The Effect of Tropical Atlantic Heating Anomalies upon GCM Rain Forecasts over the Americas

Abstract

Severe droughts occurred over eastern sections of North America and central sections of South America in 1986 and 1988. We summarize data suggesting that both periods were characterized by above-normal tropical Atlantic sea surface temperatures and convection, and investigate the response of a general circulation model to positive heating anomalies in the tropical Atlantic sector. An eight-case control ensemble of 30 day global predictions is made starting from the atmospheric state observed on 1 January of each year from 1977 through 1984. The same eight cases are integrated in a second experimental ensemble that is identical to the first control ensemble, except that a heating term is added to the thermodynamic equation in a region centered at 30°W, 6.6°N. This is intended to simulate the latent heating of enhanced tropical Atlantic convection. The third ensemble is identical to the second, except the heating is centered at 6.6°S. Both heated ensembles produce reductions of forecast precipitation over most of North and South America, but these appear to have greater statistical significance over North America. Here the greatest precipitation reductions are forecast over the southern and eastern United States, and this response does not change substantially between the two experiments. The South American response is more sensitive to the placement of the heating anomaly. When the anomaly is located north of the equator, drying occurs over northeast Brazil; meanwhile this region receives increased rainfall when the anomaly is located south of the equator. Both experiment ensembles display a region of reduced rainfall over the Andes Mountains, and over southern portions of Brazil. However, only the former region is statistically significant above the 95% confidence level. The present usage of real initial data and an ensemble of cases permits us to draw quantitatively meaningful estimates of the time scale of response and case-to-case variability. For presently tested cases, the South American response is evident by day 5, but exhibits substantial intersample variability, and the North American response is fully established by day 10, and exhibits less intersample variability. The model drying effects can be explained only partly by enhanced local subsidence; much of the rainfall reduction appears to be related to a reorientation of the synoptic scale wave pattern in which the lower tropospheric circulation is unfavorable for water vapor inflow from source regions over the tropical Atlantic and Amazon Basin.