A change that appears most likely is that global average precipitation will increase as global temperatures rise. Evaporation potential will increase with warming because a warmer atmosphere can hold more moisture. This capacity is governed by the exponential Clausius-Claperyon equation, which states that for a one-degree Celsius increase in air temperature, the water-holding capacity of the atmosphere is increased by about seven percent.
A simple-minded explanation for the resulting intensification of the hydrologic cycle is that “what goes up, must come down.” Of course, it really is not that simple, but the overall scientific consensus is that globally the Earth will be warmer with higher globally averaged precipitation. Exactly how much global average precipitation will increase is less certain. On average, current climate models suggest an increase of about 1–2 percent per degree Celsius due to warming forced by CO2 (Allen and Ingram 2002). An increase in global average precipitation does not mean that it will get wetter everywhere and in all seasons. In fact, all climate model simulations show complex patterns of precipitation change, with some regions receiving less and others more precipitation than they do now. The local balance between changes in precipitation and changes in actual evaporation will determine the net change in river flows and groundwater recharge.
In general, the models agree in projecting precipitation increases over high-latitude land areas, much smaller and less certain increases over the equatorial regions, and decreases over some subtropical areas. Elsewhere, precipitation changes are more variable across models (Carter et al. 1999; IPCC WGI 2001). Wigley (2004) has developed a statistical summary of the spatial distribution of precipitation change seen in scenarios generated by various climate models. Figure 1 displays these results in the form of normalized signal-to-noise ratios, where noise represents scatter among model projections. In other words, at the red end of the spectrum, the models tend to agree on increased precipitation. At the opposite end, where the map is shaded in blue tones, they tend to agree on reduced precipitation. However, in the middle of the color spectrum (corresponding to the green and yellow-green tones), the various projections give differing results regarding whether annual precipitation will increase or decrease. This suggests that mid-latitude areas such as the continental and much of Europe and face an especially uncertain future regarding changes in average annual precipitation.
Figure 1. Inter-model signal-to-noise ratios for annual-mean precipitation (mean precipitation change per 1°C global-mean warming, averaged over 17 AOGCMs, divided by the inter-model standard deviation). This is a measure of both the sign and strength of the expected precipitation change and the level of agreement between models. Values between –1 and +1 indicate considerable uncertainty in the expected change. Source: The Benefits of Climate Change Policies: analytical and Framework Issues, p. 243, OECD 2004
The major difficulty is that although different model simulations are fairly consistent in regional temperature changes, they often display very different regional precipitation patterns. To understand why this occurs and what it implies for the usefulness of climate model projections, it is helpful to begin with an explanation of what climate models are and how they are used to simulate present and future climates.
Coupled Atmosphere-Ocean General Circulation Models (AOGCMs) are currently the primary tool used to analyze the potential impacts of increased greenhouse gases, aerosols and other factors on global climate. To be useful for the analysis of climate change, the atmospheric model must be coupled to models of other components of the climate system, such as the oceans, the sea ice, and the land surface. The major climate models include tens of vertical layers in the atmosphere and the oceans, dynamic sea-ice sub-models, and effects of changes in vegetation and other land surface characteristics (Gates et al. 1996; Washington 1996). The atmospheric part of a climate model is a mathematical representation of the behavior of the atmosphere based upon the fundamental, non-linear equations of classical physics. A three-dimensional horizontal and vertical grid structure is used to track the movement of air parcels and the exchange of energy and moisture between parcels.
Despite tremendous technological advances in computing capability, it is still very time-consuming and costly to use these models to simulate future climates. One of the most important choices for achieving model results in a reasonable amount of time is to increase the model’s horizontal resolution. This limitation means that it is prohibitively costly to run full coupled-climate models at a spatial resolution that would accurately depict the effects of mountains and other complex surface features on regional climates.
The problem with such a coarse horizontal resolution is that important processes that occur at finer scales are not well resolved. Topography, for example, is very important in determining the location of precipitation. As moist air rises over mountains or hills, the moisture condenses, producing clouds and if conditions are right, precipitation occurs. Although there has been marked improvement over the last three decades, the coarse horizontal resolution of typical climate models still tends to smooth out important landscape features that affect atmospheric processes. At the resolution of most AOGCMs, the models see the mountains of the western United States as a large set of ridges and do not resolve finer-scale features that influence regional climate. Clearly, that spatial resolution is too coarse to reproduce the effects of topography on the region’s precipitation and runoff patterns (Grotch and MacCracken 1991; Giorgi and Mearns 1991; Pan et al. 2004). For example, coarse-resolution models would see the Great Basin area as being located on an upslope. They would therefore predict it to be wet, when it is actually a desert. The global-scale models cannot adequately capture the actual rain-shadowing effect of the Sierra Nevada Mountain Range. In short, raw AOGCM output will put the precipitation in the wrong places and perhaps at the wrong time.
Recognition of limits imposed by the relatively coarse horizontal scales of AOGCMs has led to the application of “downscaling” as a means of trying to understand how local-scale processes, of greater interest to water utilities, might respond to larger-scale weather and climate changes (Wilby, et al., 2004). Downscaling includes statistical methods and the use of regional climate models run at a relatively high resolution over a limited area with boundary conditions (and sometimes interior domain information as well) prescribed from the lower resolution AOGCM. Like global climate models, regional climate models will vary in their precipitation projections depending on the downscaling method, the model specifications, and the AOGCM scenario that is downscaled. While it is possible for a downscaled model to resolve some limitations of general circulation models for a specific region, they are still limited in their capabilities to give reliable projections for future precipitation. Downscaling can produce more sub-regional detail but not necessarily more information.