6.1 Climate change in response to variations in total solar irradiance

The simplest approach to simulating global average temperature changes in response to climate forcings is to use a 1-D Energy Balance Model (EBM). Figure 32View Image presents some of the results from one such study of the past millennium, showing the response estimated to a combination of forcing factors, along with estimates of surface temperature derived from observations. The results suggest that the gross features of the global temperature record are determined by volcanic and solar drivers until the twentieth century when human-induced factors, especially greenhouse gases, dominate.
View Image

Figure 32: Global average surface temperature calculated using an EBM. Observational record (in red), model calculations with natural (solar and volcanic) forcings (in blue), difference when anthropogenic greenhouse gases are included (in green). After Crowley (2000).

Simulations with full 3-D computer models of the general circulation of the atmosphere and oceans (GCMs) have also been carried out with time-evolving natural (solar and volcanic) and anthropogenic (greenhouse gases, sulphate aerosol) forcings. The GCMs are generally able to reproduce the temporal variation of surface temperature over the past two centuries (see Figure 33View Image). Before the mid twentieth century it is not possible to distinguish the all-forcing simulations from those only using natural forcings but since that time the models can only reproduce the observed warming if anthropogenic factors are included. This conclusion applies over each continent and the oceans as well as in the global or hemispheric means.

View Image

Figure 33: Annual and decadal mean surface temperature from observations (black line) and calculated using GCMs. Each panel shows calculations using all forcings (pink) and only natural forcings (blue), the spread indicates uncertainties in the estimates. From IPCC (2007).

However, there is a large amount of natural variability and noise in both the model and observational datasets which makes detecting component causes of climate change difficult. An alternative approach is “optimal fingerprinting” in which it is assumed that the geographical patterns of response to particular factors are known, that the time-dependences of the forcing factors are known but that the amplitudes of the responses are unknown. The task is then essentially to perform a multiple regression analysis on a dataset to find which weighted combination of the response patterns best matches the data, taking into account known errors/uncertainties in both the data and patterns. An example of the results of one such analysis, using a dataset of surface temperature observations on a latitude-longitude grid over the twentieth century, is shown as global averages in Figure 34View Imagea. The black curve is the observational record with the grey band representing measurement uncertainty; the red curve shows the result of using only anthropogenic forcing factors, the green only natural factors and the blue both together. A good fit is obtained when both types of forcing are included and the figure shows that increasing solar activity with declining volcanic activity were major drivers of the warming in the first half of the twentieth century and the mid-century plateau in temperatures.

View Image

Figure 34: Optimal fingerprinting technique in which geographical patterns of surface temperature change for different forcings are fitted to the observed time series. (a) results for different forcing factors, (b) derived magnitude of natural and anthropogenic forcings relative to that found from standard model runs. From Stott et al. (2003Jump To The Next Citation Point), redrawn by M. Lockwood (personal communication).

Figure 34View Imageb, however, shows the derived magnitudes of the forcings. Here the value 1 indicated that the derived magnitude equals that the model gives using standard radiative forcing estimates. The model appears to be underestimating the solar influence by a factor of 2 or 3 implying that some amplification factors of the solar influence are not incorporated into the model’s representation.

This result, however, is sensitive to the choice of TSI reconstruction. The Stott et al. (2003) work used the Hoyt and Schatten (1998) series, which has a large secular variation, although more recent work (see the discussion in Section 4.2) is suggesting much smaller long-term variability in TSI. If the latter were used then an even larger amplification of the model’s response would be required to match observations. This is an intriguing suggestion but this work needs to be reassessed using different GCMs to check that it is not an artefact of the specific one used in that study.

Nevertheless, most GCM runs which only include variations in TSI are unable to reproduce the distribution of temperature response shown in Sections 2.2 and  2.3, confirming that something is lacking in their ability to simulate the response of climate to solar activity. One recent study by Meehl et al. (2003), however, presented results from a model experiment which produced changes in SST similar to those shown in Figure 9View Image. The authors explained these in terms of a response of the mean overturning circulations to sea surface temperature gradients enhanced between cloudy and clear regions. This intriguing possibility remains to be validated.

To explain the model underestimate it is necessary to find some factors which amplifies the effect from that derived simply by consideration of total solar irradiance as the primary driving mechanism behind the impact of solar variability on climate. Potentially one such amplification mechanism is through the effects of variations in solar UV radiation on the stratosphere.

  Go to previous page Go up Go to next page