Lessons learned on short-term explicit prediction of storms through a joint IHOP_2002 retrospective study

An across-laboratory research program, named STEP (Short Term Explicit Prediction), was established at the National Center for Atmospheric Research (NCAR) in 2006 with the objective of improving the short term explicit (convection permitting) forecasting of high impact weather. The STEP program enables the coordination of NCAR's exiting wide-range of activities and expertise across the institution with emphasized themes to achieve this objective. These activities range from basic understanding of high impact weather to development and real-time demonstration of data assimilation and forecasting systems. Last year, STEP organized a joint retrospective study using IHOP_2002 data with the participation of eight projects covering topics of basic understanding of deep convective systems, high-resolution data assimilation and numerical forecasting, physical parameterization, nowcasting system, and advanced verification technique. The main objective of the joint retrospective study was to evaluate the current capabilities, which are developed at NCAR with the goal of supporting the research and operational community, for short-term explicit forecasting of storms.

The period of June 10-16, 2002 was selected for the common focus of the study. This period was chosen because of the high frequency of convection occurrence. Wilson and Roberts (2005) documented Twenty-two initiation episodes for the seven days. The statistical analysis of observed rainfall revealed that the regime of convection for the one-week period corresponded well to translating synoptic cold front pattern with significant southeastward propagation. To evaluate the impacts of various factors that play key roles for the short-term forecasts of convective systems, a baseline simulation for the period was conducted using WRF (Weather Research and Forecasting) model to set a benchmark. The baseline run was initialized at 00 UTC and 12 UTC with NAM 40 km analysis, producing 36 hour and 24 hour forecasts respectively. Simulations initialized by WRF-3DVar and RTFDDA, a data assimilation system based on observation nudging, were conducted to compare with the baseline run. Sensitivities of precipitation forecasts with respect to microphysical parameterization, land-PBL coupling, and PBL schemes are also evaluated. Due to computation limitation and the level of system readiness, WRF-4DVar and EnKF were run with only specific cases and on smaller domains. Diagnostic analyses of key dynamical and thermodynamical processes contributing to forecast success and failure, such as low-level convergence, low-level wind shear, CAPE and CIN, and cold pool, were conducted. These analyses were compared with observations or analyses resulting from data assimilation. Lessons learned from the sensitivity and diagnostic studies will be presented.