Bulk Statistics of 80-m Wind Speeds of Reforecast Runs from HRRR and HRRRNEST Models during the Second Wind Forecast Improvement Project (WFIP2). Part I

During the second Wind Forecast Improvement Project (Oct, 2015 – Mar, 2017, Columbia Gorge and River Basin area) several improvements to the parameterizations of the High Resolution Rapid Refresh (HRRR – 3km grid spacing) and HRRRNEST (750m grid spacing) NWP models were tested during four 6-week reforecast periods (one for each season).

In this presentation we will show the impact of these improved parameterizations on the forecast of 80-m wind speeds (hub height) from these two models, comparing the forecast output to the observations collected by 19 sodars and 3 lidars. These instruments were chosen because they accurately measure wind speed and direction from 20 m to few hundred meters above ground level, which is the layer of the atmosphere more relevant for wind energy production.

In this presentation we will focus our attention on regular statistics (MAE, Bias, etc.) results, while in the following presentation (Part II) the impact on the improved parameterizations on the skill of the models at forecasting ramp events will be evaluated. In particular, we will compare the improvements due to higher resolution (HRRRNEST vs HRRR) versus those due to improved physics (control vs experiments runs), we will also show the results as a function of the models initialization time, of the season of the year, and of the time of the day.