Quantile mapping enhances the sea surface temperature prediction accuracy in the northern coastal region of Penang using ROMS

Sea surface temperature (SST) is a crucial climate indicator for tracking atmospheric and oceanic interactions, particularly in coastal areas. The study focused on the simulation of SST in the region around the north coast of Penang Island, Malaysia, where the prediction of SST is challenging due to its complex atmospheric and oceanic interactions. The Regional Ocean Modeling System (ROMS), a sophisticated numerical model, is employed to predict the variation of SST in the study region. In the present study, HYCOM Global Ocean Forecasting System (GOFS) was incorporated to generate the boundary condition, initialisation, and climatology, while MERRA-2 datasets were considered as atmospheric forcing datasets. The generated SST from ROMS was compared with Aqua-MODIS observations of SST across six selected locations. Different methods, such as time series plots, linear modelling plots, and Taylor diagrams, and error estimation methods were employed to understand the accuracy of the model. The result indicates an underestimation of the SST using the ROMS model. Also, the root mean square error (RMSE) and mean absolute error (MAE) show an average of 2.58°C and 2.50°C in the study area, highlighting the requirement for bias correction. Three bias correction methods, such as Delta Change (DC), Linear Scaling (LS), and Quantile Mapping (QM), were considered to improve SST predictions. The comparative analysis of these three methods through time-series plots and statistical evaluations demonstrates that all three methods significantly reduce errors by bringing RMSE and MAE below 0.7°C. It is also noted that the best result was obtained by the QM method, as it not only reduces mean errors but also enhances correlation between the predicted and observed SST, the other two methods show no variation in the correlation value. The study confirms that the ROMS model can effectively capture the characteristics fluctuation of the SST in the dynamic regions like the north coast of Penang Island but bias correction is crucial for improving the prediction. In this case, the QM bias correction method provides the most balanced and effective adjustment compared to the other two methods.