A First Look at the Meteorological Development Laboratory's Experimental ECMWF MOS System

The skillful performance of the European Centre for Medium-Range Weather Forecasts (ECMWF) model is well documented and widely recognized in the meteorological community. While quite skillful, ECWMF model output can contain systematic bias. Further it does not directly provide forecasts for weather elements such as probability of precipitation. In order to enhance the usefulness of the ECMWF model to NWS forecasters, the Meteorological Development Laboratory (MDL) has applied a Model Output Statistics (MOS) approach to ECMWF model output. The MOS approach, originally developed by Glahn and Lowry, has been successfully employed by MDL to statistically post-process numerical model output for several decades.

A short and extended-range experimental MOS product has been developed for stations from 0000 UTC ECWMF model output. This product is both deterministic and ensemble-based. The deterministic guidance contains many of the weather elements that are found in MDL's station-based Global Forecast System (GFS) MOS for both the short-range and extended-range. The ensemble guidance provides additional statistical information that shows the uncertainty surrounding ECMWF MOS forecasts.

In this paper, we focus on the performance of ECMWF MOS deterministic temperature and dewpoint, daytime maximum temperature (MaxT), nighttime minimum temperature (MinT), and 12-h probability of precipitation (PoP12) forecasts. For comparison purposes, we have included the verification of GFS MOS forecasts for the same independent period. We have found that 0000 UTC ECMWF MOS forecasts demonstrates skill over both the 0000 UTC and subsequent 1200 UTC GFS MOS guidance when matched at the same valid time.