Impact of improved initialization of mesoscale boundaries on heavy rainfall forecasts in 10 km Eta simulations

Several methods of improving the initialization of mesoscale features such as the inclusion of cold pools, convective boundaries, and small-scale moisture tongues are being investigated using the Eta model run with 10 km horizontal resolution. Simulations are being performed over the Iowa/Minnesota region where an extensive mesonetwork of surface stations exists. Approximately 20 cases have been identified during the 1998 and 1999 warm seasons where some type of mesoscale boundary is present at either 12 UTC or 00 UTC, and is followed by later convective development that may be due in part to the mesoscale features. Many of these cases involve substantial rainfalls and flash flood events.

An ensemble of runs is being performed using two types of mesoscale adjustment to the initialization and two different convective parameterizations, the operational Betts-Miller Janjic scheme, and the Kain-Fritsch scheme. A cold pool initialization procedure developed by David Stensrud is used for some of the ensemble members, with variations in the assumed cold pool layer humidity. For other ensemble members, the model's vertical diffusion is used to adjust lower tropospheric temperature and moisture values input from mesonetwork surface observations.

Preliminary results suggest that no single technique consistently improves the QPF for these heavy rain events. Both techniques, however, do improve the forecasts over short time periods within some cases. Some evidence suggests that the mesoscale modifications have the greatest ability to improve the forecast when low-level flow is relatively weak. The cold pool adjustment may offer the greatest benefit when the Kain-Fritsch scheme is used, since that scheme includes a convective downdraft which can sustain the cold pool over a longer period in the simulation.

The mixed results obtained with the various initialization improvements suggest that ensemble guidance may be of more value to forecasters in these situations. Some discussion will be given on probability forecasts derived from the ensembles.