•  
  •  
 

Abstract

Yu and Barton proposed a non-regression-coefficient method in improving the accuracy of routine sea surface temperature retrievals through satellite data, which are then applied to sea surface temperature estimation models such as MCSST or CPSST algorithms. Their method involves a parameter called R54(=∆T5/∆T4), which is derived from the ratio of neighborhood temperature variances from two window channels in the thermal IR region. Due to the fact that statistical calculations were used in the derivation for the value of R54, the method is not applicable in various real cases. The aim of this study is to retain the advantage of their method in terms of its physical considerations and derivations, while providing an alternative in improving the calculations of R54. In this study, a spatial coherence method is also used to mask cloudy areas, and an improved three-point method was brought in to determine the clear-sky brightness temperature. A NOAA-12 AVHRR data set partially covered with clouds was used to estimate a 100 × 100 longitude-latitude SST map of the Taiwan area, which were compared with 61 ship-measured SSTs, as well as SST retrievals estimated by Chen, MCSST and CPSST. The results reveal improvements in the 0.8K RMSE and 0.1K bias SST map. The accuracy levels also grew higher, especially for partially cloud-contaminated cases. Moreover, this study can further produce maps regarding the amount of clouds present, or the temperature distribution of the cloud-tops for related research.

Included in

Engineering Commons

COinS