Estimating Spatial Variability in Atmospheric Properties over Remotely Sensed Land Surface Conditions

Abstract

Abstract This paper investigates the spatial relationships between surface fluxes and near-surface atmospheric properties (AP), and the potential errors in flux estimation due to homogeneous atmospheric inputs over heterogeneous landscapes. A large-eddy simulation (LES) model is coupled to a surface energy balance scheme with remotely sensed surface temperature Ts as a key boundary condition. Simulations were performed for different agricultural regions having major contrasts in Ts, canopy cover, and surface roughness z0 between vegetated/irrigated and bare soil areas. If AP from a single weather station in a nonrepresentative location within the landscape are applied uniformly over the domain, significant differences in surface flux estimation with respect to the LES output are observed. The spatial correlations of AP with the fluxes, the land cover properties, and surface states were examined and the spatial scaling of these fields is analyzed using a two-dimensional wavelet technique. The results indicate a significant local correlation of the spatial distributions of the air temperature Ta with the sensible heat flux H, the specific humidity q with the latent heat flux LE, and the wind speed U with z0. These relationships can be described by a general linear form, suggesting that a simple regression relation may be applicable for most agricultural landscapes to estimate spatially variable AP fields. A simple yet practical method is proposed using remotely sensed observations and the land surface scheme, based on general linear expressions derived between Ta and H, q and LE, and U and z0. The method is shown to reproduce the main spatial patterns of AP and to reduce potential errors in local and regionally averaged heat flux estimation. This approach is recommended when only local weather station observations are available.