Determining the Minimum Number of Ensemble Members Needed for Statistical Postprocessing

Ensemble forecast systems provide forecasters with an estimate of the amount of uncertainty in the future state of the atmosphere. Statistical postprocessing techniques, like MDL's Ensemble Kernel Density MOS (EKDMOS), have been used to calibrate the reliability of an ensemble forecast from global models for a number of weather elements. However, as the number of members in an ensemble forecast system increases, so do the resources needed for postprocessing. This work investigates the minimum number of ensemble members needed to reliably calibrate the ensemble mean and spread, while minimizing the impact on limited computational resources. Results from this study can also help reduce the resources required for future reforecast configurations.