A Three-measurement Model Developed for Evaluating Satellite Land Surface Temperature Product

摘要

In development of retrieval algorithm for satellite land surface temperature (LST) measurements, it is crucial yet difficult to estimate precision and accuracy of the algorithm using ground measurements. In this effort, we built up a theoretical model for estimating the random error of the satellite measurement. The method requires a series of surface temperature measurements from three independent data sources. In our case, they were collected from the GOES-8, GOES-10 Imagers and the SURFace RADiation (SURFRAD) budget network stations. SURFRAD data at five sites in the year 2001 were processed along with the corresponding GOES-8 and -10 Imager data. A manual cloud filtering procedure was applied to ensure a high quality cloud-free data set. An LST retrieval algorithm developed for the GOES-R mission was applied to the GOES-8 and -10 data, while the SURFRAD data provided the third independent LST estimation. Standard deviation errors of the three measurements were calculated from the theoretical model, and biases of the measurements were estimated with some assumptions. The method was particularly developed for evaluating GOES-R LST algorithm. It may have wider applications in remote sensing development and applications.