Comparison of atmospheric correction models for thermal bands of the advanced very high resolution radiometer over FIFE

Publisher Website
Google Scholar
Abstract
Using data from the First International Satellite Land Surface Climatology Project (ISLSCP) Field Experiment (FIFE), land surface temperature (LST) measurements from channels 4 and 5 of the advanced very high resolution radiometer (AVHRR) on NOAA 9 were corrected for atmospheric effects using five different algorithms on three days in 1987. The performances of three widely used algorithms were compared with two other correction schemes that use coincident radiosonde data. The radiosonde measurements were used (1) to derive local split window coefficients and (2) to derive linear coefficients for atmospheric correction. Each model's performance was assessed by comparing the corrected LSTs with ground measurements made using infrared thermometers. If the difference between the two measurements was within ±3°C, then the correction was considered satisfactory. The three general methods that did not use any radiosonde data failed to meet the ±3°C criterion on any of the days. Of the two techniques which used coincident radiosonde data, the split window method passed the ±3°C criterion on two days, while the linear method passed only on one day. All five techniques overestimated the LSTs in comparison with ground data. While the use of coincident radiosonde data did provide more accurate retrievals relative to the three generalized methods, the improvement is somewhat unconvincing when compared to uncorrected AVHRR channel 4 temperatures, which were much closer to ground truth than any of the algorithms on two of the three days. In addition, we performed a linear correction using the standard midlatitude summer water vapor and temperature profiles but scaled so that the total water vapor amount matched that derived from the radiosonde data. In this case, our results were significantly better with estimates on all three days within ±1.4°C. Although our results using the other methods are in reasonable agreement with ground truth, our failure to meet the accuracy criterion is perplexing, and while its ultimate causes are unknown, possible explanations are provided.