First JPL Workshop on Remote Sensing of Land surface Emissivity

--Andrew Morrison (andy@lithos.jpl.nasa.gov), Jet Propulsion Laboratory, CA

Summary

A very successful First JPL Workshop on Remote Sensing of Land Surface Emissivity was held at the Jet Propulsion Laboratory May 6-8, 1997. The workshop was organized by Anne Kahle and Dave Nichols of JPL, and Alan Gillespie of the University of Washington. Over fifty remote-sensing specialists from universities, government laboratories, and industry in the United States, France, the Netherlands, China, and Japan participated. The objectives of the workshop were to provide a comprehensive assessment of the past, current, and future technologies for thermal infrared (TIR) remote sensing, to identify current and potential uses for data in this region of the spectrum, and to identify the steps necessary to realize these objectives. Scheduled sessions addressed instrumentation, applications, and temperature-emissivity extraction. There was a great deal of information exchange both inside and outside of the technical sessions. An informal evening session covered a wide range of topics, including the trade-offs between pursuing new technologies (the 3-5-µm range, thermal and spatial resolution, algorithm development, etc.) or new science objectives. The group unanimously supported a concept introduced by Jeff Myers of Ames that NASA should fly the Aerospace Corp. high-resolution hyperspectral Spectrally Enhanced Broad-band Array Spectrograph System (SEBASS) instrument over sites of investigators to be selected via a NASA Research Announcement (NRA). Anne Kahle (JPL), Alan Gillespie (University of Washington), and Jeff Myers and John Hackwell (Aerospace) will follow up on the concept. The next workshop will be held in the summer of 1999.

Narrative Minutes

Diane Evans, representing JPL, welcomed the attendees to the Laboratory. She noted that the ASTER Project continues a long history of TIR remote sensing at JPL. She said that two objectives of the workshop are to understand the broad community's needs in thermal infrared sensing and to look for cooperative ventures to follow ASTER.

Overview Session

Alan Gillespie noted that this workshop is intended to complement other TIR workshops. He said that the rationale for holding this workshop at this time was the advent of new detectors/sensor systems, the upcoming launch of ASTER and MODIS, and the need to plan for the next decade in TIR remote sensing. He reviewed the special problems involved in land surface TIR sensing.

Anne Kahle presented an overview of thermal infrared remote sensing including:

• fundamentals such as the process of deriving the properties of surface temperature and spectral emissivity from the observed variable of spectral radiance;

• commercial and operational applications;

• science applications; and

• a survey of airborne and spaceborne multispectral thermal infrared instruments.

She emphasized the importance of letting the technology continue to improve to meet the demands of a new user community.

Alan Gillespie presented an overview of temperature-emissivity (T-E) extraction techniques. He pointed out that T-E extraction from remotely-sensed data is based on the assumption of homogeneous, isothermal, flat pixels, but that the assumptions are commonly violated. He followed with methods to obtain temperatures and emissivities as a function of the number of bands of data collected. He reviewed the multispectral and hyperspectral algorithms used to derive temperature and emissivity and noted that the large number of algorithms points out the artfulness needed to solve the problem. He also discussed the kinds of non-quantitative information that can be obtained from multispectral thermal infrared data that can be used to satisfy several different kinds of science communities without providing precise or accurate temperatures.

Stillman Chase presented an overview of thermal infrared technology. He talked about desirable TIR detector attributes and then described the characteristics of the various detectors available in the market today. His presentation emphasized the new silicon microbolometers produced by Santa Barbara Research Center (now SBRS Santa Barbara Remote Sensing) and Amber (both Raytheon subsidiaries).

Instrumentation Session

Sarath Gunapala (15 µm 128x128 GaAs/AlxGa1-xAs Quantum Well Infrared Photodetector (QWIP) Focal Plane Array (FPA) Camera--Gunapala and Bandara) spoke about the status and characteristics of the FPA Camera. He noted that four QWIP cameras have been sold to date by the manufacturer (Inframetrics). In addition, Inframetrics hopes to soon have available their new "palm-sized" QWIP camera.

Ed Blazejewski (Application of QWIP FPA Technology to the Integrated Multispectral Atmospheric Sounder (IMAS) - Blazejewski) noted that IMAS, which is in its formative stages, is currently viewed as the next-generation flight instrument after AIRS. He described the IMAS instrument and mission concepts, goals, and the instrument development approach and status.

A demonstration of the QWIP camera preceded the lunch break.

Anne Kahle (ASTER and Beyond--Kahle and Nichols) described the history, design, and characteristics of the ASTER instrument, and she presented an overview of the ASTER mission. This included operational constraints on instrument usage and the plan for allocation of instrument viewing resources. ASTER will be scheduled to collect an all--radiometer global data set, data for regional monitoring, and also targeted data for individual investigators. She then addressed the future for multispectral thermal infrared, noting that there are no TIR instruments scheduled to be launched after ASTER. She described Sacagawea and Sacagawea Light, concepts for compact high-resolution ASTER multispectral thermal infrared follow-ons being proposed by JPL.

Peter Kealy (ESA's Candidate Earth Explorer Mission for Land-Surface Processes: Its Thermal Capabilities„Kealy, Caselles, Coll, Rubio, and Valor) presented ESA's basic objectives and then explained that the Earth Explorer Mission is one of two missions planned to achieve ESA's objectives. He presented the mission requirements and instrument specifications „ TIR channels at 8.45 and 8.95 µm will each have 0.5-µm bandwidth, T_min of 212 K, T_max of 360 K, and NEdeltaT@300 K of 0.1 K. He also presented the split window algorithm that will be used to extract land surface temperature and the Vegetation Cover Method that will be used to estimate emissivity. He concluded with the schedule from Phase A kick-off at the beginning of '98 through launch at the end of 2003.

Carl Schueler (Advanced Multispectral TIR Land Imaging Sensor Concepts--Schueler and Blasius) described and compared the technologies and capabilities of two sensor concepts, multispectral uncooled microbolometers in a Pushbroom Imager vs. a hyperspectral Wedge Imaging Spectrometer (WIS). His conclusions were:

• sensors based on new uncooled microbolometer arrays promise ~0.2-0.6 K NEdeltaT at ~10% delta lambda/lambda, for Landsat-like spatial resolution, 60-200 m, by heavily exploiting Time Delayed Integration (T delta lambda) (40 samples merged);

• cooled HgCdTe detectors offer NEdeltaT ~1.5% delta lambda/lambda for 60-m spatial resolution with no Tdelta lambda (potential is there); and

• microbolometer-based sensors will offer size/mass/power/cost savings for missions with lower spatial /spectral resolution requirements.

Simon Hook (Synergy of Active and Passive Airborne Thermal Infrared Systems for Surface Composition Mapping„Hook, Cudahy, Kahle, and Whitborne) described work collecting and comparing data from the Thermal Infrared Multispectral Scanner (TIMS) and the Australian Mid-infrared Airborne CO₂ Laser Spectrometer (MIRACO₂LAS)). Data were collected over Mt. Fitton, South Australia, and processed to provide surface composition information (emissivity in TIMS and reflectivity in MIRACO₂LAS). The results compared well with field emission spectra and laboratory measurements. The authors conclude that future thermal infrared systems designed for geologic mapping should include both an imager for mapping units and a high-spectral-resolution profiler for mineral identification.

John Hackwell (LWIR/MWIR Imaging Hyperspectral Sensor for Airborne and Ground-based Remote Sensing„Hackwell, Warren, et al.) presented the Aerospace Corporation-sponsored SEBASS instrument history, design, and characteristics. The design goals (all achieved) of this line-scanner instrument included:

• Operate in 2.1-to-5.2-µm and 7.8-to-13.5-µm bands
  • Observe entire spectral range at once
  • Resolving power lambda/delta lambda of 200

• Highest possible sensitivity

• Operable from aircraft or ground-based platform

• Ground sample distance 0.5 - 3 meters from 15,000.
  • 10,000 ft AGL

The SEBASS data cube is characterized by:

• 128 pixels cross-scan

• 6000 lines in-scan (maximum)

• 256 wavelength bands.
  • 128 in LWIR: 7.8 - 13.5 µm
  • 128 in MWIR: 2.1 - 5.2 µm

He also presented data showing the optical performance of the instrument and the spectral resolution of the LWIR and MWIR channels and median Noise Equivalent Spectral Radiance (NESR) of the two arrays (NESR <10⁶ W/cm²/ster/µm at 120 Hz [8 msec]).

Anu Bowman (Hyperspectral Mine Detection„Lucey, Williams, Julian, Kokobun, Stocker, Kendall, Schaff, Winter, Schlangen, Batik, and Bowman) talked about a Defense Advanced Research Projects Agency (DARPA) -sponsored two-phase effort to detect buried land mines. Phase 1 was a non-imaging phenomenology investigation and Phase 2 was an empirical study of remote detection of buried land mines. The empirical effort demonstrated that under some conditions, the ability of the hyperspectral sensor to distinguish disturbed soils could be used as a tool to locate buried land mines.

Applications Session

Frank Palluconi (Thermal Infrared Atmospheric Correction--Palluconi, Thompson, Alley) described the ASTER thermal infrared subsystem and the TIR atmospheric correction algorithm approach. He listed the atmospheric parameters and the effects of each on derived radiance. He also provided web sources for all of the NOAA data files used by the algorithm.

James Crowley (Death Valley Field Spectra/Image Data--Crowley and Hook) used remote sensing to survey the mineral suites of the evaporites in Death Valley. They were able to characterize the chemistry of the groundwater in the basin by studying the distribution of the various evaporites.

Tsuneo Matsunaga (ASTER Simulator Flight Experiment„Matsunaga, Rokugawa, Tonooka, Kannari, and Kato) presented the results of the 1996 flight experiment at Cuprite. He said that they plan to upgrade the instrument to improve the signal-to-noise ratio and also plan to tune the algorithm to better support the ASTER Airborne Simulator (AAS).

David Ripley (The Application of Atmospheric Corrections to Arbitrary Vegetation and Temperature--Ripley and Carlson) reported that he and T. Carlson have observed that, under certain circumstances, it may be possible to omit the application of atmospheric corrections and substitute temperature and albedo to obtain values of related surface parameters such as surface soil water content and fractional vegetation cover. An example using NOAA-9 AVHRR data was presented.

Mike Ramsey (Monitoring Potential Desertification via Airborne TIR Data: Sediment Transport in the Mojave Desert, California--Ramsey and Christensen) characterized mineral distributions in the Kelso Dune field in the Eastern Mojave and correlated TIMS data with mineral samples analyzed in the laboratory. The analyses confirmed that the dune fields did contain variations observed in the TIMS data, and that the fields were less mature than previously reported. The authors demonstrate that use of multispectral TIR data can provide the geologist with a synoptic look at the entire eolian system and argue that monitoring programs using instruments such as ASTER can benefit studies of dune encroachment and desertification.

Tom Schmugge (Application of TES algorithm to TIMS Data from the HAPEX-Sahel--Schmugge) reported that he tested a version of the Temperature/Emissivity Separation (TES) algorithm on TIMS scenes from the Hydrological and Atmosphere Pilot Experiment (HAPEX)-Sahel experiment. There was excellent reproducibility on data taken on the same day, but with differences noted for data taken on subsequent days„he speculates that this may have been due to changes in soil moisture. Temperatures for the vegetated areas are in good agreement with the air temperature at the time of the flights.

Tong Qingxi (Study on Some Land Surface by Thermal Infrared Multispectral Remote Sensing--Tong) reported on the development and application of the Modular Airborne Imaging Spectrometer (MAIS), described the application of the multispectral TIR system for retrieving temperature and emissivities, and discussed the T-E extraction methods used. He also described the 'ratio-weighted' method for extracting relative emissivity.

Dave Pieri (An Overview of Volcanological Application of Infrared Remote Sensing--Pieri and Realmuto) discussed the various applications of multispectral thermal infrared to volcanological studies including:

• mapping the relative ages of lava flows based on emissivity variations;

• mapping abandoned lava tubes based on detection of alteration near fumaroles or vents;

• mapping high- and low-temperature geothermal features using the VNIR, and the short- and longwave infrared channels; and

• mapping SO2 and ash plumes based on absorption of ground radiance by the plumes.

He pointed out that, for several reasons, remote sensing is particularly well suited to monitoring dynamic volcanic processes.

John Schieldge (Comparison of Surface IR Radiometer and Near Surface Air Temperatures Over a Grass-Covered Field„Schieldge) reported that large-scale surface temperature fluctuations, due to turbulent horizontal eddy motions, could skew the results of TIR data taken during such temperature excursions. He concluded by advising that the effects of large-scale 'inactive' eddies may have to be taken into consideration in algorithms that are used to compute energy fluxes near the Earth's surface.

Kevin Czajkowski (Vegetation and Hydrology Applications„Czajkowski) listed air temperature, drought/soil moisture, and atmospheric moisture as variables that can be estimated using TIR data and land surface temperature. He gave examples of derived Temperature/Spectral Vegetation Index (TVX) and air temperatures. He presented four approaches that have been reported that can be used to estimate atmospheric water vapor: the split window technique, ratio of channel 4 and 5 variances, slope of channel 4 and 5 regression, and use of a radiative transfer model.

Alan Gillespie (Geologic Mapping in the Middle Mountains, Yuma County, Arizona, with SEBASS Hyperspectral TIR Images„Gillespie, Hackwell, Alexander, Alley, Cothern, Grove, Kahle, and Smith) reported on an experiment using the SEBASS hyperspectral TIR instrument to map rock-type distributions in Arizona. His objective was to present a 'sense' of the analytic process supported by SEBASS images. His conclusions included:

• Hyperspectral TIR imaging will allow recovery of emissivity spectra and geologic mapping and rock-type identification.

• Keys are:
  • good NEdeltaT
  • Sufficient delta lambda to resolve emissivity features (e.g., >0.5 µm)
  • in-scene atmospheric corrections (pixel-by-pixel)
  • lack of requirement for registration to Digital Terrain Model (DTM)

X. F. Gu (Short-term Fluctuations of Brightness Temperature Measured with a Thermal Video Camera„Gu, Lagouarde, Seguin, Hanocq, and Prevot) described the characteristics of their thermal video camera (256x250 pixels, FOV 7 degree, 20 degree or 80 degree, 26 images/sec, three spectral channels [8-13, 10-11.5, 10-13 µm], radiometric resolution of 0.1 K, Sterling cooled to 77 K) including a noise evaluation of the camera and the effects of digitization. He then presented data of observed temperature fluctuations of a sorghum canopy and of a forest surface. He reported that they also measured wind speed in four directions at the target and attempted to relate wind fluctuations to observed temperature fluctuations. He concluded that temperature measurements of 40-m resolution are significant, but wind fluctuations must be taken into consideration in satellite dual-angle observation, airborne measurement, and in-field sampling.

Alan Gillespie (Field TIR Imaging Experiments„Bland, Gillespie, and Kahle) used two experiments to relate 'snapshots' of temperature fields (equivalent to images taken from space) to short time series data of phenomena that vary with periods of minutes, etc. Data were taken using JPL's QWIP detector camera. False color displays (chronochromes) were produced. Conclusions drawn from a short time series of data from forests included:

• short-term fluctuations may be due to wind;

• changes due to sensible heat exchange are significant (e.g., 4 K) at the 5-minute time scale;

• short-term variability will have implications for temperature product accuracy (but still may need accurate temperature for emissivity recovery); and

• field imagery may prove useful in quantifying sensible heat exchange from canopies.

• Conclusions from a day/night forest community mapping experiment were:

• day/night DTs for conifers/deciduous stands behave as predicted; and

• deltaT can be used as a basis for community mapping.

Temperature/Emissivity Separation Session

Ken Watson (Three Algorithms to Extract Spectral Emissivity Information from Multispectral Thermal Data and Their Geologic Applications„Watson) is using remote sensing to study the geological units at Joshua Tree National Monument. He presented a historical review of the "two temperature" and "emissivity ratio" algorithms and then described his new ñInverse Waveî algorithm. The use of the new algorithm requires that you use the same algorithm when you analyze laboratory data as well. He showed the effect of removing 1/l, which is one feature of the new algorithm. One potential advantage is the removal of 1/l levels from the spectra without losing the spectral shapes necessary to identify the minerals. He also developed an algorithm for when you have two overlapping flight lines which have the same data from different angles. Finally, he juxtaposed AVIRIS and TIMS data to show the complementarity of the two data sets.

Anu Bowman (Empirical Solutions for Land Surface Temperature Estimation from Thermal Remote Sensing: The Emissivity Factor„Bowman) using a derivative analysis of the radiance equation performed an empirical assessment of J. Salisbury's database in order to determine the role of emissivity in determining surface temperature (her Master's thesis). She concluded that:

• emissivity spectrum is "flat" in portions of the thermal spectrum; and

• direct calculation of surface temperature using laboratory-derived emissivity values is possible.

Alan Gillespie (TES: The EOS/ASTER Temperature/Emissivity Separation Algorithm„Gillespie, Rokugawa, Hook, Matsunaga, and Kahle) presented the basic structure and flow diagram of the algorithm and showed examples of emissivity and temperature recovered from Railroad Valley data by the algorithm. He summarized by saying that the accuracy of TES depends on ASTER's radiometric accuracy and the accuracy of the atmospheric compensation. He also discussed the advantage of sacrificing ASTER's band 10 (as proposed by Simon Hook), the shortest wavelength of the five TIR bands, which is the most error prone due to water vapor. This approach will be adopted in the algorithm. He also said that he is going to propose that for areas where the surface appears to be a black- or graybody, we should assume that we know the emissivity and derive the temperature, and then back-calculate the emissivity.

He noted that ASTER doesn't go out to 12-14 µm, and so TES doesn't specifically address this region. When applied to hyperspectral data, such as SEBASS data, the performance of TES in this region will have to be examined.

Zheng-Ming Wan (A Multi-method Strategy for Remote Sensing of Land-Surface Emissivity„Wan, Snyder, Zhang, Feng, and Li) listed the following four elements of their strategy for remote sensing of land-surface emissivity:

• laboratory measurements

• field measurements

• airborne measurements

• satellite measurements

He described the MODIS day/night method including the underlying physics, assumptions, and new features. He also described the sun/shadow method (similar to the day/night method) including its physics and advantages. He reviewed the MODIS team's validation methods and results (from Death Valley, Railroad Valley, and Mammoth Lake) and their error analysis.

Christopher Borel (Iterative Retrieval of Surface Emissivity and Temperature for a Hyperspectral Sensor„Borel) reviewed the problem of retrieving emissivity and temperature from remotely-sensed data and then presented a new technique to retrieve emissivity spectra from hyperspectral data. Using synthetic data and a thermal model described in his presentation, he generated a series of data cubes. He analyzed the results for temperatures and emissivities using an algorithm which depends on the smoothness of the spectral emissivity and variable temperature (Adaptive Spectrally Smooth E-T Retrieval„ASSETR-deltaT) and a similar algorithm with variable emissivity (ASSETR-de). He demonstrated the effects of unknown atmospheric parameters. All of these simulations were illustrated in a video presentation.

Masao Moriyama (Sensitivity Analysis of Surface Emissivity Estimation„Moriyama) presented a summary of his work to develop a sensitivity of Quality Assurance (QA) error for ASTER pixels by varying the tunable parameters in the ASTER TES algorithm and applying an error-propagating formula to evaluate emissivity uncertainty.

X. F. Gu (Estimation of Spectral and Directional Thermal Emissivity from Multispectral Infrared and Visible Imaging Spectrometer (MIVIS) Data Acquired Over the La Crau-Camargue Region„Gu, Seguin, Guyot, Hanocq, Clastre, and Wang) described data collected by the airborne MIVIS instrument and the radiometric and atmospheric correction applied to it. He showed the process flow for emissivity determination. He also showed the spectral variation of emissivity for different surfaces and the results of normalization of the spectral variation using a body of water. He then showed the relationships between temperature, emissivity, reflectance, and vegetation index for temperature-reflectance, temperature-Soil Adjusted Vegetation Index (SAVI) and emissivity-Normalized Difference Vegetation Index (NDVI).

Open Discussion Session

A special Open Discussion Session was held the evening of the second day of the workshop. The objective of the session was to bring out ideas that might not come out of the more-structured meeting sessions. Discussion topics are summarized here.

Dave Nichols presented a draft revision to the MTPE plan from NASA HQ. It included a proposed list of measurements to meet MTPE scientific priorities. Nichols asked:

• What kind of science do we want to work on for MTPE?

• Is there science we can do that is desired by more than one agency?

• Do we want to proceed with some facility instrument to satisfy a large number of users, or should there be more focused efforts to satisfy individual users?

• Do we want to ask what kind of instrument we want to have flying up there in a couple of years, or do we first want to identify what data we want or need?

Anne Kahle added the question:

• When does an aircraft program make more sense than a spaceborne program?

  John Schott said there is a good chance that Landsat-8 will not have thermal channels, meaning that there won't be any high-resolution thermal instruments flying in the foreseeable future. He noted that industrial plumes, volcano plumes, and calibration should each justify flying thermal instruments„and just thermal; adding visible capabilities, he said, is more difficult and more expensive.

  Anne Kahle asked about the status in this country of registration software and Frank Palluconi replied that the opportunity will be there because AM-1 and Landsat-7 will be flying synchronously, and it will be needed.

  David Ripley urged building broadly capable flight hardware and making it available to the scientific community. He said that this would produce science users. Ed Winter said that HYDICE and AVIRIS are examples of instruments that have exceeded their original promise, and there's nothing like them in the longer wavelengths.

  Zheng-Ming Wan urged including the 3-5-µm range. "It's the future for vegetation, moisture, geology, etc.," he said. He added that spatial resolution is also important. He said that for a single instrument he would choose high spatial resolution. Sig Gerstl agreed about the 3-5-µm range. He said that they just measured fluoride emissions from a volcano using a Fourier Transform Infrared Radiometer (FTIR) in this range. Kevin Czajkowski stressed the more-immediate need for surface temperature. He said that currently there is not a lot going into surface temperature validation and scaling. Now it is needed from AVHRR, and it will be soon needed from ASTER and MODIS.

  John Hackwell asked about the need for a follow-on instrument. Is it that the community will fail if there isn't a Sacagawea? Anne Kahle replied that the future depends strongly on a follow-on. "The funding could dry up without it." She added that hyper-spectral is the wave of the future, and it should be explored to see what opens up.

  Ken Watson said that the USGS is now struggling with the new USGS role. He feels certain that remote sensing will play an important part there, but it has to be demonstrated that it relates to the new USGS problems.

  Jeff Myers proposed an integrated joint ground campaign using current instruments, including flying SEBASS on a NASA aircraft. John Hackwell agreed, saying that we should explore new applications„determining a direction is not a technology issue. Ed Winter also expressed his support for the proposal.

  Simon Hook suggested that investigators should propose to participate in such an activity. He noted that investigators will participate if data are collected over their own sites. He thought that 20-30 investigators might propose to a hyperspectral TIR campaign in the U.S. Southwest.

  The attendees of the workshop unanimously endorsed the concept of a NASA NRA to solicit investigators for an integrated joint air and ground hyper-spectral TIR campaign that would fly SEBASS on a NASA aircraft as well as other multispectral TIR instruments to collect data over the selected investigators' sites. John Schott, Alan Gillespie, Anne Kahle, and Jeff Myers will take the action to pursue this concept with NASA HQ.

  Next Meeting

  It was generally agreed that the next meeting should be held during the summer of 1999, after preliminary data from ASTER and the results of the next ground campaign are received, and that it should be held jointly with Zheng-Ming Wan's Land Surface Temperature Workshop.