Water, Climate and Uncertanity: An Introduction to the Issues, Language, Literature and the Conference Materials June 11-13 2003

Home Browse Search

set of professionals and considerations. For our purposes, it is useful to identify six stages, described below. (1) Global -Scale Projections. The foundation of most long -term climate projections are massive computer programs called General Circulation Models (GCMs) used to estimate future temperature and precipitation averages at large scales (grid boxes) collectively covering the Earth's surface. A wide variety of GCMs exist producing different projections, fueling intense debate within the scientific and political communities. One major source of differences in projections are the various assumptions about future carbon dioxide (and other greenhouse gas) emissions, which in turn is a function of assumptions about global economic development, population growth, energy policies, and the likely degree (and scheduling) of policy responses to global climate change. Most modelers have much more confidence in temperature projections than in precipitation estimates, something reflected in the much greater diversity seen in the precipitation projections? (2) Regional Downscaling. The projections of the GCMs are calculated at the scale of "grid boxes" that differ in size according to the model, but in most cases, are much larger than the scales necessary to evaluate impacts on particular water systems. A variety of models and statistical techniques are used to "downscale" the GCM output to smaller regions of concern, where local topographic and microclimate forces can significantly impact temperature and precipitation. (3) Streamflow Estimates. Once future temperature and precipitation conditions are projected for a given basin, hydrologists can translate this data into streamflow estimates. This can be particularly difficult in the and and semi -arid West, where even minor changes in precipitation can have disproportionally large impacts on runoff. Additionally, at this stage, it is a challenge to consider the competing natural processes that influence streamflow. For example, in most US basins, the majority of projections call for increased temperatures and increased precipitation, but whether or not this translates to more runoff is often determined by whether or not rising evapotranspiration (which increases with temperature) will offset additional precipitation. Despite these and many other complications, generating streamflow estimates is generally considered a more precise exercise than either of the two preceding steps. If similar assumptions about future temperature and precipitation regimes are used as input (rather than the diversity of GCM projections), resulting streamflow estimates are generally consistent. (4) Water System Simulations. Water managers typically employ simulation models to describe the movement of water into, through, and out of the developed water infrastructure of reservoirs and related facilities. These models were generally built and calibrated using historic data, but nonetheless, can often accommodate the analysis of streamflow inputs associated with different future climate scenarios. This is a very useful approach for evaluating how changes in the magnitude and timing of inflows might resonate through a water system, and is the first stage in estimating impacts to water systems. (5) Vulnerability Assessments. A closely related next step is determining whether or not changes in inflow characteristics have a discernable impact on water yields, reliability, costs, and other key management parameters given the synergistic influence of climate with other variables, 2 The challenge in projecting precipitation is evident by comparing the two GCMs used in the National Assessment studies (mentioned later). For the period 1990 to 2030, the Canadian model projects runoff in the Upper Colorado Basin to decrease by 36 percent, while the Hadley (UK) model suggests an increase of 7 percent. Even more divergent are the projections for the Lower Colorado, where the models predict a decrease of 38 percent and an increase of 23 percent, respectively. Some regions —such as California — feature much more consistent projections, but whether consistency equals accuracy is an open question.