Abstract
Ordinary least-squares is used to estimate the accuracy of monthly averaged sea surface temperature products in the eastern near-equatorial Pacific. Daytime multichannel sea surface temperature (MCSST) data, nighttime MCSST data, Climate Analysis Center (CAC) in situ temperatures, and CAC blended temperatures are all compared to monthly averaged, equatorial, 1 m moored-buoy temperatures at 110°W, 124°W, and 140°W. In addition, reduced least-squares (RLS) is used to develop regression equations between the CAC in situ temperature and the MCSST data. Bootstrap methods are used to estimate the RLS regression statistics. new regression equations are used to convert the MCSST to equivalent in situ temperatures prior to combining the two datasets with a one over distance-squared gridding algorithm. Data used in this study are from the period January 1983 to December 1985. When data from the 1982/1983 El Niho are excluded from analysis, the MCSST data are not significantly different from the moored-buoy temperatures at the 5% significance level. The CAC in situ and blended temperatures have warm biases of 0.85°C and 0.70°C when compared to the moored-buoy temperatures. These differences are significantly different from zero at the 5% level. The bias between the daytime MCSST data and the CAC in situ data is 0.65°C when data from 1983 are excluded. The bias between the nighttime MCSST data and the CAC in situ data is 1.04°C. This difference is attributed to diurnal temperature fluctuations. The blended temperature product developed in this study is 0.94°C warmer than the moored-buoy temperature data. The shape of this blended product is similar to the CAC blend, but some differences exist. These differences are discussed.
