The data set group consists of: a) statistical values calculated from numerous satellite and aircraft images covering portions of the FIFE study area and b) values for use in performing atmospheric correction of portions of the imagery.

The satellite images represented in the data sets were collected by the following instruments:

• NOAA 9, 10, and 11 AVHRR-LAC
• GOES 7 VISSR
• Landsat 4 and 5 Thematic Mapper (TM)
• SPOT 1 High Resolution Visible (HRV)

The aircraft images represented in the data sets were collected by the following instruments:

• NS001 Thematic Mapper Simulator (TMS)
• Advanced Solid-state Array Spectrometer (ASAS)

The atmospheric correction information (Satellite Image Value Conversion Coefficients) was used in calculating the site reflectances and is provided for use in correcting other portions or all of the visible and near-infrared bands from most of the AVHRR-LAC, Landsat TM, and SPOT HRV image data. Note that some of the corrections are based on actual optical thickness measurements but when data were not available a default value was used or provided which represents the median of all such measurements taken in 1988 and 1989 over the area.

The extracted image data includes the following features:

• Sensor-corrected radiance statistics for the available spectral bands at varying look angles and illumination conditions
• Exoatmospheric reflectances (ratio of total incoming to outgoing radiation of the earth-atmosphere system within the sensor bandpass)
• Atmospherically corrected surface reflectance at varying look angles and under various illumination conditions
• Surface temperatures (both apparent at-sensor and retrieved surface temperatures)
• Daily incident solar insolation and PAR
• Average hourly incident insolation, PAR, and surface albedo

The data in the Advanced Very High Resolution Radiometer (AVHRR) Local Area Coverage (LAC) data set consists of mean and standard deviation statistics derived from the pixels within each image that fell within the FIFE site area. Average at-sensor radiances from the acquired images are available for the five or four (NOAA 10) sensor wavebands. Exoatmospheric and surface reflectances are available for the visible (Band 1) and near-infrared (Band 2) bands. Surface temperature values derived using a split-window technique are available for the NOAA 9 and NOAA 11 images.

The entire FIFE area is designated by the SITEGRID_ID of "FIFE-LAC" in the data set and covers the area between 38 degrees 52 minutes 30 seconds and 39 degrees 07 minutes 30 seconds North latitude, and between 96 degrees 22 minutes 30 seconds and 96 degrees 37 minutes 00 seconds West longitude. The AVHRR-LAC extracted data provides spectral coverage from February 3, 1987 through October 13, 1989. During this period, relatively cloud-free data from 331 days are included. The periods of February 1987 through January 1988 and July through September 1989 contain the densest temporal coverage with approximately 16 days-per-month of coverage. The period of November 1988 through May 1989 does not contain any data. The original image data were acquired through the USGS EROS Data Center and the Naval Research Laboratory (NRL) and were processed by the FIFE Information Staff.

The data in the Geostationary Operational Environmental Satellite (GOES) Visible Infrared Spin-Scan Radiometer (VISSR) data set consists of tables of the following parameters which were derived from the pixels in each image that fell within the designated FIFE site area during the 57 days of the 1987 IFC periods:

• Daily integrated incident solar insolation and PAR (Photosynthetically Active Radiation)
• Daily average cloud cover estimates
• Daily average atmospheric water vapor estimates
• Daily average ozone estimates
• Half-hourly statistics of actual and clear-sky solar insolation and PAR, atmospheric liquid water content, and cloud albedo
• Surface albedo statistics calculated for half-hourly intervals across all days in a given Intensive Field Campaign (IFC)

The FIFE area covered by this data set is as follows:

• NW corner: 39 07 24 N, 96 37 02 W
• NE corner: 39 07 10 N, 96 25 56 W
• SW corner: 38 58 46 N, 96 37 19 W
• SE corner: 38 58 32 N, 96 26 15 W

These data were derived from the original GOES imagery by Dr. Robert Frouin and his staff at the Scripps Institute of Oceanography in La Jolla, CA, USA and delivered to FIS.

The data in the Landsat Thematic Mapper (TM) data set consists of mean and standard deviation statistics derived from the pixels within each image that fell within 39 FIFE ground measurement sites. The location and boundaries of these sites can be found in the FIFE Landsat Thematic Mapper (TM) Extract Data document. Average at-sensor radiances from acquired images are available for the seven TM sensor wavebands on the Landsat 4 and Landsat 5 satellites. Exoatmospheric and surface reflectances are available for the visible and near-infrared bands. Surface temperature values are available for the single thermal channel. The original image data were acquired through the EOSAT Corporation and were processed by the FIFE Information Staff.

The Landsat TM extracted data provides spectral coverage of the 39 FIFE ground measurement sites from April 9, 1987 through August 4, 1989. During this period, data from 14 relatively cloud-free images are included. The period of April 9, 1987 to October 18, 1987 contains seven image acquisitions; the period of April 27, 1988 through September 26, 1988 contains six images; and one image acquisition occurred on August 4, 1989.

The data in the SPOT (Satellite Pour l'Observation de la Terre) High Resolution Visible (HRV) data set consists of mean and standard deviation statistics derived from the pixels within each image that fell within 39 FIFE ground measurement sites. The location and boundaries of these sites can be found in the FIFE SPOT High Resolution Visible (HRV) Extract Data document. Average at-sensor radiances from acquired images are available for the three HRV sensor wavebands in multispectral mode or for the one HRV sensor waveband in panchromatic mode on the SPOT 1 satellite. Exoatmospheric and surface reflectances are available for all bands. The original image data were acquired through the SPOT Image Corporation and were processed by the FIFE Information Staff.

The SPOT-HRV extracted data provides spectral coverage of the entire FIFE site from March 20, 1987 through August 9, 1989. During this period, relatively cloud-free data from 42 three-band multispectral (XS) images and four single-band panchromatic images were acquired. The period of March 20, through November 9, 1987 contains 27 acquisitions; the period of April 13, 1988 through October 23, 1988 contains ten acquisitions; and the period of June 8, 1989 through August 9, 1989 contains five images.

The data in the Satellite Image Value Conversion Coefficients data set consists of coefficients needed for deriving surface reflectance values from the visible and near-infrared bands of the FIFE AVHRR-LAC, Landsat TM, and SPOT HRV Level-1 image products. These coefficients were used by FIFE staff in calculating site reflectances from pixel values extracted from the Level-1 imagery. In most cases, measured optical thickness data acquired at the FIFE site area were used to derive the coefficients. For the image acquisition times when no optical thickness measurements were available, a default value calculated from the median of all such measurements taken in 1988 and 1989 was used. The original optical thickness data were acquired through both individual FIFE investigators (see Atmospheric Optical Properties Data Group Summary) and FIFE staff efforts. The data were processed to coefficients using the algorithm of Fraser, et al. (1992) by the FIFE Information Staff. For the correction of the thermal channel data, LOWTRAN-7 (Kneizys, et al., 1988) was used.

The data in the ASAS data set consists of mean and standard deviation at-sensor radiance and atmospherically corrected reflectance factor statistics derived from the pixels in multiple subsites at various FIFE ground measurement sites. These data were processed and provided to FIS by Dr. James Irons and his staff at NASA GSFC in Greenbelt, MD.

The ASAS extracted data provide coverage on seven dates: July 10-11, 1987; August 10, 15 and 17, 1987; October 11, 1987; and August 4, 1989. The 1987 data include 16 different acquisitions (observation times), all from ASAS Site 45 (FIFE Site Grid ID 4439). Results from four different image subareas at this site were made for each acquisition (16 acquisitions x 4 subareas = 64 files). The 1989 data contain six acquisitions, three from Site 916 (FIFE Sitegrid ID 4439), and three from Site 906 (FIFE Sitegrid ID 2133). Site 916 datasets provide values for four image subareas, while for Site 906, two of the datasets have five subareas and one set has six subareas. This makes a total of 28 files for 1989.

The data in the NS001 data set consists of mean and standard deviation at-sensor radiance statistics derived from the pixels in each image that fell within 39 FIFE ground measurement sites. The location and boundaries of these sites can be found in the FIFE NS001 Extract Data document. Average at-sensor radiances from acquired images are available for the eight NS001 sensor wavebands. The original image data were acquired by the Medium Altitude Aircraft Facility at NASA Ames Research Center and were subsequently sent to and processed by the FIFE Information Staff.

The NS001 extracted data provides spectral coverage of the 39 FIFE ground measurement sites during the five IFCs in 1987 and 1989 on the following dates:

• 04-JUN-87 Afternoon
• 06-JUN-87 Afternoon
• 28-JUN-87 Afternoon
• 10-JUL-87 Morning
• 15-AUG-87 Morning and Afternoon
• 17-AUG-87 Afternoon
• 11-OCT-87 Morning and Afternoon
• 04-AUG-89 Morning
• 11-AUG-89 Afternoon

These scenes were selected during the NS001 processing to provide morning and afternoon coverage at least once in each IFC and to provide multi-day coverage during the dry-down periods in IFCs 1 and 3.

Data from the AVHRR-LAC instrument provides spatial resolution ranging from 1.1 km at nadir to approximately 4 km at a 55-degree view zenith. The spectral coverage (in micrometers) for the NOAA 9, 10, and 11 AVHRR-LAC instruments is as follows:

   Band            NOAA 9                  NOAA 10                NOAA 11


1          0.570 - 0.699          0.571 - 0.684          0.572 - 0.698

2          0.714 - 0.983          0.724 - 0.984          0.716 - 0.985

3          3.525 - 3.931          3.554 - 3.950          3.536 - 3.935

4         10.334 - 11.252        10.601 - 11.445        10.338 - 11.287

5         11.395 - 12.342        10.601 - 11.445        11.408 - 12.386

Processing changes over the three years of coverage by the original AVHRR-LAC data sources and occasional operational processing errors by FIFE staff are explained in the FIFE NOAA AVHRR Local Area Coverage (LAC) Extract Data document and should be reviewed in detail. The supplied atmospheric correction coefficients (which were used to derive the given reflectances) were based on median atmospheric optical properties.

Data from the GOES instrument provides spatial resolutions of 0.9 km and 6.9 km at nadir for the visible and thermal infrared channels, respectively. The spectral coverages for these bands are as follows:

• Visible: 0.55 to 0.75 micrometers
• Infrared: approximately 11.0 micrometers

The GOES extract data parameters provided were derived by Dr. R. Frouin and provided as site averages calculated from the pixels that fell within the defined FIFE site boundaries given above.

Data from the Landsat TM instrument provides nominal spatial resolution of 30 m. Although the thermal (Band 6) data in the images obtained from EOSAT have a spatial resolution of 30 meters, the original resolution collected by the sensor was 120 m. The 30-m resolution was derived by EOSAT using resampling of the original Band 6 data to a regular geometric grid. Since the FIFE site area was always within quad 2 (upper right quadrant) of the full Landsat TM scene, the view zenith angles over the extracted sites range from 3.3 to 5.6 degrees. The spectral coverage of the seven Landsat Thematic Mapper bands is as follows:

• Band 1 : 0.451 - 0.521 micrometers
• Band 2 : 0.526 - 0.615 micrometers
• Band 3 : 0.622 - 0.699 micrometers
• Band 4 : 0.771 - 0.905 micrometers
• Band 5 : 1.564 - 1.790 micrometers
• Band 6 : 10.45 - 12.46 micrometers
• Band 7 : 2.083 - 2.351 micrometers

Data from the SPOT-1 HRV1 and HRV2 instruments provides spatial coverage of 20 m in multispectral (XS) mode and 10 m in panchromatic (PAN) mode. With the off-nadir pointing capability of the SPOT satellite, the view zenith angles for the FIFE scenes range from 2.2 to 31.9 degrees. The spectral coverage of the SPOT bands is as follows:

• XS Band 1 : 0.50 - 0.59 micrometers
• XS Band 2 : 0.61 - 0.68 micrometers
• XS Band 3 : 0.79 - 0.89 micrometers
• PAN Band 1 : 0.51 - 0.73 micrometers

Sensor level radiance corrections for all three bands were made using gains calculated from information provided on the original tapes from SPOT Image Corp. Note that the at-sensor radiances included in the data set were based on a set of calibration coefficient calculation values which has since been revised. An improved set of calibration coefficient calculation values were received by FIFE from the Centre d'Etudes et de Recherches de Toulouse (CERT) in February 1993. An analysis of the effect of the new values on the radiance calculations and the new table of values is given in the FIFE SPOT High Resolution Visible (HRV) Extract Data document. Those wishing to adjust the existing values based on the improved values can do so at their own discretion.

The Satellite Image Value Conversion Coefficients data set provides coefficients needed for deriving surface reflectance values from the visible and near-infrared bands of the AVHRR-LAC, Landsat TM, and SPOT HRV imagery. The correction algorithm used for the visible and infrared channels was that of Fraser, et al., (1992). For the correction of the thermal channel data, LOWTRAN-7 (Kneizys, et al., 1988) was used. When available, in situ measurements of optical thickness were used to derive the corrections. When optical thickness measurements were not available, a default value equal to the median of all such measurements taken in 1988 and 1989 was used. These data were used since the temporal coverage of optical thickness measurements was better in 1988 and 1989 and the data were of better quality.

Data from the ASAS instrument provides spatial coverage of approximately 4.3 m from an altitude of 5000 m in the 1987 data and 6.9 m from an altitude of 8000 m in the 1989 data. The forward- and aft-pointing capability of the ASAS instrument causes the spatial resolution to vary for each look angle. The spectral band centers of the ASAS are as follows:

    Band         Center (um)           Band           Center (um)


* 2            0.468                16             0.666

* 3            0.482                17             0.680

4              0.496                18             0.695

5              0.509                19             0.710

6              0.523                20             0.724

7              0.539                21             0.739

8              0.551                22             0.753

9              0.565                23             0.768

10             0.582                24             0.782

11             0.594                25             0.798

12             0.609                26             0.813

13             0.625                27             0.828

14             0.637                28             0.842

15             0.652                29             0.857

                                    30             0.873

From a nominal altitude of 4878 m, data from the C-130-based NS001 instrument provides spatial coverage ranging from 12.2 m at-nadir to 29.5 m at a 50-degree view zenith. The spectral coverage of the NS001 bands is as follows:

• Band 1 : 0.458 - 0.519 micrometers
• Band 2 : 0.529 - 0.603 micrometers
• Band 3 : 0.633 - 0.697 micrometers
• Band 4 : 0.767 - 0.910 micrometers
• Band 5 : 1.13 - 1.35 micrometers
• Band 6 : 1.57 - 1.71 micrometers
• Band 7 : 2.10 - 2.38 micrometers
• Band 8 : 10.9 - 12.3 micrometers

NS001 calibration in the reflective channels was examined in detail, and instability problems in the on-board power supply for the lamp in the calibration sphere were discovered. The instrument was updated with redesigned circuits and a new internal lamp in 1988, and recalibrated. The thermal channel of the NS001 was not extensively evaluated as part of FIFE; however, examination of water targets at a variety of altitudes with pre-dawn data suggest that the instrument is stable and not affected by problems associated with other aircraft multispectral scanners (e.g., TIMS).

The spectral bands of the various imaging sensors provide both distinct and overlapping spectral information of the FIFE site areas at various times throughout the day and season. This information is important in characterizing both fine and coarse spatial, spectral, and temporal properties of the surface and changing vegetation cover. Extraction of the large FIFE site from AVHRR-LAC and GOES and the individual site extractions from Landsat TM, SPOT, NS001, and ASAS were performed to simplify comparisons of satellite and ground observations spectral observations by the FIFE investigators.

The image correction coefficients were derived to provide a consistent set of atmospheric correction parameters that could be used to generate corrected surface parameter images to compare with ground observations. These coefficients were also used to derive surface reflectance from the extracted at-sensor radiance statistics in the database.

These data were derived from and for use with the FIFE Level-1 imagery found on CD-ROM Volumes 2 through 5 and discussed in the Executive Summary. They can be compared to ground and helicopter measurements of spectral radiance also described in the Executive Summary.

Except for the GOES VISSR data, the data in this group are all from relatively clear (i.e., cloud-free) scenes which were screened, selected, and processed based on that attribute. Although, the at-sensor radiance corrections were performed with the most current information available at the time, subsequent review of the data and sensor systems may show that the applied corrections should be adjusted (which happens to be the case with the described SPOT site extractions and possibly with the TM thermal band). In addition, the subsequent calculation of the reflectance values from at-sensor radiances is dependent on the accuracy of the view and illumination information and the atmospheric characterization information. The precision of the view and illumination angles supplied is believed to be plus or minus three degrees or better. As noted previously, some of the atmospheric correction coefficients were calculated from actual atmospheric condition information for that day while others are median values based on observations made during 1988 and 1989.

The extraction of the site data for AVHRR-LAC data was performed in a fairly automated fashion by software that used the pixel-by-pixel latitude and longitude locations to determine if a given pixel was within the set boundaries of the particular sites. The calculated pixel locations were determined to be within 1 pixel of the actual location based on location information from various sources. Due to the large pixel size of AVHRR-LAC and the shifting overpass track of the NOAA satellites, the effects on the average responses due to edge pixels should be noted. Although the AVHRR-LAC scenes used for site extraction were relatively cloud-free, detailed checks were not performed to determine if all the pixels extracted within the FIFE site area were cloud-free. Even if the FIFE site area looked cloud-free, the coarse spatial resolution would have prevented detection of sub-pixel size clouds that may have been present. All the atmospheric correction coefficients for AVHRR-LAC are median values based on observations made from 1988 to 1989. Based on analysis of 68 FIFE LAC channel 2 (infrared) reflectances in 1987, there is a large periodic oscillation of reflectance with the 9-day orbital cycle of the satellite (NOAA 9). The bi-directional component of the reflectances (25% to 60% variation in reflectance within a given 9-day period) is substantially larger than reflectance variation (22 to 30%) due to seasonal variability in biological components of the vegetation, such as green LAI. There is also an asymmetric range of solar illumination angles for forward versus back-scatter viewing directions, which increases non-linearity in relationships between NDVI and vegetation parameters.

As with the AVHRR-LAC, the Landsat TM and SPOT data extractions were performed in an automated fashion by software that used the pixel-by-pixel latitude and longitude locations to determine if a given pixel was within the set boundaries of the particular sites. The calculated pixel locations were determined to be within 1 pixel of the actual location based on checks made by FIS staff. However, since these data were extracted from smaller areas, the effects from edge pixels may be more obvious. Although the Landsat TM and SPOT scenes used for site extraction were relatively cloud-free, detailed checks were not performed to determine if all the pixels extracted within the site areas were cloud-free. Some of the data were corrected with median atmospheric correction information while others were corrected with date and time specific atmospheric condition information. The details of this are provided in the FIFE Landsat Thematic Mapper (TM) Site Extract Data and FIFE SPOT High Resolution Visible (HRV) Site Extract Data documents.

Algorithms developed to extract sites from the NS001 imagery in an automated fashion proved to be inadequate due to local distortions in the imagery introduced by aircraft motion (pitch and yaw) for which adequate compensations could not be made using ground control points. As a result, the site extractions were done in a manual fashion by locating them in the imagery on a display system using a trackball and cursor along with a SPOT panchromatic reference image. Errors in site extractions from the NS001 are believed to be comparable to those of the satellite site extractions (e.g., 1 pixel) in most cases.

The uncertainty associated with ASAS at-sensor spectral radiance values is approximately 6%. The errors associated with the atmospherically corrected reflectance factors have not been assessed. Concerning spatial sampling of each ASAS view-angle image, since the different image segments were not geometrically registered, the geographic area sampled in each image is only approximately the same. Spatial sampling errors probably do not exceed 5%, and may typically be lower; however no quantitative evaluation of these errors has been made. Individual subareas were selected on the basis of internal homogeneity, and therefore spatial sampling variability is less important.

Users of the data are encouraged to refer to Sections 9, 10, and 11 of the individual data set documents to obtain more specific information on known errors and problems.

Detailed inter-comparison of multi-sensor surface reflectance retrieval from radiometrically calibrated and atmospherically corrected radiances suggests that both visible and infrared channel reflectances can be retrieved to within 1% absolute (at a reflectance of 6% in the visible and 30% in the infrared) under relatively clear atmospheric conditions and a wide range of view angles. Errors introduced by assumption of a default atmosphere that is significantly different from the actual atmosphere on any given day can increase these errors substantially. Problems with the NS001 calibration (section VIII) resulted in larger errors, particularly in the infrared. In the mid-infrared bands, reflectance retrieval is more problematic, with errors of 2% absolute (at reflectances of 23% in TM channel 5 and 14% in TM channel 7), and greater variability among instruments.

Inter-comparison of multi-sensor surface temperature retrieval was also done using radiometrically calibrated and atmospherically corrected radiances from Landsat-5 TM and the NS001 (as well as the helicopter MMR). Differences in atmospherically corrected brightness temperatures from near surface measurements ranged from less than 1 degree to more than 8 degrees, with the Landsat-5 TM systematically overestimating observed temperature in all cases. Corrected brightness temperatures from helicopter-MMR and NS001-TMS were in general agreement with near-surface infrared radiative thermometer (IRT) measurements collected from automated meteorological stations, with mean differences of 1.7 and 3.2 degrees for grassland targets. Much better agreement (within 1 degree) was found between the retrieved aircraft surface temperatures and near-surface temperatures acquired from a mast-mounted MMR and IRT. The NS001-TMS was also in close agreement with near surface temperatures acquired over water targets.

Fraser, R. S., R. A. Ferrare, Y. J. Kaufman, B. L. Markham, and S. Mattoo. 1992. Algorithm for atmospheric correction of aircraft and satellite imagery, Int. J. Remote Sensing 13(3):541-557.

Goetz, S. J., F. G. Hall, B. L. Markham, and R. O. Dubayah. 1993. Intercomparison of retrieved surface temperature from multi-resolution sensors at the FIFE site, In: American Society Photogrammetry and Remote Sensing, pp. 108-117, New Orleans, LA.

Goetz, S. J., R. Halthore, F. G. Hall, and B. L. Markham. 1994. Surface temperature retrieval from multi-resolution sensors at the FIFE site. J. Geophysical Research, FIFE Special Issue II (in press).

Goetz, S. J., B. L. Markham, and J. E. Newcomer. 1992. Radiometric calibration and atmospheric correction of satellite and aircraft data for FIFE. In: International Geoscience and Remote Sensing Symposium (IGARSS'92), pp. 798-801, Houston Texas.

Goward, S. N., D. G. Dye, W. Dulaney, J. Yang, and B. Markham. 1991. Critical assessment of normalized difference vegetation index measurements from the AVHRR sensor: Part 1: The AVHRR as a land observatory. Int. J. Remote Sensing.

Hall, F. G., S. J. Goetz, K. F. Huemmrich, and P. J. Sellers. 1994. AVHRR estimation of land-surface characteristics over a prairie grassland. J. Geophysical Research, FIFE Special Issue II (in press).

Halthore, R. N., and B. L. Markham. 1992. Overview of atmospheric correction and radiometric calibration efforts during FIFE. J. Geophysical Research 97(D17):18731-18742.

Kneizys, F. X., E. P. Shettle, W. O. Gallery, J. H. Chetwynd, L. W. Abreu, J. E. A. Selby, S. A. Clough, and R. W. Fenn. 1988. Atmospheric transmittance/radiance: Computer code LOWTRAN-7, Air Force Geophysics Lab, Hanscomb AFB, MA, AFGL-TR-88-0177.

Markham, B. L., J. R. Irons, D. W. Deering, R. N. Halthore, R. R. Irish, R. D. Jackson, M. S. Moran, S. F. Biggar, D. I. Gellman, B. G. Grant, J. M. Palmer, and P. N. Slater. 1990. Radiometric calibration of aircraft and satellite sensors at White Sands, New Mexico. In: International Geoscience and Remote Sensing Symposium (IGARSS), pp. 515-518, Washington DC.

Markham, B. L., R. N. Halthore, and S. J. Goetz. 1992. Surface reflectance retrieval from satellite and aircraft sensors: Results of sensor and algorithm comparisons during FIFE. J. Geophysical Research 97(D17):18785-18795.