Radar Remote Sensing for Flood Monitoring

0:10 Speaker 1: The question was for when you're using amplitude images, radar images for mapping inundations or really for anything, do you need to do atmospheric corrections? In the case of when you use the phase and interferometry, you do need to do atmospheric corrections, but is it not necessary when using the amplitude images? And the response is that, no, you don't need to do an atmospheric correction with amplitude, radar amplitude images. You only need to do atmospheric corrections when using the face.

0:51 Speaker 1: However, that doesn't mean that the atmosphere does not affect the amplitude images. It might, If you have, again, it's going to depend on the radar wavelength. Longer wavelengths, like L-band, will be less sensitive to very significant atmospheric events. But at C-band or even at X-band, you will be sensitive to large precipitation clouds and wetting on the ground. So you do need to take that into account when interpreting your images. Sometimes you might see some either attenuation of the signals, like dark areas, especially in the tropics, or sometimes you might see just very high reflectivity, high backscatter, sudden high backscatter, or limited to the area where the rain event took place.

1:54 Speaker 1: because the ground is very wet, it's going to increase reflectivity. And so if the acquisition occurred during a heavy rain event or right afterwards and things are very wet, it's going to appear brighter than previous images that didn't have that level of moisture. Entonces lo que aclaré ahorita en inglés es que a pesar de que no hay que hacer correcciones atmosféricas, al usar la amplitud con radar, eso no significa que las condiciones atmosféricas no afecten la señal. Y esto ocurre en el caso cuando hay eventos meteorológicos significativos, especialmente en el trópico donde hay células, o sea, nubes con alta precipitación.

2:45 Speaker 1: Entonces, en estos casos, si hay una adquisición de... the radar data during a significant meteorological event, one may see attenuation of the signal. In other words, there is a darkening of the signal in areas where there are clouds with a lot of water, a lot of water is falling. And it may also be that during an event, when a lot of water falls, a lot of rain, and se adquiere la imagen durante un evento así o justo después que haya llovido fuertemente, es posible ver entonces una alta reflectividad, o sea, un incremento en la señal de radar que no estaba allí en imágenes anteriores.

3:39 Speaker 1: Así que por eso siempre es bueno visualizar las imágenes because sometimes you can see these things a little strange that could have been due to strong meteorological events. And obviously this will also depend on the band, the frequency that is used, the band L less affected by these strong events than the band C, for example, or the band X, when we are reaching a wavelength that is puede interactuar más con las gotas de agua, especialmente en el trópico donde son grandes, con esas gotas de agua y por ello la señal se atenúa.

4:25 Speaker 1: Okay, esa fue una explicación bastante extensa. Okay, entonces la pregunta número dos. Question number two, how does synthetic aperture radar technology improve flood detection and monitoring of compared to optical remote sensing methods. So the improvement is that radar can see, quote unquote, through clouds. And so you have this all weather capability with radar. And that is unique to radar, obviously not with optical. And usually if you're looking at flood events that affect humans, then it's important to be able to monitor these events and really understand the impact.

5:17 Speaker 1: And this has a whole series of implications for response, for example. And so these events tend to occur, these flood events tend to occur or are characterized by cloud cover, by dense cloud cover. So that's the limitation with optical is that you cannot see through clouds and you can't really monitor these events. as they occur. The other advantage of radar is that the signal can penetrate through the vegetation canopy or through a medium. So that medium would be a vegetation, a volume of vegetation, soil, snow.

6:06 Speaker 1: In the case of vegetation, that signal can penetrate through the vegetation and detect whether the vegetation is inundated or not. And this allows or is very useful for studying, say, wetland ecosystems or if you're looking at flood events affecting humans, then it's useful to identify areas like vegetation, agricultural areas that have been flooded perhaps and affected by such an event. So with optical, you're just looking at the top of the canopy, so your signal is not penetrating through the vegetation, and you're limited to seeing whether the vegetation is inundated or not.

6:55 Speaker 1: Okay, entonces, la pregunta número dos,¿ cómo mejora la tecnología de radar de apertura sintética la detección y el monitoreo de inundaciones en comparación con los métodos de teledetección óptica? So, there are two reasons why the synthetic aperture radar has advantages over optical teledetection. The first is that SAR has the advantage that it can see the.

7:30 Speaker 2: Surface despite almost any meteorological condition and can then detect las áreas y monitorear áreas donde ha ocurrido una inundación.

7:44 Speaker 1: Y en el caso de inundaciones que afectan infraestructura o poblaciones humanas, entonces se puede monitorear la extensión de esa inundación y si las aguas están disminuyendo, etc. Lo cual ayuda a informar la respuesta a desastres. Y durante estos eventos de inundaciones, They are characterized by dense cloud coverage, and for this reason it is very difficult, or the optical images are limited to characterize these events. Now, another advantage of radar is that the signal can penetrate through vegetation or some medium, Al referirme a medio, me refiero a un volumen, digamos, un volumen de vegetación, el suelo, nieve.

8:42 Speaker 1: Así que en el caso de vegetación, la señal al penetrar la vegetación puede detectar si hay agua, si la vegetación está inundada. Y esto es ventajoso porque puede ayudar a...

8:59 Speaker 1: caracterizar o entender especialmente eventos de inundación que afectan ecosistemas o áreas de agricultura, por ejemplo, que tienen un efecto de repercusiones sobre poblaciones.

9:19 Speaker 2: Así que se puede hacer.

9:21 Speaker 1: Un monitoreo de, digamos, las áreas agrícolas que han sido inundadas.¿ Cuáles son las implicaciones entonces en términos de producción? Y finalmente, algo que no mencioné anteriormente, esa capacidad de ver a través de las nubes, también es esa capacidad de poder monitorear la superficie, ya sea de día o de noche. Okay, la pregunta número tres. Question number three. Can radar remote sensing differentiate between floodwater, wet soil, and other water-related features? And if so, what techniques are used?

10:07 Speaker 1: Okay. So this is a good question. Yes. So radar can differentiate between, okay, so floodwater. So radar, let me just make clear that when we say we monitor flooding with radar, we're monitoring two things. One is, and these two things are very distinctive with the radar signal. One is surface water. So that means what we call in the community open water. So that's water without any, say, standing vegetation on it, like lakes or just puddled water, right? And it doesn't matter if that water is five centimeters above the soil or two meters above the soil.

11:02 Speaker 1: So that we cannot tell is how much water is there with satellites like Sentinel-1, for example, or upcoming satellites like NISAR. However, you can't tell if it's inundated. And if you do a comparison with other images, then you can assess the change in inundation, right? So how much has that area, the extent of inundation compared to normal conditions. The other part that radar is very good at is detecting inundated vegetation. So That is any sort of vegetation, standing vegetation that is inundated.

11:50 Speaker 1: Again, we're not determining the amount of water, but we are determining that there is water above the soil, whether it's five centimeters or two meters. We're just detecting that there's standing water above the soil underneath the vegetation. So I think that's what is meant by floodwater. Now, what soil is different? And that one is much more challenging. Because at that point, what soils, you're talking about soil moisture. And that is another, that's a whole other algorithm. And SMAP, the Soil Moisture Active Passive Satellite, has global soil moisture products every three days.

12:37 Speaker 1: And the reason I say it's more challenging is because it's a quantitative measure of the amount of water in the soil. So it's an algorithm that is telling you how much volumetrically, how much water is in the soil. And the reason it's challenging to make this retrieval is because with radar you have a signal, because radar is an active system, the signal comes in through the vegetation and and it scatters in different directions and part of it scatters back to the satellite. And that signal contains information about everything that was on the surface, including the moisture, right?

13:22 Speaker 1: So basically you need to inform your algorithm about all these other things that might affect the signal, like the roughness, the surface roughness, the vegetation, the texture in the soil. You want to know if there was snow, if conditions were frozen. So there are many, many things. If conditions are frozen, you can't retrieve soil moisture. So there are many things that inform the algorithm, and it becomes a little more complex. But yes, there are algorithms to estimate soil moisture using radar.

14:00 Speaker 1: And the radar signal can penetrate through the soil the level of penetration is going to depend on how wet the soil is, as well as the wavelength that is being used. So which frequency? The longer the wavelength, the greater the penetration. Okay. La pregunta número tres.¿ Es posible que radar pueda diferenciar entre inundaciones y suelos... y la cantidad de agua en el suelo? Y si es el caso,¿ cuáles técnicas son utilizadas? Entonces, la respuesta aquí es que para radar, primero aclaro que con radar y al hablar sobre la detección de inundaciones con radar, realmente estamos detectando dos cosas.

14:56 Speaker 1: Uno es lo que yo llamo en español agua abierta. Es una traducción literal del inglés open water. which are basically bodies of water on the surface without vegetation. So, whether it is water that has been accumulated, lakes, etc. That is open water. And the other type of flood that can be detected with radar is flooded vegetation. Y en ambos casos estamos detectando simplemente si hay agua arriba del suelo, básicamente, ya sea 5 centímetros o 2 metros de agua. Eso no lo podemos determinar, simplemente estamos detectando que hay inundación, hay agua acumulada en la superficie, ya sea en una superficie sin vegetación o en un área donde hay vegetación y debajo de la vegetación hay there is a flood, that is, there is water on the surface.

16:06 Speaker 1: So those are the two types of floods that we can clearly identify with radar. And when we are talking about open water or water, this has a very dark appearance with radar images and

16:26 Speaker 1: When we talk about flooded vegetation, the radar signal is very high, what we call double rebound. So there are like two extremes with these two situations, or the signal is very high or it is very low. very high when there is flooded vegetation and very low when there is open water. And to identify them, usually a threshold is used, but there are other ways to be able to map them. Now, in terms of flooding, the amount of water on the ground, this is different, a different algorithm is used. And there are different algorithms that use the radar signal to estimate the amount of water in the soil.

17:26 Speaker 1: And these algorithms can be somewhat complex because it is complex to estimate the humidity of the soil, since the radar signal has two ways. Basically, it enters and there is an interaction with the surface, that is, that radar signal interacts with the components of the surface and disperses in different directions. Obviously, the direction, the signal that returns to the direction of the satellite is collected by the satellite and that signal contains information about the components of the surface, whether there was vegetation or what there would be, grass, whatever.

18:12 Speaker 1: So, basically, that signal, part of that signal contains a component that de la humedad del suelo pero hay que extraer ese componente y cómo se extrae básicamente hay que informar el algoritmo sobre las condiciones en la superficie digamos cuál es la vegetación qué

18:37 Speaker 1: tipo de vegetación cuál es el contenido de agua de la vegetación la rugosidad la textura del suelo hay como 10 12 diferentes cosas que se utilizan para informar el algoritmo. Entonces, para llegar a la parte de la señal que es relevante a la humedad del suelo. El satélite NISAR, el satélite SMAP, que se llama el Soil Moisture Active Passive, mide la humedad del suelo. Tiene productos de humedad del suelo a nivel global. Y el satélite NISAR que se estará lanzando en un satélite, una colaboración entre la NASA y la Agencia Espacial de la India.

19:20 Speaker 1: Tendrá un radar de banda L y otro de banda S. Y este satélite va a estar generando un producto de humedad del suelo a nivel global. Y se lanzará a principios del próximo año. Okay, la siguiente pregunta. La numero cuatro. Question number four, what are the key challenges in interpreting radar data for flood extent mapping and how are they mitigated? This is a great question. All right. So there are a couple. I mentioned earlier that with radar where when we say flood detection, we're really detecting two things.

20:10 Speaker 1: One is open water and the other one is inundated vegetation. So there's standing water above the soil. And how much that we're not measuring, it's just we're measuring some level of standing water above the soil. So when there's open water, the radar signal is very low. And when there is inundated vegetation, the radar signal is super high. So basically, when these two conditions occurred, you have two extremes here. to extreme signal responses. What are the challenges? Let's start with open water.

20:47 Speaker 1: The challenges with detecting open water with radar is that sometimes there are other things that can have very low return. It's called backscatter. So there are other conditions that can have very low backscatter that will look you know, on the radar image like water, but are not. And some of these things can be roads. They can be recently tilled fields, for example. And so. How are they mitigated? You need to then inform your algorithm.

21:31 Speaker 1: in some way that those are agricultural fields or maybe bringing in a roadmap and informing it that creating a mask of roads. Okay. The other challenge is that open water is usually very low. So with radar, we have The reason it's low is because radar is side-looking, okay? So the sensor is looking toward the side. It's not looking straight down. It's looking toward the side. And it's sending the signal, and it just visualizes, okay? So you're the sensor, and you're side-looking. You're looking at a chair, look at the ground.

22:27 Speaker 1: sending a signal towards the ground. And when that signal hits a surface that is very smooth, and that would be a body of water that's called a specular reflector, right? If it's very smooth, that signal will basically bounce straight away in the opposite direction. Okay, that's why open water bodies are very dark. Okay. If there's any sort of roughness, say there's vegetation there. So instead of that being a water body, there's vegetation. What happens? That signal is going to interact with the vegetation and it's going to disperse in different directions, including the direction of the satellite.

23:09 Speaker 1: And so you're going to get some signal return. All right. So now let's return. Now that you've visualized all of this, let's return to the open water body. So if it's a smooth surface, it will reflect away and it looks very dark. However, if that open water body has some roughness on it because of wind, there's going to be some scattering in different directions, including the direction of the satellite. And so what normally would appear as very dark now has some level of or some return. So it doesn't look so dark and it has some level of return, which might be confused with other things around it.

23:56 Speaker 1: So maybe areas where there are grasses or other things. So that is a challenge. The way to mitigate that is to use different bands or polarizations. Usually the cross polarizations are best for detecting open water because they are not affected by that roughness. or they're less affected by that wind roughness. Okay. The other challenge in detecting flooded vegetation, so that's the extreme, that's very bright. And with flooded vegetation, basically, you have what's called double bounce. And it's very easy to detect double bounce.

24:43 Speaker 1: flooded vegetation areas. However, there are challenges in areas where you have topography. Because of the topography, especially complex topography that is facing the direction of the satellite, you will have stronger return of the signal, higher backscatter, and that can be picked up by your classifier as areas of inundation. And so what you do is you use a DEM to mask those areas out. And other things that confuse or might be confusing are urban areas. So urban areas can be, also have high backscatter, similar to inundated vegetation, very, very bright.

25:32 Speaker 1: And How do you mitigate these? You use some sort of urban mask to mask out your urban areas. Okay,

25:48 Speaker 1: la pregunta número cuatro.¿ Cuáles son los desafíos claves en la interpretación de datos de radar para el mapeo de la extensión de las inundaciones y cómo se mitigan? Entonces, lo que expliqué en inglés es que To detect flooding, as I mentioned before, basically two characteristics are sought in the signal, either very dark, very low signal, or very high. The areas that are low are areas where there is water, and the areas that are high, that is, a high retrodispersion, means that they are areas of flooded vegetation.

26:31 Speaker 1: So we start with the areas where there is water, without any type of vegetation. In these areas, other things that can be confused with water are things that are also characterized with a low signal. And these are, let's say, roads or Campos recién arados, por ejemplo. Y la forma que se mitigan es utilizando una máscara de la infraestructura de carreteras o de campos de cultivo. Y una de las cosas que expliqué en inglés es cómo funciona radar. Entonces, hay estos mecanismos que le llamamos mecanismo de retrodispersión.¿ Cómo interactúa esa señal a llegar a la superficie?

27:31 Speaker 1: Radar de apertura sintética se caracteriza por estar mirando de lado, o sea, está orientado a mirar la superficie de la tierra de lado, no directamente hacia abajo. Así que, si se pueden imaginar, si están sentados, o sea, miren de lado y miren hacia el suelo, imagínense que ese suelo es una superficie de agua, y esa superficie de agua se caracteriza because it is very smooth. So what happens is that the signal, when reaching the surface of the water, is reflected in the direction opposite to the satellite.

28:11 Speaker 1: And for this reason, the areas where there is water appear very dark in the radar images. Now, if in that water there is a certain roughness, due to, let's say, the wind, then the signal, when reaching the surface of the water, va a tener cierta dispersión en diferentes direcciones, incluyendo la dirección del satélite, y esa superficie de agua no va a tener ese aspecto oscuro como normalmente lo tiene, y se va a confundir con otras cosas como, digamos, campos de cultivo, por ejemplo. Y la forma de mitigar esto es utilizando otras...

28:58 Speaker 1: Utilizando la polarización cruzada, lo que le llaman VH o HB, que realmente no es tan afectada por la rugosidad en el agua. En términos de.

29:12 Speaker 3: Vegetación inundada, Ahí ocurre lo que le.

29:29 Speaker 1: Llamamos el doble rebote y por eso la vegetación inundada tiene un aspecto muy alto, muy brillante en las imágenes. Pero también si... las áreas urbanas tienen este aspecto muy brillante, muy alto en las imágenes. Y se confunden. El algoritmo va a básicamente indicar que las áreas urbanas, o especialmente las áreas altamente urbanas, están inundadas cuando no es el caso. Así que la mejor forma de mitigar esto es utilizando una máscara de las áreas urbanas. Y también las áreas donde hay topografía compleja, especialmente si las montañas están orientadas en la dirección del satélite, o sea, donde está mirando el sensor.

30:22 Speaker 1: Estas áreas aparecen brillantes, las montañas, Y para mitigar esto, entonces se utiliza un modelo de elevación digital. Okay. Seguimos a la siguiente pregunta. How are SAR data and machine learning algorithms combined to improve the accuracy of flood forecasting and impact assessment? That's a super great question. And I think this is a work in progress. There are many different machine learning algorithms that are being applied. Random forest is sort of the standard, really, of what most people are using.

31:03 Speaker 1: But there are other things that I've seen in the literature, CNNs and neural networks and many different approaches to for flood mapping. Now, you're talking about flood forecasting, and that is more of, yeah, so that is, I know there are studies that are using different machine learning algorithms. I'm sorry, I can't name the approach or the methods, but they combine different conditions to to do flood forecasting. So they're pulling in information related to the moisture in the soil, related to precipitation events, related to slope, for example, and I think other things as well for flood forecasting.

32:12 Speaker 1: And then an impact assessment, and obviously the impact assessment, I think GIS has a big part there. because of pulling in social and economic data related to human population, infrastructure, agriculture, et cetera. Okay, la pregunta número cinco.¿ Cómo se combinan los datos de radar y los algoritmos de aprendizaje automático para mejorar la precisión de las... de la previsión de inundaciones y la evaluación del impacto? Y esta es una muy buena pregunta. Realmente sé que hay estudios que están haciendo esto de los prognósticos de inundaciones.

33:08 Speaker 1: No tengo tanto conocimiento sobre las metodologías que están utilizando, pero sí sé que utilizan varios datos para informar el algoritmo y que están utilizando algoritmos de machine learning para hacer esto, donde utilizan datos, digamos, de humedad del suelo, de precipitación, de vegetación, modelos de elevación digital, para entonces hacer estas predicciones de inundación. Y en términos de la evaluación del impacto, la parte de GIS, de sistemas de información geográfica son importantes para, entonces, integrar datos socioeconómicos para mejor entender estas áreas en riesgo, cuál es la población, cuál es la infraestructura, campos agrícolas, etcétera, que pueden potencialmente ser afectados, igual como líneas eléctricas, etcétera, torres de comunicación, cosas así.

34:12 Speaker 1: La pregunta número seis.¿ Cuáles son las principales técnicas para mejorar la identificación de objetos en imágenes de radar de mediana resolución?¿ Cómo se puede sacar el mejor provecho a este tipo de imágenes libres de radar? Ok, esta pregunta no sé si la entiendo con total claridad, porque cuando dice la identificación de objetos, No estoy segura a qué se refiere esa persona. Si son clases, digamos, si se hace una clasificación de cobertura vegetal, digamos, de la vegetación. Así que me voy a referir, voy a asumir que es identificar diferentes clases en la imagen de cobertura terrestre.

35:03 Speaker 1: En imágenes de radar de mediana resolución, esa también... It ends a little relative because it depends on who is asked that question, medium resolution. So, well, I'm going to assume that the radar images, well, let's talk about the images that are available openly, that are radar, are Sentinel-1 of the European Space Agency, Banda C, that has a resolution De alrededor de 10 metros están los datos de Pulsar que uno los puede acceder por medio de la agencia espacial japonesa o por medio de Google Earth Engine también, que son mosaicos anuales y tienen ciertos datos también Pulsar ScanSar, una adquisición de banda ancha.

36:02 Speaker 1: Así que, y esa tiene una resolución yo creo que de alrededor de 25 metros.¿ Cuáles son las principales técnicas para la identificación de objetos? Vamos a decir que la clasificación de datos en las imágenes de radar. Las principales técnicas son,

36:26 Speaker 1: la que más se usa, diría yo, es random forest, árboles de decisión, decision trees, por medio de random forest. Pero hay otras técnicas que también se están utilizando. Otros machine learning, CNNs, neural networks, gradient boost. Hay muchas que se están utilizando. Pero definitivamente la que más se utiliza es Random Forest.¿ Y cómo se puede sacar el mejor provecho a este tipo de imágenes libres de radar? Yo creo que totalmente depende de la aplicación y del algoritmo que estás aplicando. Y es importante entender el algoritmo y refinarlo, lo que le llaman los hiperparámetros, para obtener los mejores resultados posibles.

37:21 Speaker 1: Okay, so question number six, what are the main techniques to improve the identification of objects in medium resolution radar images and how can you get the most out of this type of free radar images? So I wasn't quite clear what is meant by identification of objects. I'm assuming this means land cover, like doing a land cover classification, identifying different classes. And then medium resolution, I think that is a relative term depending on who you ask. But I'll refer to just the openly available radar images or some of the most openly available radar images right now globally, which are Sentinel-1 from the European Space Agency.

38:06 Speaker 1: That's C-band. And then there's some PALSAR data that's openly available on Google Earth Engine. These are global mosaics on a yearly basis. So PALSAR is L-band. It's from the Japanese Space Agency. And there's also PALSAR-SCANSAR. And then the Alaska Satellite Facility has PALSAR data between 2006 and 2011. So that was the first batch, I'll say, of PALSAR data. It was a sensor that flew on ALOS-1 that operated between that time frame. And so those images, all of those are between 10 meters and 25 meters around that.

38:50 Speaker 1: And so there are many different ways of doing land cover classifications. I think one of the most used ones in terms of machine learning is random forest. So it's the decision tree-based approach. But there are other algorithms that are being used, CNNs and neural networks and gradient boost and many other things. So with radar, obviously, you just have a limited number of bands, right? Usually you just have two polarizations, which are BBBH or HHHV. And a lot of times you can derive textures from these.

39:30 Speaker 1: which do help in your classification results. So there are different measures of textures that you can derive that will tell you like the homogeneity of an area surrounding each pixel or the variance. or the relationship between a pixel and its neighbors, et cetera, as you move further away from the pixel. So there are many different texture measurements that you can derive and feed them as layers. There are also digital elevation models that you can feed into your algorithm. So there are many different things to try.

40:09 Speaker 1: Obviously, the most important one is, one, having good training data, and the other one is understanding the algorithm and really optimizing it so that you optimize the hyperparameters and you get the best results that you can achieve. Okay, so the next question. La pregunta número siete.¿ Las imágenes SAR de uso abierto tienen la capacidad de medir la humedad y salinidad del suelo? Okay, las imágenes de SAR de uso abierto tienen They have the ability to measure the humidity of the soil, because there is a significant or measurable difference in the signal, depending on the amount of water in the soil.

41:00 Speaker 1: Salinity in the soil also affects the signal, but it is really so small. So the next question

41:38 Speaker 1: Do open-use SAR images have the ability to measure soil moisture and salinity? So the response to that is that, yes, you can measure soil moisture. And the change in the radar signal due to changes in or as a function of the moisture in the soil is noticeable. It's significant. However, even though salinity has an effect on the radar signal, the signal-to-noise effect So it's such a small change that you cannot accurately measure it with openly available SAR data. So yes, you can measure soil moisture, but not salinity.

42:23 Speaker 1: That's a super tough one. And there are algorithms out there to estimate soil moisture. Pregunta número ocho,¿ cómo podríamos detectar zonas inundadas mediante el uso de la curva de decibeles?

42:51 Speaker 1: No estoy segura a qué curva de decibeles se refiere esta persona. Yo creo que es la que demostró el doctor Franz Meyer en su presentación. And this curve was, it seems to me, what he demonstrated was how the radar signal changed, the decibels or the retrodispersion, as an area flooded. So he demonstrated two different examples, flooded vegetation, so through a medida que incrementa el agua en la superficie, o sea, una superficie seca que va absorbiendo el agua, los suelos, la humedad en el suelo va incrementando hasta un punto donde se satura.¿ Cómo se ve eso en las imágenes de radar?

44:00 Speaker 1: Y lo que él enseñó en esa curva es que gradualmente, a medida que la humedad en el suelo y en la vegetación incrementa, Gradually, the retrodispersion, the radar signal, increases. And once the vegetation is flooded, that is, there is water on top of the surface already accumulated, then that signal jumps to be super high. And that's what I think this person is referring to. And Realmente, la mejor forma de detectar zonas inundadas es comenzando con un umbral y simplemente estableciendo el umbral en base a los valores de inundación, de vegetación inundada y los valores de agua.

45:01 Speaker 1: Y definir que los valores de aquí hacia arriba son vegetación inundada y de aquí hacia abajo So the question is, how could we detect flooded areas by using the decibel curve? And I'm not sure what this person means by the decibel curve. I think it might be a graph from Dr. Franz Meyer's presentation on using SAR for flood monitoring, where he showed an example of how flooded vegetation, or one of his examples was how flooded vegetation looks as things flood, right? What's the radar signal as vegetation floods?

45:46 Speaker 1: So assuming that things are dry, vegetation is dry, and suddenly you have a strong precipitation event, so things get wet. The vegetation gets wet, the soil gets wet. And as the water starts accumulating, the radar backscatter starts to increase gradually. But once you hit that point where you have inundation, underneath the vegetation, so standing water underneath the vegetation, then that signal just shoots up. It's incredibly high because you have now double bounce, what's called. And so to detect flooding, so that was just an example, but to detect flooding, the best way to do it is, at least to start out with, is to set a threshold.

46:37 Speaker 1: and look at areas that above a certain threshold are inundated vegetation and then below a certain threshold are open water. So that's one way. Another way is you can also do a difference between an image where you have normal conditions and then an image where you have inundated conditions and then set a threshold based on the differences La pregunta numero diecinueve, nueve. What are the calibration conversion considerations outside of normal preprocessing that must be considered of any when using multiple SAR images from Sentinel-1 to make continuous and reliable time series observations?

47:32 Speaker 1: Okay, so this is a good question. Outside of the normal pre-processing, you don't really need to do any calibration conversions. But it's always good to take a look at your images because, again, as I mentioned earlier, you might have had like a heavy rain event and that is going to throw off your image. Right. So if conditions in the ground are sunny, like super wet, right during the acquisition, that's going to, the signal is going to be a little different. So you need to account for that. And obviously, if you have any sort of seasonality, you need to account for that too.

48:24 Speaker 1: Because, for example, there are going to be differences in the signal, in the backscatter, when conditions are frozen, the land surface is frozen versus when the land surface is thawed. So question number nine, what are the considerations related to calibration or conversion outside of the normal pre-processing of the images that must be considered when multiple images of Sentinel-1 radar are used to make continuous observations a través del tiempo y la respuesta es que hay que siempre es bueno visualizar las imágenes en la serie temporal porque como mencioné anteriormente si hay un evento fuerte digamos de precipitación durante la adquisición de una de las imágenes entonces the images in the temporal series.

49:19 Speaker 1: Because, as I mentioned before, if there is a strong event, let's say, of precipitation during the acquisition of one of the images, then it is very possible that that image, aspects on the surface, look different. You also have to consider the stationality, especially, let's say, if there are differences over time between the condition on the surface and the condition... si hay nieve o la superficie está congelada versus descongelada. La pregunta número 10. Question number 10. How does signal-to-noise ratio affect a 3D imagery of a landscape in SAR arrays?

50:00 Speaker 1: Yeah, I'm not sure I understand that question in terms of the 3D imagery of a landscape. in SAR arrays. So perhaps that person can clarify or repost the question, reword the question.

50:29 Speaker 1: La pregunta número 10 no la entiendo, así que le he pedido a la persona que escribió esa pregunta a ver si la puede aclarar. La pregunta 11,¿ se necesita un software para procesar o visualizar las imágenes obtenidas? Sí se necesita un software. Hay softwares en línea o hay otros softwares que puedes descargar o puedes abrir las imágenes en un software de SIG, ya sea QGIS o ARCGIS.

51:03 Speaker 1: Para procesar las imágenes sí necesitas un software especial que no lo vas a encontrar en QGIS. Este software se llama SNAP y es un software de la Agencia Espacial Europea.

51:24 Speaker 1: Pero realmente las imágenes las puedes obtener ya preprocesadas de la Alaska Satellite Facility O aún las imágenes que están en Google Earth Engine ya están preprocesadas. Question 11. Do you need a software to process or view the images obtained? So, yeah, you do need a software to process the images. Now, a lot of the images are already processed. So, for example, the Sentinel-1 images that you access on Google Earth Engine or the Pulsar event images are already pre-processed. And also the ones that you can access through the Alaska Satellite Facility are also already pre-processed.

52:04 Speaker 1: However, if you want to process them yourself, you can use something like the Snap Toolbox. And this is a free software from the European Space Agency. And with that, you can go through each of the processing steps for the imagery. And then once they're processed, you can visualize the images in any software you want. including a GIS software. Question number 13. Can radar be used to estimate water level elevation for open water, inundated vegetation? If yes, what kind of accuracy can be expected for the water levels?

52:41 Speaker 1: Yes, you can. There are techniques to estimate water level, and there is actually a satellite called SWAT that does this, but it's just for open water. It's not for inundated vegetation. And I will put in more information here about that mission and the accuracy. La pregunta 13 dice,¿ radar puede estimar el nivel del agua para lo que le llamamos agua abierta o vegetación inundada? Y si sí,¿ qué tipo de exactitud se pueden esperar de los niveles de agua? Entonces, la respuesta es que sí se puede.

53:30 Speaker 1: Hay un satélite que se llama SWAT que hace precisamente esto, pero nada más sobre agua, lo que le llamo agua abierta, no en vegetación inundada. Y pondré detalles sobre esa misión aquí en esta respuesta. Catorce, la pregunta catorce. En la capacitación, in the RSATs are for detecting and monitoring flood CIs and subsidence. In the flood hazard Jupiter notebook, how do I create the RTC imagery needed to run the analysis? So in that Jupiter notebook, I believe the images were already pre-processed.

54:10 Speaker 1: So RTC means radiometrically terrain corrected. So that's the processing that needs to be done to each of the SAR images. And the Alaska satellite facility has already done this. However, I will double check and I will elaborate on this answer here in the Google Doc. La pregunta 14 dice que en la capacitación de RSET, el uso de radar para detectar y monitorear inundaciones, hielo y subsidencia por extracción de agua. En la sesión de inundaciones había un cuaderno de Jupiter y la persona aquí pregunta cómo creo las imágenes de de RTC, que son las imágenes preprocesadas, para ejecutar ese análisis con Jupyter Notebook.

55:06 Speaker 1: Y mi respuesta es que las imágenes que se utilizaron en el Jupyter Notebook ya estaban corregidas, preprocesadas. Y el Alaska Satellite Facility ya ha hecho este preprocesamiento para la mayoría de las imágenes y uno las puede acceder ahí. Pero de todas maneras voy a verificar y voy a elaborar la respuesta cuando ponemos este documento en línea. Okay. Pregunta, question 15, is there data collected over Niger River and is it available for commercial purpose? I believe there is. Yes, there is SAR data.

55:47 Speaker 1: There is global SAR data from Sentinel-1. I'm sure it covers the Niger River. Then I'm sure there's multi-temporal data. And it is openly available. And I believe there should be PALSAR data, too, because these are global yearly mosaics. And there's a PALSAR ScanStar data, which you'll have to check if there's coverage over the Niger River. And then Alaska Satellite Facility has, as I mentioned, PALSAR data going back 2006 to 2011. That is also global, and it's freely available. And you can use it for any purpose, for commercial purposes, I believe.

56:37 Speaker 1: Okay, so that's it, and we're at the top of the hour. These are, these, okay, entonces la pregunta, la última pregunta es si hay datos de radar sobre el río Niger y si se pueden utilizar para propósitos comerciales? Y mi respuesta es que sí, sí hay datos. De Sentinel-1 a nivel global, estoy segura que cubren el río Niger y deben de ser datos multitemporales también. De Sentinel-1 y también de Palsar, también deben de haber datos. Y esto los puedes... Puedes buscar estos datos ya sea en Google Earth Engine o a través del Alaska Satellite Facility.

57:31 Speaker 1: Okay, so we're at the top of the hour. Really great questions. We will be cleaning up this Google Doc and putting it online. If you have any additional questions, please feel free to reach out to me and I will respond or even include your question in this Google Doc. So before closing, I would like to let everyone know that we will be doing a radar SAR training at the end of October, early November. We will be announcing that through the RSAT webpage. And so it will be an introduction to SAR. It's basically a refresher.

58:12 Speaker 1: And we'll be discussing past, current, and upcoming SAR missions and how to access the data, different ways that you can access the SAR data. So, yes, please stay tuned. I think it will be of great interest to many of the participants here. Así que con esto llegamos al final de esta sesión. Quisiera agradecer a todos los participantes. Si tienen preguntas adicionales, por favor me pueden enviar preguntas directamente a mí. Y quisiera recordarles que Tendremos una capacitación de ARCET a finales de octubre, principios de noviembre, por ahí.

58:55 Speaker 1: Va a ser una introducción a radar. O sea, básicamente estamos rehaciendo esa capacitación, actualizándola. Vamos a hablar sobre también los diferentes sensores de radar de datos que pueden acceder. sensores históricos, operacionales y los nuevos radares que van a estar siendo lanzados al espacio y cómo acceder a todos estos datos. Al igual que, bueno, introducción a radar de apertura sintética y a radar interferométrico. Así que esperamos que puedan asistir. Pondremos los detalles una vez que finalicemos la agenda.

59:35 Speaker 1: Lo pondremos en la página de ARCET. Lo estaremos anunciando. Así que hasta entonces, gracias a todos. Gracias al equipo de ARCET por el apoyo. A Brock Blevins, a Natasha Johnson-Griffiths, Selwyn Hudson-Odoi, Jonathan O'Brien y Sarah Kuchel. So thanks to the RSAT team for their support in all of this. Special thanks to David Barbato, who's our translator. Especialmente agradecidos a David Barbato, es nuestro traductor estrella. Así que con eso, deseándoles un excelente día y hasta la próxima. Wishing everyone a great day and until next time.