Can a Global Model Accurately Simulate Land-Atmosphere Interactions under Climate Change Conditions?

Abstract:

Surface air temperature (T_a) is largely determined by surface net radiation (R_n) and its partitioning into latent (LE) and sensible heat fluxes (H). Existing model evaluations of the absolute values of these fluxes are less helpful because the evaluation results are a blending of inconsistent spatial scales, inaccurate model forcing data and inaccurate parameterizations. This study further evaluates the relationship of LE and H with R_n and environmental parameters, including T_a, relative humidity (RH) and wind speed (WS), using ERA-interim reanalysis data at a grid of 0.125°×0.125° with measurements at AmeriFlux sites from 1998 to 2012. The results demonstrate that ERA-Interim can reproduce the absolute values of environmental parameters, radiation and turbulent fluxes rather accurately. The model performs well in simulating the correlation of LE and H to R_n, except for the notable correlation overestimation of H against R_n over high-density vegetation (e.g., deciduous broadleaf forest (DBF), grassland (GRA) and cropland (CRO)). The sensitivity of LE to R_n in the model is similar to the observations, but that of H to R_n is overestimated by 24.2%. In regions with high-density vegetation, the correlation coefficient between H and T_a is overestimated by more than 0.2, whereas that between H and WS is underestimated by more than 0.43. The sensitivity of H to T_a is overestimated by 0.72 Wm^-2 °C^-1, whereas that of H to WS in the model is underestimated by 16.15 Wm^-2/(ms^-1) over all of the sites. Considering both LE and H, the model cannot accurately capture the response of the evaporative fraction (EF=LE/(LE+H)) to R_n and the environmental parameters.