At the Forecast System Lab many of our customers have a need for 0- to 3-hour precipitation forecasts to support many missions in transportation, fire weather, space operations, flash flood, and water management. These efforts frequently use small LINUX clusters. Our goal was to improve model initialization so that cloud and precipitation systems would become dynamic entities in the model fields, and yet operate efficiently in the limited computer environment. The Local Analysis and Prediction System (LAPS) developed in the 1990s is used to perform the data synthesis and analysis to initialize the mesoscale models. LAPS is unique in that in addition to analyzing state variables it performs a three-dimensional cloud and precipitation analysis using satellite imagery, radar, aircraft, and surface reports.
The diabatic scheme takes advantage of the cloud analysis by recovering cloud variables (cloud water and ice) by running a simple one dimensional cloud model through the diagnosed cloud layer. In addition, after determining the cloud type (based on satellite and radar data and the ambient conditions in the cloud environment), we parameterize a cloud vertical motion. The next phase uses a 3-dimensional variational technique that imposes two constraints on the field: 1) a mass continuity constraint that couples the horizontal winds to the input cloud vertical motion, and 2) a mass-balance constraint that forces a coupling between the momentum and mass fields. Further, the cloud environment is slightly supersaturated. The resulting analyzed fields then are interpolated to the model grid and the forecast is initiated. We find that the model rapidly spins up clouds and precipitation in the first few times steps.
The presentation will discuss the LAPS cloud scheme, the governing equations, and model preprocessing. We will show examples of how the clouds and precipitation develop in the first few model time steps. The diabatic scheme has been applied to forecasts for sub-tropical and winter precipitation with improved verification. The LAPS diabatic scheme has also been applied to typhoon forecasting in the Taiwan and China coastal areas. We will discuss efforts to develop systematic post processing to blend these forecasts seamlessly with diagnosed QPE
One exciting development is a hourly-run, time-phased, mesoscale ensemble using diabatically initialized MM5 and WRF. This allows production of probabilistic QPF in the first few hours. This ensembe has been run to provide probabilistic forecasts for road transportation forecasts. Here, the forecast spread is based on hourly varying initial conditions from the diabatic scheme. Results from these tests and applications will be presented.