Benefits and Tradeoffs of Ensemble Resolution versus Ensemble Size for Orographic Precipitation Forecasts

The west coast of the United States receives 25-50% of its annual rainfall from just a handful of powerful storms, referred to as atmospheric rivers. Water managers in the west coast rely on water captured in reservoirs from these heavy storms to provide water to California for the rest of the year. However, these storms can also present severe flood risks. As such, accurate and timely forecasts of atmospheric rivers are critical for managing water resources and flood risk in the west coast. The Center for Western Weather and Water Extremes (CW3E) has developed an in-house version of the WRF model (West-WRF) that is optimized for forecasting atmospheric river events along the west coast. During the wet season (October through April), CW3E runs an operational 200-member ensemble based on West-WRF at 9km horizontal grid spacing to provide forecasting and decision support tools to water managers. An important consideration moving forward, however, is are limited computational resources best spent on creating larger ensembles, or smaller ensembles with higher resolution? In this study we compare results from two experimental versions of the operational West-WRF ensemble, one run at 9km resolution, and one run at 3km resolution, each run with 72 members. Increasing the horizontal resolution from 9km to 3km adds skill to the forecast, but increasing the computational expense by a factor of 27. We then evaluate how the skill of the 3km ensemble degrades as we reduce the number of members to 35, 20, and 10. We find that the 35 member version of the 3km ensemble still outperforms the 72-member 9km ensemble, however as we move to smaller and smaller ensembles, the skill degrades significantly, especially for extreme events. The 10-member version of the 3km ensemble, while still being almost 4 times as computationally expensive, performs worse than the 9km ensemble in almost every metric considered. Therefore, we conclude that while there is clear value in increasing the horizontal resolution in an ensemble forecast, it is outweighed by the need to maintain a large ensemble.