How well do the current generation of General Circulation Models (GCMs) simulate present-day climate? Answering this question is an important stage in the improvement of our ability to model global climate change. The answer to the question will also determine the confidence that tiational and international bodies might place both in the projections of climate change generated by these models and in the regional climate change impact studies which use model output. This paper quantitatively evaluates the performance of a number of model precipitation climatologies, using four ‘independent’ observed climatologies. Three of the observed climatologies consist of mean monthly fields (JAEGER, LEGATES and IIASA), while the fourth is a new land-only time series of monthly precipitation from 1900 to 1992 (CRU0092). Comparisons are undertaken of global means and of the spatial pattern of the field of mean monthly precipitation. A distinction is made between global and “terrestrial” precipitation fields. CRU0092 includes interannual variability and therefore allows other validation exercises to be undertaken. The importance of the observed reference period for model validation is assessed, and observed and modelled global precipitation sensitivities are compared. These validation exercises reveal markedly different model performances and show that uncertainty about the ‘true’ observed precipitation hinder the task of model validation. Validation analyses such as these, together with more detailed validation of the interannual variability and daily precipitation characteristics of model output, must remain an essential dimension both of ‘greenhouse climate’ predictions and of climate change impact analyses which rely upon GCM experiments.
Weitere Kapitel dieses Buchs durch Wischen aufrufen
- Validation of Large-Scale Precipitation Fields in General Circulation Models
- Springer Berlin Heidelberg