18th Conference on Weather and Forecasting, 14th Conference on Numerical Weather Prediction, and Ninth Conference on Mesoscale Processes

Wednesday, 1 August 2001
Evaluation of high resolution MM5 and Eta forecasts over the Northeast U.S
Brian A. Colle, SUNY, Stony Brook, NY; and J. S. Tongue and J. B. Olson
Poster PDF (962.6 kB)
A major question before the meteorological community regards the benefits of increasing horizontal resolution in short-range (1-3 day) numerical weather prediction (NWP). Over the Northeast U.S., the Appalachians, coastal boundaries, and urban areas result in a wide variety of mesoscale phenomena, such as sea breezes, coastal fronts, heavy precipitation, and gravity waves; all of which may more realistically predicted using high resolution NWP. Since October 1999, Stony Brook, in collaboration with the local NWS forecast office, has been running the Penn State/NCAR Mesoscale Model (MM5) twice daily down to 4-km resolution for a large region surrounding Long Island (http://atmos.msrc.sunysb.edu/html/alt_mm5.cgi). The goal of this study has been to quantitatively evaluate high resolution regional NWP for the Northeast U.S. as well as improve our understanding of various mesoscale weather phenomena. Two MM5 forecasts are run for each forecast cycle (0000 and 1200 UTC) using the NCEP ETA and AVN models, respectively, for initial and boundary conditions. The MM5 is integrated 48 hours for the 36 and 12 km domains and 36-h for the 4 km domain. The MM5 was verified at all upper-air and surface sites over the Northeast, and the results are compared to the verification of the same forecasts from the Eta interpolated to a 80 km grid (104 grids). This presentation will highlight the MM5 and ETA verification statisics as well as results from tropical storm Floyd (16 September 1999), in which increased horizontal resolution dramatically improved the MM5 precipitation forecasts. The sensitivity of the 12/4-km MM5 forecasts to using a different initialization (AVN) will also be discussed. Some preliminary conclusions from our verification statistics include:

1. There is a noticable improvement in the MM5 coastal windspeed and temperature skill and bias going from 36 to 12-km resolution, while little improvement in skill is apparent going from 12- to 4-km resolution.

2. During the cool season, both the MM5 and ETA surface temperatures are 0.5-2.0 deg C too cool on over land and 0.5-1.0 degC too warm over the coastal waters. This cool bias is maximized over the urban areas. During the warm season both the MM5 and Eta are too warm over land except at urban locations.

3. The MM5 10-m winds are 0.5-1 m/s too strong over non-urban land sites and 1-2 m/s too weak over major urban areas at 12/4-km resolutions. The Eta winds are 2-3 m/s too strong at most inland stations. Both the MM5 and ETA have little speed bias over the coastal waters.

4. The 12-km MM5 has overprediction of precipitation over the windward slopes of the Appalachians and underprediction in the lee of ridges. The warm season explicit 4-km precipitation is highly sensitive to what cumulus is used in the 36/12 km domains. The mid-level warming and drying induced by the Kain-Fritsch parameterization dramatically reduces 4-km explicit convective precipitation.

Supplementary URL: