• Goetz, S. J., R. N. Halthore, F. G. Hall, and B. L. Markham, 1995: Surface-temperature retrieval in a temperate grassland with multi resolution sensors. J. Geophys. Res.,100, 25397–25410.

  • Gu, X. F., B. Seguin, J. F. Hanocq, and J. P. Guinot, 1994: Evaluation and comparison of atmospheric correction methods for thermal data measured by ERS1-ATSR, NOAA11-AVHRR, and Landsat5-TM Sensors. Proc. Sixth Int. Symp. on Physical Measurements and Signatures in Remote Sensing, Val d’Isere, France, CNES, 793–800. [Available from CNES, 18 av. E. Belin, 31055 Toulouse, France.].

  • Harris, A. R., and I. M. Mason, 1992: An extension to the split-window technique giving improved atmospheric correction and total water vapour. Int. J. Remote Sens.,13, 881–892.

  • Kalluri, S. N. V., and R. O. Dubayah, 1995: Comparison of atmospheric correction models for thermal bands of the Advanced Very High-Resolution Radiometer over FIFE. J. Geophys. Res.,100, 25411–25418.

  • Kniezys, F. X., E. P. Shettle, L. W. Abreu, J. H. Chetwynd, G. P. Anderson, W. O. Gallery, J. E. A. Selby, and S. A. Clough, 1988:User guide to Lowtran 7. Air Force Geophysics Laboratory Rep. AFGL-TR-88-0177, 137 pp. [Available from AFGL, Hanscom AFB, MA 01731.].

  • McClain, E. P., W. G. Pichel, C. C. Walton, Z. Ahmad, and J. Sutton, 1983: Multichannel improvements to satellite derived global sea surface temperatures. Adv. Space Res.,2, 43–47.

  • Schmugge, T. J., F. Becker, and Z.-L. Li, 1990: Spectral emissivity variations observed in airborne surface temperature measurements. Remote Sens. Environ.,34, 95–104.

  • Sellers, P. J., F. G. Hall, G. Asrar, D. E. Strebel, and R. E. Murphy, 1992: An overview of the First International Satellite Land Surface Climatology (ISLSCP) Field Experiment (FIFE). J. Geophys. Res.,97, 18345–18371.

  • Steyn-Ross, M. L., D. A. Steyn-Ross, and W. J. Emery, 1997: A dynamic water vapor correction method for the retrieval of land surface temperature from AVHRR. J. Geophys. Res.,102, 19 629–19 643.

  • Strebel, D. E., D. R. Landis, K. F. Huemmrich, and B. W. Meeson, 1994: Collected Data of The First ISLSCP Field Experiment. Vols. 1–3. NASA, CD-ROM.

  • Sugita, M., and W. Brutsaert, 1990: Wind velocity measurements in the neutral boundary layer above hilly prairie. J. Geophys. Res.,95, 7617–7624.

  • View in gallery

    AVHRR thermal image in raw counts for 21 August 1987 (day 233) over the FIFE site. The two reservoirs with lower temperatures appear white in this image. The distorted shape of the FIFE site results from the image not being georegistered. The numbers indicate the date, day 233 1987, and time, 2120 UTC, of the observation.

  • View in gallery

    Water vapor profiles 2 July 1987. The profiles measured before and after the satellite overpass are shown as well as the interpolated profile used as input to the MODTRAN model.

  • View in gallery

    Thermal image from the airborne NS001 sensor onboard the NASA C-130 aircraft. The image is from an east–west flight line at the 5-km altitude, pixel size is 12 m. North is to right in the image. The locations for nine of the AMS stations are shown. The squares are the approximate size of a 1-km AVHRR pixel. The road going lengthwise through the image is I-70 and the one crossing it is Rt. 177 from Manhattan, Kansas. The temperatures in this image range from the high 20’s for the black areas to the upper 40’s for the white areas.

  • View in gallery

    The flow chart for the approach used in this paper.

  • View in gallery

    Surface temperature results for Lake Milford derived using the approach given in Fig.4.

  • View in gallery

    Temporal variation of surface temperature for one of the AMS sites with the times of aircraft and satellite overpasses shown. The ΔIRT value shown is the change in surface temperature at this site between times of the NS001 overpass and that of the AVHRR. The vertical lines indicate the range of standard deviation for the remotely sensed data.

  • View in gallery

    Comparison of the average surface temperatures from the AVHRR and the averages of the ground AMS stations for 13 days in 1987.

  • View in gallery

    Histograms of the NS001 temperature distributions for a 80 × 80 pixel square (1 km × 1 km) around two of the AMS sites. The AMS reading at this time is indicated by the vertical line with the black square. The dashed histogram is for a block displaced 1 km to the south.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 93 93 82
PDF Downloads 5 5 0

Surface Temperature Observations from AVHRR in FIFE

View More View Less
  • 1 USDA Hydrology Lab, Beltsville, Maryland
© Get Permissions
Full access

Abstract

Observations of the surface radiometric temperature by the AVHRR sensor on board the NOAA-9 satellite during the First ISLSCP (International Satellite Land Surface Climatology Project) Field Experiment conducted in central Kansas during 1987 are presented. The satellite observations were corrected for atmospheric effects using a path radiance model (MODTRAN3) and radiosonde measurements. Problems with this approach include the nonsimultaneity of the soundings with the overpass and errors involved in profile measurements. For the former, soundings before and after the overpass were interpolated to the time of the overpass. For the latter, some of the errors arise from the ±0.5°C uncertainty in the dry- and wet-bulb temperatures, which can produce up to a ±14% relative uncertainty in the water vapor. To overcome this uncertainty, the water vapor profiles were adjusted until the channel 4 and 5 temperature differences over a large reservoir were reduced to zero. This adjusted profile was then used over the entire site. The results are compared to ground broadband temperature readings at 10 sites and to aircraft results from the thermal channel of the NS001 sensor on the C-130 aircraft. The AVHRR values were found to be 5° to 6°C warmer than the average of the ground measurements. This difference is attributed to the fact that the ground measurements were made preferentially on well-vegetated surfaces while the AVHRR integrates over the entire site, which includes many warm surfaces.

Corresponding author address: T. Schmugge, USDA/ARS Hydrology Lab, Bldg. 007—BARC West, Beltsville, MD 20705-2350.

Email: schmugge@hydrolab.arsusda.gov

Abstract

Observations of the surface radiometric temperature by the AVHRR sensor on board the NOAA-9 satellite during the First ISLSCP (International Satellite Land Surface Climatology Project) Field Experiment conducted in central Kansas during 1987 are presented. The satellite observations were corrected for atmospheric effects using a path radiance model (MODTRAN3) and radiosonde measurements. Problems with this approach include the nonsimultaneity of the soundings with the overpass and errors involved in profile measurements. For the former, soundings before and after the overpass were interpolated to the time of the overpass. For the latter, some of the errors arise from the ±0.5°C uncertainty in the dry- and wet-bulb temperatures, which can produce up to a ±14% relative uncertainty in the water vapor. To overcome this uncertainty, the water vapor profiles were adjusted until the channel 4 and 5 temperature differences over a large reservoir were reduced to zero. This adjusted profile was then used over the entire site. The results are compared to ground broadband temperature readings at 10 sites and to aircraft results from the thermal channel of the NS001 sensor on the C-130 aircraft. The AVHRR values were found to be 5° to 6°C warmer than the average of the ground measurements. This difference is attributed to the fact that the ground measurements were made preferentially on well-vegetated surfaces while the AVHRR integrates over the entire site, which includes many warm surfaces.

Corresponding author address: T. Schmugge, USDA/ARS Hydrology Lab, Bldg. 007—BARC West, Beltsville, MD 20705-2350.

Email: schmugge@hydrolab.arsusda.gov

1. Introduction

Several recently published papers have looked at the surface temperatures derived from remotely sensed thermal infrared data obtained during the First ISLSCP (International Satellite Land Surface Climatology Project) Field Experiment (FIFE) (Sellers et al. 1992) conducted in central Kansas during 1987 and 1989. Goetz et al. (1995) studied the data from the thermal channel (10.5–12.5 μm) of the Landsat Thematic Mapper (TM) and the corresponding channel of the NS001 sensor, which is an airborne simulator for the TM. They compared the satellite and aircraft data with ground infrared thermometer (IRT) measurements for several dates during FIFE in both 1987 and 1989 using MODTRAN (a Moderate Resolution Model for Atmospheric Radiative Transfer) and the measured profiles for atmospheric correction. In general they found good agreement for the aircraft observations, within about 3°C, but that the agreement for the Landsat data was worse, with up to 8°C differences. In the same issue of the Journal of Geophysical Research, Kalluri and Dubayah (1995) compared surface temperatures derived from the National Oceanic and Atmospheric Administration Advanced Very High Resolution Radiometer (NOAAAVHRR) sensor with ground IRT measurements using several techniques for atmospheric correction. These included several split window approaches and two approaches based on radiative transfer codes using coincident radio soundings. They found that the split window approaches did not yield agreement to the ground measurements within their ±3°C criteria, whereas the results from the two radiative transfer approaches did satisfy the criteria in half of the cases. In general they found that the land surface temperatures derived from the AVHRR data were higher than the ground measurements. This disagreement with the ground measurements was also found in an analysis done by Steyn-Ross et al. (1996) working the FIFE data from 1989. They found an rms difference of about 3°C for the daytime observations but found good agreement for nighttime observations. This latter fact led them to conclude that the observed differences in the daytime observations were due to the ground radiometers preferentially observing cooler vegetated surfaces within the enclosures surrounding each ground site, while the AVHRR with its 1-km pixel size was observing a mix of the cool vegetation with warmer, less well-vegetated surfaces. This observation was also made by Goetz et al. (1995) and Kalluri and Dubayah (1995) in their papers.

This study examines how well satellite observations of thermal infrared radiation from the land surface can be corrected for atmospheric effects. The satellite data are from the Advanced Very High Resolution Radiometer (AVHRR) sensor on the NOAA-9 satellite duringFIFE in 1987. The MODTRAN3 path radiance model (Kniezys et al. 1988) was used with radiosoundings launched during FIFE to calculate the atmospheric path radiance and transmission. The corrected temperatures were compared with ground and aircraft infrared measurements. The problems involved in doing the comparison are discussed. They include the different spectral bandpasses for the radiometers, varying view angles for both the satellite and aircraft observations, differences in the times of the observations and, most importantly, differences in the field of view of the instruments. The data used in this analysis were obtained from the FIFE CD-ROM (Strebel et al. 1994).

The radiance measured at the satellite contains the radiance from the surface (Lsurf) reduced by the atmospheric attenuation (τ) and augmented by the upwelling radiance from the atmosphere itself (Latm) and is given by Eq. (1):
i1520-0469-55-7-1239-e1
which can be inverted to give the radiance at the ground:
i1520-0469-55-7-1239-e2
where the atmospheric transmittance τ and Latm were determined with MODTRAN3 path radiance model (Kniezys et al. 1988). The measured atmospheric profiles were used in the model. The corrected brightness temperature values were obtained by inverting the Planck function using the radiances over the area of interest. The problem with this approach is that of the uncertainties in the atmospheric profile, especially the water vapor content. In this paper the values of τ and Latm were adjusted by varying the water vapor profile to yield a zero temperature difference for the two AVHRR channels over a water body.

2. Measurements

The AVHRR has two thermal channels at 10.3–11.3 μm (channel 4) and 11.5–12.5 μm (channel 5). The spatial resolution for these channels is approximately 1 km at nadir degrading to about 1.7 km at 55°. The data in the digital counts from the CD-ROM were converted to radiance using the calibration information given in the documentation for each image. An example of a scene is given in Fig. 1. Since this image portrays raw counts, where cold has higher values, the two nearby reservoirs, which are cooler in the daytime, appear aswhite and warmer land surfaces appear darker. The latitude and longitude for each pixel were provided in ancillary files for each image. This information was used to locate the ground sites in the AVHRR data. Because of the uncertainties in these location data, a 3 × 3 grid of pixels around the ground site was used to calculate the AVHRR response, with the center pixel being weighted more heavily. The days chosen for this analysis are given in Table 1 along with some of the observational parameters. Note in particular the look angle variations.

The radiosondes (Sugita and Brutsaert 1990) measure the atmosphere dry- and wet-bulb temperatures along with the pressure as the balloon ascends. These are then used to calculate the vapor pressure. In general there were radiosound ascents available before and after the times of the satellite or aircraft overpasses. The values of the atmospheric profiles were interpolated to the time of the overpasses. An example is given in Fig. 2, showing the measured profiles before and after the overpass with interpolated profile in the middle. Above the freezing level for the wet bulb, the water vapor content was assumed to decrease exponentially with height. This approach worked well in most cases. However, for situations where the radiosonde passed through a dry layer at the freezing level this approach did not include sufficient water vapor at the higher levels. This point will be discussed later. For input to MODTRAN (Kniezys et al. 1988), 34 levels were selected on the basis of equal amounts of water vapor in each layer.

The ground IRT measurements were performed at 10 Automatic Meteorological Stations (AMS). These data were taken with an 8–14-μm bandpass radiometer mounted about 2 m above the ground and viewing the ground at about a 10° view angle and having a 0.5-m field of view. The sites were generally located within enclosures to protect them from grazing animals. The instruments were calibrated after the completion of FIFE-87.

The aircraft data were acquired with the thermal channel of the NS001 sensor, which has a 10.4–12.5-μmbandpass. The sensor was on the NASA C-130 aircraft, which flew at an altitude of 15000 ft (4500 m). At this altitude the instantaneous field of view for the thermal channel is 12 m. These data were also corrected for atmospheric effects using MODTRAN using an approximately coincident radio sounding. An example of the NS001 data for 4 June 1987, day 155, is given in Fig. 3. This figure shows the type of spatial variation that is present within an AVHRR pixel. The squares on the image are 1 km × 1 km to represent an AVHRR pixel. The location of several of the AMS sites is also shown.

A fourth data source was the Barnes Modular Multichannel Radiometer (MMR), which was carried by a helicopter. The MMR has seven spectral channels in the visible and reflected infrared plus a channel in the thermal infrared with a 10.4–12.5-μm bandpass. The helicopter would hover over individual ground sites to acquire data for several minutes at a nominal altitude of 300 m, yielding a footprint on the ground of approximately 5 m. We selected data for a site that was covered at about the same time as the NS001 coverage.

3. Methods

The objective of this exercise was to test a method for correcting the satellite observations of land leavingthermal radiance, or brightness temperature, for the effects of the intervening atmosphere. This was to be done using the measured atmospheric profiles as input to an atmospheric path radiance model, MODTRAN3. The major atmospheric absorber for the 10–12-μm band is water vapor and the problem is the uncertainty in its measurement. The manufacturer’s stated accuracy for the dry- and wet-bulb temperatures is ±0.5°C, which can result in uncertainties up to ±14% in each water vapor measurement, being larger for cool, dry conditions. To compensate for this uncertainty, the brightness temperatures for channels 4 and 5, T4 and T5, were calculated for a few pixels centered on a reservoir north of the FIFE site, Lake Milford. This target was chosen because the emission from the water surface should be the same in the two channels; thus the brightness temperature for the channels would be the same at the lake surface. At the satellite, T4 > T5 because of the greater atmospheric absorption at the wavelengths of channel 5, which results in a larger contribution from the cooler atmosphere in this channel. Typically the T4T5 difference at satellite was 2° to 3°C, which typically was reduced after atmospheric correction to less than one degree. The water vapor profiles used in MODTRAN were adjusted to yield a zero surface temperature difference for the lake pixels. The adjustment was done by multiplying the water vapor content at each level by a constant factor, for example, 1.05 or 0.95. This results in an increase or decrease in the total amount of water vapor along the path to the satellite. The flow chart for this process is given in Fig. 4. This adjustment process was done iteratively until the T4T5 difference for the lake was reduced to 0.1°C or less. These adjusted profiles were then used for the atmospheric corrections for the data over the entire site. This method is similar to that published by Gu et al. (1994) and more recently by Steyn-Ross et al. (1997). A difficulty with the general application of this procedure is the uncertainty of the surface emissivities for the two AVHRR channels, thus the need for having the water bodies in the scene. Water has approximately the same emissivity in both channels and is close to one. In the FIFE case the reservoirs are of marginal size in that they are only one or two pixels wide especially at some of the larger incidence angles observed here. We are helped in case by the presence of vegetated and forested conditions along the shorelines. These surfaces behave similar to water in terms of their emissivities. The results of this procedure are presented in Fig. 5 for 13 observations during the 1987 campaigns of FIFE. The results show a consistent seasonal pattern with the highest temperatures, about 30°C, being observed during, IFC-3 (intensive field campaign 3) in August (days 219–229). Unfortunately there are no surface measurements of the lake temperature available during the 1987 experiment. Therefore the only validation of these water results is to use alternate methods for determining the water surface temperature, namely, what are called the split window approaches.These make use of the fact that the magnitude of the difference T4 − T5 is related to the size of the atmospheric correction and this is used routinely to make estimates of sea surface temperature. In Fig. 5 the +’sare temperature values derived from the McClain et al. (1983) split window approach using theequation:
TsurfT4T4T5
There is excellent agreement between the surface temperatures derived by the two methods. Similar results were obtained using a method proposed by Harris and Mason (1992). These temperatures seem a bit high but Lake Milford is a shallow lake and, as noted above, the water pixels may be contaminated by radiation from the warmer land surface. Recall from Table 1 that most of the observations were at rather large incidence angles ≥30° and would have the larger pixel sizes. The atmospheric water vapor amounts are indicated by the bars at the bottom of the figure in centimeters of H2O. This is the amount of water along the viewing path and thus can be large for the large angles. The resulting profile adjustment factors are given in Table 1. Most of the values are within the 14% stated accuracy. However, there are a couple of days when the adjustment is rather large, for example, day 279 when the factor was 2.06. This was a rather dry day and even after the adjustment there were only 2 cm of water in the atmosphere. A similar situation was observed for day 157. However, when the profile was analyzed, it was found that there had been a dry layer at the freezing level for the wet bulb, at approximately 2-km altitude. Our procedure maintained this very dry condition into the higher layers of the atmosphere. If instead we used the humidity levels given by the United States midlatitude summer atmosphere we obtained the smaller adjustment factor given in Table 1. This points out a shortcoming of the radiosondes used in FIFE in that they did not make accurate measurements of the humidity especially at the upper levels of the atmosphere.

The next step is to use these atmospheric profiles to calculate the ground brightness temperatures for the FIFE site itself and to compare these results with those obtained by the ground and aircraft sensors.

4. Results

In Fig. 6 we present a comparison of the data from the four sources on 4 June 1987 (day 155). The temporal variation of the ground data at site 4439 is presented from about 1000 to 1700 LDT (local daylight time). This site is an upland burned grass site used for extensive flux measurements and is outside the Konza preserve and thus subject to grazing. The temperature range is from 294 K to 302 K. The NS001 and MMR data were acquired at about the same time, 1530 LDT (2030 UTC), which is about 1 h before the AVHRR overpass. The NS001 data is for a 1-km square area centered on the site and the value shown is the average and plus and minus one standard deviation. It can be observed that the NS001 data is about 7–8 K warmer than the ground measurements, whereas the MMR result is about 3–4 K warmer. Both of these temperatures were corrected for atmospheric effects using the measured profiles interpolated to overflight times. The AVHRR pass was at 1630 LDT (2130 UTC) and is about 6–7 K warmer than the ground measurements. It should be noted that all the remotely sensed measurements were made with sensors operating in the 10–12-μm band and are in reasonable agreement with each other. For example if the NS001 measurement is translated by the amountthe ground measurements changed in time to the AVHRR time (ΔT ∼ 2 K), the results are in good agreement. A possible explanation for the lower reading for the broader band ground radiometers is that they were seeing some bare soil, which can have a lower emissivity in the 8–9-μm region (Schmugge et al. 1990): for example, an emissivity of 0.93 would be typical for bare soil with the broadband IRT compared with 0.97 for the 10–12-μm channels. However, considering the limited amount of bare soil showing it is unlikely that this emissivity difference could be large enough to account for the total differences observed, but perhaps 1° or 2°C of it.

Because of the difficulty in comparing the AVHRR results with an individual AMS station it was decided to compare the overall averages. This was done taking the average AVHRR radiance over the entire FIFE site—that is, over the 15 km × 15 km area—and comparing this result with the average of the AMS stations, which were operating on each day, generally 9 or 10. The results are given in Table 2, where we present the average brightness temperature for entire FIFE site: 1) at the satellite, 2) the values at the surface with the adjusted profiles, and 3) the values at the surface with unadjusted profiles. In the first pair of TB columns it is seen that T5 is always less than T4 as expected. For the TB values at the surface, we find that T5 is usually greater than T4 indicating that perhaps we overcompensated for the atmosphere. A graphical summary of the results is shown in Fig. 7. The AVHRR results for channel 4obtained before (×’s) and after (+’s) adjusting the profile are shown. The AMS readings are shown as the black squares. The adjustment factors are presented in Table 2 and except for three cases the adjustment factor was within the ±14% limits. It is seen that the AVHRR results track the AMS results very well but are consistently five or more degrees higher. The rms temperature differences between the AVHRR results and ground station average is 5.7°C. In most cases the temperature changes due to the profile adjustments are rather small, a degree or two.

To understand the possible cause of this difference we need to consider the distribution of temperatures observed by the NS001 thermal channel. An example is presented in Fig. 8 for 4 June 1987 (day 155), the histograms shown are for 80 × 80 pixels segments around two of the AMS sites (4439 and 4609). The solid line is for the block centered on the site and the dashed is for the block 1 km to the south. At each site there is about a 10- or 12-K spread of temperatures with the AMS reading, the vertical line, at the cool end. If we consider site 4439, the mean value of the histogram is 308 ± 3 K, while the AMS reading is 301 K at approximately the same time. These histograms show that the AMS readings may be significantly cooler than the average for the entire site as observed by the AVHRR.

5. Discussion

The profile adjustment process was seen to reduce the brightness temperature differences between the two channels indicating that it is reducing the errors in the atmospheric correction process. This is true for most cases; however, there were several days for which theunadjusted profiles yielded closer agreement between the two channels. An example of this is day 279, the difference at the satellite was only 0.8°C indicating a dry atmosphere. The unadjusted profile reduced that difference to 0.5°C with channel 5 the cooler. The adjusted profile had channel 4 cooler by 0.6°C indicating that we probably overcompensated for water vapor giving rise to the adjustment factor of 2. This may have arisen because of spatial differences in atmosphere between the lake and the FIFE site. Thus it would be better to have the adjustment target within the site of interest.

We believe that this adjustment procedure should be a useful tool for those cases where there is a water body or other constant emissivity surface in the scene. For example, the MODIS and ASTER sensors, to be flown on NASA’s first Earth Observing System platform, have multiple bands in the thermal infrared and it should be possible to use this technique to fine tune the atmospheric correction for these sensors.

The results presented indicate the magnitude of the problem of evaluating the accuracy of surface temperature measurements made from spaceborne platforms. The corrected AVHRR average brightness temperatures over the FIFE site were consistently about 5°–6°C higher than the average for the AMS sites. This difference may be attributed to the bias introduced by the siting of the AMS stations on vegetated sites within enclosures as was indicated by the histogram of the NS001 temperatures and to differences in the bandwidth of the radiometers.

REFERENCES

  • Goetz, S. J., R. N. Halthore, F. G. Hall, and B. L. Markham, 1995: Surface-temperature retrieval in a temperate grassland with multi resolution sensors. J. Geophys. Res.,100, 25397–25410.

  • Gu, X. F., B. Seguin, J. F. Hanocq, and J. P. Guinot, 1994: Evaluation and comparison of atmospheric correction methods for thermal data measured by ERS1-ATSR, NOAA11-AVHRR, and Landsat5-TM Sensors. Proc. Sixth Int. Symp. on Physical Measurements and Signatures in Remote Sensing, Val d’Isere, France, CNES, 793–800. [Available from CNES, 18 av. E. Belin, 31055 Toulouse, France.].

  • Harris, A. R., and I. M. Mason, 1992: An extension to the split-window technique giving improved atmospheric correction and total water vapour. Int. J. Remote Sens.,13, 881–892.

  • Kalluri, S. N. V., and R. O. Dubayah, 1995: Comparison of atmospheric correction models for thermal bands of the Advanced Very High-Resolution Radiometer over FIFE. J. Geophys. Res.,100, 25411–25418.

  • Kniezys, F. X., E. P. Shettle, L. W. Abreu, J. H. Chetwynd, G. P. Anderson, W. O. Gallery, J. E. A. Selby, and S. A. Clough, 1988:User guide to Lowtran 7. Air Force Geophysics Laboratory Rep. AFGL-TR-88-0177, 137 pp. [Available from AFGL, Hanscom AFB, MA 01731.].

  • McClain, E. P., W. G. Pichel, C. C. Walton, Z. Ahmad, and J. Sutton, 1983: Multichannel improvements to satellite derived global sea surface temperatures. Adv. Space Res.,2, 43–47.

  • Schmugge, T. J., F. Becker, and Z.-L. Li, 1990: Spectral emissivity variations observed in airborne surface temperature measurements. Remote Sens. Environ.,34, 95–104.

  • Sellers, P. J., F. G. Hall, G. Asrar, D. E. Strebel, and R. E. Murphy, 1992: An overview of the First International Satellite Land Surface Climatology (ISLSCP) Field Experiment (FIFE). J. Geophys. Res.,97, 18345–18371.

  • Steyn-Ross, M. L., D. A. Steyn-Ross, and W. J. Emery, 1997: A dynamic water vapor correction method for the retrieval of land surface temperature from AVHRR. J. Geophys. Res.,102, 19 629–19 643.

  • Strebel, D. E., D. R. Landis, K. F. Huemmrich, and B. W. Meeson, 1994: Collected Data of The First ISLSCP Field Experiment. Vols. 1–3. NASA, CD-ROM.

  • Sugita, M., and W. Brutsaert, 1990: Wind velocity measurements in the neutral boundary layer above hilly prairie. J. Geophys. Res.,95, 7617–7624.

Fig. 1.
Fig. 1.

AVHRR thermal image in raw counts for 21 August 1987 (day 233) over the FIFE site. The two reservoirs with lower temperatures appear white in this image. The distorted shape of the FIFE site results from the image not being georegistered. The numbers indicate the date, day 233 1987, and time, 2120 UTC, of the observation.

Citation: Journal of the Atmospheric Sciences 55, 7; 10.1175/1520-0469(1998)055<1239:STOFAI>2.0.CO;2

Fig. 2.
Fig. 2.

Water vapor profiles 2 July 1987. The profiles measured before and after the satellite overpass are shown as well as the interpolated profile used as input to the MODTRAN model.

Citation: Journal of the Atmospheric Sciences 55, 7; 10.1175/1520-0469(1998)055<1239:STOFAI>2.0.CO;2

Fig. 3.
Fig. 3.

Thermal image from the airborne NS001 sensor onboard the NASA C-130 aircraft. The image is from an east–west flight line at the 5-km altitude, pixel size is 12 m. North is to right in the image. The locations for nine of the AMS stations are shown. The squares are the approximate size of a 1-km AVHRR pixel. The road going lengthwise through the image is I-70 and the one crossing it is Rt. 177 from Manhattan, Kansas. The temperatures in this image range from the high 20’s for the black areas to the upper 40’s for the white areas.

Citation: Journal of the Atmospheric Sciences 55, 7; 10.1175/1520-0469(1998)055<1239:STOFAI>2.0.CO;2

Fig. 4.
Fig. 4.

The flow chart for the approach used in this paper.

Citation: Journal of the Atmospheric Sciences 55, 7; 10.1175/1520-0469(1998)055<1239:STOFAI>2.0.CO;2

Fig. 5.
Fig. 5.

Surface temperature results for Lake Milford derived using the approach given in Fig.4.

Citation: Journal of the Atmospheric Sciences 55, 7; 10.1175/1520-0469(1998)055<1239:STOFAI>2.0.CO;2

Fig. 6.
Fig. 6.

Temporal variation of surface temperature for one of the AMS sites with the times of aircraft and satellite overpasses shown. The ΔIRT value shown is the change in surface temperature at this site between times of the NS001 overpass and that of the AVHRR. The vertical lines indicate the range of standard deviation for the remotely sensed data.

Citation: Journal of the Atmospheric Sciences 55, 7; 10.1175/1520-0469(1998)055<1239:STOFAI>2.0.CO;2

Fig. 7.
Fig. 7.

Comparison of the average surface temperatures from the AVHRR and the averages of the ground AMS stations for 13 days in 1987.

Citation: Journal of the Atmospheric Sciences 55, 7; 10.1175/1520-0469(1998)055<1239:STOFAI>2.0.CO;2

Fig. 8.
Fig. 8.

Histograms of the NS001 temperature distributions for a 80 × 80 pixel square (1 km × 1 km) around two of the AMS sites. The AMS reading at this time is indicated by the vertical line with the black square. The dashed histogram is for a block displaced 1 km to the south.

Citation: Journal of the Atmospheric Sciences 55, 7; 10.1175/1520-0469(1998)055<1239:STOFAI>2.0.CO;2

Table 1.

AVHRR observations.

Table 1.
Table 2.

AVHRR brightness temperatures, TB.

Table 2.
Save