Recent analyses of a six-year time series of radar-rainfall observations have clearly established the temporal and spatial coherence of both stationary and propagating warm-season rainfall over the continental United States. However, operational numerical weather prediction models have shown only limited skill in the prediction of both the diurnal cycle and movement of warm-season precipitation over this region. In the current study we use extended (1-10 day) WRF simulations with different configurations (e.g., grid resolutions, model physics) to better understand sources of error in warm-season precipitation forecasts. Such understanding could lead to improvements in operational forecasts and help elucidate dynamical mechanisms responsible for rainfall coherence.

We are currently analyzing a 10-day simulation for a period (19-29 July 1998), during which observed heavy precipitation episodes initiating near the lee of the Rocky Mountains exhibit coherence on timescales of multiple diurnal cycles with spatial correlations over swaths greater than 1000 km. Our preliminary WRF simulation is continental in scale, utilizes 3-h updates from operational ETA model analyses as lateral boundary conditions, employs 27-km horizontal grid spacing, and a cumulus scheme. While this extended simulation produces the observed late afternoon and early evening initiation of deep convection in the lee of the Rockies with reasonable accuracy, it fails to simulate the subsequent eastward propagation in a systematic fashion. Long-lived eastward propagating convection is present in the simulation. However, both the location of the onset of propagation (generally too far east) and the speed of propagation (generally too slow) are inconsistent with observations from the 10-day period.

We plan to compare additional model produced/derived fields with their observational counterparts. We also plan to compare the simulated precipitation patterns in the current model setup with higher-resolution (cloud resolving) simulations. By applying this approach to both the 10-day period of our preliminary simulation and to additional extended periods that comprise different large-scale regimes, we expect to gain a better understanding of both the forecast errors and physical mechanisms responsible for midlatitude warm-season rainfall coherence.