Working Group I: The Scientific Basis

Other reports in this collection Climate variability and extreme events

Figure 10.9: Winter surface air temperature change (°C) over Europe at the time of CO2 doubling in (a) a transient climate change experiment with the AOGCM HadCM2 and (b) the stretched grid AGCM ARPEGE driven by SSTs and sea-ice from the HadCM2 integration. From Dèquè et al. (1998).

Enhanced resolution improves many aspects of the AGCMs’ intra-seasonal variability of circulation at low and intermediate frequencies (Stendel and Roeckner, 1998). However, in some cases values underestimated at standard resolution are overestimated at enhanced resolution (Dèquè and Piedelievre, 1995; Stratton, 1999a,b). Martin (1999) found little sensitivity to resolution in either the interannual or intra-seasonal variability of circulation and precipitation of the South Asian monsoon in HadAM3a. Extreme events have not been studied, with the exception of tropical cyclones. This subject cuts across various sections and chapters and thus is dealt with in Box 10.2.

10.4.2 Simulations of Climate Change Mean climate

Climate change simulations using ECHAM3 at T42 and T106 resolutions predicted substantially different responses for southern Europe (Cubash et al., 1996). For example, surface temperature response of less than +2ºC in summer at T42 increased by over 4ºC for much of the region at T106 and winter precipitation increased more at T106 than at T42. An important factor in generating the different responses was the substantial difference in the control simulations. Wild et al. (1997) showed a large positive summer surface temperature bias in the T106 control derived from a positive feedback between excessive surface insolation and summer dryness. This mechanism provided a large increase in the insolation, and thus the surface temperature, in the anomaly experiment. As this process was handled poorly in the control simulation, little confidence can be placed in the warming amplification simulated at T106.

A variable grid AGCM climate change experiment using the ARPEGE model and sea surface forcing from HadCM2 predicted moderate warmings over Europe, 1.5ºC (northern) to 2.5ºC (southern) in winter and 1ºC to 3.5ºC in summer (Dèquè et al., 1998). In contrast, HadCM2 predicted greater warming and a larger north-south gradient in winter (Figure 10.9). These differences result mainly from the ARPEGE large-scale flow being too zonal and too strong over mainland Europe, which enhances the moderating influence of the SSTs. The precipitation responses are more similar, especially in summer, when both models predict a decrease over most of Europe, maximum –30% in the south. Differences in the control simulations suggest that little confidence should be placed in this result.

In a similar experiment, HadAM3a at 1.25°30.83° resolution used observed sea surface forcing and anomalies from a HadCM3 GHG simulation and produced a response at the largest scales in the annual mean similar to the AOGCM (Johns, 1999). However, regionally or seasonally, many differences were evident in the two models, notably in land sea contrasts, monsoon precipitation and some circulation features. Over Europe, large-scale responses in surface temperature and precipitation were similar except for a larger winter surface warming in northern Europe in HadCM3. This was due to a greater melting back of Arctic sea ice which was too extensive in the HadCM3 control (Jones, 1999). In a 30-year ECHAM4 T106 experiment driven by ECHAM4/OPYC simulations for 1970 to 1999 and 2060 to 2089, the simulations of future climate were more similar to each other than those for the present day (May, 1999). This implies that the differences in the control simulations would determine a proportion of the difference in the responses. In these cases better control simulations at high resolution increase the confidence in their responses. Climate variability and extreme events

Due to the limited number and length of simulations and a lack of comprehensive analyses, this subject has been almost completely ignored. The only response in variability or extremes that has received any attention is that of tropical cyclones (Box 10.2).

Box 10.2: Tropical cyclones in current and future climates

Simulating a climatology of tropical cyclones
Tropical cyclones can have devastating human and economic impacts (e.g., Pielke and Landsea, 1998) and therefore accurate estimates of future changes in their frequency, intensity and location would be of great value. However, because of their relatively small extent (in global modelling terms) and intense nature, detailed simulation of tropical cyclones for this purpose is difficult. Atmospheric GCMs can simulate tropical cyclone-like disturbances which increase in realism at higher resolution though the intense central core is not resolved (e.g., Bengtsson et al., 1995; McDonald, 1999). Further increases of resolution, by the use of RCMs, provide greater realism (e.g., Walsh and Watterson, 1997) with a very high resolution regional hurricane prediction model giving a reasonable simulation of the magnitude and location of maximum surface wind intensities for the north-west Pacific basin (Knutson et al., 1998). GCMs generally provide realistic simulation of the location and frequency of tropical cyclones (e.g., Tsutsui and Kasahara, 1996; Yoshimura et al., 1999). See also Chapter 8 for more details on tropical cyclones in GCMs.

Tropical cyclones in a warmer climate
Much effort has gone into obtaining and analysing good statistics on tropical cyclones in the recent past. The main conclusion is that there is large decadal variability in the frequency and no significant trend during the last century. One study looking at the century time-scale has shown an increase in the frequency of North Atlantic cyclones from 1851 to 1890 and 1951 to 1990 (Fernandez-Partagas and Diaz, 1996). See Chapter 2 for more details on observed tropical cyclones.

Most assessments of changes in tropical cyclone behaviour in a future climate have been derived from GCM or RCM studies of the climate response to anthropogenically-derived atmospheric forcings (e.g., Bengtsson et al., 1996, 1997; Walsh and Katzfey, 2000). Recently, more focused approaches have been used: nesting a hurricane prediction model in a GCM climate change simulation (Knutson et al., 1998); inserting idealised tropical cyclones into an RCM climate change simulation (Walsh and Ryan, 2000).

In an early use of a high-resolution AGCM, a T106 ECHAM3 experiment simulated a decrease in tropical cyclones in the Northern Hemisphere and a reduction of 50% in the Southern Hemisphere (Bengtsson et al., 1996, 1997). However, the different hemispheric responses raised questions about the model’s ability to properly represent tropical cyclones and methodological concerns about the experimental design were raised (Landsea, 1997). In a similar experiment, the JMA model also simulated fewer tropical cyclone-like vortices in both hemispheres (Yoshimura et al., 1999). Other GCM studies have shown consistent basin-dependent changes in tropical cyclone formation under 2xCO2 conditions (Royer et al., 1998; Tsutsui et al., 1999). Frequencies increased in the north-west Pacific, decreased in the North Atlantic, and changed little in the south-west Pacific. A high resolution HadAM3a simulation reproduced the latter changes, giving changes in timing in the north-west Pacific and increases in frequency in the north-east Pacific and the north Indian basin (McDonald, 1999). Some GCM studies show increases in tropical storm intensity in a warmer climate (Krishnamurti et al., 1998) though these results can probably not be extrapolated to tropical cyclones as the horizontal resolution of these models is insufficient to resolve the cyclone eye. The likely mean response of tropical Pacific sea surface warming having an El Niño-like structure suggests that the pattern of tropical cyclone frequency may become more like that observed in El Niño years (see Chapter 9).

An indication of the likely changes in maximum intensity of cyclones will be better provided by models able to simulate realistic tropical cyclone intensities. A sample of GCM-generated tropical cyclone cases nested in a hurricane prediction model gave increases in maximum intensity (of wind speed) of 5 to 11% in strong cyclones over the north-west Pacific for a 2.2°C SST warming (Knutson and Tuleya, 1999). The RCM study of idealised tropical cyclones (in the South Pacific) showed a small, but not statistically significant, increase in maximum intensity (Walsh and Ryan, 2000). These results are supported by the theory of the maximum potential intensity (MPI) of hurricanes (Emanuel, 1987). A calculation using the MPI framework of Holland (1997) suggested increases of 10 to 20% for a 2xCO2 climate (Henderson-Sellers et al., 1998). This study also acknowledges physical omissions that would reduce this estimate though Emanuel (2000) suggests there is a linear relationship betweeen MPI and the wind speed of real events. Published modelling studies to date neglect the possible feedback of sea surface cooling induced by the cyclone. However, a recently submitted study using a hurricane model with ocean coupling indicates that the increased maximum intensity by CO2 warming would still occur even when the sea surface cooling feedback is included (Knutson et al., 2000).

The extreme precipitation associated with tropical cyclones can also be very damaging. The very high resolution studies discussed above suggest that increases in the intensity of tropical cyclones will be accompanied by increases in mean and maximum precipitation rates. In the cases studied, precipitation in the vicinity of the storm centre increased by 20% whereas peak rates increased by 30%. Part of these increases may be due to the increased moisture-holding capacity of a warmer atmosphere but nevertheless point to substantially increasing destructive capacity of tropical cyclones in a warmer climate.

In conclusion, there is some evidence that regional frequencies of tropical cyclones may change but none that their locations will change. There is also evidence that the peak intensity may increase by 5% to 10% and precipitation rates may increase by 20% to 30%. There is a need for much more work in this area to provide more robust results.

10.4.3 Summary and Recommendations

Since the SAR, several variable and high-resolution GCMs have been used to provide high-resolution simulations of climate change. Clearly the technique is still in its infancy with only a few modelling studies carried out and for only a limited number of regions. Also, there is little in-depth analysis of the performance of the models and only preliminary conclusions can be drawn.

Many aspects of the models’ dynamics and large-scale flow are improved at higher resolution, though this is not uniformly so geographically or across models. Some models also demonstrate improvements in their surface climatologies at higher resolution. However, substantial underlying errors are often still present in high-resolution versions of current AGCMs. In addition, the direct use of high-resolution versions of current AGCMs, without some allowance of the dependence of models physical parametrizations on resolution, leads to some deterioration in the performance of the models.

Regional responses currently appear more sensitive to the AGCM than the SST forcing used. This result is partially due to some of the model responses being dependent on their control simulations and systematic errors within them. These factors and the small number of studies carried out imply that little confidence can be attached to any of the regional projections provided by high and variable resolution AGCM simulations. The improvements seen with this technique are encouraging, but more effort should be put in analysing, and possibly improving the performance of current models at high resolution. This is particularly important in view of the fact that future AOGCMs will likely use models approaching the resolution considered here in the next 5 to 10 years.

Other reports in this collection

IPCC Homepage