Sensitivity of 0-12-h warm-season precipitation forecasts over the central United States to model initialization

Sensitivity of 0-12 -h warm-season precipitation forecasts to atmospheric initial conditions, including those from different large-scale model analyses and from rapid cycled (RC) three-dimensional variational data assimilations (3DVAR) with and without radar data, is investigated for a 6-day period during the International H₂O Project. Neighborhood-based precipitation verification is used to compare forecasts made with the Advanced Research core of the Weather Research and Forecasting Model (ARW-WRF).

Three significant convective episodes are examined by comparing the precipitation patterns and locations from different forecast experiments. From two of these three case studies, causes for the success and failure of the RC data assimilation in improving forecast skill are shown. Results indicate that the use of higher-resolution analysis in the initialization, rapid update cycling via WRF 3DVAR data assimilation, and the additional assimilation of radar observations each play a role in shortening the period of the initial precipitation spinup as well as in placing storms closer to observations, thus improving precipitation forecast skill by up to 8-9 h.