Improved quality gridded surface wind speed datasets for Australia

Gridded near-surface (2 and 10 m) daily average wind datasets for Australia have been constructed by interpolating observational data collected by the Australian Bureau of Meteorology (BoM). The new datasets span Australia at 0.05 × 0.05° resolution with a daily time step. They are available for the period 1 January 1975 to present with daily updates. The datasets were constructed by blending observational data collected at various heights using local surface roughness information. Error detection techniques were used to identify and remove suspect data. Statistical performances of the spatial interpolations were evaluated using a cross-validation procedure, by sequentially applying interpolations after removing the observed weather station data. The accuracy of the new blended 10 m wind datasets were estimated through comparison with the Reanalysis ERA5-Land 10 m wind datasets. Overall, the blended 10 m wind speed patterns are similar to the ERA5-Land 10 m wind. The new blended 10 m wind datasets outperformed ERA5-Land 10 m wind in terms of spatial correlations and mean absolute errors through validations with BoM 10 m wind weather station data for the period from 1981 to 2020. Average correlation (R2) for blended 10 m wind is 0.68, which is 0.45 for ERA5-Land 10 m wind. The average of the mean absolute error is 1.15 m/s for blended 10 m wind, which is lower than that for ERA5-Land 10 m wind (1.61 m/s). The blending technique substantially improves the spatial correlations for blended 2 m wind speed.