The Use of Multiple Reanalysis Data Sets as an Ensemble of Historical Climate for Wind Energy Resource Assessment

When a wind energy developer is considering building a project at a particular site, it is important to develop a good understanding of the wind resource at the site, both spatially and temporally, for several purposes including project financing, turbine selection, and turbine layout. To this end, a developer conducts a measurement campaign, in which wind measurements are gathered for at least one year at one or more points across the project site using anemometry on tall towers. However, these observations are limited in both spatial coverage and temporal record. Multi-decadal wind data from reanalysis datasets (RDs) provide a valuable input to help fill these gaps. The full record of a RD (either raw or downscaled with a mesoscale model) can be used as a long-term reference, to determine the long-term climate context in which the single year of observations reside. Additionally, a more limited selection of the temporal record can be used to drive very high-resolution (< 1 km) mesoscale model simulations to help fill in the spatial gaps in the observations from the tower network. This talk will focus on the former application, i.e., the use of RDs to understand the long-term historical climate context of short-term observations. The long-term climate is often distilled down to a single 'long-term adjustment factor', which is the factor one can apply to the single year of observations (which may have been gathered during an anomalous year) so that the adjusted mean is representative of the multi-decadal historical average, or what one might expect the wind farm to produce over the life time of the project (typically ~20 years). The growing set of high-quality, publicly available global RDs provides near-surface wind estimates that can be used as a ~30-year reference data set anywhere on the globe. Because the RDs are produced by independent centers with different methodologies, they necessarily give different answers at the same site. However, an attractive idea is to use the spread among the RDs to help quantify the uncertainty in the long-term adjustment, or in other words, to use multiple RDs as an ensemble of long-term references. The ensemble can potentially be improved by calibrating it to actual uncertainty for a number of sites at which measured long-term wind speeds are known. This presentation will explore the use of several existing RDs as an ensemble of long-term references at a site, and demonstrate how such a use can help quantify uncertainty in the long-term adjustment. The analysis will be based upon a large number of measurement sites for which multiple years of data are available, so that out-of-sample validation can be conducted.