Science Briefs

Improving Global Model Precipitation Patterns by Regional Model Downscaling

Data from simulations using the GISS "ModelE" atmosphere-ocean global climate model are among those provided to the Intergovernmental Panel on Climate Change for formulating its assessments. Yet flaws in ModelE simulations of the current climate strain confidence in its projections of the future climate. For example, West African monsoon rains advance northward during the summer, irrigating the pastures and fields of the Sahel. ModelE simulations of this phase of the monsoon are somewhat unrealistic, as are its simulations of the tropical rain band over the adjacent Atlantic Ocean and precipitation maximums over locally high topography in Africa. Recent research, however, shows that more realistic simulations over selected geographical areas can be achieved by using ModelE data as input to drive a regional climate model on a higher resolution computational grid.

Map plots of JJAS mean precipitation rate

Figured 1. JJAS 5-year mean precipitation rates (mm/day): a) observed by TRMM satellite, b) simulated by ModelE, c) simulated by regional model.

Time-latitude plots of early June mean precipitation rate

Figured 2. Time vs. latitude West Africa 5-year mean precipitation rates (mm/day) for consecutive 5-day periods (beginning June 1-5), averaged over 10°W-10°E: a) observed by TRMM satellite, b) simulated by ModelE, c) simulated by regional model.

One evaluation of the benefits of such “downscaling” — from ModelE's 2° latitude by 2.5° longitude grid spacing down to the regional model's 0.44°×0.44° grid — is demonstrated here for the summer (June-September: JJAS) climate. Note that the regional model does not just interpolate ModelE results to a finer grid. Rather, the regional model often simulates a more realistic evolving climate because it uses some 30 computational elements to cover the same geographical area as a single computational element of the global model. The regional model creates a new simulation of climate patterns with better spatial details, made possible by its finer grid.

Comparison of Fig. 1b to Fig. 1a shows several ModelE "mistakes". It consistently simulates an unrealistic band of heavy precipitation over the tropical South Atlantic. (The 30-year mean is almost identical to the 5-year mean shown here.) It features a West African summer monsoon rain maximum that is too far south, and it produces a poor representation of coastal precipitation maxima over the windward slopes of the Guinean and Cameroon Highlands.

Figure 1c shows the JJAS five-year mean from the GISS regional model (RM3) simulation that uses synchronous ModelE results as input data along the edges of the regional model grid. The regional model eliminates the unrealistic rain band over the South Atlantic, improves the mountain-induced maximums along the coasts and simulates a monsoon rain maximum over West Africa that reaches the Sahel near 10°N.

Figure 2a shows how the observed rainfall maximum over West Africa shifts northward from 5°N to 10°N at the beginning of July. Figure 2b shows that the ModelE rain maximum does not shift latitude at all during the entire summer, whereas in Fig. 2c we see that the regional model corrects this deficiency.

The study suggests that ModelE projections of the future climate should be downscaled by using them to drive the GISS regional climate model, thereby creating more realistic projections of the African and tropical Atlantic Ocean climates.


Druyan, L., and M. Fulakeza, 2016: Downscaling GISS ModelE boreal summer climate over Africa. Clim. Dyn., 47, no. 11, 3499-3515, doi:10.1007/s00382-015-2880-y.


Please address all questions about this research to Dr. Leonard Druyan.

➤ Return to Science Briefs Index