Impact of field-calibrated vegetation parameters on GCM climate simulations

This paper describes a study in which, for the first time, advanced systems -engineering parameter-estimation techniques were applied to data from seve ral field studies to estimate the preferred set of parameters for some of t he most com mon biomes represented in an advanced Soil-Vegetation-Atmospher e Transfer (SVAT) scheme (BATS2, a recent version of the Biosphere-Atmosphe re Transfer Scheme): the effect on modelled climate was also investigated. Observational data from field sites in Brazil, Canada, Arizona and Kansas/O klahoma in the USA, and the Netherlands were chosen as representative of tr opical rain forest, coniferous forest. semi-arid vegetation, agricultural c rops, and grassland biomes, respectively. Together, these five biomes make up 50% of the land area represented in BATS. Multi-criteria calibration alg orithms do not produce a unique set of model parameters and, when different combinations of the available objective functions at each sire are conside red. the number of solutions increases substantially. The need for a single parameter-set for each site (biome) is an important practical issue that w as necessarily addressed in this study. A procedure was defined in which op timized parameter-sets were successively discarded by successively applying a cut-off threshold to single observable objective functions following a p reference hierarchy. In this study, only the vegetation-related parameters are calibrated for each of the five biomes and implemented into BATS?; howe ver, in a separate experiment, the effect of including soil parameters in t he optimization was investigated. When the calibrated parameters are adopte d and used in BATS2, there are significant changes between the climates cal culated in an eight-year run with version 3 of the Community Climate Model and in an equivalent right-year run in which the original default parameter s were used. The overall conclusion of this exploratory study is that advan ced parameter-estimation techniques and appropriate field data can be used successfully to improve representation of surface exchanges and the modelle d climate given by a GCM, by defining appropriate values for vegetation-rel ated parameters in an advanced SVAT scheme.