Results from Climate Change Prediction Experiments

As we have seen in the previous section, there is still a considerable spread between different climate models in the equilibrium response to a given forcing. Similar differences can be found in transient experiments. The main reason is that the degree of climate change strongly depends on the dynamical response of the coupled system. The marked surface warming of the Northern Hemisphere during the past 20 years, for example, is strongly influenced by a positive phase of both ENSO (stronger El Nino events than normal) and NAO (stronger westerlies over the North Atlantic), both of which have contributed to milder winters over the land areas of the Northern Hemisphere (Hurrell, 1995; Wallace et al., 1995).

If, for example, both ENSO and NAO are chaotic events and hence unpredictable, this could cause long-term differences between models since they could then statistically correctly simulate these features out of phase with each other. Alternatively, it could also happen that both ENSO and NAO respond to the increased forcing of the greenhouse gases so that there is a systematic change in their probability distribution and then the positive phase we have seen in recent decades is a physically correct

TABLE 2

List of Experiments

Name

Forcing Due to Changing Atmospheric

Years

Concentrations of . . .

GHG

C02 and other well-mixed

1860-

■2100

greenhouse gases

GSD

GHG plus sulfate aerosols

1860-

■2050

(direct effect only)

GSDIO

GHG plus sulfate aerosols

1860-

■2050

(direct and indirect effect)

plus tropospheric ozone

CTL

Unforced control experiment

300

response. However, at present we cannot answer this important question. Some models indicate a successive increase in the positive phase of NAO; others like the MPI model do not show any distinct response at all. At the same time the MPI model (Timmermann et al., 1998) suggests a slow increase in the amplitude of ENSO events, which is not so clearly seen in other models.

It also follows from this general discourse that regional climate is even more strongly model-dependent, since small geographical changes in predominant weather patterns such as the stormtrack between different models may create huge differences. This is confirmed by Raisanen (1998), who has compared results from 12 transient coupled GCMs for Northern Europe and the eastern North Atlantic.

With these general reservations, we will now describe results from a recent series of transient experiments by Roeckner et al. (1999)(Table 2). The experiments start in the year 1860. Observed concentrations of greenhouse gases and sulfate aerosols were used until 1990 and thereafter changed according to the IPCC scenario IS92a. Tropospheric ozone changes have been calculated from precursor gases.

In the first experiment, GHG, the concentrations of the following greenhouse gases were prescribed as a function of time: C02, CH4, and N,0, as well as a series of industrial gases, including CFCs and HCFCs. The absorptive properties of each gas constituent were calculated separately. Furthermore, the radiative forcing was practically identical to the narrow band calculations. This meant an increase in the radiative forcing by some 10% compared to the actual broad band calculation in the radiation code of the model.

In the second experiment, GSD, Table 2 the greenhouse gases were treated as in GHG but with the additional incorporation of the tropospheric sulfur cycle as due to anthropogenic sources only. Natural biogenic and volcanic sulfur emissions were neglected, and the aerosol radiative forcing was generated through the anthropogenic part of the sulfur cycle only. The space/time evolution in the sulfur emissions was derived from actual emission records. The full anthropogenic sulfur cycle was integrated into the atmospheric model, including the actual geographical emission of S02, chemical transformation to sulfate, semi-La-grangian transport of the sulfate aerosols, and finally the dry and wet disposition of sulfate particles from the atmosphere.

Aerosols Transport Cycle

FIGURE 9 Evolution of the annual sulfate content in snow/ice at the Dye 3 site in southern Greenland (65°N, 43°W). (a) Observed (Legrand, 1955). (b) GSDIO simulations for the nearest grid point with prescribed natural sulfur emissions only (gray line) and total emission (natural plus anthropogenic (dash line)). After Roeckner et til. (1999).

FIGURE 9 Evolution of the annual sulfate content in snow/ice at the Dye 3 site in southern Greenland (65°N, 43°W). (a) Observed (Legrand, 1955). (b) GSDIO simulations for the nearest grid point with prescribed natural sulfur emissions only (gray line) and total emission (natural plus anthropogenic (dash line)). After Roeckner et til. (1999).

—— Observed

i

>—.. A^a ^y .jt* y ™

1860 1900 1950 2000 2050

FIGURE 10 Evolution of changes in the annual global mean surface air temperatures compared to observations (for the experiments, GHG, full thin line, GSD light gray thin line, and GSDIO, gray thin line). Observational data from 1860 until present is shown by a heavy dark line. A 5-year running mean is applied (after Roeckner et al., 1999).

1860 1900 1950 2000 2050

FIGURE 10 Evolution of changes in the annual global mean surface air temperatures compared to observations (for the experiments, GHG, full thin line, GSD light gray thin line, and GSDIO, gray thin line). Observational data from 1860 until present is shown by a heavy dark line. A 5-year running mean is applied (after Roeckner et al., 1999).

In the third simulation, GSDIO, table 2 the indirect aerosol effect on cloud albedo was added. The tropospheric ozone distribution was also changed as a result of the prescribed anthropogenic emission of precursor gases. Figure 9 shows an attempt to validate the deposition of sulfate in the wet and dry deposition in ice core measurements at the Dye 3 on Greenland. Figure 9a shows the measured concentration of sulfate (in ng g-1) according to Legrand (1995), and Figure 9b shows the results from the corresponding control integration and from the GSDIO experiment. The agreement between the calculated depositions is in broad agreement with the measurements.

The global annual mean temperature change from the three ex periments, GHG, GSD, and GSDIO, is shown in Figure 10. As can be expected, the long-term warming is largest in experiment GHG and smallest in experiment GSDIO. Until 1980 or so the simulated temperatures are more or less within the range of natural variability of the control integration (not shown). However, the simulated temperature patterns undergo large low-frequency variations on a multidecadal time scale, in broad agreement with the estimated observed temperature pattern. In the model simulations, there are pronounced ultra-low fluctuations at higher latitudes of the Southern Hemisphere, but it is not possible to say whether these fluctuations are realistic or simply an artifact of the coupled model. However, when we compare the long-term trends in observations and

Observed

180 120w 60w 0 60e 120e 180
0 0

Post a comment