Using Model Ensemble-based Uncertainty Information to Blend Modeled and Extrapolated Forecasts of Convection

The goal of this work is to more skillfully blend extrapolation and model forecasts of convection using model-based forecast uncertainty information derived from a model ensemble. This R&D effort takes place within the framework of CoSPA, which has been developed for the short-term prediction of convective hazards for aviation and is currently being integrated into the Federal Aviation Administration’s NextGen Weather Processor (NWP). The current version of CoSPA blends Corridor Integrated Weather System (CIWS)-like extrapolation forecasts of VIL (Vertically Integrated Liquid, a proxy for precipitation intensity) and Echo Tops from MIT-Lincoln Lab with those obtained from Version 2 of the High Resolution Rapid Refresh (HRRR) available from NCEP. The blending algorithm, currently run at the National Center for Atmospheric Research (NCAR), refines the model forecasts of VIL and Echo Tops via calibration and position error corrections and then combines the two forecasts using a heuristic weighted average. The initial set of weights are computed based on the relative past performance of the model and extrapolation forecast. This initial set of weights is then adjusted regionally based on the stage of convective development or regime (e.g., ongoing, initiation and growth, decay). Work presented here focuses on optimization of the weights used in the storm initiation and growth regime (IGR). The IGR weights are based on forecast uncertainty information derived from a time-lagged ensemble of successive HRRR forecasts. The relationship between forecast uncertainty and storm scale will be assessed and the optimal scale to use in the blended forecasts will be presented.

This research is in response to requirements and funding by the Federal Aviation Administration (FAA).  The views expressed are those of the authors and do not necessarily represent the official policy or position of the FAA.