Abstract
Comparing the near-surface wind speeds obtained from the most recent global circulation model (GCM) simulations to well-known benchmark datasets like the European Centre for Medium-Range Weather Forecasts reanalysis Version 5 (ERA5) and the Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA2), is necessary to make a critical assessment. Using 28 Coupled Model Intercomparison Project Phase 6 (CMIP6)-based monthly surface wind predictions, the multi-model ensemble (MME) approach in this study generates these predictions using random forest (RF) and multiple linear regression (MLR) methods over seven geographical regions in Türkiye with varying topographic complexity between 1980 and 2014, along with an offshore region. Benchmark datasets, station observations, and individual GCM predictions are used to compare the performances of MME predictions. The analysis showed that individual and the simple mean of GCM simulations are highly biased in spatial and temporal wind means. On the other hand, the MMEs formed by using groups of GCMs have significant skill for representing temporal variability in wind speed as well as for producing annual climatology and anomaly range for topographically complex regions. In MME predictions, the correlation improvements are 38–45% for RF and 22–34% for MLR. Moreover, the effect of the model group with dynamic vegetation growth on improvement remains only marginal.
DOI Link to Publication