporal resolutions as well as other key characteristics? So as mentioned, I put some platforms there. Third session will be on the Alaska satellite facility capabilities. And usually most freely available SAR data is on the order of 10 meters. Like Sentinel-1, it's posted at 10 meters. PAL-SAR data, I believe that's at 25 meters, at least on Google Earth Engine.

2:17:35 Speaker 1: On the Alaska satellite facility, you have other perhaps higher resolution PAL-SAR data. So it tends to be higher resolution. OK. So I think in the interest of time, I would love to stay here and keep answering your questions but in the interest of time let's just just finish this off just come to an end for this session I will answer the rest of the questions in the document and we will be posting it on the training web page thank you everyone Thank you for your great interest, and keep tuned. We will have another great session next Wednesday on INSAR Interferometric Synthetic Aperture Radar, and then a third session on the 20th on the capabilities of the Alaska Satellite Facility, so what sort of SAR data they have and the amazing resources that they have on their portal.

2:18:35 Speaker 1: Okay, so thanks to the RSAT team and I will see you all next time. Have a great day. Bye-bye.