Correlation between LGM and doubled CO
2temperature changes
The ensemble mean, annually averaged T2 results are shown in Fig. 1, with sub-plot A showing the CTRL results and the other three sub-plots showing the differences in temperature between the each specific climate state and the CTRL. The zonal temperature changes for December, January and February (DJF) and June, July and August (JJA) for the three experiments and also the actual average temperatures for the control run are illustrated in Fig. 2. The existence of a strong “polar amplification” (as discussed by MD06) of the temperature changes can be seen in the results from this model. Crucifix (2006) quotes the following observational estimates of climate change: Antarctica, –9±20C (Jouzel et al., 2003); Greenland, –20±20C (Cuffey and Clow, 1997; Dahl-Jensen et al., 1998); and the tropical ocean, –2.7±0.50C (Ballantyne et al., 2005; Lea, 2005). For comparison with these results we have the following mean and 1 standard deviation range for our ensemble: Antarctica, –9±1.30C;Greenland, –18±20C; tropical ocean, –3.0±0.50C. Here we quote the average 2m temperatures over the Greenland and Antarctic land masses and the tropical region includes the ocean grid boxes between latitudes 30 0S and 30 0N. While interactive vegetation models are included in the PMIP2 protocol, the PMIP2 boundary conditions for AGCMs and AOGCMs exclude the effects of some vegetation and dust forcings which are thought to be significant and negative. This supports our belief that our ensemble of models has an overall bias towards high sensitivity. That is, with a more complete set of forcings, our models would have shown a clearly stronger cooling than that observed.
Global and tropical analysis
Figure 3 shows scatter plots for the globally averaged T2 changes for both LGM and LGMGHG verses 2×CO2, illustrating the smaller T2 changes (as expected) for LGMGHG. The correlation between the T2 changes is clear. The correlation coefficients for these results and some others are given in Table 2. The correlation coefficient is stronger between 2×CO2 and LGM climates when looking at the tropics only. The LGMGHG global T2 change is more highly correlated with LGM than 2×CO2 climates. This is perhaps surprising since, if the response to changes in greenhouse gas forcing was linear across the range covered by the 2×CO2 and LGMGHG states, then one would expect the correlation between these two states to be the stronger, because the LGM climate state is also strongly influenced by the large ice sheet and to a lesser extent by changes in solar forcing. It seems, therefore, that a large proportion of uncertainty in the model response is due to a nonlinearity in the response to positive and negative forcings, which we discuss further in Sect. 4.
Zonal analysis
Here we consider how the globally and zonally averaged patterns of temperature change are correlated for the different experiments.
Doubled CO2 experiment
The dashed line in Subplot A of Fig. 4 shows that, unsurprisingly, the correlation between the temperature changes for global and zonally averaged temperature change for the 2×CO2 climate is strong at all latitudes, although there is a notable drop in the southern sea-ice region (around 65 0S). Small changes in sea ice extent cause large localised temperature changes due to the positive feedback of the albedo effect. Even with ocean heat fluxes calculated to reasonably reproduce the present day climate, the ensemble members have somewhat different sea-ice extents in the modern climate, which results in substantially different temperature changes in this region when the ice extent shrinks (vanishes) in the warmer climate. Thus, the temperature change is strongly influenced by small biases in the initial sea ice extent.
LGM experiment
There is also also generally a high correlation between the global and zonally averaged temperature change at the LGM. Figure 4 subplot B shows this result split into DJF and JJA seasons. Both polar sea-ice regions (but not the poles themselves) show markedly lower correlation in the summer seasons, falling away to nothing during JJA for northern high latitudes, where the northern hemisphere ice sheets and sea ice are located.
With a lack of identical models run for both 2×CO2 and LGM conditions, the relationship between the two was assumed by MD06. Here we examine the relationship between the LGM and climate sensitivity by looking at the correlation between the magnitudes of the zonally averaged LGM and globally averaged 2×CO2 temperature changes. This result is shown as the solid black lines in subplots A and C of Fig. 4. While still high in places (including Antarctica), it is considerably lower (especially at northern latitudes) than the correlation between global and zonally averaged LGM temperature change. Areas of strong correlation include both central Antarctica and the tropics. The other two lines in subplot C of Fig. 4 shows the annually averaged results for the same correlation, split into land (magenta) and ocean (cyan). This shows a generally better correlation with the temperature over the ocean than the land between the latitudes of 50 0S and 50 0N. Also shown are the values of the correlation coefficients for the averages over the Antarctica and Greenland land areas. These show that while the MD06 conclusions are supported with a high correlation for Antarctica, in MIROC3.2 there is not such a high correlation for Greenland. The MD06 results were for central Greenland (>1300 m) and central Antarctica (>2500 m). Although there are differences (0C) in the magnitude of the temperature change, there is not a significant difference in the correlation coefficients evaluated using the central values rather than the averages. Due to the coarse resolution of our model we show the average land mass values since these are more likely to be robust. Our results suggest that the tropics, particularly the ocean regions, may also be good places for calibrating and improving models which are then to be used for prediction of future climate change caused by increased greenhouse gas levels. The existence of this particular correlation in the same model has already been used in previous work (Annan et al., 2005), where we attempted to constrain estimates of climate sensitivity using tropical SST data from the LGM. Our results here show that including Antarctic temperature estimates from ice cores into the calculation could potentially improve the result from such an experiment. Despite the small area at the poles, the data there may be less noisy than at the tropics due to the fact that the total temperature changes (Figs. 1 and 2) are much greater for the polar regions than the tropics in the winter months (and for the annual mean) for both 2×CO2 and the LGM. Similar conclusions have also been drawn by Schneider von Deimling et al. (2006). Correlations in the sea ice regions and over zones where the Northern ice sheets are situated at the LGM are weak, suggesting that, at least in our model, these regions are less informative of future climate changes.
Understanding and predicting climate change at smaller scales than global is obviously desirable. In this context we would like to know to what extent LGM climate changes can be used to validate the predictive models at the regional scale. As a step towards this we have calculated the correlation between the magnitude of the zonally averaged temperature changes for LGM and 2×CO2 climates. The resulting variation of the correlation coefficient with latitude is similar in shape to that obtained from analysing the globally averaged 2×CO2 and zonally averaged LGM changes. The correlation in the tropical regions is stronger, while insignificant in the southern sea-ice region. This strengthening in those areas that were strongly correlated with global changes might be expected, while the weakening in the sea-ice region indicates that, further to the discussion in Sect. 3.2.1, the large non-linear albedo feedback is such that the small differences in the modelled extent of sea ice leads to large differences in the local temperature response to forcing changes.
Answers to all your Biology Questions
