Hi-res gridded MOS 6-h qpf guidance

A new Global Forecast System (GFS)-based MOS model has been developed by the Meteorological Development Laboratory to produce 6-h quantitative precipitation forecasts (QPFs) to 120 hours (five days) on a 4-km grid over the contiguous United States (CONUS). The basic form of the products is probabilistic QPF (PQPF) for categories, but, in addition, categorical and continuous precipitation variables are derived from the multi-category probabilities. The CONUS grid together with the continuous QPF product is compatible with that contained in the National Digital Forecast Database (NDFD) of the National Weather Service (NWS).

A signature feature of the MOS QPF products is their high spatial resolution (“hi-res”), which is most pronounced in mountainous regions of the CONUS. An added hi-res dimension in the multiple QPF products is their enhanced precipitation intensity resolution. The added detail arises largely from fine scale precipitation climatology and topography predictor inputs, crafted interaction of these topo-climatic predictor inputs with large scale predictor variables from the GFS, and geographical regionalization of the underlying PQPF regression equations.

Another defining feature of the hi-res QPF model is the adoption of a new geographical regionalization technique, which was designed to retain sharp PQPF gradients along “natural” regional boundaries and prevent non-meteorological discontinuities along “quasi-arbitrary” regional boundaries. The new technique involves area overlap of neighboring model regions and objective weighting of multiple PQPFs in region overlap zones. While the technique prevents non-meteorological discontinuities, it retains forecast skill exhibited by the conventional non-overlapping regions approach.

Preliminary subjective and objective verification of the hi-res QPFs against comparable model and human-produced QPF products indicates the former should have substantial guidance value to forecasters. Beyond their competitive skill to the best model and human-produced QPFs, they exhibit superior bias properties.

Presently, the QPF products are produced at and made available from the NWS National Centers for Environmental Prediction Central Computing Facility twice daily in an experimental mode. Conversion to operational production and dissemination is anticipated within the next year.