Simulating surface energy fluxes using the variable-resolution Community Earth System Model (VR-CESM)

Abstract

Recent advances in variable-resolution (VR) global models provide the tools necessary to investigate local and global impacts of land cover by embedding a high-resolution grid over areas of interest in a seamless and computationally efficient manner. We used two eddy covariance tower clusters in the Eastern USA to evaluate surface energy fluxes (latent heat, $λE$; sensible heat, $H$; net radiation, $R_{\textrm{n}}$; and ground heat, $G$) and surface properties (aerodynamic resistance to heat transfer, $r_{\textrm{aero}}$; Bowen ratio, $β$; and albedo, $α$) by uncoupled point simulations of the land-only Community Land Model (PTCLM4.5) and two coupled land–atmosphere Community Earth System Model (CESM1.3) simulations. The CESM simulations included a 1° uniform grid global simulation and global 1° simulation with a 0.25° refined VR grid over the Eastern USA. Tower clusters included the following plant functional types—broadleaf deciduous temperate (hardwood) forest, C3 non-Arctic grass (grass), a cropland, and needleleaf evergreen temperate (pine) forest. During the growing season, diurnal cycles of $λE$ and $H$ for grass and the cropland were simulated well by PTCLM4.5 and VR-CESM1.3; however, $λE$ ($H$) was biased low (high) at the hardwood and pine forested sites, contributing to biases in $β$. Growing season $R_{\textrm{n}}$ was generally well simulated by CLM4.5 and VR-CESM1.3; however, modeled elevated albedo (indicative of snow cover) persisted longer in winter and spring leading to large biases in $R_{\textrm{n}}$ and $α$. The introduction of a VR grid does not adversely impact surface energy fluxes compared to 1° uniform grids and highlights the usefulness of this approach for future efforts to predict land–atmosphere fluxes across heterogeneous landscapes.