Estimating rainfall changes by means of GCMs

E. Linacre and B. Geerts


Cloudiness and precipitation are notoriously difficult to predict, not only by numerical weather prediction models but also by general circulation models (GCMs) used to assess climate change. Twenty-nine GCMs have been compared as regards their ability to estimate rainfalls resulting from given various observed scenarios of sea-surface temperatures and sea-ice boundaries (1). Using the average annual cycle of ocean conditions, the GCMs yield global average rainfalls of near the measured 2.3 mm/day, but a quarter of the models showed significant errors in calculating the global water balance. There was reasonable model agreement with the observed frequencies of heavy precipitation (due to deep convection in the atmosphere), but consistent underestimation of (mostly non-convective) rainfalls less than 1 mm/d, especially for areas of low rainfall, such as deserts and to the west of continents. Most models suggest that a warmer climate implies a wetter climate, on average, however there are large regional differences. The higher precipitation totals are due to in an increase in rainfall intensity (when it rains), rather than to an increase in rainfall duration.

The models are successful in reproducing the large-scale effects of ENSO events, such as the eastward migration of the area of the highest equatorial rainfalls in the Pacific. However, there is still too much difference between the outcomes of various models for reliance to be put on them for estimates of regional rainfalls, especially under different SST scenarios which may occur as a consequence of global warming.



(1) Lau, K.-M., J.H. Kim and Y. Sud 1996. Intercomparison of hydrologic processes in AMIP GCM’s. Bull. Amer. Meteor. Soc. 77, 2209-27.