Theoretical aspects and operational results of physical deterministic sea surface temperature retrieval

Abstract

Physical deterministic sea surface temperature (PDSST) retrieval scheme is built on radiative transfer forward model and a mathematically deterministic approach to the solution for inverse problem. This requires atmospheric profiles information from Numerical Weather Prediction (NWP), which offers the prospect to account for local retrieval conditions and yields a more uniform product with superior accuracy. One of the unprecedented capabilities of the PDSST scheme is that it can use aerosol profiles in addition to atmospheric profiles information for the forward modeling, and also allows for adjustment of the aerosol burden by including it as a retrieved element. Cloud detection is a vital part of SST retrieval processing. An innovative cloud and error masking (CEM) algorithm has been developed, combining the functional spectral differences and radiative transfer based cloud detection tests, especially the functional double difference tests are unique. These advancements have led to substantial improvements in information retrieval from expensive satellite measurement. This improvement refers to a dual benefit of increased data coverage (reduced false alarms) and detection of actual cloud contamination (improved detection rate). The PDSST retrieval suite, is combining the PDSST retrieval scheme and CEM, demonstrates the superiority of this approach with an overall ~3-4 times information gain when implemented on data from MODIS-Aqua and GOES Imager. For example, RMSE reduction from 0.52 K to 0.35 K and data coverage enhanced from ~9% to ~19% as compared to NASA operational MODIS-AQUA SST products.