Some of the basic physical processes that relate global warming to changes in water availability are accessible to the layperson: warm air can hold more water vapor than cool air, and ice melts when heated above a critical temperature. The higher water content of a warmer atmosphere implies changes in atmospheric transport of water to and from a given region, with direct consequences for surface runoff and groundwater recharge. The decreasing prevalence of ice and snow in a warmer world has fundamental implications for seasonal storage of water and for response of soil to precipitation.

The foregoing considerations lead to the admittedly vague prediction that a warmer world is a world in which water availability differs from that of a cooler world. Approximate quantitative expressions of this prediction have been produced by a series of increasingly complex global climate models over a period of many years. However, the water-availability predictions differ from one model to the next, and they depend on uncertain projections of human activities. Furthermore, global climate models address mainly physical processes and tend to ignore various biological and chemical feedbacks of potential importance (e.g., changes in plant structure and functioning, changes in soil carbon balance). Accordingly, skepticism toward projections of changing water availability is not inappropriate.

The projections of the most recent generation of climate models differ even in the direction of change of mean annual runoff (i.e., tendency toward wetter vs. drier conditions) in many regions. Nevertheless, the global pattern of projected change is broadly consistent across most models. Furthermore, the global pattern of multi-decadal trends in streamflow that was observed during the 20th Century bears a striking similarity to the average pattern simulated by the climate models forced by estimated historical drivers of climate (both anthropogenic and natural). This similarity appears too great to be explained readily by chance, but seems rather to indicate that the models have predictive skill for estimating future water-availability trends. The more robust projections of these same models driven only by possible anthropogenic forcings for the 21st Century imply decreasing water availability in southwestern North America, southern Europe, the Middle East, and southern Africa, and increasing water availability in high-latitude North America and Eurasia, the La Plata basin of South America, eastern equatorial Africa, and Indonesia. In general, regions of projected decreasing water supply tend to be regions of contemporary water-supply stress, and regions of projected increasing supply tend to be regions where water shortage is not a major issue for water managers.