Sampling Errors in Satellite-derived Infrared Sea Surface Temperatures

Abstract:

Sea Surface Temperature (SST) measured from satellites has been playing a crucial role in understanding geophysical phenomena. Generating SST Climate Data Records (CDRs) is considered to be the one that imposes the most stringent requirements on data accuracy. For infrared SSTs, sampling uncertainties caused by cloud presence and persistence generate errors. In addition, for sensors with narrow swaths, the swath gap will act as another sampling error source. This study is concerned with quantifying and understanding such sampling errors, which are important for SST CDR generation and for a wide range of satellite SST users. In order to quantify these errors, a reference Level 4 SST field (Multi-scale Ultra-high Resolution SST) is sampled by using realistic swath and cloud masks of Moderate Resolution Imaging Spectroradiometer (MODIS) and Advanced Along Track Scanning Radiometer (AATSR). Global and regional SST uncertainties are studied by assessing the sampling error at different temporal and spatial resolutions (7 spatial resolutions from 4 kilometers to 5.0° at the equator and 5 temporal resolutions from daily to monthly). Global annual and seasonal mean sampling errors are large in the high latitude regions, especially the Arctic, and have geographical distributions that are most likely related to stratus clouds occurrence and persistence. The region between 30°N and 30°S has smaller errors compared to higher latitudes, except for the Tropical Instability Wave area, where persistent negative errors are found. Important differences in sampling errors are also found between the broad and narrow swath scan patterns and between day and night fields. This is the first time that realistic magnitudes of the sampling errors are quantified. Future improvement in the accuracy of SST products will benefit from this quantification.