In this last lesson of Unit Five, we're going to touch on two different data products: thermal data and LiDAR. You might say, "aren't we in a unit on land use and land cover classifications?" To answer your question, Yes, we definitely are. But both of these data products are really commonly used within land use and land cover classifications, and therefore, we want to make sure you understand the data. Now, beyond that, these are data that are rising in popularity because there's such utility in what they provide. And so even if we weren't in a unit on land use and land cover classifications, it's really ideal to understand these data sets. [Slide] Let's touch on thermal remote sensing over the course of this first video. What we mean by thermal remote sensing is literally we're just detecting surface temperature differences, and that is done through measurements in the thermal infrared portion of the electromagnetic spectrum, which we'll look at here in a second. We know from the early weeks of our class, that we have incoming solar radiation. It's reflected off the surface of the Earth, and we know about the different wavelengths associated with this reflection. We've talked a lot about visible portion of electromagnetic spectrum. Now we're talking more so in the Thermal Infrared-- NOT Near Infrared, which we've heard a lot about the semester, but Thermal Infrared. [Slide] If we look at the electromagnetic spectrum, we've seen this a few times the semester before. We're understanding our radiation, our wavelengths in the visible portion from violet here all the way to the longer wavelengths in the red area. But then right adjacent to our visible portion electromagnetic spectrum is the IR region. And with the IR region, we actually have a few different sub areas. We call the Near IR, the Mid IR, and the Thermal. You've seen a lot with the Near IR this semester. That Near IR region is basically right in here. I'm going to show you a more detailed image in a second. That's what we use for some of our compositing. We can use that as inputs to our land use and land cover classifications. When we get to vegetation indices, you'll see its use there. There's a lot of ways that we use near infrared. What we're going to actually focus on this lecture is this region here and (ignore that I covered radio waves here-- we're not going into radio waves). It's this portion of the infrared part of electromagnetic spectrum. [Slide] Rather than looking at my first diagram, here's a better breakdown of it. So we know as we're going across this chart, we're increasing in wavelengths from gamma rays having our shortest to TV waves having our longest wave lengths. So if we look in the interior portion of the infrared section, the Near IR have the shorter wavelengths, the Thermal IR have the longer wavelengths of the infrared region. Okay. And it's right there adjacent to our visible portion of the electromagnetic spectrum. [Slide] What is the difference between thermal and traditional optical remote sensing? Well, first off, thermal is dealing with emitted heat radiation temperature. Whereas optical, we're dealing with obviously re surface reflection measurements. Additionally, we know now from looking at the previous two slides, what portions of the electromagnetic spectrum they're dealing with. The thermal IR section is dealing with just that longer wavelengths of infrared radiation, whereas the optical can cover visible to near infrared portions. What are some of the applications? Well, there are some overlaps here. But we can think of temperature monitoring, urban heat studies, changes in, say, moisture content within different land covers. Those will have different patterns of heat or thermal emissions. Additionally, with the optical side, we've seen a bunch of examples, of land use land cover mapping, vegetation health assessments. We've seen dozens of examples this semester. [Slide] If we go back to our pretty sure our first and second units of this semester, when we were talking about black body radiation, the basics of remote sensing, weans law, thing Stefan- Boltzmann Law. We actually started to touch on some of the physical principles of thermal remote sensing then. If you don't remember that, you can go watch the videos again. But for this part of the lecture, just keep in mind that one of the core things we defined was a black body entity or black body radiation. Meaning black body radiation is the principle that an idealized object (it doesn't exist in real life, so just say idealized object) that is a perfect absorber of energy. Doesn't matter if it's any portion of electromagnetic spectrum, doesn't matter the incidence angle, doesn't matter the frequency of the data of the radiation. It's a perfect absorber. Now, the reason I bring this up is because one of the grounding principles of thermal remote sensing is this thing called Planck's Law. In order to explain Planck's Law, it's best to start with this diagram here. Let's look at this diagram. It's got spectral reflectance here versus wave length. What we see is obviously the visible portion electromagnetic spectrum and infrared portion electromagnetic spectrum. Now, remember, here's basically our near IR, and then we have our mid IR getting into our thermal IR right in this area here. It goes a little bit beyond two. Now, with Planck's Law, there's a really complex definition, but what it essentially says is that as temperature increases, a black body emits more radiation at shorter wavelengths. Let's look at this diagram here where we have three different temperatures. As temperature increases, here's our highest temperature to our lowest temperature. A black body emits more radiation at shorter wavelengths. Look at what happens to the distribution of these as you get into the shorter wavelengths. This is true, and this is how we can differentiate some thermal properties by different land covers. Different land covers have different heat content, that different heat content will therefore emit different patterns of thermal radiation, right? So just make sure we're getting this, the warmer the object, the more radiation is emitted at a shorter wavelength. So you can see here versus 3,000 Kelvin versus 5,000 Kelvin, what are the patterns of radiation that is emitted? 5,000 has higher emissions, 3,000 has lower emissions, right? [Slide] That leads us to the term "Emissivity." Emissivity is a measure of a materials ability to emit. Because remember, we said black body is an idealized. It's not a true surface. It's not a perfect black body-- it is not a realistic thing, I guess is what I'm saying. Emissivity says, with a realistic surface, with a realistic entity, what are going to be the patterns of emission, of radiation? And it can be ranging from zero, meaning no emission, all the way to one, meaning it's a perfect emitter. Now, the factors that are influencing emissivity depend on things such as the color of the surface, the texture, the temperature adjacent to it, as well as the wavelengths of the radiation itself. When we talk about the significance of this in terms of remote sensing, we know that accurate knowledge of a surfaces emissivity or surface temperature is crucial because differential patterns of surface temperatures imply different aspects related to soil moisture usage related to land cover, land cover thickness, vegetation thickness, vegetation health. There's a huge number of applications of this data, and therefore, there's a large significance in understanding the thermal properties. [Slide] What sensors can we use for thermal remote sensing? Well, there's quite a few. There are thermal sensors or thermal measurements on Landsat, especially Landsat 8 and 9, MODIS, ASTER, VIRS. A lot of our big satellites have these thermal sensors associated with them. Now, not all of them are highly or commonly used. Some of them are very complex to process and so aren't used as commonly. [Slide] But one of the most common for a very long time was this MODIS product, MOD11, and this is a modus land surface temperature, so they've already done all the conversions for you. It is a daily data set that also shows you emissivity. The pixels are 1 kilometer, so rather large in size. This can be shown as a night measurement, sometimes you'll have a day measurement, but you can get a range of temperatures across the different MODIS observations. [Slide] Beyond that, MODIS also provides some thermal anomaly data. They take it one step even further. We can measure fire, for instance. That's another great thing about using thermal data. We can detect wildfire extent, wildfire severity, and the MODIS products provide those for you as well. So MOD14A1, 14A2 those are either daily or eight-day fire products-- that's what they call them are fire products, but truly all they are thermal anomalies. Again, they're at 1 kilometer, so they are rather coarse, but given the scale of most of these events, that's typically fine. [Slide] In terms of Landsat 8 (onboard Landsat 8), we have the TIRS sensor-- and this is a thermal infrared sensor, that's why it's called TIRS. TIRS is nowadays, one of the most commonly used because it's not only has actually two infrared channels that it's measuring, but it also is higher resolution than a lot of the other thermal sensors. So you can get a 30-meter pixel here. Now, the reason it has two channels is it really allows for what they call a split-window technique to study the thermal properties. What this does is this distinguishes between the temperature at the earth's surface and the temperature at the atmosphere. They can measure it in a dual-modality essentially. [Slide] With data acquisition and processing, how does this work? How do we make thermal data into temperatures on the ground temperatures? The first thing obviously is you need to have imagery that has been atmospherically corrected. And we've now downloaded and seen landsat and modus data that have been atmospherically corrected? We know what that means. Next, we need to convert our digital numbers into radiance. And radiance is obviously something we've seen. Surface reflection is something we've seen already. Digital Numbers provide some measure for us of the surface properties, but they're not formulated to be converted into temperatures. We do need to have surface reflectance and radiance. Then what we're going to do is convert that radiance value to at sensor temperature. That at sensor temperature means atmospheric temperature, we can then distinguish and differentiate and convert to land surface temperature. If you're going out and you're looking for thermal data products, look for the term LST, Land Surface Temperature. You don't want Atmospheric Temperature. You want land surface. Atmospheric temperature really doesn't provide you much about your land use and land covers temperature patterns. Keep that in mind, that's the workflow. Now, that being said, even though this is the workflow, it's not really common for you to do this yourself. So many of our datasets come like this already for us, because there's so much error that can happen in between here that a lot of these big agencies have like U.S.G.S. and NASA, they have done the work for you because they don't want you to use their data incorrectly. We can get this land surface data for the globe at least every eight days. Sometimes we can get it daily, though, too. [Slide] What are some applications of thermal remote sensing? I've mentioned a few. But if we can think about environmental monitoring, think of say urban heat islands, which that's the figure you see here on the right. We know that over urban surfaces, we see a rise in temperature, over more vegetated surfaces, we see lower temperatures across those gradients. And so that urban heat island, we can monitor even within an urban area, are there regions that are having a higher impact of the urban heat island versus maybe in the suburban areas where there's more vegetation, tree cover, and parks, we would have a lesser urban heat island effect. Beyond environmental monitoring. We can use this for crop applications. Things like crop health assessments can be done. I'm going to show you an example in the upcoming slides on one of those. Then lastly, really not lastly, there's a lot of other applications, but lastly on this slide is to talk about disaster management. I already showed you that we can detect fire, and that's fire extent, fire severity. We can also look at post-fire recovery like burn marks and things of that nature. We can also look at flooding extent, flood severity, we can look at sea surface temperature patterns, of things like like La Nina / El Nino those are based on upwelling of cold water or downwelling of cold water in the Pacific, and we can monitor that through remote sensing of thermal data. [Slide] Let's go through an example of thermal data's usage within an agricultural perspective. And I'm not going to talk about wildfires. I'm not going to talk about vegetation per se. I'm going to instead talk about cold air and specifically cold air drainage. We know across different topographic gradients. We look at the topography in the area, we have highs, we have lows. We have gradients of our elevational ranges. We also have our pools or areas where there's sinks in the ground. There's areas that are concave on the ground. [Slide] And with thermal remote sensing, we can start to map what happens in those areas. Imagine if you had these sinks or these concave areas where resources water, for instance, or air could pool in an agricultural setting. What that would do to the crops in those areas? They would be more vulnerable, right? We've done a project in Sparta, Michigan, where we looked at apple orchards. And within these apple orchards, we were interested in seeing what percentage of their area was covered in these depressions that could hold a lot of cold air. And then also how they were managing for the cold air, how they were trying to fix that problem. [Slide] We used a UAV, with a thermal sensor attached to it. We could measure the air temperature and the ground surface temperature using this thermal sensor. We could see what was happening across a broad portion of the landscape. We also did this very early in the morning when the air temperature was getting to its lowest so that we could truly see this was the critical time for the fruit that would be on the apple trees. We could see that critical time and see how the air would be impacting that fruit. [Slide] Again, we can map this air pattern. What you see actually in the polygons are where we have those concave areas, areas where the air would pool. The lines that you see there are showing you the pass that air would more easily flow around those. This is a lot like hydrology. If you're familiar with hydrology, that's how water moves overland, air would move in a very similar way, and it's going to follow that path of lease resistance and it's going to pool where there's blockages. And that's what we've mapped here. [Slide] What we've also mapped with our thermal sensor is how growers are trying to mitigate the impacts of this cold air drainage. They use something called wind fans. These are very expensive. They're large. They cost about $15,000 each, and they have to place numerous of these fans across the landscape. What happens when it gets really cold, the farmer will go out turn on this fan because it basically pushes this cold air out of these concave areas and redistributes it. What you can see on this image on the left is actually one of these fans that's running. This is it right here, and you can see it's actually pointing this direction. The fan is moving the air that's in this area. You can totally tell just by looking at this image because this is a colder portion of landscape, the fan isn't impacting here. This portion of landscape is slightly warmer because the air has been disrupted and moved around. You've lost that sinking of the air there. What you also see here though is the impact of the fan itself. It's got, you know, some energy behind it. It's got some heat built up into it. So what you see is that red blob anomalously is because of the fan. You can see the impact of the fan just by looking at the thermal image, but we can go beyond that. [Slide] What we did is we mapped out the patterns of the thermal data. Let's first talk about this figure here on the right, this one here. And each of these, each of these little regression setups that you see here are representative of a fan, and they are fans across different elevations. And what it's saying is, as you go away from the fan in certain distances, we're showing distance in one by the shape, we're showing the distance from the fan in another in the color. As you move away from the fan, how fast is the temperature changing? As you can see, there are some areas like clear, this fan was not placed correctly. It is not moving the air around very well. Conversely, this fan is having a great impact. As it's turned on, as it's moving that air, that air around it is rapidly changing in almost all directions except for the South, which is the area directly behind the fan. Similarly, these two fans here are having great impact, not as strong of an impact, as the previous example, but still a strong impact. If we go to this figure here on the left, what we first see is that this is showing the temperature change from different distances as well. Again, really similar to what you're seeing here, displayed in slightly more spaced out manner. We're not bringing elevation into this at all. If we look at a single fan across the area, you can see that this fan does not have much of an impact on the air temperature. That's this fan here. These other three (one, two, three), you can see how the temperatures are for the most part increasing as you move away from the fan. If you put a line through here, that line would increase as you move away from the fan. Now there are some differences. This one definitely goes down and that's because that's behind the fan. Again, I don't care if you don't understand the study. That's not the point. The point is to show you that all of this was purely done with just thermal data. That's all we had no multi-spectral, no LiDAR, nothing else besides thermal data. We could tell these grower, for instance, this fan is in a bad location, or the orientation is incorrect, or the way is set up, the height maybe of it is incorrect. Versus these three get the check mark here. They're actually being effective. Then what we could also do is say, are your apples healthier in this region versus this region? I can tell you by talking to this farmer, absolutely. The apples were not healthy here, and it was because he didn't have the fan oriented right. [Slide] As we wrap up this lesson, let's talk about the challenges of using this data and where we're going with the use of such data. Big issues associated with this data are the spatial resolution. You saw the 1 kilometer. Also the temporal resolution. Sometimes it's daily, that's ideal. Sometimes it's only every eight days. If you're doing something like mapping wildfires, that's not going to be really ideal. But conversely, if you're wanting to map heat waves, 8 days is fine. Similarly, just like when we talked about land surface reflectance, we want to make sure we're understanding land surface temperature, not atmospheric temperature. This is typically not data you want to process on your own. You want to let the organizations process it and you use the products so that you know that you're getting a proper calibrated land surface temperature or temperature anomalies product. Additionally, we have some differences in the sensors between this thermal sensor and a more multi spectral sensor. And what that can mean is some of our platforms, some of our satellites, we can't mount one of these sensors on because either the angle that is flying at the viewing angle is not ideal for it or maybe the altitude that's going to fly out is not ideal for a thermal sensor. That's why not every satellite will have a thermal sensor available. [Slide] Where are we going with this? What is the future? This technology is constantly developing. The thermal drone data that I showed you for the apples example, that thermal sensor cost about $11,000. That was also three to four years ago. Nowadays, that sensor would cost about $9,000. We're definitely improving the technology and improving the cost of the technology. We're also seeing an integration of this data, more so with other datasets. So things like LiDAR combined with thermal, which you haven't seen LiDAR yet. That's the next video. But as you're looking at that lecture, think about, well, when you have this data, what would happen if you would have these point clouds in it as well. This thermal data combined with LiDAR data. That will make more sense what you see the next video. So this integration with not just LiDAR, but multispectral data is becoming more and more common as are our research applications with it. I do think we struggled for a bit to include it at first in our typical remote sensing analysis. But over the last three or so years, we've really seen a boom of thermal data popping up in what are seen as traditional remote sensing studies.
GEO 324, Video Microlecture: Unit 05 (Classification Approaches), Lesson 06a (Thermal)
From Beth Weisenborn October 28th, 2024
17 plays
0 comments
|
From Erin Bunting January 8th, 2023 |
Add a comment