Further Evaluation of the HRW Model using Conditional Sampling

A revisit to the previous evaluation of the high resolution wind model (HRW) was undertaken recently with the use of conditional sampling. Our preliminary evaluation of HRW using the MADONA data base showed a very good agreement of the simulated wind speed and direction values versus the observed surface station values for 39 cases. Correlation coefficients for direction and speed were 0.97 and 0.86 respectively. HRW is US Army Research Laboratory's 2-D, diagnostic, microscale, time independent, high resolution wind model to calculate wind fields over complex terrain. The model is usually run with a computational grid size of 100 m and a domain of 5 km by 5 km. Physically, calculations are performed on an array of air parcels in a pressure field such that accelerations of these parcels are determined as they negotiate the changing slopes of the terrain and the added thermal lift or suppression component imparted by buoyancy. The procedure makes use of Gauss' Principle of Least Constraints that requires the forces to be minimized in order to satisfy the equations of motion. Mass is also conserved during the calculations. Numerically, the direct variational over-relaxation method is used to obtain a minimum of the combined acceleration forces in a pressure field. . These calculations are completed in an iterative manner. The solution is reached when the internal constraints forces imposed by the warped terrain surface, thermal structure, and requirements for flow continuity are minimized. Recently, a more in-depth analyses was performed on the same data set of paired observed and simulated values by dissecting the results with a specific set of sampling conditions. The paired data set was conditionally sampled with respect to wind direction sectors, ranges of wind speed, and atmospheric stability. Correlation remained high ( r=0.93 to 0.98) for all classes of wind direction. With regard to sampling on stability, the correlation was better for wind direction (r=0.88 to 0.98) while agreement for wind speed was somewhat less (r=0.85 to 0.87). For the analysis based upon wind speed, agreement continued to be good (r=0.66 to 0.89) except for the correlation value (r=0.16) of the highest class of observed speeds. In general, the detailed correlations substantiated the overall results previously attained even though there was a low correlation specifically at the highest wind speed range (speed=6 to 11 m/s). This low correlation is thought to be related to gustiness at higher wind conditions and how it violates the steady state assumptions of HRW