An Experiment in Mesoscale Weather Forecasting in the Michigan Area

Abstract

During an experiment in mesoscale weather forecasting in the Michigan area, consensus improved over NWS guidance in maximum/minimum temperature and probability of precipitation forecasts out to 24 hours. Forecasts were generally best in the vicinity of the forecast site. Climatology, persistence, NWS guidance and consensus forecast errors in maximum/minimum temperature and precipitation probability forecasts were divided into synoptic and mesoscale contributions. The errors of the climatology and persistence forecasts resulted substantially more from the synoptic scale than the mesoscale. NWS guidance and consensus forecasts improved over climatology on the synoptic scale with much less or no improvement on the mesoscale. For temperature, consensus showed improvement over guidance only on the synoptic scale. For precipitation, no relationship to scale was evident. In terms of remaining errors, errors were distributed approximately equally between the synoptic scale and mesoscale for temperature. For precipitation, the errors were significantly greater from the mesoscale than the synoptic scale. The implications of these results for zone forecasts and future improvements in mesoscale forecasting accuracy are discussed.