NASA ARSET Foundational Concepts

Speaker 1: Welcome everyone to our RSET training series. Introduction to Thermal Remote Sensing and Applications in urban heat Island Mapping Today is part one of our training series, Foundational Concepts of Thermal Remote Sensing. My name is Savannah Cooley. I'm a researcher at NASA, AIMES research center with the Bay Area Environmental Research Institute in Mountain View, California, and I'm a trainer with the RSET Ecological Conservation Team. I'll start with a few quick words about the RSET program. The NASA Applied Remote Sensing Training or RSET program provides cost free training on the use of remote sensing observations, analysis methods, and tools.

Speaker 1: We provide training in several thematic areas, including agriculture, climate and resilience disasters, ecological conservation, health and air quality, water resources, and wildland fires. Our SET provides trainings both online or in person. Our online trainings are delivered in two formats, live an instructor led like Today's training, or asynchronous in self paste like NASA's free and accessible data. All of our trainings are offered at no cost. We try to offer trainings in more than one language whenever we can, and we only use no costs in open source software and data.

Speaker 1: We offer our trainings at a range of levels, so you can find a training series that fits your level of experience and need. Please visit us at our website to learn more. We'll start today by giving an overview of this training series introduction to thermal remote sensing, foundational concepts and applications. Here we are looking at a thermal image of the oxets in Toulouse, France. In the thermal infrared, we measure both the effect of temperature and emissivity. Emissivity relates to the composition of the material.

Speaker 1: Because the cross is made of bronze and copper, it has a different emissivity and temperature and therefore a different thermal emission than the surrounding material of the pavement. Thermal infrared radiometer instruments such as NASA JPL's Airborne High Test Instrument and NASA JPL's spaceborne Ecostress Mission measure land surface temperature. You'll learn to use some of this data in part two of this training. First, and foremost, for part one of the training, you will gain foundational knowledge about the principles and concepts that underpin all thermal remote sensing instruments as well as learn about some key applications.

Speaker 1: This training has seven learning objectives. Learning objectives one through five focus on the underlying science of sensor capabilities and general applications, and will be covered in Part one today. Learning objectives six and seven are specific to workflows for the use of ecostress data for urban heat islands and will be the focus of Part two. We have fundamentals of remote sensing as the only prerequisite for this training. However, we also have several suggested trainings that we highly encourage participants to take.

Speaker 1: The first is satellite remote sensing for measuring urban heat islands and constructing heat vulnerability indices. The next is a brief introduction to the Ecostress mission, which is another our set training on the ecostres Us Mission website. And then finally, there's a number of tutorials available on the Land Processes doc that also provide more details on ecostress and how to access ecostress data. There are two parts in this training. In Part one today will introduce the theoretical background of thermal remote sensing.

Speaker 1: Part two will cover hands on exercises focused on urban heat island applications. The homework opens on the last day of the training on June second. It is due two weeks later on June sixteenth, twenty twenty six. A certificate of completion will be awarded to participants who attend all live sessions and complete the homework assignment before the given due date. We'll start today by giving an overview of this training series introduction to thermal remote sensing, foundational concepts and applications.

Speaker 1: We have two instructors for Part one. I'm a Savannah Cooley and I'm a research scientist in the Biospheric Sciences Branch within the Earth's Science Division at NASA Ames Research Center with the Bay Area Environmental Research Institute. Our other instructor is Glenn Holly, who is a research scientist in the Earth Science Section of NASA Jet Propulsion Laboratory, California Institute of Technology. By the end of Part one, participants will be able to identify the fundamental concepts and physical principles of thermal infrared remote sensing, define the role of emissivity retrievals and ensuring the accuracy of satellite derived land surface temperature products, distinguish key differences between thermal and optical remote sensing approaches, including emission versus reflection, day versus night capability, and atmospheric window considerations.

Speaker 1: Identify applications of thermal remote sensing data for ecosystem stewardship, agricultural management, climate adaptation, and urban planning. Finally, compare the characteristics of current and upcoming thermal missions in the context of their suitability to different application use cases.

Speaker 1: Let's take a moment to consider this knowledge check. This training builds on the concepts from the prerequisite for this training, fundamentals of remote sensing, so a basic knowledge of remote sensing and satellite and sensor characteristics is assumed. I'll pause for a few moments to give everyone a chance to think for yourself about which of the options A through D listed below best completes the sentence. Thermal remote sensing data can be used for dot dot dot As you can see, options A, C, and D all describe real remote sensing applications, but they use fundamentally different parts of the electromagnetic spectrum.

Speaker 1: Thermal infrared remote sensing, which is the focus of today's training, works by measuring energy that is emitted from the surface itself, not reflected from the sun therm more remote sensing, in particular, detects emitted radiation at wavelengths in the thermal infrared bands between eight and fourteen microns, which we will discuss more in depth in the next section.

Speaker 1: B is the closest to what we'll be focusing on. By the end of today, you'll be able to explain exactly why B is correct and how it differs from the others. If you have questions during today's training, you can put them in the Q and A box within WebEx at any time. We encourage you to put your questions into the Q and A chat as they arise. The earlier you share your questions in the chat, the more likely it is that we will be able to respond to them in real time during the Q and A session. If there are questions we do not have time to address during the live Q and A session, we will make sure to collect these and post the written answers to the training page within a week after the training.

Speaker 1: In other words, we will try to address as many questions as we can today. All of these questions and answers will be posted, along with the additional questions and written answers to the training page within a week after the training. Let's dive into section one, the theoretical basis of temperature measurements. This diagram shows the full electromagnetic spectrum from long wavelength radio waves on the left to short wavelength gamma rays on the right. The visible light that our eyes can see occupies just a small sliver of this spectrum, roughly point four too point seven micrometers.

Speaker 1: Most optical remote sensing works in the visible, near infrared, and short wave infrared regions, spanning abouto point four to two point five micrometers. The thermal infrared region extends from approximately three to fourteen micrometers. Within this we distinguish two main atmospheric windows. Mid wave infrared or MWR, which is the window that spans roughly three to five micrometers. This region has complications from mixed solar reflection and thermal emission during daytime. The thermal infrared or TIR window spans roughly eight to fourteen micrometers.

Speaker 1: This is our primary working range for land surface temperature retrieval because it contains purely emitted thermal radiation with no solar contamination. Notice the atmosphere absorption features shown here. Water vapor absorbs strongly throughout much of the infrared. Carbon dioxide has absorption bands around four point three and fifteen micrometers. Ozone absorbs around nine point six micrometers, right in the middle of our thermal infrared window, which is why some sensors avoid that specific wavelength.

Speaker 1: A key point is that thermal infrared remote sensing measures emitted energy radiation that originates from the surface itself based on its temperature, rather than reflected energy from an external source like the sun.

Speaker 1: This fundamental difference has major implications for how we acquire, process, and interpret thermal data. Before we discuss how we measure temperature remotely, let's establish what temperature actually means from a physics perspective. Is fundamentally a measure of the average kinetic energy of molecular motion within a substance. Every atom in molecule in matter is in constant motion, vibrating, rotating, and translating. The faster these particles move on average, the higher the temperature for an ideal gas, The average kinetic energy of molecules is directly proportional to the absolute temperature.

Speaker 1: The average kinetic energy equals three halves times Boltzmann's constant times temperature. So here in this equation k E represents the average kinetic energy, so the mean energy of motion per molecule in jewels, measured as one half m. The squared the three halves factor represents the three degrees of freedom corresponding to the spatial dimensions for movement in a three D space where it's there's also a contributing half sigma T to each to energy for each dimension. K represents the Boltzmann constant, which relates temperature to energy, holding a constant value of one point three eight times ten to the minus twenty three jewels per Calvin T is absolute temperature, the temperature in kelvin.

Speaker 1: The equation twells us that energy is directly proportional to the temperature in kelvin, meaning that zero kelvin, which is absolute zero, corresponds to zero kinetic energy. This relationship tells us that temperature is not some abstract property. It's a direct measure of how much energy is stored in the random motions of atoms and molecules. We measure temperature and several scales. Kelvin is the absolute scale, starting at absolute zero, where all molecular motion would theoretically cease zero Calvin equals negative two hundred and seventy three point one five degrees celsius.

Speaker 1: In remote sensing, we typically work in kelvin because the radiation laws require absolute temperature. The key insight here is that temperature reflects the internal thermal energy of matter. An object at three hundred Calvin has molecules moving at a certain average speed. An object at three hundred and ten Calvin has molecules moving faster with more kinetic energy. This kinetic energy is what ultimately gets converted into the thermal radiation we measure from space. Now we arrive at a crucial connection.

Speaker 1: How does the kinetic energy of molecular motion become electromagnetic radiation that we can detect from space. The answer lies in the behavior of charged particles. Atoms in molecules contain electrons and protrons, charged particles. When molecules vibrate, rotate, or collide with each other, these charges are accelerated. And here's the physics. According to Maxwell's equations of electromagnetism, any accelerating charge emits electromagnetic radiation, So the process works like this. First, molecules at any temperature above absolute zero are in constant random motion.

Speaker 1: Vibrating, rotating, colliding. These motions cause the charge particles within atoms to accelerate. Accelerating charges radiate electromagnetic energy. The hotter the object, the more vigorous the molecular motion, and therefore the more radiation emitted. This is the fundamental bridge between the thermodynamic definition of temperature kinetic energy of molecules, and what we measure remotely electromagnetic radiation. You can think of it then, thermal radiation is the electromagnetic exhaust, if you will, of molecular kinetic energy.

Speaker 1: When we point a thermal sensor at the Earth's surface, we're detecting the electromagnetic signature of all that molecular motion happening at the surface. The intensity and spectral distribution of this radiation follow precise physical laws, the Plank function and Stefan Boltzmann law, which we'll discuss next. These laws quantify exactly how much kinetic energy converts to radiated energy. This connection is why thermal remote sensing works. By measuring the electromagnetic radiation emitted by a surface, we can infer its temperature because that radiation is a direct consequence of the surface's internal thermal energy.

Speaker 1: To understand thermal radiation quantitatively, we need to introduce a concept of a black body. A black body is a theoretical ideal, a perfect absorber and perfect emitter of radiation. A black body has two defining properties. First, it absorbs all incident radiation at all wavelengths. Nothing is reflected or transmitted. We express this as absorptivity alpha equals one. Second, black bodies emit the maximum possible radiation for its temperature. We express this as emissivity epsilon equals one.

Speaker 1: The term black body comes from the fact that an object that absorbs all visible light would appear perfectly black. But don't let the name confuse you. A black body at high temperature actually glows brightly. Think of the heating element on an electric stove. As it heats up, it goes from black to red, to orange to yellow. In the real world, no material is a per black body. Real materials are sometimes called gray bodies. They have amissivity less than one, meaning they emit less radiation than a black body at the same temperature.

Speaker 1: The closest approximation to a black body is a cavity with a small opening. As the diagram shows, radiation entering the opening, bounces around inside and is almost completely absorbed. This is why laboratory black body calibration sources use cavity designs. Why does the black body concept matter and remote sensing because the radiation laws we use Plank's law, Stefens Stephen Boltzmann's law describe black body radiation. To apply these laws to real surfaces, we must account for their emissivity, which tells us how efficiently they emit compared to a black body.

Speaker 1: Water, for example, has a massivity of around onero point nine to nine in the thirdsmal infrared, very close to a black body, but dietation is similar around point ninety seven to point nine nine, But bear soil might be point nine two to point ninety six, and dry sand can be as low as point nine. These differences matter for accurate temperature retrieval. Now we come to the fundamental equation of thermal radiation Plank's radiation law. This equation, developed by Max Plank in nineteen hundred, describes exactly how much energy a black body emits at each wavelength for a given temperature.

Speaker 1: The equation is b of lambda and t equals two hc squared over lambdra to the fifth power times one over the quantity e to the power of HC over lambda ktnus one. Let's break this down and define each term so B as a function of lambda and T is the spectral radiance the power emitted per unit area per unit solid angle per unit wavelength. Its units are watts per meter squared per stridian per micromemeter. Lambda is the wavelength in micrometers. T, like we've seen before, is the absolute temperature in kelvin.

Speaker 1: H is Plank's constant equal to six point two six times ten to the negative thirty four duel seconds. This is the fundamental quantum of action. C is the speed of light approximately three times ten to the eight meters per second. K like we saw before, is Boltzmann's constant, and this constant links temperature to energy at the molecular level. If you remember a few slides back, and E is just the base of the natural logarithm approximately two point seven one eight. The graph on this slide shows Plank curves for three different temperatures relevant to Earth's surfaces.

Speaker 1: So three hundred calvin, three hundred and ton Calvin, and three hundred and twenty Calvin. Roughly twenty seven, thirty seven, and forty seven degrees celsius.

Speaker 1: Notice several important features. First, higher temperatures produce more radiation at all wavelengths, the curve shifts upward. Second, higher temperatures shift the peak towards shorter wavelengths. Will quantify this with Wine's law. Next. Third, in the eight to twelve micrometer range, the curves are well separated. A ten Calvin time difference produces a clearly measurable radiance difference. This is why this wavelength region is so useful for temperature sensing. Plank's law is the foundation for all thermal remote sensing.

Speaker 1: When we measure radiance at a specific wavelength, we can invert this equation to solve for temperature. If we know the emissivity. Willen's displacement law tells us where a black body's emission spectrum peaks at what wavelength the most energy is emitted. The equation is elegantly simple. Where we have lambda max so the maximum wavelength of the peak emission in micrometers, which is equal to the term b divided by t. Where b is Wien's displacement constant equal to two ninety eight My michrometer calvin.

Speaker 1: This is a derived constant that comes from finding the maximum of planks function. The inverse relationship means that hotter objects emit most strongly at shorter wavelengths. This graph visualizes Ween's displacement law by comparing the emission spectra of the Sun versus the Earth. First notice the logarithmic skills on both axes. This allows us to compare the massive energy output of the Sun with the relatively tiny output of the Earth on a single plot. The core principle here is the equation lambda max equals b divided by TEA.

Speaker 1: Where the Sun at around six thousand calvin is incredibly hot, so its peak emission is pushed to the left at zero point four eight micrometers. This falls squarely the visible light spectrum, which is why we see sunlight. The Earth at around three hundred ish Telvin is much cooler, so its peak shifts to the right

Speaker 1: at about nine point sixty six micrometers. This is the thermal infrared. This shift explains why we design satellite sensors differently based on the target. We use the eight to fourteen micrometer range for Earth observation because that's where terrestrial surfaces emit the most signal. While Plank's law showed us the shape of the emission curve, the Stephen Boltzmann law tells us the total area under that curve the total energy emitted. The relationship is defined by this equation M equals epsilon sigma T to the fourth.

Speaker 1: Let's break that down. M is the radiant exitance, which is the total power emitted per unit area measured in watts per square meter. Epsilon is emissivity, which, as we discussed before, is the efficiency factor ranging from zero to one. For a perfect black body, that value would be one, and like we said, real materials are usually less efficient, so we will be diving into this more in the next section. And sigma is the Stephen Boltzmann constant, So this is a derived constant that roughly equals five point sixty seven times ten to the minus eight.

Speaker 1: And then again T is the absolute temperature in Calvin. The power of T to the fourth is the most critical feature here where you know. This basically shows a massive nonlinearity, meaning that energy output doesn't grow with temperature, it grows really really fast. Look at the graph on the slide. Here we are comparing a standard Earth surface temperature of three hundred kelvin to a slightly warmer three hundred and ton kelvin, this is only a three point three percent increase in temperature. However, if you look at the energy jump from on the y axis, the output rises from roughly four hundred and fifty nine to five hundred and twenty three watts per square meter.

Speaker 1: This is a fourteen percent increase in radiated energy. So why does this matter? This extreme sensitivity has multiple important implications for remote sensing in general. We can see this as great news in the sense that it means that our thermal sensors are incredibly sensitive. We can detect very subtle differences in surface temperature because the signal changes so dramatically. For climate physics, this fourth dependence relationship with temperature acts as a stabilizing feedback valve. As the Earth worms, it sheds heat much more rapidly into space, which helps prevent runaway heating.

Speaker 1: For data analysis, it does complicate things in a certain sense because the relationship isn't linear. So if you have a pixel that contains both hot and cold objects, the hot objects will dominate the signal disproportionately. This is a problem we call subpixel mixing. We apply the stuff in Boltzmann law to reconcile the difference between a real surface and a perfect black body. Kinetic temperature t KIN is what we can think of as the ground truth, the actual thermal state of the surface determined by molecular motion, but satellite sensors don't measure this directly.

Speaker 1: Instead, they measure radiance and convert it to brightness temperature t B by inverting Plank's law. Brightness temperature is defined as the temperature a perfect black body would have to be to emit the radiance we observe. We call the radiometric temperature t RAD, which is the temperature that integrates brightness temperature across all wavelengths. Via the stuff in Boltzmann constant, t RAD can be related directly back to kinnetic temperature with the equation Here, t RAD equals the emissivity to the one fourth power multiplied by kinetic temperature.

Speaker 1: Since emissivity is less than one for all real surfaces, t RAD is less than t kinetic t kin. The dryer and sandier the surface, the larger the gap between them. Let's go into a review question to convert measured radiance to kinetic temperature. What additional information is needed. I'll pause here and let folks think a little bit on your own. The correct answer is b surface and massivity. Remember, satellite sensors measure radiance and convert it to brightness temperature, which is defined as the temperature a perfect black body would have to be to emit the radiance we observe.

Speaker 1: The problem is that real surfaces aren't perfect black bodies. They have emissivities less than one. So if we assume everything has a massivity of one, will systematically underestimate surface temperature. The key takeaway is that the lower the emissivity, the larger the gap between what a sensor measures and the surface. The surface is true temperature. This is why accurate emissivity is so critical and why we will spend time delving into the topic in the next section. In section two, we will focus on emissivity.

Speaker 1: Here you can see radiance curves of surface emission, surface emission and reflection, and at sensor radiance for a quartz amissivity spectrum. Atmospheric corrections are necessary to isolate surface features which are obscured by atmospheric attenuation.

Speaker 1: The red curve shows what the surface is actually emitting. You can see the characteristic quartz and massivity feature as a dip near eight point five to nine micrometers. The blue nerve adds in the reflected downwelling component and atmospheric attenuation.

Speaker 1: By the time we get to the black curve what the sensor actually sees. The Coretz spectral feature is still visible, but it is shifted under atmospheric structure. The ozone absorption feature near nine point six micrometers is cutting right through the middle of our tir window. Water, vapor and CO two are carving out additional structure throughout. This is a core challenge. We have one measurement at sensor radiance, but we are trying to solve for two unknown simultaneously, temperature and emissivity.

Speaker 1: The rest of the signal is atmospheric noise we have to characterize and remove before we can even get to the surface. This is precisely why atmospheric correction is not optional and why am acivit retrieval algorithms exist. We need to mathematically untangle all of these contributions to get back to the surface. Temperature emissivity is defined as the ratio of the actual emission to black body. Emission emissivity varies with material roughness, angle, and wavelength. The top graph illustrates the differences in radiances between a black body, which is shown in the solid black line, versus quartz material, which has lower emissivity than a black body.

Speaker 1: As a result, we can see that emissivity of quartz on the lower graph is lower than one across all wavelengths, especially in the eight to ten micron window. Emissivity for most natural surfaces varies from point five to five to point nine to nine. Here are some examples from different surfaces, including soils, rocks, sands, and a gray body, which, unlike a black body, is an imperfect emitter, though as you can see, compared to these other surfaces, is closer to one across all wavelengths.

Speaker 1: There are two main approaches for estimating amissivity. First is split window and the second is temperature emissivity separation or tests algorithm retrieval. The split window method used in Modus data products, for instance, uses the brightness temperature difference between two adjacent thermal channels typically around eleven and twelve micrometers to simultaneously correct for atmospheric water vapor and estimate surface temperature. It works by assuming a fixed or land cover base a massivity assigned by vegetation class lookup tables.

Speaker 1: This approach is computationally efficient and works well over vegetated surfaces and water where amissivity is high and relatively stable between around point nine seven to point nine nine. Typically. However, it is sensitive to errors when the assumed emissivity is not correct, particularly particularly over arid, bare, and geologically complex surfaces. The temperature emissivity separation or TESTS method, which is used in aster and ecostress data sets, simultaneously retrieves both temperature and spectral emmissivity from multiple thermal bands.

Speaker 1: Rather than assuming amissivity from a lookup table. Tests uses a semi empirical relationship based on the minimum and messivity and the spectral contrast across bands, the MMD, which is the minimum maximumum difference in order to constrain the solution. This yields both a temperature estimate and a full emissivity spectrum, providing much richer information about surface composition and substantially reducing errors over bare and heterogeneous surfaces. Split window errors will occur in a systematic fashion anytime that emissivity is incorrectly assigned, leading to systematic biases.

Speaker 1: The Mona Loa caldera example here provides a clear illustration. Basaltic lava flows have strongly variable and spectrally complex emissivity in the eight to twelve micrometer range, dropping below point nine in some channels. The ten percent emissivity difference shown in the graph arises in the eleven micron channel. When comparing the standard level to modus split window approach to the tests algorithm retrieval from the same modus input, the split window algorithm retrieves the temperature of three one hundred and ten calvin, approximately twelve to fourteen calvin cooler than the aster tests, which was at three hundred and twenty two on average, and the modus tests at three hundred and twenty four calvin, both of which explicitly retrieve amissivity.

Speaker 1: This is not a subtle calibration issue. It's a physically meaningful error driven entirely by the wrong emissivity assumption. The rule of thumb shown here is that a one point five percent emissivity error propagates to roughly one calvin of LST error. This means that over rocky, sandy, and geologically diverse surfaces. Split window errors can reach three to ten calvin for applications such as urban heat island analysis and ecosystem thermal stress, where one or two Calvin differences even are scientifically meaningful.

Speaker 1: Here are more comparisons between the LST retrievals based on split window emissivity values versus the tests algorithm emissivity. There is similar accuracy over vegetation and water of about one calvin. However, we see a split window cold bias over bear regions in this case, ranging from three to five calvin. This is especially important for thermal sensing applications of urban heat islands. Cities are highly heterogeneous, with asphalt next to grass and other surfaces in small areas. Split window isssumes a massivity based on land cover, but tests calculates it.

Speaker 1: If the urban land cover map is outdated, the LST will be incorrect. Tests is more robust for rapidly changing urban environments. Now that we've covered the how the physics of missivity and temperature, let's look at the what, where and when. In section three, we survey the landscape of current and upcoming thermal infrared missions and highlight the key application areas that drive this science. This will help you connect those physical equations to the actual data products you'll use in Part two.

Speaker 1: Thermal remote sensing is a diverse field with sensors tailored for different scales. Mode is an aster on the Terra platform have provided a global record since two thousand. Lansat offers an incredible historical archive, though it has fewer thermal bands compared to multi spectral TIR sensors like ast Ecostress. Our focus for this training is unique because it's on the International Space Station. This allows us to sample the diurnal cycle, measuring heat at different times of day rather than just once in the morning.

Speaker 1: The ability to observe temperature at different times of day is advantageous for many applications, including the urban heat island focus that our training will delve into in Part two. Finally, high Test is an airborne instrument used for high resolution campaigns, providing one to thirty meter resolution that is invaluable for validating spaceborne data. We are entering a golden age of thermal remote sensing. While ast led the way in nineteen ninety nine, we now have Trishna, which is scheduled to launch in twenty twenty six, Eagle Ter which was formally called Speech and that this mission has been in reformulation and does not currently have a new launch date, and then LSTM, which is scheduled to launch in twenty twenty eight.

Speaker 1: The trend is moving toward higher spatial resolution around fifty to sixty meters and more frequent repeat cycles, which will revolutionize how we monitor urban heat and water use. This figure illustrates why sensor band placement matters for emissivity retrieval. Each panel compares how well three different sensor configurations can capture the missivity spectral shape for two geologically distinct surface types felsic minerals like gypsum on the left and mapic minerals like pyroxene on the right.

Speaker 1: The blue shaded region represents the uncertainty envelope for tr A, TRB, and tr C band configurations, while the redline shows laboratory measured and massivity. This is what we would refer to as the ground truth that we're trying to match. Notice that felsic minerals like gypsum have a pronounced emissivity minimum near eight point five to nine micrometers. This feature falls at a wavelength where some sensor configurations have poor band coverage, leading to mist or smooth spectral structure.

Speaker 1: Mathic minerals like pyroxene show a different, more subdued spectral shape, but the same principle applies. If your sensor bands don't straddle key spectral features, you lose compositional information. The takeaway here is that aster ecostress and speary SBG now EGO TAR those configurations with their strategic band placement across the eight through twelve micrometer window, do a much better job of capturing these spectral features than coursercuit figurations. This has direct implications not just for geology, but for urban surface mapping.

Speaker 1: Pavement, concrete, and roofing materials all have distinctive spectral and massivity signatures that better placed bands can resolve. Now, let's take a step back and compare thermal remote sensing to optical remote sensing, which is what many of you may be more familiar with. If you are coming from a traditional optical remote sensing background and are most familiar with sensors like sentinel to or lansat visible neurinfrared, there are three key differences to remember. This table captures an important conceptual distinction in Earth observation.

Speaker 1: Optical sensors are fundamentally passive receivers of reflected sunlight. They have no signal at night and are limited to daytime observations. Thermal sensors, by contrast, detect energy that the surface itself is generating twenty four hours a day, driven purely by the surfaces temperature. This is why Ecostress can observe nighttime land surface temperature from the International Space Station. The surface is always emitting. Notice also the difference in what each type tells us about the surface.

Speaker 1: Optical data is rich in biochemical information, for example, plant pigments, water content, soil organics, while thermal data tells us directly about thermodynamic state, energy balance, evaporative cooling, heat stress, and surface composition through a massivity.

Speaker 1: A key practical trade off is that achieving fine spatial resolution is fundamentally harder in the thermal compared to the optical, because the emitted photon flux is much lower than reflected solar This is why even state of the art missions like Ecostress have seventy meter pixels while sentinel to achieves ten meters. This resolution gap is one of the major motivations for the downscaling exercise we'll do in part two of this training. There are many different applications of LST and emissivity in earth science.

Speaker 1: For hydrology, we use thermal data to estimate of appotranspiration or ET, which is very useful for agricultural water management. LST is a key input to land surface models for drought monitoring, soil moisture and ET estimation. LST exerts control over the partitioning of energy into latent and sensible heat flexes and provides an indicator of surface warming trends from anthropogenic climate change. Smaller massivity errors can lead to massive errors in energy balance models, making accurate tir retrievals essential for climate projection.

Speaker 1: For example, an error of point one in the emissivity will result in climate models having errors of up to seven watts per meter squared in their upward long wave radiation ITS estimates. This is a much larger term than the surface radiative forcing of about two to three watts permeter squared. Due to an increase in greenhouse gases. In the field of ecology, we can identify ecosystem thermal stress before plants show visible signs of wilting using satellite based LST. In tropical forests, existing studies have documented the increased thermal stress of burned forest relative to logged and intact forests, even over a decade after the forest fire occurred.

Speaker 1: For urban heat applications, of course, we use these maps to identify heat vulnerable neighborhoods, which is the key focus of Part two. This slide illustrates two compelling applications of long term LST records for detecting climate signals that would be invisible to shorter observational windows. The map on the left shows Modus Land Service temperature across the northern Hemisphere for August two thousand and four. This is a snapshot of what the global thermal landscape looks like at the peak of northern summer.

Speaker 1: This kind of global coverage, updated every one to two days, is what makes motives such a powerful climate modern monitoring tool. The upper right panel shows a result from Schneider and Hook twenty ten. Surface water temperature at Lake Tahoe increased at a rate of point one two degrees celsius per year over a roughly two decade period. This is a lake that sits at over six thousand feet elevation in the Sierra Nevada, a system that we might expect to be buffered from rapid warming. The thermal record tells a different story, and it's a story we can only read because we have a consistent, well calibrated spaceborne thermal record going back to the early nineteen nineties.

Speaker 1: The lower right panel shows a different type of trend analysis from Holy and Do set twenty twenty the number of annual average nighttime days exceeding twenty degrees celsius in Anaheim, California. The trend of ero point six additional days per year with n our square to point nine three is remarkably strong. Nights that stay hot and are a key public health indicator. Nighttime heat prevents the body from recovering from daytime heat stress, and this is one of the primary drivers of heat related mortality during extreme events.

Speaker 1: Together, these two examples show that LST isn't just a snapshot measurement. It can also provide a long term climate variable. The consistency of the Modus record across more than two decades is what makes these trend analyzes possible, and it's exactly why maintaining continuity in thermal missions matters. LST can be used for deriving many other biophysical properties of the land surface, including of appo transporration ET. As mentioned before, ET is the total amount of water going into the atmosphere from the surface of aporation plus transpiration from leaves and canopy interception.

Speaker 1: Here we see average of apple transpiration at local time for June through August twenty nineteen. One thing to emphasize again is the importance of overpastime. Due to its sun synchronous orbit, Landsat observations always occur at ten thirty am before p KEAT. In contrast, ecostress can observe a location at three PM when the heat island is most intense. This justifies why we use ecostress specifically for this training, So stay tuned uned for Part two. Now we will go into a summary for Part one, Foundational concepts of thermal remote sensing.

Speaker 1: Let's take a moment to consolidate what we've covered in Part one before moving into the Q and A. We've built up from first principles from what temperature means at the molecular level through the radiation laws that govern how the internal energy becomes measurable electromagnetic radiation, through the practical complications of emissivity and atmospheric effects, and finally to the missions and applications that put all of this to work. The through line in all of this is that thermal remote sensing gives us a direct window into the thermal dynamic state of Earth's surface in a way that optical data simply cannot.

Speaker 1: When we point a thermal sensor at the Los Angeles Basin, we're not measuring reflected sunlight. We're measuring the electromagnetic signature of every parking lot, every park, every neighborhood's thermal energy. Part two will take you from this conceptual foundation into hands on analysis of exactly that.

Speaker 1: Part two will go from theory to data accessing real ecostress observations, working through quality screening, and running a machine learning workflow to downscale thermal data for fine scale urban analysis. The case study we'll use is write in the Los Angeles Basin, which you may recognize from the header image in our slides. Remember that the homework opens on the last day of the training on June second. It is due two weeks later on June sixteenth. I would like to acknowledge my fellow colleagues in the Arset Ecological Conservation Team who have supported the creation of this training Sativa Cruz, Juan Torres, Berres and and Fain.

Speaker 1: I would also like to acknowledge the RSET program staff who made this training possible, including Selwyn Hudson, Odoy, Maria Maravito, Suzanne Monty, and Melanie Follette.

Speaker 1: We encourage you to check out our website as well as the RSET YouTube channel linked here in this slide. You can also join our quarterly newsletter to stay up to date on our latest trainings by sending an email with no subject to our set dash join at lists dot NASA dot gov, and then follow the instructions sent in the response. The contact information for myself and our other guest instructor, Glenn Holly, is also here if you'd like to email us directly with specific questions about this training.

Speaker 1: Right, we will now transition to Q and A session, So let's start with question one. We have. Question one states something I've always been a bit confused about is what we mean when we talk about light being absorbed? Where does it go? Thank you for this question, So light that is not transmitted or reflected is absorbed by the material and converted into thermal energy the objects that gets and so and as a result, object gets hotter as light is absorbed, but the light's energy doesn't go away. So, in other words, light that is absorbed occurs when the electromagnetic energy is transferred directly into the atoms or molecules of the material itself that it hits.

Speaker 1: Question two, which satellite platform collection and spectral bands are being used for deriving LST measurements? Are there LST values sourced directly from a standard product eg. Modus Landset or derived through a custom retrieval methodology, And perhaps, Glenn, I'll maybe let you speak to this question since you were your answer was the one that we're showing here.

Speaker 2: Sure.

Speaker 3: Yeah, So there are several NASA satellites flying right now that actually have both thermal bands and also visible shortwaven for air bands, and probably the most common or most well known are the Motus series of satellites, so Modus Terra and Modus Aqua, and more recently we have the Beers follow on instruments which are also have very similar thermal bands. And so as soon as we have more than if we have more more than three or more thermal bands on one of these sensors, then we can directly apply the temperature emissivity separation algorithm that Savannah was going over ino slides.

Speaker 3: So for example, with the three Motus bands, we can apply tests and then directly retrieve the land service temperature and emissivity from those three thermal bands. So we don't technically measure you know, the land service temperature directly, we measure more the thermal radiance and from that we derive what the temperature should have been to produce that thermal radiance.

Speaker 2: And then another.

Speaker 3: Example yeah I had was lance at eight and nine, which in fact have two thermal bands on bands ten and eleven, and so we can't use the test algorithm in that case to produce the temperature emissivity. Instead we use a split window approach and also direct single channel approach to derive the temperature, and we use an external emissivity database. And then of course e can stress was mentioned by Savannah as well in our slides, and that in fact us five thermal bands and actually gives us the highest temperature accuracy because in general, the more thermal bands we have, the more accurately we can retreat themessivity and then also the surface temperature.

Speaker 1: Great, thank you Glenn. Let's go into question three, which asks about eco if is the ecostress data currently accessible? And the short answer is yes. There are a number of different ways to access ecostress data. So we'll list a few of the most common ways here, the biggest ways, and then we will in part two, you know, provide some some examples of of of accessing the data. So the Land Processes Distributed Active Archive Center or doc lp DOC is the the where the data of the Ecostress mission are stored.

Speaker 1: And this link here provides some UH tutorials that are public by the lp DAC and that they are a great one great place to start for for learning to work with ecostress data. Another is the NASA's Earth Data Search, which has you know a number of different filters, filtering options and kind of ways of refining the location and temporal bounds of of the data that you're interested in using. And so you can this this UH provides, you know, the link here will link directly to that Earth Earth Data Search where you can just type in ecostress and a number of different ecostress products will become available so include not including lands surface temperature and amissivity, as well as other of the products that are derived like I mentioned, such as evappo transpiration.

Speaker 1: And then the subset tool known as appeers is also one way to obtain ecostress data both over areas as well as in point sample locations, and there are also

Speaker 1: a video available demonstrating how to access eco data with the peers. So in Part two, I did use appeers for the first demo and we went over very briefly because there's a more extensive video already available that we will link to that provides kind of a step by step demonstration of how to use a peers for obtaining ecostress data. And then finally Google Earth Engine, which will be the primary focus of Part two, the second demo of Part two, which Glenn, I don't know if you want to say any words about that right now, but you know, stay tuned for all participants for next for next time to to do a deeper dive there.

Speaker 1: But yeah, Glenn, I'll let you maybe just say a few words now just to let folks know how that will work, you mean.

Speaker 2: Part two right out via the tutorial training.

Speaker 1: Yeah, yeah, the next week yeah.

Speaker 3: So we'll be going through a tutorial on how to what we call thermal sharp sharpen the data, which means we take our observed land surface temperature at our native resolution and then actually sharpen it or downscale it to higher spatial resolutions that are more similar to what we get from visible shortwave infrared data and products, which assreciate about the ten to thirty meter scale. So we'll be going through a demo of actually how to do that within Google Earth Engine, and then we'll also go through a demo on how to actually visualize the data within a GIS software software platform called GIS.

Speaker 3: So we'll be going through some of the physics of how we do that, the modeling aspects, and actually step you through step by step one on how we can do that.

Speaker 2: Within Google and Change.

Speaker 1: Great, So for now we'll leave. We'll leave you know, question three as is we are. We will provide you know, we will publish this question and answer with document with the links available, and you know, after this session we will kind of also flesh out some of our answers for them to be a bit more complete and comprehensive, so look out for that, and then also stay tuned for Part two, where we will provide some live demonstrations of accessing ecosource data. All right, Question four, how are land service temperature data products validated?

Speaker 1: So this is a great question, and we'll start by talking about ground based valalidation approaches and then we'll talk about airborne campaigns which are also used. So from the ground, we use thermal radiometers that measure surface leaving radiance, so and then that's converted to surface temperature, and then we compare those temperatures measured on the ground to the satellite value at the exact same time and location. So one example of a validation site that is routinely used for land surface satellite based land surface temperature validations is at Lake Tahoe in California, Nevada and the United States.

Speaker 1: And that is a site that's been in operation for uh several decades now and provides and and and the focus on water bodies and partainicular for validation is useful because water bodies are reliable in that their emissivity is very stable and known a priori isolating atmospheric correction accuracy. So that is one example of one such site we will link to in the published Q and a document to a website hosted at NASA Jet Propulsion Lab which provides an overview of some of the most commonly used sites that NASA uses for validating satellite based LANDSERFCE temperature.

Speaker 1: And so that can be a place for folks to look further into if they want more information about the validation network that is ground based. And then the other kind of major way that we validate and compare satellite based lancer force temperature is through high resolution airborne campaigns. So one instrument which we mentioned today in in the slides is high tests, which Glenn has actually been recently I think working this year on on on a a new the newest validation campaign I believe, and Glynn, if you'd like to speak to I know you said that that work is in the process of publication, but maybe at some point we can link that study as well as other re existing studies that have been done with specifically with high tests, for folks to learn more about that instrument and see how how those kinds of validations are are done.

Speaker 2: Yeah.

Speaker 3: So the we have this hyperspectral airborne sensey called high tests that we fly out of JPL and it can be also used indirectly to validate and to evaluate the satellide sensors. So, for example, if we fly high tests during or near an overpass of Ecostress, for example over Los Angeles, we can directly match up the high test pixels to the Ecostress pixels and upscale them and then use that indirectly as a way to validate, to we even calibrate the sensor that is flying over us at that time. In general with high tests, actually we use it mostly just for algorithm analysis, and it's a hyperspectral sensors, so we can convolve the wavelengths to any other sensors spectral resolution, which is a really powerful way to test and validate various algorithms.

Speaker 3: So it has a multi use type of instrument, and it's also used for actually science on the ground as well, so it has multiple uses.

Speaker 1: Thank you. Let's go to question five here, I understand since thermal sensing relies on naturally emitted weak thermal radiation, the sensors have a coarser resolution to capture thermal energy effectively. Is NASA or the European Space Agency working towards making thermal imagery sensor resolution even higher? Than the existing ones. What level of resolutions achievable in the near future since urban heat island and local heat hotspot studies require much higher resolution, so this is an exciting question, and yes, the short answers that both NASA and and ISSA are actively pushing towards finer resolution thermal observations, with the next generation of agency missions delivering you know, roughly fifty sixty meter thermal infrared data within day to three day revisits, and as Glenn mentioned, airborne thermal instruments provide even finer spatial resolution data as well as higher spectral resolution, which can be very useful for certain applications if you know the fifty ish meter range is not meeting the needs of a given application.

Speaker 1: And the second part to this answer is that longer wavelengths in the thermal infrared require larger optical detector elements to collect enough photons and cryo cooling adds mass, power and cost, so this has historically kept thermal infrared sensors or satellites at sixty two hundred meters, with the example of aster Ecostress landst versus ten to thirty meters, which is the visible Nearonfred sensors. Is there anything you'd like to add to that, Glenn.

Speaker 2: Yeah.

Speaker 3: In fact, the ISA future censor called LSTM, they've actually pushed that down to fifty meter resolution, so that should probably be the highest thermal resolution data that we'll ever acquire in space by the end of this decade. So that's pretty exciting, But of course it's highly cost constraints, so going any higher than that just really ramps up the cost in terms of the size of the telescope needed, the fry cooling, et cetera, et cetera. So it's really cost cost driven.

Speaker 1: Yeah. Yeah, So fifty meters is what we have, that what we can is in the foreseeable future of this decade in terms of the spatial resolution, which kind of emisity retrieval algorithm will these new mission and use. So for my understanding, and I you know, I'm not I'm not as familiar with with TRISHNA and LSTM and Glenn, maybe maybe you have been to some of these meetings and can speak to that, But for my understanding, the temperature emissivity separation retrievals, which you know, as Glenn mentioned, requires multiple multiple thermal bands to be able to really work and the more bands we have, the better the estimates are.

Speaker 1: And so I from my understanding, all of these new missions are going to be focused on able to use tes retrievals or test retrievals rather than relying on the split window approach. Okay, and then Glenn, I actually didn't say that you added this part. So do you want to speak to the LSTM and other missions?

Speaker 2: Sure? So, yeah, I'm going to go back up to the question.

Speaker 1: Yeah, question six. Yeah, maybe justin actually at this point, but yeah, go ahead.

Speaker 3: Yes, there are other algorithms being explored that we actually have been working with directly with those teams on LSTM and Trishna is the other French Indian mission that will actually be.

Speaker 2: Going up within a year, I believe.

Speaker 3: But we are exploring other algorithms that are becoming more modernized, if you will. One is called optimal estimation, and it's a basin theory algorithm where given our existing climatology, if you will, or existing measurements of emissivities and temperature over the past twenty years that we've already acquired, we can utilize that knowledge, that existing knowledge that we have of the surface and use that in a more of a constrained retrieval where we can start off with our first gas and then pivot off of that first gas based on what our measurement is telling us as well.

Speaker 3: And so some of these are being explored and they're a little bit more stable than the existing algorithms, but the word is still at out and whether they will actually be more accurate because they haven't actually been implemented operationally yet with any current algorithms.

Speaker 1: Great, thank you. Question seven, how do these data sets differ from landstuh data sets which are already available from USGS, so by these days that I'm guessing maybe these questions referring to ecostress data and how they how it differs from LANDSAT. So both the spatial resolution and the temporal resolution and the overpast time are the key differences between Ecostress in LANDSAT, And so I'll just highlight that Slide thirty six of the presentation that was given today provides a you know, a table where different instruments including LANCEAT and ecostress are compared.

Speaker 1: But you know, the brief The brief answer is that LANSAT has both larger spatial resolution of one hundred meter native resolution per pixel and then also a longer repeat time and then observation location time of day in the morning ten thirty am, whereas Ecostress has a seventy meter pixel resolution has a varying temporal repeat pattern based on the orbit of the International Space Station that is mounted on and so for in any given location once every three to five days, and then also the time of day of an Ecostress overpass also differs.

Speaker 1: And so those are I think the key differences. Question eight, how can we validate the thermal infrared remote sensing drive lands Service temperature data if we don't have in situ data.

Speaker 1: So one of the things that we mentioned so if you if you don't have ground based thermal radiometer available for validation, there are ways to compare other land surface temperature results over the same time and place. And if that and if those sensors are then validated to other ground based networks and others, then that is a way of doing a validation right, So for example, you could look at ecostress versus landsat and how they compare and then indirectly validate those Lancet LST from previously validated Ecostress measurements.

Speaker 1: Is there anything you'd like to add, Glenn to that.

Speaker 3: That's pretty much the any way I can think that you could do it indirectly, but you really need direct, ground based, independent set of measurements to validate the space of one data.

Speaker 1: Question nine asks. In part two, we are going to learn to downscale native seventy meter ecostress LST to ten meter using a random force model for a neighborhood level urban heat vulnerability analysis in London. Do you recommend use the ten meter downscaled product or is it safer to stick to the native seventy meter resolution to avoid introducing model artifacts when aggregating to neighborhood polygons and Glenn, I'll let you maybe speak to that if I see you. You've written on an answer here, but maybe yeah, I'll let you speak to that.

Speaker 3: Yeah. Well, this will really be eliminated in the tutorial that I'm going to give on the sharpening tutorial, because what you can see at ten meters versus seventy meters is just an incredible a lot amount of detail. For example, you can see what we call street level temperatures, so you can see you down through the scale of a rooftop or a small street. You can see differences in temperatures of a stand of trees on a street, for example. At seventy meters, it's a lot harder to discriminate.

Speaker 2: You can kind of.

Speaker 3: See blobs of cool and warm temperatures in the urban environment. But again, you know that the one thing to keep in mind again is that we're modeling that ten meters. We're not directly observing it from space or from able data, so there is an inherent uncertainty associated with that, you know, ten meter downscaling or thermal sharpening. So it really depends on your application.

Speaker 2: If you want, if you need temperatures at a certain.

Speaker 3: Level of accuracy, I would probably stick with the seventy meter data. If you want to, you know, better discern at the temperature of one rooftop from another, you know, then I'll go with a ten meter scale resolution. But we do provide in the thermal sharpening. We actually do provide an output of the accuracy estimate of the thermal sharpened data, and that's included within the Google Earth package. And oh yeah, the other thing is that we're actually not technically making up temperatures because if we apply this radiometric normalization that we call it.

Speaker 3: Whether if you averaged up the individual ten meter or twenty and thirty meter components to the original seventy meter data, you would still end up with the same temperature, So we're not modifying that original, original retrieved temperature.

Speaker 1: Thank you. Yeah, So let's and you know, after part two, I think there might there will be opportunity for more discussion, you know, along along these lines of questions about you know, particular applications and then thinking about, yeah, considerations of accuracy and suitability for particular applications. So thanks for that question, and we'll move to question ten. How can we select the appropriate remote sensing data for our LST based studies considering the trade off between temporal and spatial resolution within complex heterogeneous urban cities, And yeah, I think that really does come down to what your application requires in terms of spatial resolution versus temporal resolution.

Speaker 1: And then also, as I mentioned, you know, time of day, right, so if you want observations at night, for example, ecostress, you know as of now, really is the sensor that will allow for delivering that for various times of day? Right? Yeah? And then modus on the other hand, which provides a daily repeat, which is more a higher repeat than ecostress, but has comes at a cost of lower spatial resolution of one kilometer, So it varies on what depending on what you want to prioritize. Question eleven. Can land service temperature be used on a small scale to see underground structures?

Speaker 1: For example, due to the differences in temperature between normal soil and soil that has stone or marble under it near the surface. Right, So if there are different temperatures that you know are due to the different materials a different substrate, then that will be detected by the sensor. But right, you won't be able to necessarily determine that there's a structure there. It won't tell you what you know, what is causing that difference in temperature, But if there is a gradient in temperature, then that gradient will be detected within you know, the accuracy that we have of the sensor.

Speaker 1: So that's the short answer to that, I think Question twelve, Oh, yeah.

Speaker 3: Did you want to add Yeah, Just just a quick example would be if you have a thicker slab of concrete or asphalt on a road, we could tell the difference in the temperature because the thickert slab would have a higher thermal mass and higher heat storage and stare the temperatures store that energy.

Speaker 2: Over the longer time periods.

Speaker 3: And so for example, at the nighttime data, we'd be able to see that as an increased temperature of that thicker slab. And so, you know, while we wouldn't know exactly, you know what the reason is, we could infer that, well, this area of the concrete or asphalt is lovelock likely thicker than than as for.

Speaker 4: Example, Yeah, great, let's go to question twelve, where.

Speaker 1: The participant asks could you repeat the example he mentioned regarding the extent to which incorrect emmissivity values can effect of resulting LST value. So the example we gave was the Mona Loa Caldera, which in which features basaltic lava flows that have variable and spectrally complex emissivities in the eight to twelve window. So and in some of these bands that emissivity goes below point nine. The graph which we will we will refer to the specific slide number that showed this, but there was a ten percent emissivity difference in that graph in the eleven micron band.

Speaker 1: When comparing the standard level to modus split window approach to the temperature and emissivity the test algorithm approach from that same modus input. So the split window algorithm or treated a temperature of three hundred and ten calvin, and that was approximately fourteen degrees for fourteen calvin cooler than the tests algorithm. And so the apples to apples comparison here is just is the modus input with the test algorithm versus the modus input with this slit window algorithm. The graph also showed the aster tests algorithm.

Speaker 1: But if you're really if you're wanting to think about kind of the comparison between the two algorithms, that those are the two that you would want to compare. So the rule of thumb also on the following side that we showed was basically just saying that that there's about one point five percent amissivity error that propagates to roughly one calvin of LST error. So that means that over rocky, sandy and geologically diverse surfaces, split window errors can reach three to ten calvin. For applications such as urban heat island analysis and ecosystem thermal stress, where you know even one to two calvin differences are meaningful.

Speaker 1: This you know is a really important consideration is what is the massivity retrieval algorithm that's being used and what does that mean in terms of the error that's propagated into the land surface temperature product. Great, let's move into question thirteen. How well does the ecostress temperature and massivity separation algorithm perform over highly hydrogen as complex urban materials. So, Glenn, would you like to get started with that.

Speaker 3: Sure, it's actually a good question because it's something that we're still working on to improve and it gets into a little bit of the nitty gritty of what the algorithm actually does. So it's a little tough to explain in words without actually looking at the ated or other documents. But the test algorithm works on using what we call an emissivity calibration curve that is based on a set of lab spectra that we measure in the lab of various types of natural.

Speaker 2: And also some man made surfaces.

Speaker 3: And based on that calibration curve, we can use that as a constraint to separate the temperature and emssility components from the thermal measurement. Now, the problem is that over urban areas that consist of mostly man made materials, think of metals, different types of brick, asphole, et cetera. And so the calibration curve is not directly optimized for those man made materials, and so we've been talking in the future of maybe deriving and calibration curve that is optimized only for urban materials.

Speaker 3: And we've applied only that calibration curve over cities and apply the regular calibration curve over natural materials. But the issue that arises there is that you start developing artifacts where you have to classify an area as urban FIR and then apply that part of the cove and then appy.

Speaker 2: Different codes to national materials.

Speaker 3: So you could end up with some strange artifacts coming through in the in the retrials.

Speaker 2: But it's definitely something that we're working on to prove.

Speaker 1: Thank you question fourteen is. And actually before we move on, I will also just mention that we will link to what Glenn. Glenn mentioned the ATBD. That's the Algorithm Theoretical Basis document which describes the algorithms that are used, and so that that will be linked to in the question for further reading, and it's it's also linked in our resources slide for this training Question fourteen asks is there a straightforward way to determine the smallest temperature change that can be detected across an urban environment with known emissivity.

Speaker 1: So I'll start and then maybe Glenn, if there's anything you want to add, we can you can work work at that question that way, because I think this is a really interesting question, and from my understanding, the kind of floor of this is set by the sensors noise equivalent delta temperature, So for ecostress, that's under point three kelvin per pixel with emissivity known in atmospheric correction handled, we can you know, detect single pixel pixel temperature differences of you know, about zero point three or point five calvin and I'll have to verify the exact number of that, but it's you know, less than one Calvin and then you can do meaningfully better through spatial or temporal averaging.

Speaker 1: So that's Glenn. I don't know if there's more that you'd want to add to that.

Speaker 3: Yeah, you're pretty much got it right. It all depends on the actual accuracy of the measurement itself, right, which is constrained by both the radiometric noise of the sensor, and then the actual retrieval algorithm error, which is the model that takes us from the thermal measurement down to the lands of temperature. So if you combine all those errors together, that is your precision of your measurements. So typically we strive for an accuracy of one degree or less. Often it's much higher than that.

Speaker 3: You know, we can get tense of a degree accuracy, but on average, you know, in general we want a degree or less. And so if you have two pixels next to each other, maybe with a very similar type of material, and they vary within that one degree level, then it would be very difficult.

Speaker 2: For us to determine those differences.

Speaker 3: So it really all depends on the actual accuracy of a measurement at any given time.

Speaker 2: And we actually do provide an.

Speaker 3: Uncertainty estimate, and this is a pixel by pixel based a certainty estimate of our retrievals for the ecostress products or the landside products. And given that uncertainty estimate, you can you know, convince yourself either way of whether you believe.

Speaker 2: The accuracy of of any given pixel.

Speaker 1: Yeah, thank you. Question fifteen asks would LST be kinetic temperature or radiometric temperature? And so let's go back to you know, the earlier slides of this of this training where we learned that true kinetic temperature is the you know, is the actual temperature of the surface, right, and so thermal instruments that we use don't measure this directly, but LST is intended to recover kinetic temperature by applying the missivity correction that converts the radiometric measurement into the physical temperature.

Speaker 1: So conceptually that retrieval pipeline would be the measured radiance to the brightness temperature, to the radiometric temperature which is the integrated you know, multi band measurement, and then the kinetic temperature, which is the emissivity corrected LST product that we know and that we will be working with in part two of this training. Yeah, so I think I'll just from the interest of time, move on, but we will we will flesh out some of these answers in more detail and kind of link back to specific slides where we're where we went over this material.

Speaker 1: Question sixteen asks is there a way to distinguish between land surface temperature and the temperature of the air near the land surface? Great? So, yes, we in fact in the Lake Tahoe site that I mentioned. We actually have instruments, you know, designed designed to do this, where we have a radiometer measuring the surface, the surface of the of the water the skin temperature, whereas we also have temperature probes at various depths of the water measuring bulk temperature. And then there's also an air temperature measure that we that we use two that we include in the in the validation site.

Speaker 1: And is there a way to distinguish Glenn, maybe do you want to maybe add that, I guess I'm not really sure about the distinguishing part between the answer temperature.

Speaker 3: We can measure that the differences between the two and they are linked, but they can also differ because various different types of circumstances. For example, if there's wind blowing or urine diirect sunlight versus in the shade, the two types of temperatures can be quite different. Also, think of on a really hot day, the surface is going to heat up a lot quicker than the air temperature above it, and that's mostly due to the different thermal mass of air versus the ground. And this is actually one of the problems that we deal with a lot is how do we derive the air temperature, which is, you know, typically what we feel as humans walking around, and which is what city planners and decision makers really want to know.

Speaker 3: That that value more than just the skin surface temperature. And so we actually spend a lot of time modeling the air temperature from the land surface temperature. And we can do that using empirical models or land surface models, and it's typically has a pretty high incertainty because it's it's a difficult it's it's a difficult modeling exercise.

Speaker 2: Think of in an urban environ.

Speaker 3: There's a lot of different things going on, but that, in a sense is the temperature that we really want to get to in terms of burden studies.

Speaker 2: Is the air temperature.

Speaker 1: Yeah, and maybe we could actually just think about an example here in an urban setting, for instance, a parking lot that might show for instance, sixty degrees celsius LST, while the air temperature you know, two meters above might be as low as thirty five degrees celsius. And you know, both of these are correct, and they're just measuring different things. Where the LST is the you know, dependent on the surface property itself, the albedo, the heat capacity, the moisture and will vary you know, a lot across land cover types within a city, and that's something that we'll look into, you know, in part two and the demos, and we'll at green spaces, specifically green spaces that have access to water or you know, one of them is actually around a reservoir area and the other is a park a nature preserve that has access to a sufficient water that the vegetation is able to basically stay cool with evaporative cooling, whereas you know, the air temperature would integrate heat flexes from the surface over the surrounding boundary layer, and then that's going to be kind of a more spatially smooth by atmospheric mixing than the LST itself.

Speaker 1: Hopefully that's that helps, And I think maybe after this session we can also kind of maybe provide a few examples from the literature that look into kind of that problem that Glenn was mentioning about how to kind of model air temperature. I think that might be nice for us to provide a few a couple examples of that to give some military illustration.

Speaker 1: All right, but yeah, great question. Let's go into question seventeen. Given the known uncertainties in retrieving near surface temperature, humidity, and radiative fluxes from satellite data, especially due to surface emissivity, selective gas absorption, and cloud radiation interactions. What approaches are NAS emissions exploring to better capture the combined human relevant exposure burden in dense environments. So I would say this is sounds like kind of an active, large area of multiple maybe sub questions are problems within this bigger question.

Speaker 1: And so NASA's strategy combines multimission earth observations, right, so there's an integration with reanalysis and institute networks and physiologically relevant indices that go beyond temperature alone. So I would say the general direction is towards actionable heat exposure products that integrate temperature, humidity, radiation, air quality, you know, all of these various factors at urban relevant spatial scales. And I would say that integration and consideration of these different

Speaker 1: these different factors is something that is really important and something that we, I would say, have a lot of opportunity to hone and refine and perhaps in future trainings could could go more into because certainly for this training we're really focused on land service temperature itself as kind of the core measurement that we're considering. But I appreciate the you know, the the inquiry in the invitation, and I'm not sure, Glenn, if you might have anything to add at this time, but this is a question that I'd like to maybe circle back to also and maybe yeah, think think more about and once we publish the Q and a document, we can update what we provide here.

Speaker 1: But Glenn, do you have anything off off the cuff to say about that?

Speaker 3: Yeah, I mean it's a pretty loaded question, but a good one. There are no single mass emissions that are directly, you know, aimed at trying to answer, you know, what is the human relevant exposure about and in urban environments but specifically, but there are different groups and different research groups that are coming up with the diferent kinds of models that are able to explain what that human relevant temperature is, you know, in urban environments for example.

Speaker 2: And one of those.

Speaker 3: Is the UTCI or the Urban Thermal Comfort Index, which has been published in the liverature in several different research groups right now. But it combines different variables that explain the physiological temperature that we experience as humans. So the air temperature, combines, humidity of course, direct thermal radiation, wind speed, and various other other factors. And you know all of those inputs are you know a lot of those are provided by by NASA through different missions, And so I guess the NASA's mission here would be to provide the highest resolution input data that we possibly can that are used in these models.

Speaker 1: Yeah, thank you. Question eighteen asks could you please explain the physics behind the difference between split window and the tests algorithm? So maybe I'll provide a brief summary in Glenn. If you want to do a deeper dive, that's great. But basically, the fundamental difference is that the differences in what each algorithm is assuming on the onset. So the split window algorithm assumes an emissivity is known based on a land cover classification lookup table, and it uses the brightness temperature difference between two channels to correct for atmospheric water vapor.

Speaker 1: Meanwhile, the test algorithm assumes nothing about amissivity to start with, and uses at least three or more channels thermal channels plus an empirical spectral construt to solve for the emissivity and temperature at the same time. And Glenn, if you want to provide a more in depth explanation, that would be great. We can also circle back to it and in our write up of this final document before we publish it to the website, we can do we can add in the details. So up to you on how in depth you want to go for now.

Speaker 3: Yeah, I mean you could have an entire seminar just on this very topic, so I won't go into too much more depth then you just explained. But split window algorithms generally are employed with the sensor that has two thermal bands, and so the two bands are used two different parts of the thermal window to account for the atmospheric absorption between those two bands, and then use an assumed emissivity to derive the temperature. Whereas the test algorithm, as you said, is able to do that all in one go.

Speaker 3: We're able to to both retrieve the temperature and the emissivity. That it requires three or more bands, and that is the constraint that is needed to apply the test algorithm.

Speaker 1: Great. Thanks, and we'll link to again the algorithm theoretical basis document and other documents that are relevant for kind of a deeper dive into understanding how these algorithms work. Question nineteen, Will we need qtas, Google Earth Engine or are slash our studio for part two? The answer is yes if you if you would like to follow along with the with the two demos that will be that will happen in part two, participants will need qjas, Google Earth Engine and are our studio for part two. So all of these are available and between now and next week in part two, I highly encourage participants who maybe had don't have those set up to get those set up if you would like to follow along with the demos.

Speaker 1: Question twenty. Given that diseases such as epola, hendivirus and these COVID nineteen and NIPA are strongly influenced by near surface temperature, humidity, aerosol load, and air quality interactions, what steps are NAS emissions taking to improve retrievals of these human relevant ground level exposure parameters, especially where uncertainties and emissivity, selective gas absorption, and cloud radiation interactions limit current satellite products. Yeah, so I think this is maybe a similar kind of big picture question that we were that we spoke to earlier, where you know, all of these I would say, uh, you know, retrievals of these different variables that that we know, you know, matter in disease outbreaks, and you know that is one very important application of NASA data.

Speaker 1: So so this this training in particular is focused on surface temperature. Right, there are other trainings that our set provides that look into other data products that you know. For instance, our set has an entire thematic area focused on air quality, and so I would suggest to the participant who asked this question to start by looking at these different thematic areas and that's how you know, this information is currently organized where we systematically go through these different thematic areas and provide trainings to address these yeah, different different factors that are that are that together influence complex systems such as disease outbreak.

Speaker 1: So thank you for the work that you're doing on this important application. And I'm glad to hear that that land service temperature you know, is one of the data products that you're currently using in your work and that you're invent that you're invested in wanting to continually improve those those estimates. So that's that's great to hear, and thank you for emphasizing that. Glenn, is there anything that you'd like to add to this question.

Speaker 2: No, that that was a good answer.

Speaker 3: I mean, the one thing to keep in mind is that NASA cannot produce high resolution near surface temperature and humidity right now directly from satellite data.

Speaker 2: These are all sort.

Speaker 3: Of model derived variables that are usually provided from numerical way, the models and so forth, and it's really difficult to get those values, you know, down at the sub kilometer scale right now, just because of the difficulty in retrieving nia surface said temperature and you know, humanity measurements.

Speaker 2: So I would.

Speaker 3: Say that it's more more of the modeling communities, either for the NASA, GMAO or UK me MET office that are striving to get to get us down to that sort of very fine scale surface temperature, humility measurements to be able to answer some of those questions.

Speaker 1: Yeah, thank you. So question twenty one will be the last question that we will answer live. The remaining questions that we didn't have time to get to live will be answered offline and we'll share We will share this Q and a document on our training page within one week of the training. So question twenty one asks is ecostress data available globally, how do we access as data? So for the data access question, I've pasted you know some of the links that we included in the earlier question and the ECO and for the first part of the question of coverage, Ecostress is near global, but due to the orbit of the International Space Station, it includes plus or minus fifty one point six degrees latitude.

Speaker 1: So basically the polls are not included are not covered by Ecostress.

Speaker 1: So with that we will conclude today's live Q and A session. Thank you all so much for your participation and engagement, and we look forward to seeing you next week at part two of our training series. Thank you all and we will see you next week

The Story