Towards the Exclusive Use of the Global Ensemble Prediction System for Global Forecasting in Canada

Several upgrades were implemented in 2024 for the Canadian operational Global Ensemble Prediction System (GEPS). These upgrades resulted in significant forecast improvements both for the ensemble forecast and the control member forecast (i.e. a deterministic forecast initialized with the ensemble mean analysis). Included among the upgrades are the use of scale-dependent localization and purely ensemble-based covariances for the 4D Ensemble-Variational (4D-EnVar) analysis used to partially recenter the analysis ensemble produced by the Local Ensemble Transform Kalman Filter (LETKF) algorithm. The horizontal resolution was also increased from 39km to 25km grid spacing. As a result of these changes and others, the GEPS configuration is now quite similar to the Canadian operational Global Deterministic Prediction System (GDPS), which also uses 4D-EnVar and has slightly higher horizontal resolution with 15km grid spacing. A thorough comparison of the quality of the GEPS control member forecast with the GDPS forecast was conducted. To better understand the differences between GEPS and GDPS, additional experiments were performed, including a GDPS at the same 25km grid spacing as GEPS and a different approach in GEPS for recentering the ensemble on the 4D-EnVar analysis. It was found that, when using the same spatial resolution, the GEPS control member produced better forecast quality than GDPS for many variables, regions, and lead times. As a result of this comparison, additional modification to GEPS are planned to further reduce the differences between GEPS and GDPS with the goal of eventually using GEPS exclusively for all operational global prediction applications.