|Working Group I: The Scientific Basis|
|Other reports in this collection|
Precipitation drives the continental hydrology, and influences the salinity of the ocean. The distribution of the precipitation in space and time, is therefore of key importance. The amount of precipitation also constitutes a measure of the latent heat release within the atmosphere. The long-term global mean precipitation of 984 mm/yr implies a vertically integrated mean heating rate of 78 Wm-2 (Arkin and Xie, 1994; Kiehl and Trenberth, 1997). The local precipitation rate on horizontal spatial scales of 1 to 10 km are 10 to 1,000 times the global number so that the associated heating rates often dominate all other effects and strongly influence local and global circulations. Precipitation also acts to remove and transport aerosols and soluble gases both within and below clouds, and thus strongly affects the chemical composition and aerosol distribution. The suspended and precipitating condensates provide sites for aqueous and surface chemical reactions.
The highly variable rain rates and enormous spatial variability makes determination of mean precipitation difficult, let alone how it will change as the climate changes. For instance, a detailed examination of spatial structure of daily precipitation amounts by Osborn and Hulme (1997) shows that in Europe the average separation distance between climate stations where the correlation falls to 0.5 is about 150 km in summer and 200 km in winter. This complexity makes it difficult to model precipitation reliably, as many of the processes of importance cannot be resolved by the model grid (typically 200 km) and so sub-grid scale processes have to be parametrized.
Precipitation is usually considered to be of stratiform or convective nature, or a mixture of the two. In stratiform precipitation the vertical velocity of air, usually forced by developing low pressure systems, monsoonal circulations, or underlying orography, is comparable to or smaller than the fall speed of snow and ice crystals. Stratiform precipitation dominates in the extra-tropics except over continents in summer, and there is substantial spatial coherency on scales up to and beyond about 100 km. In contrast, convective precipitation systems are associated with vigorous latent-heat-driven vertical circulations on horizontal scales of a few kilometres. Convection is responsible for most of the precipitation in the tropics and middle latitude continents in summer. In many cases, convective and stratiform precipitation interact or occur together, for instance as convective cells are embedded within areas of stratiform precipitation.
There is increasing evidence that the use of one or more prognostic cloud water variables for the modelling of stratiform precipitation is fairly successful in simulating continental and sub-continental scale precipitation distributions, including orographic precipitation, provided the synoptic-scale circulation is properly accounted for. This is the case for numerical weather predictions in the short-term range of 1 to 2 days (Petroliagis et al., 1996), and regional climate models driven by observed lateral boundary conditions (e.g., Jones et al., 1995; Lüthi et al., 1996; Christensen et al., 1998; Giorgi et al., 1998). Cloud schemes that include an explicit cloud water variable and some parametrization of the ice phase (either by carrying an explicit cloud ice variable or by including a temperature-dependent formulation of microphysical conversion rates) appear able to credibly reproduce some of the major features of the observed intensity-frequency relations (Frei et al., 1998; Murphy, 1999).
In contrast, the simulation of convective precipitation, which is fully parametrized, is of substantially poorer quality over continental regions. In a recent intercomparison study that attempted to simulate dry and wet summers over the continental US, large inter-model and model-observation differences were found (Takle et al., 1999), although the larger-scale atmospheric circulation was prescribed. One specific difficulty is the strong diurnal cycle of convective precipitation over land, which is accompanied by the build-up of a well-mixed boundary layer in response to solar heating prior to the onset of convection. Recent studies (Yang and Slingo, 1998; Dai et al., 1999) find that moist convection schemes tend to initiate convection prematurely as compared with the real world, and instability does not build up adequately. Premature cloud formation prevents the correct solar heating from occurring, impacting the development of the well-mixed boundary layer and continental-scale convergence at the surface, which in turn affects the triggering of convection. Scale interactions between convection that is organised by somewhat larger scales also seem to underlie the difficulties all GCMs have in simulating the Madden-Julian Oscillation of intra-seasonal variations in the deep tropics (Slingo et al., 1996).
With increasing temperature, the surface energy budget tends to become increasingly dominated by evaporation, owing to the increase in the water holding capacity of the boundary layer. The increase of evaporation is not strictly inevitable (Pierrehumbert, 1999), but it occurs in all general circulation models, though with varying strength. Simulated evapotranspiration and net atmospheric moisture content is also found to increase (Del Genio et al., 1991; Trenberth, 1998), as is observed to be happening in many places (Hense et al., 1988; Gaffen et al., 1992; Ross and Elliot, 1996; Zhai and Eskridge, 1997). Globally there must be an increase in precipitation to balance the enhanced evaporation but the processes by which precipitation is altered locally are not well understood. Over land, enhanced evaporation can occur only to the extent that there is sufficient soil moisture in the unperturbed state. Naturally occurring droughts are likely to be exacerbated by enhanced potential evapotranspiration, which quickly robs soil of its moisture.
Because moisture convergence is likely to be proportionately enhanced as the moisture content increases, it should lead to similarly enhanced precipitation rates. Moreover, the latent heat released feeds back on the intensity of the storms. These factors suggest that, while global precipitation exhibits a small increase with modest surface warming, it becomes increasingly concentrated in intense events, as is observed to be happening in many parts of the world (Karl et al., 1995), including the USA (Karl and Knight, 1998), Japan (Iwashima and Yamamoto, 1993) and Australia (Suppiah and Hennessy, 1998), thus increasing risk of flooding. However, the overall changes in precipitation must equal evaporation changes, and this is smaller percentage-wise than the typical change in moisture content in most model simulations (e.g., Mitchell et al., 1987; Roads et al., 1996). Thus there are implications for the frequency of storms or other factors (duration, efficiency, etc.) that must come into play to restrict the total precipitation. One possibility is that individual storms could be more intense from the latent heat enhancement, but are fewer and farther between (Trenberth, 1998, 1999).
These aspects have been explored only to a limited extent in climate models. No studies deal with true intensity of rainfall, which requires hourly (or higher resolution) data, and the analysis is typically of daily rainfall amounts. Increases in rain intensity and dry periods are simulated along with a general decrease in the probability of moderate precipitation events (Whetton et al., 1993; Cubasch et al., 1995; Gregory and Mitchell, 1995; Mearns et al., 1995; Jones et al., 1997; Zwiers and Kharin, 1998; McGuffie et al., 1999). For a given precipitation intensity of 20 to 40 mm/day, the return periods become shorter by a factor of 2 to 5 (Hennessy et al., 1997). This effect increases with the strength of the event (Fowler and Hennessy, 1995; Frei et al., 1998). However, estimates of precipitation and surface long-wave radiation suggest that the sensitivity of the hydrological cycle in climate models to changes in SST may be systematically too weak (Soden, 2000). Accordingly, it is important that much more attention should be devoted to precipitation rates and frequency, and the physical processes which govern these quantities.
Other reports in this collection