Coupled estimation of global 500m daily aerodynamic roughness length, zero-plane displacement height and canopy height

•A practical method is proposed for estimates of global 500 m daily z0m, d and h.•An optimized scheme of d/h and z0m/h is intercompared with five classical schemes.•A BRF model is constructed with observations at 271 flux-sites and by Lidar.•Model estimates are in good agreement with observations and reanalysis products. Aerodynamic roughness length (z0m) and zero-plane displacement height (d), highly related to canopy height (h), are pivotal parameters to characterize the exchange of momentum, heat and water between the land and atmosphere and widely used in many remotely sensed evapotranspiration models, hydrological models and land surface models. In this study, we developed a practical method for coupled estimates of global 500 m daily z0m, d and h by combining the measurements over 2000–2019 at 271 sites worldwide, collocated remote sensing and reanalysis datasets, an optimized scheme, balanced random forest (BRF) and wind profile equation. The h estimated by the optimization (Opt) scheme, in comparison to that by five classical schemes, was in the highest agreement with the dynamic observations at 109 of 271 sites, with a root mean square error (RMSE) of 1.14 m (varying between 0.17 m and 2.58 m over different vegetated surfaces) and a coefficient of determination (R2) of 0.99. Meanwhile, the validation of h estimated by the BRF model against the inferred h at 271 sites and the Lidar-based observations showed an RMSE of 0.13 m - 4.84 m and an R2 of 0.79 - 0.95 over different vegetated land cover types. The z0m estimated by combining the BRF model and the Opt scheme presented a similar global spatial distribution compared to the CFSv2 and ERA5 products in 2019, with the estimated z0m exhibiting the greatest temporal-spatial variability. Moreover, the estimated z0m was more similar in magnitude to the ERA5 product but lower than the CFSv2 product. The generated d product showed similar spatial distribution and temporal-spatial variability to the z0m estimates, but was about 4 to 15 times larger than z0m in magnitude. This study could benefit the improvement of simulation of turbulent flux transfer between the land and atmosphere.