Mean growing season soil PCO, data were obtained for 19 regions of the woi.1 in nine countries. Bivariate anL multiple linear regression analysis with soil log(PC0,) as the dependent variable and TEMP, PRECIP, log (AET), and log(PET) as the four climatic independent variables demonstrated that AET was the best independent predictor of soil PCO, .An improved soil PC0,-AET model was developed by assuming (1) that as AET approaches zero, soil PCO, approaches the atmospheric value and (2) that there is an upper limit to soil PCO, at very high AET. This model has the form log(PC0,) = -3.47 + 2.09 (1 -e-0'00172 AET) where AET is in mm. It explains 67 per cent of the initial variation in the soil PCO, data, predicts a soil log(PC0,) of -3.47 at AET = 0, and an upper limit of 3.5 per cent (log(PC0,) = -1.45) for mean growing season soil PCO, at AET values of 2000mm and above. The results of this study suggest that soil PCO, levels in tropical areas are, on average, higher than those in temperate, alpine, and arctic regions.