Precision Test of the Estimated Sea Surface Temperature by NOAA/AVHRR Data-Comparison with the Mutsu bay marine environment buoy system data-

Abstract

The sea surface temperature (SST) detected by the AVHRR sensors of NOAA-7 and NOAA-9 was compared with the in situ SST observed by the automatic buoy system of Mutsu bay, which is situated in the northern Honshu, Japan. The buoy system is composed of six fixed buoys and has been monitoring SST at every hour. By screening cloud free AVHRR data in the period of 1984-1987, the total of 141 brightness temperatures at the buoys were recovered. To each AVHRR SST, the nearest time buoy SST at the corresponding position was matched up. The spatial and temporal consistency in each match-up is excellent to be within three pixel resolution and 30 minutes. The regression analysis was applied to the data set by specifying the object variable and the explanation variable to be the buoy SST and the AVHRR SST, respectively. In the results of the single variate regression analysis, the standard errors were distributed between 0.5-1.2^°C depending upon the grouped conditions of the data set. But in the double variate regression by using both ch. 4 and ch. 5 SSTs, the residues were improved remarkably. The matchups under the large temperature difference conditions between the air and the sea surface, however, were accompanied by large residues. This is meant that data strongly disturbed by air-sea interacting effects are difficult to be compensated by the MCSST method. In the regression by eliminating these specific match-ups, the standard errors in the grouped data sets of NOAA-7 and NOAA-9 were reduced to 0.47 and 0.48^°C, respectively. When our data set was applied to the SST estimation functions suggested by McClain and Strong, there appeared biases of -0.54--0.41^°C and standard errors of 0.94-0.96^°C.