Comparative verification of experimental HRMOS QPFs with model and forecaster-prepared QPFs in the NWS

Experimental High Resolution MOS (HRMOS) 6-h quantitative precipitation forecast (QPF) guidance, produced in real time twice daily to 192 hours over the contiguous United States (CONUS) since June 2008, is comparatively verified against selected model and forecaster-prepared QPF elements in the National Weather Service (NWS). Two QPF elements are scored: categorical QPFs (CQPF) and probabilistic QPFs (PQPF), each for multiple precipitation thresholds. The verifying precipitation observations are NWS Stage IV Quantitative Precipitation Estimates (QPE) for which supplemental quality control is applied at the NWS Meteorological Development Laboratory (MDL).

The CQPFs comparatively scored with the HRMOS CQPFs were from various sources. The sources consist of model QPFs from the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS) and the MDL Gridded MOS (GMOS) systems along with forecaster-prepared QPFs from the NCEP Hydrometeorological Prediction Center (HPC) and the NWS National Digital Forecast Database (NDFD). Relative to a given forecast valid time, the issuance time of each model QPF is such that the forecast could be used as guidance for the human QPF. Also, for each QPF source, the categorical form was obtained by categorizing QPFs issued in continuous form. The scoring was performed on the QPE 4-km grid, which necessitated grid interpolation of the continuous QPFs from all sources except HRMOS.

The PQPFs comparatively scored with HRMOS included PQPFs from station-oriented MOS forecasts and the recently inaugurated experimental gridded (model) PQPFs developed by HPC. (The MOS PQPFs are not available operationally, but they are used to produce the operational MOS CQPFs.) The HRMOS versus MOS scoring was performed at irregularly-spaced stations, while the HRMOS versus HPC scoring was performed on the 4-km QPE grid.

The performance scoring used conventional summary measures, which include threat score and bias for the CQPFs, and Brier skill score and reliability for the PQPFs. In the case of the CQPF scoring, day 1 to day 3 scores for recent warm and cool season samples indicated HRMOS performed as well or better than the best model or human QPFs included in the verification. The corresponding PQPF scoring showed that HRMOS and HPC had about the same skill (when averaged over all precipitation thresholds), which was better than that for MOS. The conference presentation will include scores to justify these statements.