Abstract

Global wind reanalysis datasets (ERA5) are commonly used in offshore wind energy (OWE) studies, especially in offshore zones where ground-based observational data (GBOD) are scarce and impossible. In tropical coastal regions such as Vietnam, however, the dataset still has systematic bias. That bias can distort wind resource estimates if it is not corrected. This study analysed the monthly zonal (U) and meridional (V) wind components from ERA5 against GBOD from the Con Dao and Tho Chu island stations for 2010-2020. ERA5 overestimated wind speed at the stations. We corrected the bias with a linear regression method applied separately to U and V at the monthly scale. The method was then tested with a Leave-One-Month-Out (LOMO) cross-validation scheme. The results showed a clear improvement. Mean Bias Error (MBE) dropped from an overall average of 3.91 m/s to a statistically negligible value, and Root Mean Square Error (RMSE) fell by 55.0%, from 4.78 to 2.15 m/s. This indicates that a local, component-wise correction is needed in this region. The method gives one practical way to prepare wind data for renewable energy investments (REIs) in special zones such as Con Dao and Tho Chu. It may also be used in similar coastal regions.

Statistics