Assessment of predictability of the Loop Current in the Gulf of Mexico from observing system experiments and observing system simulation experiments

Abstract

This study presents results from numerical model experiments with a high-resolution regional forecast system to evaluate model predictability of the Loop Current (LC) system and assess the added value of different types of observations. The experiments evaluate the impact of surface versus subsurface observations as well as different combinations and spatial coverage of observations on the forecasts of the LC variability. The experiments use real observations (observing system experiments) and synthetic observations derived from a high-resolution independent simulation (observing system simulation experiments). Model predictability is assessed based on a saturated error growth model. The forecast error is computed for the sea surface height fields and the LC frontal positions derived from the forecasts and control fields using two metrics. Estimated model predictability of the LC ranges from 2 to 3 months. Predictability limit depends on activity state of the LC, with shorter predictability limit during active LC configurations. Assimilation of subsurface temperature and salinity profiles in the LC area have notable impact on the medium-range forecasts (2–3 months), whereas the impact is less prominent on shorter scales. The forecast error depends on the uncertainty of the initial state; therefore, on the accuracy of the analysis providing the initial fields. Forecasts with the smallest initial error have the best predictive skills with reliable predictability beyond 2 months suggesting that the impact of the model error is less prominent than the initial error. Hence, substantial improvements in forecasts up to 3 months can be achieved with increased accuracy of initialization.