Land-surface characterization in climate models: biome-based parameter inference is not equivalent to local direct estimation

Soil-Vegetation-Atmosphere Transfer (SVAT) schemes in atmospheric general c irculation models (AGCMs) require land-surface information. SVATs need this information to simulate the interactions between the atmosphere and the bi osphere, in general, and to determine how the radiation absorbed by the sur face is partitioned into sensible and latent heat, in particular. When inve stigating future climates, parameters can either be inferred from the tabul ated, average characteristics of predicted biomes or computed locally, i.e. cell by cell. A null hypothesis is formulated to test if AGCMs can disting uish between the two approaches: ''For AGCMs, biome-based parameter inferen ce is equivalent to cell by cell estimation''. The hypothesis is tested wit h a terrestrial biosphere model (TBM), the Ecological ModUle (EMU). EMU use s a set of primary and composite generic vegetation types to calculate loca l vegetation characteristics and to identify dominant assemblages. For the test, EMU is run with two different climate scenarios: a present climate an d the Climate Change scenario for an equivalent doubling of the atmospheric concentration of carbon dioxide generated with the United Kingdom Meteorol ogical Office AGCM. Results indicate that the two approaches are not equiva lent. Consequently, the null hypothesis is rejected. Moreover, the differen ces in Vegetation characteristics are of a magnitude which should have noti ceable impacts on exchanges of energy, water, and momentum between the atmo sphere and the surface. This suggests that estimating vegetation locally sh ould increase the credibility of AGCM investigations of Global Change. (C) 1998 Elsevier Science B.V. All rights reserved.