People associate global warming with temperature, but the phrase is misleading-it fails to mention the relevance of water. Nearly every significant indicator of hydrological activity-rainfall, snowmelt, glacial melt-is changing at an accelerating pace (one can arbitrarily pick any point of the hydrological cycle and notice a disruption). One analysis pegged the increase in precipitation at 2 percent over the century. In water terms this sounds auspicious, promising increased supply, but the changing timing and composition of the precipitation more than neutralizes the advantage. For one thing, it is likely that more of the precipitation will fall in intense episodes, with flooding a reasonable prospect. In addition, while rainfall will increase, snowfall will decrease. Such an outcome means that in watersheds that depend on snowmelt, like the Indus, Ganges, Colorado river basins, less water will be stored as snow, and more of it will flow in the winter, when it plays no agricultural role; conversely, less of it will flow in the summer, when it is most needed. One computer model showed that on the Animas River an increase in temperature of 3.6 degrees Fahrenheit would cause runoff to rise by 85 percent from January to March, but drop by 40 percent from July to September. The rise in temperature increases the probability and intensity of spring floods and threatens dam safety, which is predicated on lower runoff projections. Dams in arid areas also may face increased sedimentation, since a 10 percent annual increase in precipitation can double the volume of sediment washed into rivers.

The consequences multiply. Soil moisture will intensify at the highest northern latitudes, where precipitation will grow far more than evaporation and plant transpiration but where agriculture is nonexistent. At the same time, precipitation will drop over northern mid-latitude continents in summer months, when ample soil moisture is an agricultural necessity. Meanwhile the sea level will continue to rise as temperatures warm, accelerating saline contamination of freshwater aquifers and river deltas. The temperature will cause increased evaporation, which in turn will lead to a greater incidence of drought.

Perhaps most disturbing of all, the hydrologic cycle is becoming increasingly unpredictable. This means that the last century's hydrological cycle-the set of assumptions about water on which modern irrigation is based-has become unreliable. Build a dam too large, and it may not generate its designed power; build it too small, and it may collapse or flood. Release too little dam runoff in the spring and risk flood, as the snowmelt cascades downstream with unexpected volume; release too much and the water will not be available for farmers when they need it. At a time when water scarcity calls out for intensified planning, planning itself may be stymied.