FireSat and the Global Monitoring of Biomass Burning

The report Global Change and Our Common Future published by the National Research Council (1989) concluded (p.1):      Our planet and global environment are witnessing the most profound changes in the brief history of the human species. Human activity is the major agent of those changes--deplection of stratospheric ozone, the threat of global warming, deforestation, acid precipitation, the extinction of species, and others that have not become apparent. All five global environmental changes identified above have one important thing in common--they are all caused by biomass burning.       The burning of the world's living and dead biomass for land clearing and land-use change is a significant global source of atmospheric gases and particulates that impact the chemistry of the troposphere and stratosphere and the climate of our planet (Levine 1991; Levine et al. 1995). In addition to the production of significant amounts of gases and particulates to the atmosphere, biomass burning impacts Earth's atmosphere/biosphere system through the following processes: (1) the reflectivity and emissivity of the land and hence the global energy budget of out planet, (2) water run off and evaporation and the global hydrological cycle, and (3) the biogeochemical cycling of compounds from the biosphere to the atmosphere (Levine 1991). Biomass burning also impacts the stability of ecosystems and leads to the extinction of species (National Academy of Sciences 1990).       Biomass burning is not restricted to one country or to one region. Biomass burning is a regular feature of the world's tropical, temperate, and boreal forests, the savanna grasslands, and agricultural fields following the harvest. Biomass burning is a regular feature in the tropical forests in Brazil, Indonesia, Colombia, Ivory Coast, Thailand, Laos, Nigeria, Philippines, Burma, and Peru, the temperature forests of the United States and Europe, and the boreal forests of Siberia, China, Canada, and Alaska, the savanna grasslands of Africa, and the agricultural lands of the United States and Europe. It is generally believed that the vast majority

of this burning is human-initiated (>90%) and that biomass burning has increased significantly over the last 100 years (Houghton 1991; Hao and Liu 1994).      Biomass burning is a significant global source of the following atmospheric gases: (1) greenhouse gases, carbon dioxide (C02), and methane (CH4), that lead to global warming, (2) chemically active gases, nitric oxide (NO), carbon monoxide (CO), and hydrocarbons (HC), which lead to the photochemical production of ozone (03) in the troposphere. In addition, NO leads to the chemical production of nitric acid (HNO3), the fastest growing component of acid precipitation, and (3) methyl bromide (CH3Br), a major atmospheric source of bromide, which leads to the photochemical destruction of ozone in the stratosphere (Andreae 1991; Mano and Andreae 1994). Biomass burning is also a significant global source of atmospheric aerosols, which impact the transfer of incoming solar radiation through the atmosphere and hence impact both global climate and tropospheric chemistry (Penner et al. 1991).      To quantify the role and importance of biomass burning as a global source of gases and particulates to the atmosphere, information is needed on the global strength of biomass burning as a source of these environmentally significant compounds. To provide the needed information, international biomass burning research activities have been initiated over the last few years. The International Global Atmospheric Chemistry (IGAC) Project, a core activity of the International Geosphere-Biosphere Program (IGBP) has initiated two biomass burning research activities--The Biomass Burning Experiment (BIBEX) and the Global Emissions Inventory on Biomass Burning, part of the Global Emissions Inventory Activity (GEIA). In addition, over the last few years, international biomass burn field experiments have taken place in diverse ecosystems, including the South African Fire-Atmosphere Research Initiative (SAFARI-92), a ground-based and airborne measurement program in the savannas of southern Africa (Andreae et al. 1994), the Transport and Atmospheric Chemistry near

the Equator-Atlantic (TRACE-A), an airborne mission over Brazil and southern Africa (Andreae et al. 1994), the Dynamics and Chemistry of the Atmosphere in Equatorial Forest (DECAFE) in the forests of equatorial Africa, the Biomass Burning Airborne and Spacecraft Experiment-Amazonia (BASE-A) in the Brazilian Amazon, and the Fire Research Campaign Asia-North (FIRESCAN) in the boreal forests of Siberia. Smaller biomass burn experiments have taken place in the chaparral ecosystem in the San Dimas Experimental Forest in southern California, in the wetlands in the Merritt Island National Wildlife Refuge, Kennedy Space Center, Florida, in the tropical rain forests of the Yucatan Peninsula, Mexico, and in the boreal forests of Canada.      In the last five years, biomass burning and its global environmental impacts were the subject of two international conferences sponsored by the American Geophysical Union--The Chapman Conference on Global Biomass Burning: Atmospheric, Climatic, and Biospheric Implications (March 1990) and the Chapman Conference on Biomass Burning and Global Change (March 1995). &nbps;    More than a decade ago, Seiler and Crutzen (1980) showed that as a first approximation, the total amount of biomass burned (M) (in units of grams of dry biomass material per year) in a particular ecosystem may be given by the following equation:       M = A x B x a x b          (12.1) where A is the total land area burned annually (m2/year), B is the average biomass material per unit area in the particular ecosystem (grams of dry biomass material per m2), a is the fraction of the aboveground biomass material relative to the total average biomass B, and b is the burning efficiency of the aboveground biomass. Parameter A, the area burned during a fire in a particular ecosystem, is the major uncertainty in solving equation (12.1). Parameters B, a, and b in equation (12.1) have been determined during a series of international and national biomass burn field experiments including SAFARI-92, DECAFE, BASE-A, and FIRESCAN, and the smaller scale experiments in California, Florida, Mexico, and Canada.      Once M is known, the total mass of carbon [M(C)] in grams released to the atmosphere during biomass burning may be calculated using the following equation:       M(C) = 0.45 M          (12.2) Since about 45% of biomass by weight is carbon (the remainder of the biomass weight is due to water (about

50%) with smaller amounts of nitrogen (about 1%), and still smaller amounts of sulfur, chlorine, and bromine. The carbon released to the atmosphere during biomass burning takes the form of several gaseous and particulate compounds, including C02, CO, CH4, nonmethane hydrocarbons (NMHCs), and particulate carbon. The ratio of any carbon compound (such as CO, CH4, NMHC) or nitrogen compound (such as NO, or N20) to carbon dioxide produced in biomass burning can be determined by knowledge of the emission ratio (ER). The emission ratio is the amount of any compound X produced during biomass burning normalized with respect to the amount of C02 produced during biomass burning. The emission ratio is usually normalized with respect to C02 because C02 is the overwhelming carbon species produced during biomass burning and it is a relatively easy gas to measure. The emission ratio is defined as         ER = /CO2          (12.3) where X and CO2 are the concentrations of the species X and C02 produced by biomass burning and are equal to (X* - X) and (CO2* - C02) where X* and CO2* are the measured concentrations in the biomass burn smoke plume and X and C02 are the background (out-f-plume) atmospheric concentration of the species. Information on the emission ratios for various gaseous and particulate compounds produced during biomass burning has been obtained during the serics of international and national biomass burn field experiments including SAFARI-92, TRACE-A, DECAFE, BASE-A, and FIRESCAN, and the smaller scale experiments in California, Florida, Mexico, and Canada.      Over the past five years, as a result of a series of national and international biomass burn field experiments, we have gathered considerable information on all of the parameters shown in equations (12.1) to (12.3), including B, a, b and ER.      As previously noted, the major uncertainty in our current understanding of the global impact of biomass burning concerns the parameter A in equation (12.1), the total land area burned annually. The Global Emissions Inventory on Biomass Burning, a research activity of the IGBP/IGAC/GEIA, held an open meeting during the Chapman Conference on Biomass Burning and Global Change in March 1995 to discuss the uncertainties and unknowns in the development of a global emissions inventory on biomass burning to better assess the environmental impact of global biomass burning. There was unanimous agreement that the major uncertainty was accurate information on the

spatial and temporal distribution of burning in the world's ecosystems.       At the present time, estimates of A are based on one of two sources of information-statistics of the occurrence of fires in each country tabulated and compiled by that country and collected and disseminated by the United Nations Food and Agricultural Organization (FAO) on an annual basis, and on certain satellite measurements, such as measurements obtained with the Advanced Very High Resolution Radiometer (AVHRR) on the NOAA polar orbiting operational meteorological satellites. The satellite systems used to deduce information about biomass burning were originally developed for other purposes and are not ideal to deduce information about the area consumed during burning. Each existing satellite system has its own set of unique problems that make it difficult to determine the needed information on the spatial and temporal variation of biomass burning. Problems include very poor or no coverage of the high latitude boreal forests, poor temporal and spatial coverage of the burned areas in the tropical and temperate regions, and low saturation temperature for the thermal channels, which makes detection of fires ambiguous.      The question of the rate of deforestation and the use of satellite measurements to deduce global burning was addressed in a National Academy of Sciences (1990) report (p.117): The rates of deforestation vary widely, mainly because Countries use different survey procedures....    and because satellite images of the entire globe are expensive and difficult to analyze. Simply put, we do not have reliable and up-to-date information on how much of the earth's surface is Covered by forests and how fast it is being cut down.      It is ironic that there has never been a dedicated space experiment to study the geographical and temporal distribution of biomass burning. To determine the spatial and temporal distribution of biomass burning On our planet, the NASA Langley Research Center has proposed a dedicated satellite experiment, FireSat. FireSat Science Mission and Requirements      A few very broad design goals have served as the underlying basis for the FireSat design and enhanced the Science mission objectives. In order to make FireSat a smaller and less expensive instrument when Compared to conventional imaging instruments, the latest technological advances have been carefully examined to determine their suitability for space-based application, to assess cost, and to minimize risk. From

the outset, FireSat has been intended to incorporate new technology and provide insights and lessons that would be useful for the design of future Earth imaging systems. To minimize risk, the new technology chosen will facilitate an instrument design that requires no moving parts for operation. In addition to hardware advances, the FireSat design has the ability to provide a calibrated dataset for Earth interdisciplinary studies. FireSat will also provide a cloud-filtering algorithm and a real-time downlink to ground stations within the spacecraft's horizon. Cloud-filtering the data will reduce the data volume that requires archiving and, by only retaining the desired clear-sky data set, shorten computer processing time of the archived imagery.      With these fundamental instrument design goals always in mind, the FireSat instrument concept will meet the goals of a well-defined science mission. This mission is to develop a continuous global mapping of vegetation fires in forests (tropical, temperate, and boreal), grasslands, and agricultural fields. Further, after the analysis of the data set provided by FireSat, there will be the ability to better assess the environmental impact of fires on the atmosphere, climate, and land to determine the couplings between fire and global change. Given the established philosophy of this to~level definition of the science mission, the science mission has been further detailed in a to~down fashion in order to arrive at those aspects of the science mission that will drive the FireSat instrument design. The FireSat science mission will need to be about five years in length in order to gain better insight into the interannual variations of fire. During the five-year mission, FireSat data would be reduced to (1) map active vegetation fires globally and determine firefront temperatures, (2) map the geographical extent of global vegetation fires, (3) monitor vegetation stress for indicating fire susceptibility, (4) estimate the gaseous and aerosol emissions of the fires released to the atmosphere, and (5) identify aerosol source regions for comparison with ground- and space-based radiation measurements.      Given the definition of the science mission above, a set of first-order requirements that drive the instrument design has been determined. In order to map active fires and gage the fire intensity along a firefront, the saturation of the thermal channel used for active fire detection should be at a temperature of at least 1000°K. Further, in order to spatially depict an active firefront and to accurately assess the area that has been burned, the spatial resolution of the instrument at nadir should be no greater than 500 x 500 m. In this case, the spatial resolution is defined as the actual

Table 12.1     Science products
Product	Archived format	Wavelengths (µm)	Unit of measure
Active fires	Geolocated fire pixels3.7,85.	Temp. (K)
Burn scars (clear sky)	Geolocated burned area pixels and regional area burned estimates	0.5, 1.0, 2.2, 8.5	Area > 25 km2 (5%)
Burn scars (through smoke aerosol)	Map recently burned areas	3.7,8.5	Area > 25 km2 (5%)
Smoke aerosol	Geolocated smoke-filled pixels	0.5, 1.0, 2.2, 8.5	Over-water optical depth
Vegetation changes	Geographical map	1.0, 1.65, 2.2	TBD
Surface temperature	Geographical map	8.5	Temp. (K)
Cloud masking	Bitmap	0.5, 1.0, 1.38, 8.5	NAa
a. Not applicable

surface area contributing energy to an individual observation including all instrument effects that would broaden the instantaneous field of view (IFOV) geometric footprint. In order to maxize usefulness to the scientific community and provide a quantitative means for evaluating the data, the data set will be calibrated. Calibration will be evaluated over time in order to characterize instrument degradation and stability, which can impact the science mission. The geographical range of fires on Earth is roughly between 60°S to 75°N latitude. Given the north/south extent of fires to be monitored and a desire to minimize the difference in solar/spacecraft geometry between successive days, a sun- synchronous orbit of 830 km has been selected for the FiresSat design studies. Since meteorological conditions are more conducive to early afternoon fire development, a trade-off has been made between observing fire activity at its diurnal peak and the potential of increasing cloud cover in the afternoon when selecting a mid-afternoon local overpass time. Even though afternoon cumulus fields may obscure much of the ground, smoke aerosol can still likely be identified in the cloud gaps and permit establishing time histories of regional fire episodes. For the geographical mapping of fire activity and for co-registering images acquired from different orbital passes, the pointing knowledge of the instrument should be no worse than 250 m to assure that the imagery can be accurately mapped.      Only one of the science mission requirements has been relaxed somewhat. The requirement was that FireSat observe the entire planet at least twice daily. FireSat does only marginally worse than this by missing a narrow gap between successive orbits, in a 24-hour period, only in tropical regions. These gaps would be filled every other day. For the northern forest ecosystems there is at least four times daily coverage.

The reason for relaxing the requirement is to avoid imposing additional demands on the optical and detector designs when the gain in geographic coverage does not really merit the additional costs. The refined science requirement now specifies that the FireSat instrument has a 96-degree total field of view. Beyond this point the footprint size begins to grossly overlap with adjoining footprints by as much as 50%. The net result is blurring the imagery beyond the point that anything other than questionable interpretation of the data could result.       Based on the science goals, a fundamental set of science products has been oudined (table 12.1). In order to produce the specified science products, spectral bands have been chosen. The selection of the spectral bands has placed additional requirements on the FireSat instrument's conceptual design. Active fire detection will primarily be accomplished with a spectral channel centered at 3.7 µm. The 3.7-µm channel will have the high saturation of no less than 1000°K with sensitivity of no greater than 1°K so that small surface temperature changes can be detected through the smoke of active fires, which is indicative of burned surface. The 8.5-µm channel will be used in conjunction with the 3.7-µm channel for isolating active fires, but its prime role is to provide the surface temperature with peak efficiency over the range of land and ocean temperatures. The accurate detection of burned scars is accomplished by evaluating the spectral response in bands centered on 0.5, 1.0, 2.2, and 8.5 µm. The monitoring of smoke aerosol would use the same set of channels. Monitoring changes in vegetation would principally be through the combined use of bands centered at 1.0, 1.65, and 2.2 µm. A 1.38-µm channel has been added, and when used in combination with the other channels, will be used for masking clouds out of the imagery. In total, two infrared channels and five

Table 12.2     Comparisons of instruments
Platform	Resolution at nadira(m)	Global coverage (days)	Thermal channel saturation for active fire detectiona(°K)	Spatial coverage	In-flight cal.	Cloud-filtered archive
FireSat	260	2	1100	Global	Yes	Yes
MODIS	500/1000	1	500	Global	Yes	No
AVHRR	1000	1	320	Global	No	No
DMSP	600(2700	1	No channel	Global	No	No
Landsat	30	14	No channel	Global	No	No
GOES	1000 (4000)	.02	335	Hemispheric	Yes	No
a. Bold values pertain to active fire detection

visible-to-near infrared channels have been specified to meet the science objectives and are included in the FireSat instrument conceptual design.      This is a good point to assess how the FireSat design concept compares to other Earth imaging instruments, both operational and under development. The FireSat instrument, with its focused mission of fire detection and monitoring, exceeds the resolution (the IFOV footprint), and the thermal saturation for detecting active fires is above that of any other instrument reviewed. A summary of this comparison is given in table 12.2. The bold-valued figures are the characteristics than pertain to active fire detection. Other instruments saturate below 500°K in the active fire detection band (3.7 µm) and the IFOV footprint of the other instruments is, at a minimum, 1000 m for active fire detection. The Defense Meteorological Saternte Program (DMSP) Operational Linescan System (OLS) instrument and Landsat have been included in the comparison even though they contain no channels for active fire detection. The DMSP OLS instrument can be used to assess fire activity at nighttime by monitoring the low-light emissions in the visible part of the Spectrum. Landsat has been included because its spectral band selection offers the potential for monitoring and assessing the areal extent of burned vegetation. Even though Landsat can be used to monitor burn area, the repeat coverage of every 14 days and very high spatial resolution are not practical for continual continental scale monitoring. The FireSat instrument concept has been developed by making several tradeoffs in regard to spatial resolution, repeat coverage, and costs. As demonstrated by this comparison, the FireSat instrument has been optimized for global fire monitoring and is unique among the instruments.      In addition to defining the channel selection, it is necessary to further define the minimum acceptable precision of each of the channels that will meet the science mission requirements. Multiple-scattering radiative

transfer calculations were made while varying the solar zenith angle, the viewing angle, the atmospheric aerosol type and loading, the atmospheric model, and target characteristics. From these radiative simulations a reasonable value for the maximum and minimum detectable radiance was estimated for each of the spectral channels. Also, the smallest radiance change required to achieve the science mission goals was derived. As a result of these simulations, the quantization requirements were specified for each channel. In practically all of the channels, with the exception of 2.2 µm and 8.5 µm, 10-bit data is required. For the 2.2-µm and 8.5-µm channels, the required quantization is nine and seven bits, respectively. Radiance Modeling for FireSat      A modeling and simulation effort, utilizing the MODerate resolution atmospheric TRANsmission model (MODTRAN) and other atmospheric radiative transfer software, was carried out to calculate the levels of total radiance at the aperture of the FireSat instrument, as the first step in the end-to~end instrument design process. This effort consisted of the following steps: 1. Determining which regions of the electromagnetic spectrum, visible and infrared, should be used to glean information on fire properties, while reducing interference from the atmosphere to a minimum in what is fundamentally a surface-reconnaissance task. 2. Determining optimal bandpasses within the chosen spectral regions, to further minimize atmospheric effects (e.g., avoid prominent water vapor absorption lines) as much as possible, consistent with maintaining adequate signal levels for sensitive measurements. The findings of steps 1 and 2 resulted in the definition of the seven FireSat channels described in the previous section.

Table 12.2   Parameters modeled for each FireSat simulation scenario Location ('atitude/longitude) Season, date Model atmosphere (either built-in or from rawinsonde data): temperature, pneesure and molecular constituent concentration profiles, from sea leve] to 100 km altitude Atmospheric aeroso] content/scattering phase functions, using built-in or user-provided models Surface temperature and spectral emittance Solar illuntination and sensor view geometries Sensor and terrain altitudes Solar zenith angle Nadir view angle Relative azimuthal angle ~tween sensor line-of-sight and solar azimuth direction) Cloud cover (type and altitude) Smoke particle-size distribution and concentration Smoke tnansmission Fire temperature and emissions models Flame radiance and transmittance 3 Determining through simulations the range of signal levels likely to be encountered, under the full anticipated range of climatic, seasonal, atmospheric, sun illumination, and viewing geometry conditions. 4. Determining the sensitivity of measurement to changes of surface temperature, reflectance, aerosol content, solar illumination and viewing gebmetry, and the presence of clouds. 5. Developing a fire radiance model for assessing the relative contributions of the flame- and hot-surface components of the fire signature. 6. Modeling the transmittance and scattering properties of fire-generated smoke.      To perform the simulations, the parameters in table 12.3 were specified for each FireSat scenario. The PCModTRAN model (ONTAR 1992), a PC computer-based commercially available implementation of the U.S. Air Force Phillips Laboratory's MODTRAN model, was the main simulation tool used in this study. It uses a two-parameter band model of atmospheric transmittance, with an as-fine-as 2 cm-1 spectral resolution over the spectral range from ultraviolet to microwave. Calculations of transmittance, as well as of single- or multiple-scattered atmospheric and surface radiance components, are readily carried out with this model, whicb is in widespread use. MODTRAN is a development of the long-used LOWTRAN atmos-

pheric radiative transfer model (Kneizys et al. 1988). and offers increased accuracy and finer spectral resolution than the latter. Other models used in the current simulations included BACKSCAT (Hummel et al. 1992), and HITRAN-PC (Killinger and Wilcox 1992). BACKSCAT, also developed under Phillips Laboratory sponsorship, is normally used to model laser back-scattering, but was used here in developing a preliminary model of smoke transmittance. HITRAN-PC, a line-by-line gas radiative transfer model developed at the University of South Florida, was used in calculating fire flame transmittance and absorption.      The main emphasis in the simulation effort was in applying PCModTRAN to calculate, for each FireSat scenario, the total radiance at the top of the atmosphere. This quantity is the input signal to the instrument; its computation is the first step in the end-to-end signal-modeling process. Calculation of the total spectral radiance and its subcomponents was carried out for several scenarios and for all seven FireSat channels.      For ease in comparing signal 1evels ainong the channels all initial simulations, except where indicated in the discussion, were performed using the following conditions: a 1976 Standard Atmosphere, which is broadly representative of mid-latitude springlfall climatic conditions a rural haze aerosol model with 23-km visibility at the surface and a Mie scattering function 830-km satellite altitude, viewing at nadir 45-degree solar zenith angle 2-cm-1 spectral resolution no clouds or smoke 280°K surface temperature Table 12.4 is a summary of some results for the seven channels, for this simulation condition-in~ommon. The table lists the bandpass for each channel, the total radiance in this bandpass incident upon the instrument aperture, and the surface reflectance value used in the simulation. The surface reflectances used for each channel were modeled after those in Bowker et al. (1985). The bandpass-averaged clear-sky transmittance from sea level vertically to satellite altitude is also listed in the table. It is noted that this transmittance value varies from nearly zero for channel 3, the clouddetection channel, to over 0.9, for channels 4 and 5.      Figures 12.1 through 12.6 are samples from the simulation results.

Figure 12.1    Total radiance and components calculated by PCMod TRAN for FireSat channel 1 (0.45-0.55 µm). Simulation conditions: 1976 Standard Atmosphere, nadir view, 23-km rural aerosol, 45-degree solar zeneth angle, 0.07 surface reflectance, 280°K surface temperature. Reflected component dashed curve; atmospheric scattering component (dotted curve); thermal component -- atmospheric plus surface -- (dot-dashed curve, negligible for this case and lying on the x-axis). Total radiance (sum of components) (solid curve)

Except where indicated, all the figures use the same logarithmic ordinate range, extending from 1.0 x 10-2 to 1.0 x 10-2 spectral radiance units (W/(cm2* Sr* µm)). Plotting all the components together is instructive for determining which components are dominant and which are negligible in a given wavelength region, and is very useful in deciding which parameters need to be varied, and which can be ignored in sensitivity studies.      Figure 12.1 is for FireSat channel 1 (0.45-0.55 µm). The results for this channel in the visible region show that most of the radiance from this low-reflectance scenario (typical of vegetated surfaces) Stems from atmospheric scattering. The reflected component is secondary in importance, at only one half or less of

the magnitude of the scattered component, and the thermal component is negligible.      Figure 12.2, for channel 3, the FireSat cirrus cloud detection channel (1.35-1.40 µm), shows a dramatically different result. Here, very strong absorption features (water vapor) are evident in the reflected component; this strong absorption is also obvious in table 12.4, where the average transmittance in the channel is seen to be only 0.004. With the thermal component still negligible in this band, the total radiance is almost totally dominated by the atmosphericscattered component. Because of the very strong atmospheric absorption, the spectral radiance is only around 1.0 x 10-5 W/(cm2*Sr* µm), two to three orders

Figure 12.2    As in figure 12.1, but for FireSat channel 3 (1.35-14.0 ~m); surface reflectance 0.60, cloud-free

of magnitude less than for channels I and 2. Figure 12.3 shows the result when a 1-km-thick cirrus cloud is inserted at 10km altitude. In this case, the level of total spectral radiance is increased by one order of magnitude from that in Figure 12.2. Thus, the existence of a high value of radiance in a pixel in this channel means that high-altitude clouds are present. Simulations for thinner cirrus, not shown here, also show strong increases of radiance from the no-cirrus value. Therefore, channel 3 is shown to be a sensitive detector of high clouds' presence.      The radiance in the FireSat fire-detection channel (channel 6, 3.60-3.80 µm) was modeled under normal (280°K) and hot (600°K) surface-temperature conditions, both for cloud-free and overcast cloud scenarios. Figure 12.4 shows that, even at the baseline 280°K surface temperature condition, the thermal component dominates. The reflected component contributes about one third of, and the scattered only around 3% of the total radiance. Figure 12.5 models a cloud-free condition with a 600°K hot surface present, as in a low-intensity fire. In this case, the total radiance level is increased by around three orders of magnitude from that in figure 12.4; the thermal component dominates even more than in figure 12.4 (the ordinate maximum is increased to 1.0 x 10-1 radiance units for figure 12.5). Figure 12.6 simulates a stratus cloud deck overlying the 600µK hot surface. The total radiance in this cloud-obscured case is much reduced from that of the

cloud-free case (figure 12.5), but is about 8O% above that of the normal temperature case (figure 12.4). Table 12.5 presents more detail on the effects of three cloud types (stratus, cumulus, and cirrus) and different surface reflectance values on the level of total radiance, for the common simulation condition. It shows that the level of radiance for the stratus-covered case is approximately equal to that for a cloud-free case with a 298°K surface temperature and 0.02 surface reflectance, or to that of a cloud-free case with a 280°K surface temperature and 0.11 surface reflectance. This result points out the need for a surface temperature estimate, and for a method of detecting clouds. The table shows that, as expected for thick clouds, the radiance is invariant with changes in underlying surface temperature and reflectance, for stratus and cumulus overcast conditions. Thus, surface temperature cannot be assessed through a stratus or cumulus cloud. However, the cirrus simulations do show a change with surface temperature and reflectance variability. This may indicate that it may be possible to detect hot surface through thin cirrus clouds. Both these findings again demonstrate the need for having a clouddetection channel to indicate unambiguously when clouds are present.      Simulations for nighttime cloud-free viewing conditions were also carried Out for all FireSat channels. The results, not shown here, demonstrate that for the thermal-sensitive channels 6 and 7 the surface thermal

Figure 12.3    As in figure 12.2, but with 1-km-thick cirrus cloud present, with base at 10 km altitude

Figure 12.4    As for preceding figures, but for FireSat channel 6 (fire-detection channel, 3.60-3.80 µm); surface reflectance 0.02, surface temperature 280°K

Figure 12.5    As for figure 12.4, but with 600°K hot surface present; surface reflectance 0.02. Note that total radiance signal is three orders of magnitude higher than for figure 12.4

Figure 12.6    As for figure 12.5, but with MODTRAN default stratus deck present, with cloud base at 0.33, top at 1.0 km altitude

Table 12.5 Simulated total radiance levelsa in FireSat channel 6b, for various typesc of overcast cloud cover conditions Tsfc,°K Sfc. refl. No cloud Stratus Cumulus Cirrus 280 0.02 3.970E-6 7.339E-6 1.233E-5 4.900E-6 298 0.02 7.481E-6 7.339E-6 1.233E-5 7.942E-6 280 0.11 7.452E-6 7.339E-6 1.233E-5 7.799E-6 600 0.02 4.379E-3 7.339E-6 1.233E-5 3.795E-3 600 0.11 3.981E-3 7.339E-6 1.233E-5 3.450E-3 a. Total radiance unites: W/(cm2*Sr) b. Viewing conditions: Nadir, 45° solar zenith angle Atmosphere: 1976 Standard, with 23-km visibility rural haze Date: 3 April (Earth-solar distance = 1 a.u.) c. Cloud models (per MODTRAN) Stratus: base 0.33, tops 1.0 km Cumulus: base 0.66, tops 2.0 km Cirrus: 1.0 km thick, base at 10.0 km Extinction coefficient: 0.14 km-1 component so dominates the total radiance signatures that fire scenes can be detected readily under clear-sky nighttime conditions.      Calculations similar to those described above were performed to bound the range of expected total radiance signals. The total radiance values were also scaled to photon flux values for use in choosing detectors during the instrument design process. (Radiance is scaled to photon flux by multiplying the radiance value by 5.031447 x 1018 times the wavelength in µm.) Selected results are given in table 12.6, to give an indication of the range of signal levels to be encountered in nadir viewing. (All simulations used a 1976 Standard Atmosphere with a 23-km rural haze model.) The tabulation is in terms of condition simulated (surface reflectance for all channels except for channel 3, where cloud cover condition is described), surface temperature, and solar zenith angle (SZA).      While these results are adeqttate for assessing the major contributors to the total radiance, and for bounding the range of signals to be encountered in the FireSat channels for instrument design purposes, further study was deemed desirable to better understand the modeling of fires themselves. Therefore, modeling and simulation efforts were initiated in two areasmodeling the emission of fires. including that from flames as well as the hot surface, and modeling the transmittance of smoke. Preliminary results show that, for the cases simulated, the radiance from a fire scene stems predominantly from the heated surface; the flame radiance is small by comparison. Therefore, just simulating the radiance emitted by the fireheated surface gives a good approximation (in smokefree conditions) to the overall thermal signature, for

instrument design purposes. A preliminary smoke-transmittance model was also developed, using a smoke model given in BACKSCAT. Initial results show, as expected, that smoke transmittance varies widely with sensing wavelength, smoke particle size distribution, and concentration profile. It is planned to extend these simulations, using other smoke models that may be available, to additional transmittance and scattering studies, and to perform more detailed sensitivity studies in support of the FireSat design and data-interpretation process. Systems Engineering Application to the FireSat Study      The systems engineering process has been applied from the inception of the FireSat study. The study process began with the establishment of the top-level goal for the FireSat instrument and continued with development of lower-level, supporting goals. These supporting goals were broken down further until system performance requirements were established. The criteria were met for inclusion in the list of performance requirements when a goal was deemed quantifiable, verifiable, and objective. This process developed a hierarchy of project goals that tied performance requirements directly to the top goal of the system. Technical as well as programmatic requirements were included.      The system performance requirements developed for FireSat included both constraints and performance measures. Constraints were defined as requirements whose numeric targets had to be obtained, but no premium was placed on exceeding the target values. Performance measures were those requirements in which more (or less) was considered to be better and would justify the expenditure of some other resource. Thus, the constraints were used to bound the problem and the performance measures to perform system4evel trade-offs. Only after the development of these performance measures was consideration given to possible system architectures. System variations were constructed and each characterized by projections of the performance values for that option. Any option failing to meet established constraints was eliminated. This led to the most-preferred set of options currently under consideration. FireSat Orbit Design and Coverage Analysis      Integral to the FireSat mission is the design of the orbit. There are three key factors in the design of the orbit: instantaneous spatial, long-term spatial, and

Table 12.6    Selected results from signal range modeling for FireSat
Channel	Condition (refl.)	SZA, deg.	Surface temperature (°K)	Radiance W/(cm2*Sr)	Photon flux/ (s*cm2*Sr)
1	Nominal (0.07) Max. radiance (0.9) Min. radiance (0.02)	45 0 45	280 280 280	5.73E-4 4.91E-3 4.19E-14	1.44E15 1.24E16 1.06E15
2	Nominal (0.07) Max. radiance (0.9) Min. radiance (0.05)	45 0 45	280 280 280	1.08E-4 1.73E-3 8.92E-5	5.58E14 8.95E15 4.62E14
3	Min. radiance (clear scene) Max. radiance (1 km-thick cirrus cloud) 0.5 km-thick cirrus cloud	45 0 45	a a a	3.65E-7 5.21E-6 1.48E-6	2.54E12 3.62E15 1.03E13
4	Nominal (0.25) Max. radiance (0.6) Min. radiance (0.05)	45 0 45	280 280 280	1.62E-4 5.54E-4 3.46E-5	1.40E15 4.79E15 3.00E14
5	Nominal (0.25) Max. radiance (0.6) Min. radiance (0.05)	45 0 45	280 280 280	6.10E-5 1.76E-4 6.76E-6	6.74E14 1.94E15 7.48E13
6	0.1 Refl. 0.1 Refl. 0.1 Refl. 0.1 Refl. 0.1 Refl. 0.1 Refl. 0.1 Refl.	45 45 45 45 45 45 45	260 300 340 600 800 900 1000	5.45E-6 1.08E-5 3.30E-5 4.03E-3 2.04E-2 3.53E-2 4.47E-2	1.02E14 2.01E14 6.15E14 7.49E16 3.81E17 6.57E17
7	0.1 Refl. 0.1 Refl. 0.1 Refl. 0.1 Refl. 0.1 Refl.	45 45 45 45 45	260 280 300 320 340	4.16E-4 5.70E-2 7.74E-4 1.04E-3 1.36E-3	1.78E16 2.44E16 3.31E16 4.43E16 5.80E16
a.    Channel 3 (1.38 µm) is opaque to the surface

periodic-temporal coverage. The instantaneous spatial coverage is a simple function of mission specifications and instrument performance. FireSat requires a spatial resolution size of lesS than 500 m. The physical limitations explained in the Instrument Concept section below yield an optimal altitude of 830 km. The chosen altitude is a trade between instrument performance (resolution and signal strength) and instantaneous spatial coverage.      The long-term monitoring mission of FireSat (globally between 75°N latitude and 60°S latitude) drives the minimum orbital inclination to greater than 60°. The long-term monitoring issues, along with the science request for constant sun angles (constant over a period of 30 to 70 orbits) drive the design orbit's

inclination to a sun-synchronous inclination at an altitude of 830 km.      Another requirement is to be able to view as much of the area of regard (75-60° latitude) as possible while the ground is lighted and in a 24-hour period. The frequency with which fires occur escalates during a hemisphere's summer season. Choosing an ascending nodal crossing time (for an altitude of 830 km circular sunsynchronous orbit) of 1400 local time will afford a few more minutes of light on the downward pass, thus increasing the covered area during the prime fire season.      The FireSat coverage analysis presented here was done for an instrument with a full field of view (FOV) of 96.0°. Table 12.7 shows the initial spacecrafi ephemeris used for all of the analysis presented here.

Table 12.7 FireSat ephemeris Altitude (km) Inclination (deg) 98.730 Eccentricity 0.0 Longitude of Asc Node (deg) -149.65 True anomaly (deg) 0.0 Epoch (yyyy/ddd/hh:mm:ss) 1998/172/00:00:00 Sensor FOV (deg) 96.0 Minimum ground elev. (deg)      The coverage of the design orbit is illustrated graphically in figures 12.7 through 12.9. Each of the figures is plotted on an equidistant cylindrical projection of the earth. For the purposes of the analysis herein, the term lighted ground swath refers to a satellite-to-sun angle of greater than 0.02 degrees, which in turn implies that at the subsatellite location the local sun angle is also a small positive number (after local sunrise). Figure 12.7 shows the entire earth (180 to -180° longitude and 90 to -90° latitude) along with FireSat's sensor swath for only the lighted portions of 24 hours worth of ephemeris. The swaths cover from approximately 60°S latitude northward to nearly over the northern pole and then descend to approximately 58°N latitude. There are 15 swaths in this figure. This includes the stunted swath which starts at the initial point of the ephemeris (see table 12.7). The gaps in the coverage are from -41 to 41° latitude in height by a maximum of approximately 5° in longitude at the equator. The coverage gaps have two striking results. First, the areas below and above -41 and 41° latitude respectively are covered 100% in a 2-hour period. This is significant because the number of viewing opportunities in the temperate-boreal areas of the globe is substantially less than those in the tropical regions (almost daily near the equator). Second, the gaps that appear in 24 hours of coverage are completely gone when the time limit for the evaluation is lengthened to 48 hours (because of nodal regression of the orbit). Also, the number of viewing opportunities ranges from two to eight above 41 0N latitude.      The scale of figure 12.8 has increased; here the bounds of the plot are from -180 to -20° longitude and 0 to 75° latitude (again, only lighted ground swaths are plotted over a cylindrical projection of the earth). The continental United States is approximately centered in the plot. Two of the gaps in coverage exttend into the United States. The relative sizes of the gaps are small with respect to the size of the United

States. One can conclude that there is a low probability of any particular area of the United States being missed in any single day (24 hours). Furthermore, because full global coverage is obtained in two days (48 hours), no portion of the globe would be without coverage in that period. Also, from about 59°N latitude there are four opportunities to view any one location.      The last orbital analysis figure (figure 12.9) shows the ground swaths for a single spacecraft for 24 hours of ephemeris. The ground swaths shown in this figure are for both local daylight and local night. The coverage gaps are very small--only about 150 in latitude by 5° in longitude and occur between -30 and 30° latitude.      The orbit designed for FireSat (table 12.8) meets most of the requirements as set forth in the science section of the chapter. The areas where the orbit design falls short are only minor and are easily solved when a more realistic time value is used. The methodology is as follows: The near-infrared (NIR) and the middle-infrared (MIR) channels of the instrument do not require the presence of daylight. Therefore, for the purposes of fire detection and short-term monitoring of "hot spots," the solid 2-hour ground swath model can be considered valid. For long-term monitoring of "fire scars," the 48-hour lighted ground swath is more than sufficient. FireSat Measurement Technique      FireSat obtains data by imaging the surface of Earth using a technique similar to pushbroom radiometry. In pushbroom radiometry, a linear array of detectors (an array with only one dimension) is moved over the surface to be imaged. Periodically, the detectors are sampled and the resulting data are stored. If the sampling occurs at the correct time interval, the data from sample to sample are contiguous and a map of the surface can be generated by reassembling the data contiguously. In FireSat, the linear arrays of the pushbroom radiometer are replaced by areal arrays, that is, arrays that have two dimensions and can obtain an image which is itself an area map of the surface scanned. Since the areal array has two dimensions, much more surface area is sampled in a single sample, and much more data is contained in a single sample. On the other hand, the time interval between samples is much greater because the spacecraft has to move farther to get to the proper location for the next sample. The net data rate is the same for both techniques, but the areal array allows time between samples to do any required on-board processing of the data.

Figure 12.7    FireSat single spacecraft lighted ground swath

Figure 12.8    FireSat single spacecraft lighted ground swath (North America)

Figure 12.9    FireSat single spacecraft 24-hour groun swath

The FireSat instrument is a hybrid instrument with both multichannel radiometer and imager characteristics. There are seven areal arrays (radiometer channels) covering a range of wavelengths from 0.5 µm to 8.5 µm. For optical simplicity the three longest wavelengths are imaged on one focal plane and the remaining four wavelengths share a second focal plane. The focal plane consists of an assembly of areal arrays arranged very precisely with respect to one another, called a mosaic, and appropriate spectral band-pass filters. The seven radiometer channels each have spectral bandwidths on the order of a few tenths of a µm. Each spectral bandwidth was chosen to maximize the signal content without undue interference from the intervening atmosphere. Each array contains the detector cells, or pixels, and the necessary readout electronics. The FireSat instrument composite field of view (FOV) is composed of images from three such mosaics concatenated electronically to form a single image. Some of the principal FireSat experiment parameters are shown in table 12.8. FireSat Instrument Concept The proposed baseline configuration for the FireSat instrument is shown in figure 12.10. The instrument is

composed of two instrument modules and a cryogenic cooler. The cryogenic cooler is mounted to the longwave instrument module to provide cooling for the focal plane to a temperature of about 80°K. Each instrument module contains three telescopes with their associated focal plane arrays. The center telescope points toward nadir while the outboard telescopes point off nadir at an angle of 32°. The composite field of regard formed by concatenating the images from the three telescopes is approximately 96°.      As FireSat orbits Earth at an altitude of 830 km, the field of regard sweeps across the surface in three slightly overlapping swaths. The geometry of the swaths and the variation of the footprint size is shown in figure 12.11. The width of the center swath on the surface is 480 km and the outboard swaths are each 772 km wide, for a total swath width of 2024 km. These values take into account the curvature of the earth, which falls away 80.9 km at the outer limits of the field of regard from a plane tangent to Earth at nadir.      The instantaneous field of view (IFOV) is constant at 0.3173 milliradians, but the size of the footprint of each pixel on the surface of Earth varies with the viewing angle off-nadir. At the center of the composite field of view the footprint is smallest, about 263 m2,

Table 12.8    Parameters of the FireSat experiment
Altitude	830 km	Small-Sat mission; five years or more
Data frequency	Daily revisit; 75oN-60oS	Sun-sync, Equator crossing: ~ 1400 h ascending
Spatial resolution	500 x 500 m	Experiment critical parameter
Footprint	263 x 263 m	Needed for > 30% MTF response at 500 m
IFOV per pixel	0.32 x 0.32 mr
Total Field of View	3 x 32°
Aperture	30 mm	Driven by channel 7 diffraction limit
Optical Speed	f/3
Optical efficiency	66%
Ch. 1 0.45-0.55 µm	Detect smoke + particulate	Range: = 1 to 100%, also see table 12.1
Ch. 2 0.98-1.07 µm	Map recently burned area	Range: = 1 to 100% also see talbe 12.1
Ch. 3 1.35-1.40 µm	Cirrus detection	Range: = 1 to 100%, also see table 12.1
Ch. 4 1.60-1.72 µm	Vegetation moisture	Range: = 1 to 100%, also see table 12.1
Ch. 5 2.1-2.3 µm	Fire scars	Range: = 1 to 100%, also see table 12.1
Ch. 6 3.6-3.8 µm	Active fire + scar w.smoke	Range: T = 260-1100°K, also see table 12.1
Ch. 7 8.0-9.0 µm	Surface Temperature	Range: T = 240-340°K, also see table 12.1
Split-focal plate FP-1: ch.1-5 FP-2: ch.6 and 7	SW channels are 7000 pix wide. Ch.2-5, use 7 arrays of 1000 pix each, ch. 1 uses two arrays each 3500 pixels; ch. 6 is 630 x 190 pixels; ch 7 is 380 x 190 pixels; ch 1 and 2 at 300°K; ch. 3, 4, and 5 at 150-180°K; ch. 6 and 7 at 80°K.
Array dwell time	5.8 sec (40.6 km/7 km/sec)	integration time may be adjusted over wide range Smart Sensor: adaptive-gain or instant-IFC
Data rage	2 Mbits/sec	TRW (RIM) flight-qualified recorder is capable of 22 Gbyte, 80-160 MB/sec
Data volume	3.6 Gbits, or < 1 Gbytes	2 Mbits/sec down-loaded every 30/min
Pointing knowledge	0.15 x 0.15 mrad	Pointing is equal to IFOV/2
Spacecraft jitter	µ ± 0.5 degree/sec (3I)	Equals IFOV/dwell time, or 0.3 mr/30/msec

but due to increasing slant range from the spacecraft to the surface with increasing look angle, the footprint size increases gradually, which is clearly shown in figure 12.11. Figure 12.12 shows the variation of the footprint multiplier with look angle, which increases to approximately 1.64 times the nadir value at 48° off nadir. In addition to experiencing a magnification due to increasing slant range, the footprint also becomes elongated due to the incidence angle being nonorthogonal to the surface. At 48° the footprint of a single ptxel becomes roughly rectangular, measuring approximately 431 m x 795 m. FireSat Optical Design The telescopes used in the FireSat instrument are approximately 30 mm diameter, are f/3 speed, and are of lightweight design, each telescope weighing less

Figure 12.11    FireSat measurement geometry and effective footprint on Earth's surface

Figure 12.12    Footpritn multiplier compared to look angle off-nadir

than 0.5 kg. Each telescope covers a 32° swath width with excellent spatial resolution. The optical performance, and radiation throughput, is optimized by using separate sets of telescopes for the visible/NIR and the MIR spectral regions. The multieletnent design is color~orrected over the wavelength range of interest. The spatial resolution of the visible/NIR telescopes is limited by aberrations, and their spatial resolution at Earth's surface is better than the 500 x 500 m spatial resolution science requirement. For the longer wavelength channels, diffraction effects predominate to degrade spatial resolution. Radiometer channel 7, which operates at 8.5 µm nominal wavelength, has an effective footprint of 500 x 500 m. FireSat Radiometric Performance      The radiometric sensitivity of the detector channels in FireSat is determined by the time allowed for integration of photons from the scene that are incident on the pixels. The integration time required for the FireSat channels is on the order of milliseconds. Additional time is required for subsequent readout of the data from the arrays. The total time required is on the order of milliseconds; this time is called one array data cycle time. The arrays are sized so that the time it takes to move one array length along Earth's surface is significantly shorter than one array data cycle time. To this end the arrays are each 200 pixels in length, resulting in a 40-km-long footprint per array. Thus, at orbital speed, it takes 5.8 sec to move the array footprint a distance equal to its size. Note that each 5.8-sec period contains several array data cycle times. Thus, multiple data cycles are possible while the instrument travels the distance equal to the size of an array footprint. The FireSat experiment adds to its operational flexibility by using several different photon integration times per 5.8 sec period. Table 12.9 shows the predicted radiometric performance of channels 1 through 5. The table is a spreadsheet, showing several rows of information for these channels, culminating in signalto-noise determination. The values shown in these rows bracket the instrument performance for the range of input radiance values expected for these channels. (The input radiance values are taken from table 12.6) Table 12.10 summarizes the performance of the 3.7 µm fire-detection channel. Table 12.10 shows parameters for a high-gain and a low-gain channel, which result from two integration times for the output of the same detector array. This use of two separate and different integration times per footprint sample has several advantages. A desirable very large dynamic measure-

ment range is possible without exceeding available detector full-well capacity. Also, the instrument can use analog-to-digital converters with a reasonable-to-implement 14-bit range. In fact, if three ranges are used (i.e., low-, medium-, and high-gain ranges), then the dynamic range for each channel allows the use of more common 12-bit analog-to-digital converters. Table 12.11 shows the performance of the 8.5-µm background measurement channel. FireSat Radiometric Calibration      The biomass burning experiment requires the measurements in the 3.7- and 8.5-µm channels to be calibrated to ± 5%. The preflight primary calibration can draw on techniques that were proven for several previous space-based remote sensors such as LIMS, HALOE, and SABER.      In-flight radiometric calibration can be viewed as a check on both the offset (i.e., y-axis intercept) and the gain (i.e., slope) of the (straight-line) calibration curve. Typically the offset is checked by letting the sensor view space to provide a "zero-radiance" input to the instrument. Then, while still viewing space, a longterm stable source is input to check the instrument channel gain.      The in-flight calibration for the FireSat sensor presents special issues. The sensor is nadirlooking and to control reliability, weight, and cost, uses no moving optical parts. Thus no "space-look" can be obtained to check sensor output offset drifts and in-flight calibration must be performed while the sensor receives nonzero radiance from Earth. The FireSat instrument "smart sensor" capability, mentioned above, enables separating the instrument offset from Earth input radiance while viewing uniform Earth scenes that are slowly changing. It is worth noting that for the 3.7-µm fire-measuring channel the predicted offset due to total cavity radiation is less than 1%.      An in-flight check on the slope of the calibration curve (i.e., the monitoring of the end-to~end radiometer channel gain) can be accomplished using known radiance inputs to the instrument. FireSat uses diode lasers as sources for its in-flight calibration of the visible and NIR channels. For the longer wavelength channels the calibration sources will be thermo~electrically cooled, quaternary, type IV-V1 diodes. Current research has identified candidate material systems for these diodes and commercially available devices are expected to be available in the near term (McCann 1995). The diode sources, in turn, are viewed by diode monitors to check the long-term source stability. The

Table 12.9    Predicted radiometric performance for FireSat channels 1 through 5
Channel	Conditiona	Photon Flux/ (S*cm2*Sr)	Photonsb/ (S*pixel)	Photon/ pixel	Q.E., %	Signal e-1	Dark noise, e-1	FPA noise, e-1	Photon noiose, e-1	Total noise, e-1	SNR
1	Nominal Max. Radiance Min. Radiance	1.44E15 1.24E16 1.06E15	4.10E7 3.53E8 3.02E5	1.64E5 1.41E6 1.21E5	25 25 25	4.10E4 3.53E6 3.02E4	4 4 4	16 16 16	203 594 174	204 595 175	201 593 173
2	Nominal Max. Radiance Min. Radiance	5.58E14 8.95E15 4.62E14	1.59E7 2.55E8 1.32E7	6.36E4 1.02E6 5.27E4	3.5 3.5 3.5	2.23E3 3.57E4 1.84E3	4 4 4	16 16 16	47 189 43	50 190 46	45 188 54
3	Clear Scene Max. Radiance	2.54E12 3.62E13	7.24E4 1.03E6	2.90E2 4.13E3	60 60	1.74E2 2.48E3	173 173	176 176	13 50	177 183	1 14
4	Nominal Max. Radiance Min. Radiance	1.40E15 4.79E15 3.00E14	3.99E7 1.37E8 8.55E6	1.60E5 5.46E5 3.42E4	65 65 65	1.04E5 3.55E5 2.22E4	173 173 173	176 176 176	323 596 149	368 621 231	283 571 96
5	Nominal Max. Radiance Min. Radiance	6.74E14 1.94E15 7.48E13	1.92E7 5.53E7 2.13E6	7.68E4 2.21E5 8.53E3	60 60 60	4.61E4 1.33E5 5.12E3	173 173 173	176 176 176	215 365 72	278 405 190	166 328 27
a. See Table 12.6 for definition of condition.
b. Phot/pix/sec: [Wcmsr1]. [Aoptics]*]Ph/sec/watt] = [L]*[2.85 E-8]*[5.0314E18*].       Standard front-side illuminted silicon CCD at room temperature 27 x 27 pixels.       NICMOS-type HgCdTe with switched FET multiplexer, at 150-180°K.       Dynamic range of 9 bits is sufficient for all SW channels.       Inherent CCD dynamic range = full-well electrons + noise electrons, or 175000/20 = 8750, or ~ 13 bits.       Integration time for channels 1 through 5 is 4 ms and Read-noise for chnnels 1 through 5 is assumed to be 15 electrons.

Table 12.10  &nsp; Radiometric performance for FireSat channels 6a and 6b; bandpass: 3.6-3.8 µm
Scene, Temp/	Lscene, Wcm-2sr-1 a	L/°K, Wcm-2sr-1 a	Scene, e-1/pixel b	Scene + cav, e/pixel c	e-1/°K, e-1/pixel b	Noise, (e-1scene+cav), (1)	NET, °Ks, (1)	SNR, (1)
High gain channel (260-500 kelvins)
260/0.9	5.45E-6	5.00E-8	3.70E+4	1.01E+5	340	316	0.93	117
350/0.9	4.40E-5	1.27E-6	2.99E-5	3.62E+5	8636	602	0.070	497
500/0.9	1.10E-3	1.70E-5	7.48E+6	7.48E+6	115600	2746	0.024	2734
Dynamic range 22000 (14+ bits)
Low gain channel (450-1100 kelvins)
450/0.9	4.68E-4	8.90E-6	1.92E+4	1.96E+6	365	140	0.38	137
700/0.9	1.02E-2	8.00E-5	4.18E+5	4.18E+5	3280	647	0.20	646
1100/0.9	7.87E-2	2.60E-4	3.22E+6	3.22E+6	10660	1794	0.13	1795
Dynamic range 8790 (13 bits)
a. Scene input obtained from MODTRAN: (76 STD; 830 km alt; 45° SZA (3 PM); = 0.1; visibility = 23 km.)
b. Integration time for high-gain channel, = 1.75 msec; for low-gain channel, = 10.5µ sec; also, optics = 50%. So that for the high-gain channel: e-1/pixel = [Wcm-2sr-1]. [Aoptics].[Ph/sec/Watt].[].[t] = [L].[3.2E-7].[5.031E18*].[0.65].[1.75E-3] = [L].[6.8E+9]. While, for the low-gain channel: e-1/pixel = [Wcm-2sr-1].[Aoptics].[Ph/sec/Watt].[].[t] = [L].[3.2E-7].[5.031E18*].[0.65].[1.05E-5] = [L].[4.1E+7].
c. Cavity e-1300K optics) = [L].[A].[effective].[Photons/sec/Watt].[].[t] = [2.6E-5].2.3E-6].[0.05].[5.031E18*].[0.65].[t] = 3.6E+5.[t] or 6.3E+4 electrons and 3.78 electrons for high-gain and low-gain, respectively
The detector is assumed to be a NICMOS-type HgCdTe at 80°K and 27 µm pixels with switched MOS-FET multiplexer and a quantum efficiency of 0.65.
The high-gain channel requires a detector D* = 2.85E+12 cm.Hz1/2.W-1; while D*BLIP = 3.7E+12 cm.Hz1/2.W-1.
The low-gain channel requires a detector D* = 2.05E+11 cm.Hz1/2.-1; while D*BLIP = 3.7E+12 cm .Hz1/2.W-1M.
Full-well: The high-gain channel MUX for a 500°K scene has 7.5E+6 electrons; while the low-gain channel MUX for a 1100°K scene has 3.25E+6 electrons.
Array readout time: 0.4 sec (239400 pixels transferred at a speed of 598 K pix/sec); read noise less than 50 electrons.

Table 12.11    Radiometric performance for FireSat channels 6a and 6b; bandpass: 3.6-3.8 µm
Scene, Temp/	Lscene, Wcm-2sr-1 a	L/°K, Wcm-2sr-1 a	Scene, e-1/pixel b	Scene + cav, e/pixel c	e-1/°K, e-1/ pixel b	Noise, (e-1scene+cav), (1)	NET, °Ks, (1)	SNR, (1)
240/0.9	3.08E-4	4.5E-6	1.37E+6	2.77E+6	20025	1664	0.083	823
260/0.9	4.16E-4	6.6E-6	1.85E+6	3.25E+6	29370	1803	0.061	1026
300/0.9	7.74E-4	1.2E-5	3.44E+6	4.14E+6	53400	2200	0.041	1564
340/0.9	1.36E-3	1.8E-5	6.05E+6	7.45E+6	80100	2724	0.034	2216
Dynamic range 306 (i.e. 9 bits)
a. Scene input from MODTRAN: (76 STD; 830 km alt; 45° SZA (3 PM); = 0.1; visibility = 23 km.)
b. e-1/pixel = [Wcm-2sr-1. [Aoptics] . [Ph/sec/Watt] . [] . [t] = [L] . [3.2E-7] . [5.0314E18*] . [0.65] . [0.5E-3] = [L] . 4.45E+9; where, intergration time, t = 0.5 msec; wavelength, = 8.5 µm; and quantum efficiency, = 0.65.
c. Cavity e-1(300 K optics) = [L] . [A] . [effective] . [Photons/sec/Watt] . [] .[t] = [9.5E-4] . [2.3E-6] . [0.05] . [5.0314E18*] . [] . [t] = 1.58E+6 electrons.
The detector is assumed to be a NICMOS-type HgCdTe at 80°K and 27 µm pixels with switched MOS-FET multiplexer and a quantum efficiency of 0.65.
The channel requires a detector D* = 5.9E+10 cm. . Hz1/2 . W-1; while D*BLIP = 9.6E+11 cm . HZ1/2 . W-1.
Full-well on the MUX for a 340°K scene equals 7.5E+6 electrons.
Array readout time: 0.4 sec (144400 pix transferred at a speed of 57760 pix/sec); read noise less than 500 electrons.

output of one set of these sources illuminates the focal plane directly. These sources will he electronically controlled to provide two levels of input to their respective radiometer channels. The time response of the sources is very rapid; calibration can therefore be checked as often as desired. Another set of diodes will illuminate a row of pixels at the edge of each array, just outside the experiment field of regard. In a Jones method calibration (Jones 1960), the energy from these diodes will pass through the entire measurement channel optical system. This allows a check of the optics and detector response, assuming the average response of all array pixels changes proportionally with the average of the edge-row pixel responses. Firesat Photodetector Arrays      Fortunately, the wavelength ranges of FireSat are coincident with other popular applications, including visible-NIR imaging (0.5 µm-1 µm), astronomical spectroscopy (1 µm-2.2 µm), terrestrial thermal imaging for military applications such as night vision systems for tanks (8 µm-9 µm), and other tactical military thermal imaging (3.7 µm). Accordingly mature photo-detector and focal plane array technologies suitable for FireSat already exist. For the visible (0.5 µm) and first NIR (1 µm) channel, silicon charge-coupled device (CCD) technology, which allows near-photon counting capability, can provide focal plane arrays (FPAs) with small pixel sizes (as low as 9 µm). Such arrays are available in megapixel formats at relatively low cost, high yield, and high reliability due to the excellent material properties of silicon and the high level of

maturity of processing techniques, which is a result of the enormous commercial market for other devices based on metallic oxide semiconductors (MOS) such as computer memory chips (Sze 1988).      For the other wavelengths, the material problems are somewhat more diiiicult, because silicon does not respond to wavelengths longer than 1 µm or so. Because of the considerable importance of infrared (IR) detection, especially for military purposes, extensive resources have been devoted to development of IR imaging technology over the past 30 years. The most widely used and successfully developed material system is mercury cadmium telluride (HgCdTe). By changing the relative proportions of mercury and cadmium (in a stoichiometric sense), the wavelength response of these detectors can he tuned from NIR to FIR (Wolfe and Zississ 1978). This is particularly useful since photodetector leakage currents, which create noise, increase rapidly with the cut-off wavelength and temperature. Thus, a device with a cut-off wavelength as close as possible to the desired detection point will require less cooling for satisfactory operation than one with a cut-off wavelength that is greater than that of the useful channel. Furthermore, HgCdTe has several desirable material properties that lead to a lower dark current for a given cut-off wavelength than for indium antimonide (InSb), making it the material of choice for most shortwave infrared-medium-wave infrared (SWIR-MWIR) applications (DeWames et al. 1992).      Several important detector/FPA constraints and trade-offs have led to the current FireSat detector concept. First, satisfactory operation at the FireSat wavelengths can be obtained with temperatures

achievable by thermoelectric cooling (185-200°K) in the visible to SWIR (0.5-2.20 µm) channels. The other two channels (3.7 µm and 8.5 µm) require cooling to 80°K or so, mandating an active cooler. Higher temperature detectors with background limited performance (BLIP) ability close to at relatively low background fluxes might alleviate this situation and make radiative cooling feasible, but this will require substantial technological advances.      Furthermore, FireSat requires coverage over a large field of view, which in turn requires very wide detector arrays. FPA size is driven by two constraints. First, although IR detector technology is fairly mature, limitations of fundamental material and device physics preclude fabrication of complicated, high-performance transistor structures that are needed to perform readout and signal-conditioning functions on materials such as HgcdTe (Barbe 1980). Thus, current state-of-the-art IR FPAs consist of a photodiode array fabricated on HgCdTe hybridized to a Si readout chip, or multiplexer (MUX), which performs the necessary signal processing (Jenkins 1987). Although highly developed, fabrication limitations prohibit die sizes of greater than approximately 2 x 2 cm (Kozlowski 1995). In addition, the extreme difficulties in processing IR photodetector materials such as HgCdTe place a limit on the size of the photodiode arrays themselves. This difference depends on the substrate on which the material is grown. For cut-off wavelengths of less than 5 µm or so, sapphire substrates are possible and FPA sizes of about 1.7 x 1.7cm can be obtained. For longer wavelengths, however, cadmium zinc telluride (CdZnTe) must be used with a resulting size limit of 1 x 1 cm or so (Kozlowski 1995). The intersection of these limits, of course, determines the maximum available FPA size. For Si CCDs, a similar limit exists, but it is much larger, on the order of 3 x 3 cm or even slightly larger (Janesick and Elliot 1991).      The desire for as small an FPA as possible, driving the design to smal] pixel sizes, collides with the other limitation, that of full-well capacity. Both CCDs and hybridized IR FPAs work by integrating photogenerated charge, which is read out sequentially by an on-chip amplifier. The full-well capacity associated with each pixel, which is the maximum amount of charge that can be integrated without saturation, determines the maximum signal handling capacity of the FPA. To accommodate a large dynamic range, this should be as great as possible. Since full-well capacity scales with pixel area, one way to improve it is to use larger pixels. For a given number of pixels needed to accommodate a given field of view at a given resolution

this will lead to the requirement for a physically larger array. A careful trade-off study for FireSat has led to the choice of 27-µm pixels for the thermal channels and 60-µm pixels for the visible NIR channels. This in turn determines the number of arrays of a given size that must be used to obtain the desired coverage. FireSat Focal Plane Cooling      The FireSat instrument is divided into two thermal zones. The short-wavelength detectors operate near the spacecraft's ambient temperature. The long-wavelength detectors require cooling to 80°K to achieve sufficient sensitivity. FireSat will use a mechanical cryogenic refrigerator to provide focal plane cooling below 80°K. Lightweight, low-power miniature refrigerators have been developed that provide 400+ milliwatts of cooling at 80°K. Innovative techniques such as fiber support technology (Batty et al. 1994) can be employed to thermally isolate the cooled focal plane from its surroundings, reducing the cooling requirements 9f the refrigerator. FireSat Mechanical Design      The FireSat instrument's mechanical design is lightweight and simple in nature yet very functional. It is packaged in a single module containing all sensor mechanical, optical, and electronic subsystems. Structurally it consists of an aluminum monocoque structure designed primarily to provide and maintain the optical relationship required by the placement of the focal plane assemblies (FPAs) and telescopes, and to maintain the stringent thermal conditions resulting from the use of cooled FPAs and a mechanical cryocooler. In addition, the structure provides the load path required to withstand all launch loads and orbital thermal loads by interfacing with the spacecraft system. The instrument structure is also used to house and package many of the instrument's electronic subsystems A sophisticated thermal design provides rigidity and thermal isolation for the FPAs and efficiently conducts heat from the FPAs to a spacelook radiator via a Stirling-cycle refrigerator, thus providing the optical alignment and its maintenance required by the six telescopes and their respective FPAs. FireSat Thermal Control      The estimated heat load of the FireSat sensorcraft is 100 watts, of which 30 watts is produced by the mechanical cryocooler and its control electronics. Due to

the small size of the instrument, much of its external surface area might have to be devoted to radiators for thermal control. Having the spacecraft bus handle the instrument's heat rejection may be a more effective alternative. FireSat Spacecraft System      Given the FireSat instrument's mass and power estimates of 23 kg and 100 watts, respectively, and the orbital requirements of 830 km circular sun-synchronous at 98.7° inclination, a state-of-the art "small-sat" bus design should surffice. The power, attitude control, and thermal control requirements appear to be reasonable and well within the capabilities of existing bus designs produced by several aerospace companies. Assuming a direct injection by a Pegasus-class launch vehicle, the mass fraction of the FireSat instrument is 12%, which is well within the range normally achieved. The cost of a bus wil~l be from $5 to $15 million depending on redundancy requirements and manufacturer. FireSat Launch Vehicle System      Two expendable launch vehicle systems are being considered for the FireSat mission. The first is a Pegasus-class vehicle capable of placing approximately 200 kg into a circular 830-km sun-synchronous orbit by direct injection. Alternatively, the use of an 185-km parking orbit and an integral propulsion system to perform a Hohmann transfer to an altitude of 830 km will increase the payload capability to more than 260 kg. There would be a 10- to 15-kg weight penalty required for the propulsion system. Launch systems in the LLV-1-class could provide higher performance (i.e., larger payloads or higher orbits). A direct injection of 290 kg to 830 km sun-synchronous can be accomplished. Alternatively, the use of a 185-km parking orbit and a Hohmann transfer to 830 km will deliver 388 kg to orbit. New Technology for FireSat      The FireSat concept is based on available technology; the performance required from each individual component or subsystem has been demonstrated. However, there is significant challenge in combining these components and subsystems into a space-based sensor to provide the challenging FireSat science measurements. Additional technology development could significantly reduce the instrument development risk

and improve performance. We have identified several technological developments for the FireSat mission.      Benefit would be derived from the availability of highly integrated NIR-MIR detector readout multiplexers, for low-power compatible cryogenic temperature focal plane application. For maximum benefit this subsystem should simultaneously provide (1) onchip correlated double sampling; (2) onchip analogt~digital converters with 14-bit dynamic range, capable of >1 MHz conversion speed; (3) onchip smart-sensor-like adaptive sensor functions; (4) fastreadout (>1 MHz) with low readout noise (<20 electrons); and (5) full-well capacity in excess of 20 million electrons for a pixel size of approximately 25 x 25µm.      Ideally, the large format FPAs (MUX plus detector) for FireSat would also have the following characteristics: (1) dimensions of 3 x 1 cm, or larger for channels in both the 3 to 5 µm and 8 to 10 µm range; (2) capability of high temperature, near-BLIP operation (>1 100K) at low backgrounds and with extremely low drift characteristics; and (3) high-performance NIR and Mid-IR 'twocolor' detector arrays.      For FireSat to benefit from its inherent Smart Imager capabilities the validation of the following techniques would be very useful: (1) multiple and adaptive gain concepts for full-well optimization; (2) demonstration of smart-sensor offset determination; and (3) for a high capability in-flight calibration subsystem of the 3.7- and 8.5-µm channels that used no moving parts, demonstration of quaternary IV-VI (e.g., PbSnSeTe) light-emitting diodes (LEDs) for operation at thermo-electric cooler temperatures.      FireSat development would also benefit from development of the following on-board data-processing technology: (1) boresighting of the seven radiometric channels by removing time delay between channel outputs; (2) automated cloud filter to remove most clouds for reduced data transmission; and (3) real-time ground station capability to automatically process FireSat data for fire monitoring.      The FireSat microspacecraft would benefit from improvements in several hardware disciplines including (1) lightweight, thermally stable structures; (2) low-power, high-speed digital electronics for data down-linking; (3) improved spacecraft attitude control system to provide IFOV pointing knowledge to 0.12 milliradians; (4) high-capacity, efficient, lightweight, reliable cryocooler compressor to allow added heat from smart-sensor, on-focal plane processing; and (5) for alternate methods of focal plane cooling, efficient high thermal load crysogenic temperature radiators.

References Andreae. M. 0.1991. Biomass Burning: Its History, Use, and Distribution and Its Impact on Environmental Quality and Global Change. Global Biomass Burningg: Atmospheric, Climatic. and Biospheric Implications (J. S. Levine, editor), The MIT Press, Cambridge. Mass.. pp.3-21. Andreae. M. 0., J. Fishman, M. Garstang, J. G. Goldammer, C. 0. Justice. J. S. Levine, R. J. Scholes, B. J. Stocks, A. M. Thompson, and B. van Wilgen. 1994. Biomass Burning in the Global Environment: First Results from the IGAC/BIBEX Field Campaign SATARE/TRACE-A/SAFARI-92. Global Atmospheric-Biospheric Chemistry (R. 0. Prinn, editor), Plenum Press, New York, pp.83-101. Barbe. D. F. (editor). 1980. Charge-Coupled Devices, Springer-Verlag, Berlin, pp. 126-128. Batty, I C., B. 0. WIlliams, W. W. Burt, and W. A. Roettker. 1994. Cooling SABER: Interface, Performance, and Design Issues, SPIE Proceedings Vol.2227-24, International Symposium on Optical Engineering in Aerospace Sensing, Orlando, Florida. Bowker, D. E., R. E. Davis, D. L. Myrick, K. Stacy, and W. T. Jones. 1985. Spectral Reflectance of Natural Targets for Use in Remote Sensing Studies, NASA Reference Publication RP-1139, 184 pages. DeWames, R. E., I M. Arias, L. I Kozlowski, and G.M. Williams 1992. An Assessment of HgCdTe and GaAs/AlGaAs Technologies for LWIR Infrared Imagers, Infrared Detectors: State of the Art, Society of Photo-Optical Instrumentation Engineers (SPIE) Vol. 1735, pp.2-16. Hao, W. M., and M.-H. Liu. 1994. Spatial and Temporal Distribution of Tropical Biomass Burning. Global Biogeochemical Cycles, 8, pp.495-503. Houghton, R. A. 1991. Biomass Burning from the Perspective of the Global Carbon Cycle. Global Biomass Burning: Atmospheric. Climatic, and Biospheric Implication (J. S. Levine, Editor), The MIT Press, Cambridge, Mass., pp.321-325. Hummel, J. R., D. R. Longtin, N. L. DePiero, and R. I. Grasso. 1992. Backscat LIDAR Simulation Version 3.0: Technical Documentation and Users Guide. Scientific Report No.3, SPARTA, Inc. Lexington, MA. (Available as report PL TR-92-2328 from Phillips Laboratory, Directorate of Geophysics, Air Force Materiel Command, Hanscom Air Force Base, Mass. 01731-3110). Janesick, J. R., and T. Elliot. 1991. History and Advancement of Large Area Array Scientific CCD Imagers, Astronomical Society of Pacific Conference Series '91, Tucson, Ariz. Jenkins, T. E. 1987. Optical Sensing Techniques and Signal Processing, Prentice-Hall, Englewood Cliffs, N.J., pp.89-95. Jones, R. Clark. 1960. Proc. Infrared Information Symposium (IRIS) 5(4). pp.35. Killinger, D. K., and W. E. Wilcox. 1992. USF HITRAN-PC Installation and Software User Manual. (Commercial Version available from ONTAR Corp, North Andover, Mass. 01845-2000). Kneizys, F. X., E. P. Shettle, L. W. Abreu, J. H. Chetwynd, G. P. Anderson, W. 0. Gallery. I E. A. Selby. and S. A. Clough. 1988. Users Guide to LOWTRAN7. (Available as Report AFGL-TR-88-0177 from Phillips Laboratory, Directorate of Geophysics, Air Force Materiel Command, Hanscom AFB, Mass. 01731 - 3110). Kozlowski, J. T. 1995. Personal communication.

Levine, J. S. (editor). 1991. Global Biomass Burning: Atmospheric, Climatic. and Biospheric Implications. The MIT Press, Cambridge, Mass., 569 pages. Levine, J. S., W. R. Cofer, III, D. R. Cahoon, and E. L. Winstead. 1995. Biomass Burning: A Driver for Global Change. Environmental Science and Technology, 29, pp. 120A-125A. Mano, S., and M. 0. Andreae. 1994. Emission of Methyl Bromide from Biomass Burning. Science, 263, pp.1255-1257. McCann, P. J., L. Li, J. E. Furneaux, and R. Wright. 1995. Optical Properties of Ternary and Quaternary IV-VI Semiconductor Layers on (100) Barium Fluoride Substrates, Applied Physics Letters, 66, 1355. National Academy of Sciences. 1990. One Earth, One Future: Our Changing Global Future. National Academy Press, Washington, D.C., pp. 116-130. National Research Council. 1989. Global Change and Our Common Future. National Academy Press, Washington, D.C., p.1. ONTAR, inc. 1992. PCMOdTRAN User's Manual. Available from the ONTAR Corp., North Andover, Mass. 01845-2000. Penner, J. E., M. M. Bradley, C. C. Chang, L. L. Edwards, and L. F. Radke. 1991. A Numerical Simulation of the Aerosol-Cloud Interactions and Atmospheric Dynamics of the Hardiman Township, Ontario, Prescribed Burn. Global Biomass Burning: Atmospheric Climatic, and Biospheric Implications (J. S. Levine, editor), The MIT Press. Cambridge, Mass., pp. 420-426. Seiler, W., and P.J. Crutzen. 1980. Estimates of Gross and Net Fluxes of Carbon between the Biosphere and Atmosphere from Biomass Burning. Climatic Change, 2, pp.207-247. Sze, S. M. 1988. VLSI Technology, 2nd edition. McGraw-Hill Book Company, New York, NY, pp.1-7. Wolfe, J., and 0. Zississ (editors). 1978. The Infrared Handbook. The Infrared Information and Analysis (IRIA) Center, Environmental Research Institute of Michigan, pp.11.91-11.93.