Evaluating The Accuracy of Gridded Sea Surface Temperature Datasets in Central Java

Abstract

This study evaluates the accuracy of three gridded sea surface temperature (SST) products, OSTIA, RAMSSA, and GAMSSA, against in-situ observations from the in-situ SST Quality Monitor (iQuam) in the waters of Central Java, Indonesia, for the period 2008–2024. Validation is performed for three spatial aggregations: a combined domain (Northern + Southern waters), the Northern Sea, and the Southern Sea. Performance is assessed using standard error metrics (bias, MAE, MAPE, RMSE), scatterplots, annual cycle patterns, and Taylor diagrams to synthesize correlation, variability, and centered RMSE. Results show that OSTIA and RAMSSA consistently outperform GAMSSA across all zones, with the strongest agreement in the Southern Sea, where correlations are highest, and errors are lowest in the Taylor diagram summary. The annual-cycle analysis indicates that all products reproduce the timing of the seasonal SST evolution, including the pronounced cool season in the Southern Sea during the southeast monsoon. However, variability is generally damped in gridded products, particularly in nearshore/complex waters of the Northern Sea. These findings support the operational use of OSTIA and RAMSSA for regional marine monitoring and climate services in Central Java, including anomaly-based advisories and early warning information for ocean-related climate hazards.