Comprehensive hydrodynamic model validation using satellite observations around the Maltese islands

Abstract

Sea surface temperature (SST) constitutes one of the most important environmental parameters and is a critical factor mainly in climatology meteorology and oceanography. Changes in SST affect the Earth's atmosphere, the underlying water currents and the balance of the biological component of the earth ecosystem. They also contribute to the global ocean energy exchange. Observing and predicting SST is fundamental since SST plays a pivotal role in the ocean-atmosphere coupled system and therefore is an important indicator of the Earth’s climate. The Malta Shelf model ROSARIO-II hydrodynamical is an eddy-resolving, primitive equation, sigma level shelf-scale numerical model. It is developed, maintained and run in operational mode by the Physical Oceanography Research Group (University of Malta), providing daily forecast of sea salinity, sea currents and sea temperature around the Maltese Islands. Through this project, the modelled SST was validated against observed SST using remotely-sensed satellite measurements (Copernicus Marine Environment Monitoring Service products (Sentinel-3). In this study, the validity of the model is assessed in order to demonstrate that the model possesses a satisfactory range of accuracy consistent with the intended application of the model. The analysis includes the comparison of available data in terms of the face validity, by generating scripts to visually compare properties such as seasonal and monthly SST and to compare variability in terms of time series. Distributional and value validity is studied by comparing modelled SST with satellite derived estimates, means, standard deviations, and spatial patterns. The validation methodology is not limited to pointwise comparing the datasets, but the research extends to exploring possible time-lags in SST predictions, SST zone intensity variation and high- and lowSST spatial displacement relative to the observations. Patterns and features observed in both modelled and observed SST are analysed. Methods include the generation of specific scripts for basic statistics analysis, pointwise analysis, cluster analysis, time lag analysis, and other specialized analysis. Two new validation indicators are generated: The Histogram Intersection Coefficient and the Performance Index that assesses the model performance based on clustering analysis.