Monday, 10 February 2003
Validation and Diagnosis of AMIP II Land-surface Simulations
AMIP Diagnostic Subproject 12 on Land-surface Processes and Parameterizations is validating and diagnosing continental simulations from some 20 current-generation global atmospheric models. Because of the present dearth of quality global-scale terrestrial observations, we are validating each simulated process against available sets of large-scale estimates provided by merged in-situ/satellite products, climate reanalyses, and off-line simulations of land-surface schemes that are driven by observed forcings. Aggregated spatio-temporal differences between simulated processes and chosen reference data are quantified by root-mean-square error statistics. Differences among the alternative validation data sets are similarly computed, as both an estimate of current observational uncertainty and a benchmark for assessing collective model performance. Of the continental processes considered, the simulations of two-meter air temperature collectively agree most closely with the reference data, while simulations of precipitation agree least closely. The error statistics of all continental processes are sensitive as well to season and latitude sector.
The AMIP II models also conserve continental energy and moisture to varying degrees: Violations of energy conservation tend to be more acute in high-latitude regions,
while moisture nonconservation seems to be related to inadequate initialization of soil moisture in certain models. In addition, the models show considerable variation
in their partitioning of continental energy (into latent vs. sensible heat) and moisture(into evaporation vs. runoff); but, for the most part, these inter-model differences
appear to fall within the current observational uncertainties.