Deducing Aerodynamic Roughness Length from Abundant Anemometer Tower Data to Inform Wind Resource Modeling

There are tens of thousands of anemometer towers currently being built for wind resource assessment. In this study, we show these towers provide a precious opportunity to improve wind resource modeling, which is the basis for the development of wind energy industry. In atmospheric models, aerodynamic roughness length (z₀) is a critical parameter for the simulation of wind speed in the near-surface layer (0-200 m). However, current gridded z₀ datasets in atmospheric models are usually estimated from land cover types and may have large uncertainties. Although some efforts have been made to produce accurate gridded z₀ datasets using machine-learning methods, their accuracy and applicability remain unknown. In this pilot study, we enriched z₀ ground truth from wind profile data of 101 anemometer towers in China and assessed the uncertainty of existing gridded z₀ datasets and their effects on wind speed simulations.

Specifically, we show that although the latest gridded z₀ dataset obtained with a machine-learning model performs better than z₀ reanalysis datasets (i.e., ERA5 and CFSv2), all of these datasets contain considerable uncertainty and fail to capture the evident variability of z₀ observed within each land cover type. Furthermore, the errors in gridded z₀ datasets do map to systematic biases in the simulated near-surface wind speed. For example, we find that z₀ in ERA5 is overestimated in wind-rich regions of China, causing an underestimation of near-surface wind speed, which is contrast to its widespread overestimation on wind speed in urbanized areas of China. Our results suggest that there is an urgent need for better gridded z₀ datasets, and the tens of thousands of anemometer towers currently being built for wind resource assessment may already provide a solution to this problem.