Consistency of Reanalysis Data for Wind Resource Assessment

Most wind resource assessment campaigns involve collecting on-site wind speed data for a period of 2-5 years, which is not entirely sufficient for determining the long-term wind resource at the site. Thus, the wind industry relies on reanalysis datasets to provide a source of long-term reference data to support the over 28 gigawatts of new build under development today (AWEA). Reanalysis data offers both good spatial coverage and high temporal resolution with global availability by assimilating as many observations as possible to recreate the atmosphere at a given point in the past. Wind resource assessments rely on a consistent relationship between the observed data at a site and reanalysis data to accurately estimate the long-term wind resource at a particular location. Although adjustments are applied to maintain reanalysis data consistency, discontinuities can still arise as the result of major observing system instrumentation changes or additions. Inconsistencies in the reference timeseries can induce bias in the forecast of long-term wind plant production which could ultimately influence the plant’s expected revenue generation.

This presentation will compare reanalysis data from both MERRA2 and ERA5 to address their suitability as consistent, long-term datasets for wind resource assessment. Trends in mean wind speed will be analyzed at the monthly level to compare the two datasets over time with special attention paid to points in time where the reanalysis datasets begin ingesting new observations or an observing suite exhibited a large change. Results indicate that ERA5 and MERRA2 show divergence at a range of locations, especially before the year 2000, as the number of observations in that period is significantly lower than what is available today. The impact of reanalysis data inconsistency is demonstrated by using different historical periods of both ERA5 and MERRA2 to predict wind plant production for a real wind farms in the United States.