A Revised Land Surface Parameterization (SiB2) for Atmospheric GCMS. Part II: The Generation of Global Fields of Terrestrial Biophysical Parameters from Satellite Data

The global parameter fields used in the revised Simple Biosphere Model (SiB2) of Sellers et al. are reviewed. The most important innovation over the earlier SiB1 parameter set of Dorman and Sellers is the use of satellite data to specify the time-varying phonological properties of FPAR, leaf area index. and canopy greenness fraction. This was done by processing a monthly 1° by 1° normalized difference vegetation index (NDVI) dataset obtained farm Advanced Very High Resolution Radiometer red and near-infrared data. Corrections were applied to the source NDVI dataset to account for (i) obvious anomalies in the data time series, (ii) the effect of variations in solar zenith angle, (iii) data dropouts in cold regions where a temperature threshold procedure designed to screen for clouds also eliminated cold land surface points, and (iv) persistent cloud cover in the Tropics. An outline of the procedures for calculating the land surface parameters from the corrected NDVI dataset is given, and a brief description is provided of source material, mainly derived from in situ observations, that was used in addition to the NDVI data. The datasets summarized in this paper should he superior to prescriptions currently used in most land surface parameterizations in that the spatial and temporal dynamics of key land surface parameters, in particular those related to vegetation, are obtained directly from a consistent set of global-scale observations instead of being inferred from a variety of survey-based land-cover classifications.