Development and Sensitivity Analysis of High Resolution Land Surface Parameters from Satellite Data and their Use in a Mesoscale Model - LDEO

The objective of our research is to better understand the effects of anthropogenic land cover modification on regional climate and to improve mesoscale climate model predictive power by incorporating detailed information about land surface properties derived from remotely sensed observations. We accomplish this by eliminating the unnecessary loss of information and introduction of error inherent in low resolution thematic land cover classifications. We focus on urban and suburban areas because the scale and variability of land cover in these areas is not well represented in thematic classifications currently used to drive mesoscale land surface models. Estimation of physical parameters from Landsat imagery is based on the use of linear spectral mixture models to represent the land surface in terms of water, vegetation, rock, soil and impervious substrates. The spectral mixture model provides decameter (10–100 m) scale estimates of the areal abundance of these biophysical components of land cover in a form that can be validated vicariously with meter scale imagery. By extending the mixture model into the temporal dimension we are able to quantify seasonal to interannual changes in land surface properties and incorporate these changes into the Land Surface- Model (LSM). We have developed a systematic methodology for multi-scale, multi-temporal spectral mixture analysis that accounts for the effects of viewing and illumination geometry, topography, sub-pixel shadow as well as atmospheric scattering and absorption. The resulting endmember fractions are used to derive scaleable parameters for the LSM. These parameters are being tested in two very different LSMs in the Ocean Land Atmosphere Model (OLAM), an Earth System Model with mesoscale and microscale mesh refinement capability. The first LSM, LEAF3, which previously relied on thematic land cover classifications to define physical parameter values, has been modified to use input parameters directly The second LSM, Hi- SVAT, does not differentiate between the above surface components, but instead lumps them together into a single “surface” with properties of albedo, emissivity, roughness, stomatal conductance, leaf area index, and storage capacities for energy and water. The performance of both LSMs with the new high resolution parameters is evaluated by comparing between LEAF3 and Hi-SVAT, and by validating OLAM simulation results with observations.