Inter-comparison of NOAA-AVHRR and IRS-P4 (MSMR) derived sea surface temperatures

Due to the limitations of infrared remote sensing, sea surface temperature (SST) can be derived only under clear sky conditions from the infrared channels like those in National Oceanic and Atmospheric Administration (NOAA) advanced very high resolution radiometer (AVHRR), where as microwave radiometers can provide SST even under cloudy conditions. However, the accuracy of SST derived from the microwave sensor is less with a poor spatial resolution. In this study, SSTs over the Arabian Sea derived from NOAA-AVHRR and Indian Remote Sensing Satellite (IRS-P4) multi-frequency scanning microwave radiometer (MSMR) observations have been compared on weekly basis with a view to blend these two observations, so that SST can be provided continuously even under cloudy conditions. The NOAA-AVHRR derived pathfinder SSTs with spatial resolution of 18km were averaged to 1.5 degrees x1.5 degrees grid resolutions to match with MSMR observations. The analysis was carried out during 2000. Statistical analysis of the NOAA-AVHRR SST shows that the spatial variation of SST within 1.5 degrees grid is negligible compared with the MSMR accuracy of 1.52 degrees C. Thus the comparison of SST from these two sources with different spatial resolution is reasonable. The RMS difference is 1.55 degrees C with a correlation coefficient of 0.73. After removing the seasonal bias, the RMS difference reduced to 0.66 degrees C and the coefficient of correlation improved to 0.89. The correlation coefficient between the two observations has further improved to 0.90 and the RMS difference reduced to 0.53 degrees C when the averaging was done using 5 degrees x5 degrees grid resolution. The accuracy of satellite derived SSTs are also evaluated with the moored buoy observations over the Arabian Sea. The accuracy of MSMR SST observations have improved if 95% confidence level data is considered. The results indicate the possibility of replacing the data gaps in AVHRR SSTs with MSMR estimations after adjusting for the seasonal biases.