Comparative statistical analysis of sea surface temperature time series modeling in the Gulf of Mexico

This work provides an analysis of sea surface temperature (SST) data in the Gulf of Mexico (GoM), and compares modeling methods to assess their performance through statistical simulation. The GoM is a region of global economic and environmental importance, including the notable Flower Garden Banks National Marine Sanctuary’s coral habitats. However, the area has largely not been focused on in terms of advanced SST modeling. The objective of this work is to provide insight into SST variability in the GoM by modeling time series data, concentrating on seasonal patterns and the overall linear warming trend. In the study, varying methods compared include a dynamic seasonal mean model, a dynamic sinusoidal regression model, a cyclic spline model, and a seasonal mean regression model with sandwich estimator. At the 95% nominal confidence level, methods ignoring the autocorrelation of error terms maintain empirical coverage between 60% and 80% to capture the linear yearly warming trends in the simulation. The recommended modeling approach, a dynamic sinusoidal regression model, provides such coverage of approximately 91.6%. It also offers a balance between accuracy, simplicity, and continuity, especially for frequently collected data. Analysis estimates the East Flower Garden Bank has an annual warming trend of approximately 0.028 ± 0.016°C. This comparative analysis of modeling techniques contributes to a better understanding of SST dynamics in this critical marine ecosystem.