Ensemble quantification of short-term predictability of the ocean dynamics at a kilometric-scale resolution: a Western Mediterranean test case

Abstract. We investigate the predictability properties of ocean dynamics using an ensemble short-term numerical regional simulations forced by prescribed atmospheric conditions. In that purpose, we developed a kilometric-scale, model for Western Mediterranean sea (MEDWEST60, at 1/60∘ horizontal resolution). A probabilistic approach is then followed, where stochastic parameterization uncertainties introduced in this setup to initialize experiments. set three experiments (20 members and 2 months) are performed, one with deterministic initiated perturbed initial conditions two model, different amplitudes perturbations. all experiments, spread shown emerge from smallest scales (kilometric scale) progressively upscales largest structures. After months, variance saturates over most spectrum, small (<100 km) have become fully decorrelated across members. These can provide statistical description dependence between accuracy forecast time lags 1 20 d. The assessed cross-validation algorithm (i.e., alternatively each member as reference truth remaining 19 forecast) together given score characterize accuracy. From joint distribution final scores, it possible quantify probability or reciprocally derive on obtain target misfit quantified terms overall (CRPS score), geographical position structures (location score) spatial spectral decorrelation surface height 2-D fields (decorrelation score). With approach, estimate example that, region period interest, location required (necessary condition) perfect (no uncertainty) 10 km 95 % confidence about 8 d forecast, 4 5 1.5 requirement cannot be met 15 longer forecast.