Impact of Model Error Techniques on the Forecast Skill of the Navy ESPC Ensemble
Abstract
The research presented here illustrates the impact of several model error techniques on the fidelity of extended-range global coupled ensemble forecasts produced by the Navy ESPC model. The presented methods aim to improve the forecasts by reducing errors in the forecasted state both at initial time and during the model integration. To address error in the initial state, we explore the use of Relaxation To Prior Perturbations (RTPP)which will aim to better capture model uncertainty in the initialization of the ensemble forecast by relaxing the initial state from the analysis(produced by the data assimilation system) back toward the prior (or forecasted state). During the model integration, we explore the use of two methods; 1) Analysis Correction-based Additive Inflation (ACAI) and Stochastic Kinetic Energy Backscatter (SKEB). Both of these methods will act as a representation of stochastic model error intended to increase the divergence of the ensemble; however, in the case of ACAI, there is also an explicit term in the perturbations aimed at reducing systematic errors. On the other hand, we have also found that the SKEB perturbations can act to modify the mean state resulting in improvements to the bias. All three model error techniques present clear improvements to the skill of the forecasts both in the short-range (weeks 1 and 2) and at extended range time scales (weeks 3-6). The Navy ESPC model is used operationally to generate 45-day forecasts, and these research findings present a clear pathway to improve the skill of our global coupled ensemble forecasting system.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 28, 2021
- Accession Number
- AD1130118
Entities
People
- William J. Crawford
Organizations
- United States Naval Research Laboratory