The Moderate Resolution Imaging Spectrometer-Nadir (MODIS-N) is one of two MODIS sensors being planned as a facility for the Earth Observing System (Eos). The Eos is scheduled for operation in the late 1990's as a major observing facility to study the Earth as a system. The MODIS-N is a 40-band scanning system designed, principally, to provide observations that would facilitate studies of the interactions between the land and ocean surfaces of the Earth with the atmosphere. At the planned orbital altitude of Eos near 824 kilometers, the swath width of the instrument is presently 1790 kilometers thereby providing nearly complete two-day coverage of the Earth. The spatial resolution of the instrument is 250 and 500 meters for bands designed primarily for land surface processes studies and 1000 meters for bands applied to oceans and atmosphere studies. Thirty-one of the bands are baseline and nine are optional. The inclusion of the optional bands will depend upon the strength of scientific proposals submitted to the Eos Program to make use of the bands for Earth science studies. At present a Request for Proposals (RFP) to perform Phase-B, detailed design studies has been released. Two contractors from private industry will be selected for one-year studies. A single Phase C/D contract will be awarded following the completion of Phase-B studies.

The Moderate-Resolution Imaging Spectrometer (MODIS) remote sensing system is designed to operate on the polar-orbiting platform and provide global surface sensing with high spectral and temporal resolution. MODIS consists of two synergistic sensors with a total of 104 spectral bands in the 0.4 to 14.2 micrometer range. MODIS-T (Tilt) is one of these sensors. It is a 64 channel imaging spectrometer with a required 10 nanometer spectral resolution (FWHM) in the 400 to 1040 nanometer spectral range. The system will have a 1 kilometer IFOV, a wide scan for global coverage in three days, and be capable of being pointed fore and aft of nadir by ±50 degrees. Optical systems analyses and trade-off studies will be discussed with respect to the science requirements. A preliminary baseline MODIS-T optical system design has been chosen based on these studies.

The High-Resolution Imaging Spectrometer (HIRIS) is a JPL facility instrument designed for NASA's Earth Observing System (Eos). It will have 10-nm wide spectral bands from 0.4-2.5 pm at 30 m spatial reso-lution over a 30km swath. The spectral resolution allows identification of many minerals in rocks and soils, important algal pigments in oceans and inland waters, spectral changes associated with plant canopy biochemistry, composition of atmospheric aerosols, and grain size of snow and its contamination by absorb-ing impurities. The bands will have 12-bit quantization over a dynamic range suitable for bright targets, such as snow. For targets of low brightness, such as water bodies, image-motion compensation will allow gains up to a factor of 8 to increase signal-to-noise ratios. The sensor will be able to point ±24° crosstrack and +60°/-30° downtrack. In the 824-km orbit altitude proposed for Eos, the crosstrack pointing capability will allow 4-5 views during a 16-day revisit cycle.

The purpose of this paper is to briefly describe the operational characteristics of the Airborne Bathymetric System Multispectral scanner (ABS) and the Airborne Multispectral Pushbroom Scanner (AMPS)', two airborne multispectral sensors being developed at NASA's Earth Resources Laboratory (ERL). The ERE, located at the National Space Technology Laboratories (NSTL) in southern Mississippi', has been designing, fabricating', testing', evaluating, calibrating, operating, and maintaining airborne remote sensing systems since 1971. Included in this overall activity are the Thematic Mapper Simulator (TMs)', Thermal Infrared Multispectral Scanner (TIMS) , Calibrated Airborne Multispectral Scanner (CAMS)', NS001 Multispectral Scanner (based at the NASA/Ames Research Center)', as well as the State of Louisiana's Daedalus 1260 Multispectral scanner.

The design of a multispectral polarized laser system for characterizing the depolarization properties of the Earth's surface is described. Using a laser as the light source, this airborne system measures the Stokes parameters of the surface to simultaneously arrive at the polarization degree, azimuthal angle, and ellipticity for each wavelength. The technology will be studied for the feasibility of expansion of the sensor to do surface polarization imaging. The data will be used in support of solar polarization studies and to develop laser radiometry as a tool in environmental remote sensing.

A procedure has been developed to improve optical characteristics at the butted edges of linear multichip, focal plane, imaging arrays. The focal plane array is a backside illuminated pushbroom sensor that has been developed for application in remote sensing instruments. The gap between sensor modules making up the array represents a discontinuity in the refractive index of the substrate material through which optical radiation must pass. This discontinuity can produce vignetting and optical crosstalk in the vicinity of the butted regions. Artifacts observed in imagery obtained from a prototype multichip focal plane array suggest that the optical crosstalk in the butt region can be significant. Theory indicates the optical crosstalk can be significantly reduced by filling the gap with a material that has an index of refraction more closely matched to that of the substrate. This is supported by the ex-perimental data. The device butt edge performance with and without epoxy filler(s) has been measured and the technique(s) for epoxy application have been examined.

An optical line sensor is presented based on a Junction Charge-Coupled Device. The gates are diffused areas, this results in a high sensitivity and good anti-blooming properties. A quantum efficiency of more than 0.8 at a wavelength of 540 nm has been measured. The resolution of the realized device is 40.5 μm.

A 244X190-element monolithic interline transfer CCD infrared imager (the 2441) has been developed which combines PtSi Schottky barrier photodiode technology with proven CCD imager technology that has been used successfully for years in the production of visible imagers. The unit cell of this device is 36um high and 60um wide. The diagonal of the television format array is 14.4mm. Each Schottky barrier photodiode (SBPD) has an active area of 430um2. Assuming quantum efficiency of 1% at 4um and given f/2.5 optics, this device should passively image a 300°K scene with a noise equivalent blackbody temperature difference of approximately 0.08°C. This sensitivity makes this television sensor useful for such applications as general passive night viewing at moderate to high scene temperatures (50°F) and dynamic thermal data acquisition from scenes. The pixel count provides approximately half the resolution of standard television in both vertical and horizontal direc-tions. The 14.4mm diagonal is relatively small for IR imagers, allowing use of smaller and light-weight lenses than those required by imagers with larger pixels. The wafer fab process and the design rules used to make this device are almost entirely the same as those used to manufacture the Fairchild-Weston 488X380-element visible imager. The primary differences are the addition of the PtSi steps and low-temperature processes for subsequent processing. The processing employs 3-level poly-Si, buried channels and an ultra-high-vacuum process for forming a thin PtSi layer. The design employs 4-phase vertical registers, a 2-phase output register and a resettable-gate preamplifier. The evaluation camera incorporates a 12-bit video digitizer, a 16-frame-averaging reference frame memory and a real-time reference frame subtractor. It functions as a one-point offset uniformity correction. Performance tests to date, using an f/1.8 lens, have given MRTs as low as approximately 0.1°C and NEDTs as low as approximately 0.2°C or better. In general the imagers tested to date show excellent uniformity and few cosmetic defects.

An alexandrite laser is spectrally narrowed and tuned by the use of three optical elements. Each element provides a successively higher degree of spectral resolution. The digitally controlled tuning and scanning control servo system simultaneously positions all three optical elements to provide continuous high resolution laser spectral tuning. The user may select manual, single, or continuous modes of automated scanning of ranges up to 3.00 cm^-1 and at scan rates up to 3.85 cm^-1/min.. Scanning over an extended range of up to 9.999 cm^-1 may be achieved if the highest resolution optic is removed from the system. The control system is also capable of being remotely operated by another computer or controller via standard RS-232 serial data link.

The role of calibration for the Earth Observing System is identified as supporting the scientific requirements of the mission. The difference between this and other missions is identified as long time series, global data sets suitable for multidisciplinary analysis. Instrument calibration must be responsive to these issues in Eos. This document deals with the management aspects of achieving the required calibration. The approach is detailed in the NASA Eos Announcement of Opportunity, and insight is provided on how the major components of the management approach were developed. Case Studies highlighting the application of these approaches are summarized to support the inclusion of many of the components. The technical aspects of reaching the Eos calibration goals are not addressed here.

A technique is presented for absolute radiometric calibration of longwave infrared satellite systems. The technique involves a combination underflight technique and radiometric models to estimate the radiance field reaching a satellite sensor. The radiance field can then be compared to the radiance observed at the satellite to evaluate the sensor's post launch calibration. The technique was applied to the Thematic Mapper band 6 sensor on board Landsat 5. Results are presented for three underflight dates. These results indicate that the TM band 6 sensor can be calibrated to yield an expected error (1 standard deviation) in surface temperature of 0.9K. The radiometric propagation models used to achieve these results are presented along with estimates of potential sensor calibration errors. The final radiometric propagation models developed can be applied independent of underflight requirements and represent a general approach to computation of kinetic surface temperatures. The parameters included in the analysis encompass internal calibration, sensor spectral response, atmospheric transmission, upwelled radiance, downwelled radiance, and sample emissivity.

The status of the SPOT 1 calibration after nearly two years of flight is discussed in terms of measurements in the preflight period, validation in the post launch assessment and monitoring during flight. The results obtained with preflight sources, with an on board solar sensor and with test site measurements with ground truth on White Sands (New Mexico) are compared. Mention is also given to measurements with an on board internal lamp for calibration monitoring, intercomparisons between the two HRV's instruments by histogram matching, and analyses of images of snowy test sites for multiband calibration. The data are discussed and interpreted in this paper and finally lead to an estimation of the most likely calibration coefficients as well as an estimation of their accuracy.

A primary goal of the radiometric calibration efforts for the First ISLSCP (International Satellite Land Surface Climatology Project) Field Experiment (FIFE) is to provide consistent radiometric calibration of the NASA-provided aircraft and field instruments, particularly in the reflective (solar) portion of the spectrum. To this goal, a common source and traceability were chosen for use by all instruments for the primary calibrations: a 122 cm diameter hemisphere maintained by the Standards and Calibration office at GSFC. Among the primary multispectral reflective band instruments used in FIFE are the NS001 scanner on a C130 aircraft operated by Ames Research Center and several Barnes Modular Multispectral Radiometers (MMR's) operated on a helicopter and the ground. These instruments have 7 bands in the reflective portion of the solar spectrum: approximately the TM bandpasses and an additional band at about 1.2 um. The NS001 has a continuously variable gain setting; calibration is maintainable only by reference to its internal light source. Calibration of the NS001 data is affected by drift in the dark current level of up to 6 counts during a mirror scan at typical gain settings. Use of both samples of the dark current recorded by the instrument in performing the calibration, as opposed to the one currently used, generally decreases the uncertainty to less than 1 count, with the possible exception of channel 7. The apparent radiance of the internal source degraded an average of 4% over the 11 month period of January 1987 to December 1987 relative to the 76 cm sphere used to monitor it; this 76 cm source in turn apparently degraded 5-10% during the same period, suggesting an overall 10-15% degradation in the internal source. The MMR instruments are being used in their 1° degree field-of-view (FOY) configuration on the helicopter and 15° FOV on the ground. Pre- and post-season laboratory calibrations were supplemented by daily stability checks using a 30 cm integrating sphere source. The changes in calibration of the MMR instruments were related to the extent of their use. In the silicon channels of the MMR's , instruments used throughout the 4 field campaigns showed degradations of 3-4% between pre- and post-season calibrations; the one instrument used for 2 of the field campaigns degraded approximately half those values; and the instrument used just one day showed pre- and post-season calibrations within one percent of each other. Strong temperature sensitivity in the lead-sulfide channels (PbS) channels, about ±25% over a 15° C range, which was reduced to ±4% or less with temperature correction, led to greater uncertainty in these channels calibration, although pre- and post-season calibrations differences were no larger than for the silicon channels.

Spectral and radiometric calibrations of a pointable airborne spectroradiometer called ASAS are discussed. ASAS employs a 512-by-32 element detector array to acquire 29-channel multispectral digital image data. A monochrometer is used to characterize relative spectral responses and to determine spectral band centers with an uncertainty of ±1 nm. The spectral bands are approximately 15 rim wide and the band centers range from 469 nm to 873 nm at approximately 14 nm increments. A laboratory integrating hemisphere is used to characterize the radiometric responses of ASAS detectors. Radiometric responses are linear except for an initial build-up lag in response to low levels of radiance. Assuming radiometric stability in flight, raw ASAS digital counts can be transformed to absolute spectral radiance values with an uncertainty of 5.5% attributable to the laboratory calibration. The calibrations are being applied to radiometrically correct ASAS data acquired from multiple view directions over a tall grass prairie during the 1987 growing season.

A radiative transfer model is presented suitable for calibration of satellite instruments sensing in the solar spectrum. The model consists of two clear layers sandwiching a plane-parallel cloud layer. Clear-sky optical effects are treated with modified Beer's Law relationships and cloud optical effects are treated with the delta-Eddington method. A model inversion process is used to determine an effective cloud optical depth from surface measurements of global irradiance. The model is then used to calculate the corresponding upward irradiance at the top of the atmosphere. Upward irradiances are converted into directional radiances using empirical bidirectional reflectance factors. For sensor calibration, the model radiances are then compared to simultaneous satellite measurements of radiances from the cloud tops over the surface site. This technique has been tested by using observations of the GOES Visible and Infrared Spin Scan Radiometer. Results indicate that the VISSR-band (0.55-0.75 pm) radiance is close to nominal (slope 0.9601 compared to the expected 1.0). Significant, approximately cubic-form non-linearity of response is found (consistent with the amount of non-linearity exhibited on pre-launch calibrations). Equivalent broad-band radiance also suffers from the same type of non-linear response, with the best linear slope being about 80% of nominal, a result consistent with that anticipated for narrow-band to broad-band conversion. Observed root-mean-square errors for sensor calibration results are comparable to those found in direct comparison between uncalibrated and calibrated sensors on different satellites, indicating that the cloud-calibration approach has merit as a means of quantitative satellite sensor calibration.

Sensor spectral responses and their calibrations influence investigations that use satellite multispectral data. Analysis efforts to date have concentrated on responsivity calibration. A simulation approach is followed here to explore the feasibility of one technique for detecting post-launch changes in sensor spectral characteristics. If the possible system changes can be modeled with a single parameter for spectral change and another for responsivity change, measurements on only two appropriate test targets can measure and differentiate between the two types of change. The suitability of two commonly occurring terrain materials (bare soil and vegetation) for measuring possible shifts in spectral bands of five sensors (MSS, TM, AVHRR, CZCS, and HRV) is assessed. Effects of sensor noise are included in the formulation.

Ultraviolet solar radiation in the spectral interval 120 to 300 nm is almost completely absorbed in the earth's middle atmosphere. Small changes in the amount of radiation incident at these wavelengths will result in corresponding changes in the photochemistry and energy balance of the stratosphere, mesosphere and lower thermosphere. Recent measurements indicate that solar cycle variability is far smaller than previous estimates. Although the Sun varies by nearly a factor of two at Lyman α (121.6 nm), at wavelengths from 120 to 170 nm the Sun varies by less than 20%. From 180 to 300 nm the variability quickly decreases to less than 1%. The challenge for future observations is to make spectral measurements with a long term accuracy far better than 1%. One approach is to directly compare the solar irradiance to the UV output of a number of bright early-type stars. The SOLar-STellar Irradiance Comparison Experiment (SOLSTICE) will first be flown on the Upper Atmosphere Research Satellite (UARS) and will make the first comparison of this type.

The need for independent, redundant absolute radiometric calibration methods is discussed with reference to the Thematic Mapper. Uncertainty requirements for absolute calibration of between 0.5% and 4% are defined based on the accuracy of reflectance retrievals at an agricultural site. It is shown that even very approximate atmospheric corrections can reduce the error in reflectance retrie-val to 0.02 over the reflectance range 0 to 0.4.

The success of the Coastal Zone Color Scanner'-3(CZCS) in measuring the concentration of phytoplankton pigments in the world oceans, has led to several proposals for follow-on systems.^4-6 Although all of the proposed systems have increased radiometric sensitivity, the methodology for determining the actual engineering specifications has never been published. In this paper, the radiometric requirements for ocean color instruments in general are developed from a knowledge of the inherent accuracies of the atmospheric correction and bio-optical algorithms used to estimate the pigment concentration from the sensor-measured spectral radiance. The issues to be examined include (1) the noise equivalent radiance, (2) the saturation radiance, (3) polarization sensitivity, and (4) calibration and stability. It is assumed that the instrument senses the ocean in six spectral bands: 443, 500, 560, 665, 765, and 865 nm. The last three bands are used for atmospheric correction of the first three bands, which are then used to estimate the pigment concentration.

Spectral and radiometric calibrations of the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) were performed in the laboratory in June and November, 1987, at the beginning and end of the first flight season. This paper describes those calibrations and the changes in instrument characteristics that occurred during the flight season as a result of factors such as detachment of the optical fibers to two of the four AVIRIS spectrometers, degradation in the optical alignment of the spectrometers due to thermally-induced and mechanical warpage, and breakage of a thermal blocking filter in one of the spectrometers. These factors caused loss of signal in three spectrometers, loss of spectral resolution in two spectrometers, and added uncertainty in the radiometry of AVIRIS. Results from in-flight assessment of the laboratory calibrations are presented. The paper concludes with a discussion of improvements made to the instrument since the end of the first flight season and plans for the future. Improvements include (1) a new thermal control system for stabilizing spectrometer temperatures, (2) kinematic mounting of the spectrometers to the instrument rack, and (3) new epoxy for attaching the optical fibers inside their mounting tubes.

A reflectance-based method was used to provide an analysis of the in-flight radiometric performance of AVIRIS. Field spectral reflectance measurements of the surface and extinction measurements of the atmosphere using solar radiation were used as input to atmospheric radiative transfer calculations. Five separate codes were used in the analysis. Four include multiple scattering, and the computed radiances from these for flight conditions were in good agreement. Code-generated radiances were compared with AVIRIS-predicted radiances based on two laboratory calibrations (pre- and post-season of flight) for a uniform highly reflecting natural dry lake target. For one spectrometer (C), the pre- and post-season calibration factors were found to give identical results, and to be in agreement with the atmospheric models that include multiple scattering. This positive result validates the field and laboratory calibration technique. Results for the other spectrometers (A, B and D) were widely at variance with the models no matter which calibration factors were used. Potential causes of these discrepancies are discussed.

Three different approaches are described for the absolute radiometric calibration of the two reflective channels of the NOAA AVHRR sensors. Method 1 relies on field measurements and refers to another calibrated satellite sensor that acquired high-resolution imagery on the same day as the AVHRR overpass. Method 2 makes no reference to another sensor and is essentially an extension of the reflectance-based calibration method developed at White Sands for the in-orbit calibration of Landsat TM and SPOT HRV data. Method 3 achieves a calibration by reference to another satellite sensor, but it differs significantly from the first approach in that no ground reflectance and atmospheric measurements are needed on overpass day. Calibration results have been obtained using these methods for four NOAA-9 AVHRR images and for one NOAA-10 AVHRR image. A significant degradation in NOAA-9 AVHRR responsivity has occurred since the prelaunch calibration and with time since launch. The responsivity of the NOAA-10 AVHRR has also degraded significantly compared to the prelaunch calibration. The suitabilities of using Method 2 with the Rogers Dry Lake site in California and using Methods 1 and 3 at White Sands are discussed. The results for Method 3, which requires no field measurements and makes use of a simplified atmospheric model, are very promising, implying that a reasonable calibration of satellite sensors may be relatively straightforward.

Radiances obtained from the NOAA-9 Advanced Very High Resolution Radiometer (AVHRR) have been compared with those derived from a coaligned aircraft-mounted double-pass Fastie-Ebert spectrometer when both instruments were simultaneously observing the dunes region of White Sands, New Mexico. The radiance calibration of the airborne spectrometer was arranged to be traceable to NBS standards through the procedures that were used for the prelaunch calibration of the AVHRR. Data from August 1985 and October/November 1986 have been analyzed, and suitable corrections made for the effects of the atmosphere between the U-2 aircraft (flying at a pressure level near 6 N/m2 (60 mb)) and the satellite. For this purpose, atmospheric altitude profiles of temperature, pressure and water vapor were measured from a radiosonde, the most probable ozone profile was derived from statistically inter-preted Nimbus-7/SBUV measurements, and likely stratospheric aerosol conditions were assumed to calculate the ratio (as a function of wavelength) of the radiances expected at each instrument. The calculated ratio was used to adjust the measured U-2 spectra to their equivalent at the altitude of the NOAA-9 AVHRR. The calculations indicated an absolute accuracy of +5% in the derived sensitivity of the AVHRR visible channels relative to the sensitivity measured prelaunch, and that some improvement in absolute accuracy is possible with modest additional effort. The analysis showed that in August, 1985, the sensitivities of channel 1 (570-700 nm) and channel 2 (710-1000nm) of the NOAA-9 AVHRR were indistinguishable from their prelaunch values. In October/November, 1986, sensitivity losses were calculated to be approximately 12% for channel 1, and 19% for channel 2. Flights on three days during the period 24 October to 5 November 1986 reproduced these results to within +1.5% in sensitivity loss. Results for the GOES-6 VISSR visible channel showed a 9-14% loss in sensitivity after 42 months in orbit. For LANDSAT-5 TM a 12% loss in band 4 after 17 months in orbit was indicated, with no discernable change in the sensitivities of bands 1-3.

Drifts in satellite observations hinder the detection of trends in climate parameters measured from space. This problem has become particularly acute in the search for trends in atmospheric ozone. Satellite instrument performance can not be uniquely characterized from ground based observations since these and observations from space are not directly comparable. The Space Shuttle will however, provide an opportunity to assess satellite instrument performance by direct comparison of measurables. An instrument nearly identical to the SBUV/2 ozone sounder flying on the NOAA operational satellites will be flown periodically on the Space Shuttle. The engineering model to the SBUV/2 series of Instruments has been modified for Shuttle compatibility and employs dedicated power, data, and command systems. An in-flight radiometric calibration system, solar and nadir aspect sensors, and a transmission diffuser have been added making the Shuttle SBUV (SSBUV) a stand alone experiment for solar irradiance and albedo observations in the ultraviolet. A model has been developed which demonstrates that with regular SSBUV observations, the long term ozone record from the NOAA satellite series can be stabilized with sufficient precision to detect the predicted ozone trend. The model takes into account instrument precision, laboratory calibration repeatability over the long term, flight frequency, atmospheric variability, and the monitoring period.

Self-calibrated detectors have been under development for the past ten years and are now capable of measuring optical power with uncertainties on the order of 0.1%. They have now been combined with conventional and new optical systems to build instruments that are capable of making 0.25% measurements of various radiometric quantities.

A method used for calibrating field reflectance panels in the visible and shortwave infrared wavelength range is described. The directional reflectance factor of painted barium sulfate (BaSO4) panels is determined. The reference for this method is the hemispherical reflectance of pressed polytetrafluoroethylene (halon) powder prepared according to National Bureau of Standards (NBS) directions. The panels and a radiometer are mounted on rotation stages to measure the reflectance factor at different incidence and view angles. The sensor can be any laboratory or field filter radiometer small enough to mount on the apparatus. The method is used to measure the reflectance factors of halon and BaSO4 panels between 0.45 and 0.85 micrometers. These reflectance factors are compared to those measured by a field apparatus. The results agree to within 0.013 in reflectance at incidence angles between 15 and 75 degrees.

Measurements of surface reflectance factors in the field are usually made under conditions of total (direct and diffuse) irradiance. However, the reference panel reflectance factor R(r/e) used to convert the target measurement to reflectance is frequently determined using only direct irradiance. A method for determining the diffuse-irradiance reflectance factor, 11,1.4(0°,/e), of a calibrated field-reference panel from a knowledge of the direct-irrA.ance reflectance factor, Rd.,(0°,/0), is described. The magnitude of the error involved if only the direct irradiance reflectance factor is known, usually from laboratory calibration, was found to vary from -2% to +5%, depending on zenith angle, atmospheric conditions, and the particular reflectance panel used.

Earth scientists are increasingly attempting to utilize spacecraft observations to measure and predict change in the Earth's surface characteristics and surface-related phenomena, such as climate. The task is not easy. The inherent dynamics in Sun, Earth and space platform orbits, together with a diversity of sensor viewing geometries on existing and future orbiting sensors, result in radiometric data that can be expected to vary independent of changes in the surface properties. Conversely, appreciable changes in consequential surface attributes may not be discernable due to compensatory variations in sunsensor-target geometries coupled with prevailing sky and surface conditions. Thus, field studies to determine the nature and magnitude of the natural surface reflectance and sky irradiance variations and to subsequently improve our ability to interpret satellite data are essential. A unique field radiometer, called the PARABOLA, was developed to accurately measure the multi-directional incoming and outgoing spectral radiances for a variety of earth surface types, ranging from rangeland vegetation to ice and snow. Improvements in computational procedures have been recently developed and 3-D graphics capabilities have been implemented for aiding in the analysis of the directional radiances. Field measurements conducted with the PARABOLA for a variety of surface cover types reveal considerable diversity in the magnitude and character of their directional reflectances, which are complicated further for some vegetation types by seasonal variations and ambient sky conditions. Changes due to solar position under "clear" skies appear to be more predictable, however, such that radiometric correction models may soon be developed for some cover types and condition classifications.

An integrating hemisphere illumination system has been developed to facilitate the collection of spectral reflectance factor measurements of targets of interest in a laboratory environment. One of the most significant advantages associated with such an illumination source is that repeated measurements can be made over an extended period of time, for a variety of targets, under nearly identical illumination and viewing angle conditions. The illumination system consists of a 76 cm (30 in.) aluminum hemisphere coated internally with barium sulfate paint. Illumination is provided by sixteen 62 watt quartz halogen bulbs with tungsten filaments. A simple metal structure has been developed to hold the hemisphere and all peripheral equipment, such as spectrometers, radiometers, and cameras, in place during data collection. The entire set up can be easily disassembled and packed in airline-approved shipping cases to facilitate transportation to laboratory facilities located near any study area. The illumination system is described briefly, and numerous plots of radiance and spectral reflectance are provided to illustrate the performance and utility of the appanitim.

To accurately model the mixed signatures present in remotely sensed datasets, both the reflectance spectra of the components present and the characteristics of the various types of illumination must he known. The Portable Instantaneous Display and Analysis Spectrometer (PIDAS) was used to characterize the different sources contributing to the total illumination in a Ponderosa Pine (Pinks ponderosa) forest con'imunity. While direct illumination from the sun contributes the major portion of the total, the amount of illumination from diffuse skylight and from light reflected off the surrounding forest canopy are shown to be significant. A method was developed that allowed the determination of the visible and near infrared radiometric characteristics of each of these sources of illumination relative to the total illumination radiance. For each of the materials present, this data was used to compute apparent reflectance spectra, (the reflectance of that material with ambient illumination relative to a standard fully illuminated by the sun), for the range of illumination conditions encountered. Measured reflectance spectra of materials illuminated by a mixture of all of the sources can then be modeled as a mixture of the apparent spectra associated with each of the individual illumination sources. This technique, when used to model high spectral resolution. datasets such as Airborne Visible and Infrared Imaging Spectrometer (AVIRIS), can be used to extract not only information about the abundances of the materials present in the image, but also information about the architecture or geometric relationships of these materials.

The Kuiper Infrared Technology Experiment (KITE) program has operated over the last two years (1985-1987). Characteristics of atmospheric background radiance levels related to seasonal, diurnal, geographic and altitude effects are currently being studied because these effects are of interest to programs where the atmospheric background limits sensor performance. Correlating data from the KITE program with archived spectral data and model code predictions has established a number of important facts. First, the 11 pm window in the infrared appears to be "filled-in", i.e., the predicted troughs are not seen. This contributes to the second point: the background radiance levels predicted by the LOWTRAN model code in a window band centered at 11.2 pm appear to be too low by a factor of 2-3 or more. The high backgrounds observed are for the quiescent atmosphere in the south Pacific (Marshall Islands); clouds are highly variable and very difficult to predict due to convective processes (the intertropical convergence zone (ITCZ) borders the area). There appears to be little diurnal variability in the 11 pm window band in the tropics; however, seasonal effects are plausible, i.e., variation of the ambient atmospheric background radiance levels in response to the seasonal temperature dependence. Moreover, there appears to be a seasonally dependent linear relation between the 7.2- and 11.2- μm data. Persistence of "holes" in high altitude .thin cirrus cloud layers is a feature of primary concern in the area of forecasting. These optically thin cirrus clouds, sometimes called "cirrus evadus", may be the prime contributor to elevated background radiance levels by providing a means for thermal scattering of energy radiated by the Earth. They are often detected by means of a ground-based lidar; however, prediction capabilities do not exist at present. A full-scale climatology study to determine the growth of cloud formations and to establish a statistical database is recommended.

Topographical planimetric features include natural surfaces (rivers, lakes) and man-made surfaces (roads, railways, bridges). In conventional planimetric feature extraction, a photointerpreter manually interprets and extracts features from imagery on a stereoplotter. Visual planimetric feature extraction is a very labour intensive operation. The advantages of automating feature extraction include: time and labour savings; accuracy improvements; and planimetric data consistency. FEX (Feature EXtraction) combines techniques from image processing, remote sensing and artificial intelligence for automatic feature extraction. The feature extraction process co-ordinates the information and knowledge in a hierarchical data structure. The system simulates the reasoning of a photointerpreter in determining the planimetric features. Present efforts have concentrated on the extraction of road-like features in SPOT imagery. Keywords: Remote Sensing, Artificial Intelligence (AI), SPOT, image understanding, knowledge base, apars.

Traditional computer assisted image analysis techniques in remote sensing lag well behind human abilities in terms of both speed and accuracy. A fundamental limitation of computer assisted techniques is their inability to assimilate a variety of different data types leading to an interpretation in a manner similar to human image interpretation. Expert systems and computer vision techniques are proposed as a potential solution to these limitations. Some aspects of human expertise in image analysis may be codified into expert systems. Image understanding and symbolic reasoning provide a means of assimilating spatial information and spatial reasoning into the analysis procedure. Knowledge-based image analysis systems incorporate many of these concepts and have been implemented for some well defined problem domains. Geographic information systems represent an excellent environment for this type of analysis by providing both analytic tools and contextual information to the analysis procedure.

Image processing techniques are described that assist an operator in delineating planimetric features in digital imagery. The techniques are simple, robust and highly efficient, allowing the processing to occur at interactive rates. The operator defines a general path for the feature, for example a road, by moving a cursor along the road. Simultaneously, the computer processes a moving window of pixel data along the direction defined by the operator and produces a centerline for the road. These data are converted into vector form and entered into line sets for subsequent GIS Processing or mapping applications. By combining manual and computer-assisted feature delineation, an efficient, less labor-intensive, more objective feature extraction approach results. An example application of this approach to road mapping in a rural area is presented.

This paper presents work in progress on the application of a self-organizing neural network to vector quantization (VQ) training. A modified version of Kohonen's self-organizing feature map algorithm was applied to simulated data and to digitized Synthetic Aperture Radar image data. Preliminary results indicate that the network-based algorithm is potentially more robust than the traditional LBG training algorithm, especially when confronted with multimodal input data distributions.

A methodology has been developed and demonstrated for semi-automated generation of high fidelity terrain databases suitable for flight simulator applications. The technique has been demonstrated using electro-optic (EO), IR, and Synthetic Aperture Radar (SAR) sensor imagery. Extensions of the methodology are described for the generation of sensor image-based representations of vegetation terrain features. In contrast to polygon based databases in use by many simulators today, the simulator databases described here are gridded in nature. Finally, a realtime image generation architecture capable of exploiting this new database technology is currently in development, and is briefly summarized here.

The definition and design of an operational environment-monitoring system is described. The system supports the principal information demands of the Food and Agriculture Organization of the United Nations (FAO) by monitoring ecological conditions in Africa.

The use and application of multi-sensor data sets are an absolute necessity for remote sensing to fulfill its promise. Initial efforts of registering dissimilar images have been made, but results are not widely available. This paper describes some initial work to better understand the difficulties of multi-sensor registration. Accuracies achieved are in excess of a pixel diameter.