The Regional Impacts of Climate Change

Other reports in this collection

B.5. Conclusions

Analysis of surface air temperature and precipitation results from regional climate change experiments carried out with AOGCMs indicates that the biases in present-day simulations of regional climate change and the inter-model variability in the simulated regional changes are still too large to yield a high level of confidence in simulated change scenarios. The limited number of experiments available with statistical downscaling techniques and nested regional models has shown that complex topographical features, large lake systems, and narrow land masses not resolved at the resolution of current GCMs significantly affect the simulated regional and local change scenarios, both for precipitation and (to a lesser extent) temperature (IPCC 1996, WG I). This adds a further degree of uncertainty in the use of GCM-produced scenarios for impact assessments. In addition, most climate change experiments have not accounted for human-induced landscape changes and only recently has the effect of aerosols been investigated. Both these factors can further affect projections of regional climate change.

Compared to the global-scale changes due to doubled CO2 concentration, the changes at 104-106 km2 scale derived from transient AOGCM runs are greater. Considering all models, at the 104-106 km2 scale, temperature changes due to CO2 doubling varied between 0.6 and 7C and precipitation changes varied between -35 and 50% of control run values, with a marked inter-regional variability. Thus, the inherent predictability of climate diminishes with reduction in geographical scale. The greatest model agreement in the simulated precipitation change scenarios was found over the South East Asia (about -1 to 30%), Northern Europe (about -9 to 16%), Central North America (about -7 to 5%), and East Asia (about 0.1 to 16%) regions in summer, and Southern Europe (about -2 to 29%), Northern Europe (about 5 to 25%), and East Asia (about 0.5 to 18%) in winter. For temperature, the greatest model agreement in simulated warming occurred over Australia in summer (about 1.65 to 2.5C, when excluding one outlier) and the Sahel in winter (about 1.8 to 3.15C, when excluding one outlier). Regardless of whether flux correction was used, the range of model sensitivities was less than the range of biases, which suggests that models produce regional sensitivities that are more similar to each other than their biases.

The latest regional model experiments indicate that high-resolution information, on the order of a few 10s of km or less, may be necessary to achieve high accuracy in regional and local change scenarios in areas of complex physiography. In the last few years, substantial progress has been achieved in the development of tools for enhancing GCM information. Statistical methods were extended from the monthly/seasonal to the daily time scale, and nested model experiments were extended to the multi-year time scale. Also, variable- and high-resolution global models can be used to study possible feedbacks of mesoscale forcings on general circulation.

Regional modeling techniques, however, rely critically on the GCM performance in simulating large-scale circulation patterns at the regional scale, because they are a primary input to both empirical and physically based regional models. Although the regional performance of coarse-resolution GCMs is still somewhat poor, there are indications that features such as positioning of storm track and jet stream core are better simulated as the model resolution increases. The latest nested GCM/RegCM and variable-resolution model experiments, which employed relatively high-resolution GCMs and were run for long simulation times (up to 10 years) show an improved level of accuracy. Therefore, as a new generation of higher resolution GCM simulations become available, it is expected that the quality of simulations with regional and local downscaling models will also rapidly improve. In addition, the movement towards coupling regional atmospheric models with appropriately scaled ecological, hydrological, and mesoscale ocean models will not only improve the simulation of climatic sensitivity, but also provide assessments of the joint response of the land surface, atmosphere, and/or coastal systems to altered forcings.

Other reports in this collection