Research Abstracts 1997-1998: Report No. 98-01

Barbour, P.L. and S. N. Walker, “Site Wind Forecasts from a Regional Weather Prediction Model," OSU Wind Research Cooperative, WRC Report No. 98-01, September 30, 1998, 40 pp., $28.50

An evaluation is performed to examine the use of a regional, mesoscale weather prediction model in preparing wind energy forecasts for time periods up to thirty-six hours.  Forecasts on this time scale would allow utilities to incorporate predicted wind energy availability into their daily schedule of resources and, would in turn, improve the economic benefits associated with a resource that can vary substantially over time.  The regional approach has a number of benefits including the use of a much higher spatial resolution.  This enhances resolution should improve the representation of the topographic influences as well as a number of physical processes that can have a significant influence on near surface winds.  The overriding goal of this project is to evaluate this approach as a method of obtaining site specific wind forecasts in areas of high winds.

The regional mesoscale model used in this study is the Advanced Regional Prediction System (ARPS), developed at the Center for Analysis and Prediction of Storms (CPAS) at the University of Oklahoma. The ARPS model is a three-dimensional, nonhydrostatic model that includes equations for Cartesian wind components, potential temperature, pressure and mixing ratios of water vapor, cloud water and other moisture quantities on a generalized terrain following coordinate system. Initial conditions and time dependent boundary conditions are obtained from real time output from the National Centers for Environmental Predictions (NCEP) ETA model. Twelve case studies were generated using the 00 GMT model cycle and run for 36 hours on telescoping grids of 36 km, 12km and 4 km resolution. Only a limited number of runs on the 4 km grid were conducted due to the computational requirements involved. The telescoping grids were centered over a location on the south coast of Oregon. This region was selected as a study region because a number of well exposed high wind data sites exist in the area and because the region is influenced by several types of topographically influenced flow patterns. 

To evaluate the use of this regional weather prediction approach, grid point model predictions are compared with surface observations and examined to determine the magnitude of the errors, the presence of any model bias and to estimate the degree of correlation. Results indicate that performance can vary substantially from site-to-site and from case-to-case. Much of this variation was found to be associated with the type of case examined, whether the winds were the result of a quasi-stationary topographically influenced flow or from the presence of a broad scale synoptic disturbance. This has important implications and suggests that a regional model may be an effective tool, especially in areas in which the winds are generated by quasi-stationary features. The results also show a negative bias in the wind forecasts at each site examined, suggesting a general tendency towards under-prediction of the observed winds. Possible reasons for this are discussed.