Mapping the land surface for global atmosphere‐biosphere models: Toward continuous distributions of vegetation's functional properties

Global land surface characteristics are important boundary conditions for global models that describe exchanges of water, energy, and carbon dioxide between the atmosphere and biosphere. Existing data sets of global land cover are based on classification schemes that characterize each grid cell as a discrete vegetation type. Consequently, parameter fields derived from these data sets are dependent on the particular scheme and the number of vegetation types it includes. The functional controls on exchanges of water, energy, and carbon dioxide between the atmosphere and biosphere are now well enough understood that it is increasingly feasible to model these exchanges using a small number of vegetation characteristics that either are related to or closely related to the functional controls. Ideally, these characteristics would be mapped as continuous distributions to capture mixtures and gradients in vegetation within the cell size of the model. While such an approach makes it more difficult to build models from detailed observations at a small number of sites, it increases the potential for capturing functionally important variation within, as well as between, vegetation types. Globally, the vegetation characteristics that appear to be most important in controlling fluxes of water, energy, and carbon dioxide include (1) growth form (tree, shrub, herb), (2) seasonality of woody vegetation (deciduous, evergreen), (3) leaf type (broadleaf, coniferous), (4) photosynthetic pathway of nonwoody vegetation (C 3 , C 4 ), (5) longevity (annual, perennial), and (6) type and intensity of disturbance (e.g., cultivation, fire history). Many of these characteristics can be obtained through remote sensing, though some require ground‐based information. The minimum number and the identity of the required land surface characteristics almost certainly vary with the intended objective, but the philosophy of driving models with continuous distributions of a small number of land surface characteristics is likely to be applicable to a broad range of problems.