Impacts of MM5 Model Data on the Performance of the NCAR Auto-Nowcaster System

The Auto-Nowcaster System (ANC) is a software system that produces time- and space-specific, routine short term nowcasts of storm location and intensity. A set of predictor fields which are based on observations (radar, satellite, sounding and mesonet), a numerical boundary layer model and its adjoint, forecaster input, feature detection algorithms as well as the operational RUC20 (Rapid Update Cycle model at 20 km resolution) output has been identified. The predictor fields include both storm scale features derived from radar and other sensors and large scale environmental variables derived from RUC20 model. Each predictor field is associated with a membership function which converts the predictor field into an interest field. All the interest fields are fused together to produce a likelihood field using a fuzzy logic algorithm according to their weights. A final forecast of the convective activity is generated by filtering and smoothing the likelihood field.

Since the RUC20 model is the only model which was ever used in ANC and it has a relatively coarse space resolution of 20 km, it would be interesting to know how the performance of ANC will be affected if the RUC20 model was replaced by a model with better spatial resolution and different model physics such as MM5 (3.3 km resolution). Another advantage of the MM5 model used in this study is its ability of assimilating radar reflectivity data using a nudging technique. The ANC performance using both RUC20 and MM5 model data will be compared and future usage of MM5 in ANC will be discussed in detail.