Water has always been and will continue to be crucial to the development of deserts. But opinions about water development have changed. Fifty years ago there was immense faith in engineers and in state investment to support them (in the western, socialist, oil-rich, and Third World countries alike). But this faith closed planners' ears to warnings about the long-term impacts and real costs of water projects (Reisner 1986). Today, this legacy considerably constrains opportunities in three ways. First, the need to clean up the consequences, as discussed; second, the need to extricate policy from the past and to construct better policy for the future; and third, the depleted and degraded state of whatever water remains.

Building more dams and drilling for more groundwater still tempt the policy-maker, and in many cases the temptation is undeniable. The ongoing debate in Pakistan about a proposal for a dam at Kalabagh on the Indus system illustrates the problems of this course. The water in the Indus and its tributaries is already thoroughly utilized; demand for water is rising with population and modest increases in prosperity; thus the distribution of water is increasingly controversial. The huge dams that were built with foreign aid after the signing of the Indus Waters Treaty in 1960, as at Tarbela (on the Indus itself) and Mangla (on the Jhelum), are filling with sediment (Tate and Farquharson 2000). The first dams took the best sites, leaving sites that are less than optimal. Large, impounded reservoirs lose large amounts of water by evaporation. Dam-building requires huge investment, with questionable long-term returns, and no guarantee that the water they save will not lead, as the water from earlier dams did, to salinization. Climate change will probably decrease the flow of the snow- and ice-fed rivers that are Pakistan's main sources of water.

A high proportion of desert river water is already used. The Colorado River in the USA, the Nile and the Mesopotamian rivers are now nearly completely utilized. Siltation of reservoirs happens worldwide; in the next few decades, sediment may clog the outlets of the Glen Canyon dam (inaugurated only in 1963) on the Colorado River in the U.S.; some NGOs are already calling for its decommissioning. Almost all the sediment in the Colorado and the Nile is retained; the figure is 60-80 per cent in the Tigris-Euphrates system (Vörösmarty and others 2003). The time between the completion of large dams and the point where siltation threatens to close them is running at about 30 to 40 years. If so, many more will soon be threatened.

Returns on investment from water schemes are debatable. In the Central Valley of California, state subsidies on water, in one case, and at one time, amounted to about $217 per acre per year, for land that yielded crops with a value of only $290. The crop was cotton, a highly water-consumptive crop, in national surplus and grown more cheaply in other parts of the USA, let alone elsewhere in the world (Reisner 1986). Of course, dams yield more than water, particularly hydroelectric power and recreation, and these must always feature in cost-benefit analyses, but all of these benefits are threatened by siltation, an almost inevitable consequence of dam-building. There are many other examples in which the underpricing of water, or the non-collection of fees for water (for whatever reason) has encouraged profligate use, which in turn may accelerate salinization and waterlogging (Ray and Williams 1999). Many of these stories would have applied to Australian water management in the past, but there has now been radical reform aimed at the more effective and less environmentally-damaging use of water (Turral and others 2005).

Groundwater, with some exceptions (Box 5.2), also has problems. Saudi Arabia's trajectory on groundwater illustrates some of these. The city of Riyadh is close to a large supply of groundwater, yet the government chose to sell this water cheaply for agriculture (to promote food security), and to supply the people of Riyadh with desalinated water pumped up from the ROPME Sea Area, at much greater cost. As in the history of desert water development in the USA, the cheapness of the water allowed Saudi farmers to grow highly waterconsumptive crops and even livestock (Figure 5.6; Allan 2005).

Water supply can be improved only by combining technology and management. Some technologies (Box 5.3) may make a great impact locally (as for new hotels or isolated settlements), and could play a greater part in this role. Some are only feasible at large expense and high consumption of energy. A small number, like the desalinisation of brackish water, are both cheap and widely applicable. Better policy will depend on learning the lessons of the many twentieth-century water policies that squandered rather than conserved water. Given the escalating water crisis in many deserts, caused not least by the old policies, better policy is urgent.