Using proxy reconstructions to constrain model parameters and detect the role of the various forcings

— Reconstructed (1900-1980 calibration, SNR = 0.4, p = 0.32) with uncertainties

— Reconstructed (1900-1980 calibration, SNR = 0.4, p = 0.32) with uncertainties

The time evolution during the past millennium of the annual mean temperature averaged over the Northern Hemisphere generally closely follows variations in the external forcing. This is probably the main reason for the success of relatively

Figure 7.6 Annual mean temperature anomaly averaged over the Northern Hemisphere in simulations performed with ECBILT-CLIO-VECODE using only one forcing at a time: greenhouse gas forcing (red), aerosol forcing (green), land-use changes (dark blue), volcanic forcing (orange), and solar forcing (light blue). The curves shown are the average over an ensemble of 10 simulations performed with the model using one of the forcings described in Figure 7.2. A 25-year running mean has been applied to all the time series. The reference period is 1500-1850.

1100 1200 1300 1400 1500 1600 1700 1800 1900 2000 Time (years)

simple models in simulating this variable. At this spatial scale, the response of the climate system to the total forcing is, in good approximation, equal to the sum of the responses to all the individual forcings. It is thus possible to disentangle the role of the various forcings by performing simulations in which only one forcing is applied at a time (e.g. Crowley 2000; Bertrand et al. 2002; Bauer et al. 2003; see also Figure 7.6).

Such simulations have allowed that, as expected, a higher than average solar irradiance is associated with higher temperature and vice versa. As a consequence, the solar forcing alone, as reconstructed in Figure 7.2, induces similarly high temperatures during the 12th, 13th, and 20th centuries, whereas lower temperatures are simulated during the 16th and the period covering the late 18th to early 19th centuries. The large volcanic forcing is responsible for the simulated cold periods after large volcanic eruptions, in particular those of 1258, 1452, 1600, 1641, and 1815. The land-use changes are associated with a long-term cooling trend and the simulated cooling due to aerosols is mainly restricted to the 20th century. Greenhouse gas forcing mainly induces a large warming during the past 150 years, and if the forcing is excluded it is not possible to simulate the large temperature increase observed during the 20th century.

The temperature response averaged over the Southern Hemisphere is more complex than in the Northern Hemisphere. Between 50°S and 70°S, the Earth's surface is nearly exclusively covered by oceans, implying that the effective heat capacity at these latitudes is much larger than in the Northern Hemisphere, inducing a damped and delayed response to the forcing and thus a less clear imprint of the volcanic signal than in the Northern Hemisphere (Figure 7.7). Furthermore, the Southern Ocean experiences large-scale upwelling of relatively old water masses that acquire their characteristics decades to centuries before they reach the surface in the Southern Ocean. As a consequence, the surface temperature at one particular time in the Southern Ocean is not only influenced by the radiative forcing at that time but also by the previous history of forcing changes, which has left its imprint on the characteristics of the older water masses that upwell there. This

Figure 7.7 Annual mean temperature anomaly averaged over the Northern Hemisphere (red) and over the latitude band 50-70°S in simulations performed with ECBILT-CLIO-VECODE driven by natural and anthropogenic forcings. The time series are grouped by 25-year average and divided by their standard deviation to facilitate the comparison. The reference period is 1500-1850.

Figure 7.7 Annual mean temperature anomaly averaged over the Northern Hemisphere (red) and over the latitude band 50-70°S in simulations performed with ECBILT-CLIO-VECODE driven by natural and anthropogenic forcings. The time series are grouped by 25-year average and divided by their standard deviation to facilitate the comparison. The reference period is 1500-1850.

1100 1200 1300 1400 1500 1600 Time (years)

1700 1800 1900

1100 1200 1300 1400 1500 1600 Time (years)

1700 1800 1900

results in a much more complex link between the forcing and the response than in the Northern Hemisphere. For instance, this has led to the hypothesis that the Medieval Warm Period was delayed in the Southern Ocean by 50 to 200 years compared with the Northern Hemisphere (Goosse et al. 2004).

The previous discussion is mainly based on model results. The data are only used to show that the models are providing results in agreement with the reconstructions if all the forcings are included and that the conclusions deduced from model simulations appear reasonable. The results could thus be model-dependent and strongly influenced by uncertainties in the reconstruction of one particular forcing. It is possible, however, to combine directly the model results and observations in order to detect the influence of various forcings on Northern Hemispheric temperature during the past millennium (Hegerl et al. 2003), using techniques similar to the ones applied for the 20th century, and to distinguish the anthropogenic and natural forcings (e.g. Tett et al. 1999; Stott et al. 2001). The first step is to obtain, from model simulations, the signal associated with a forcing, i.e. the time evolution of the temperature change when only one forcing is applied. This is usually called the "fingerprint" of the forcing. A multiple regression is then used to obtain the linear combination of those fingerprints that provide the best agreement with the reconstructions. In other words, the model provides the shape of the response to a forcing whereas it is the fitting of the linear combination of the fingerprints with observations that gives the amplitude of the response to this forcing. The conclusions of such analyses are relatively insensitive to the magnitude of the response of the model to a particular forcing or to uncertainties in the amplitude of the forcing. Indeed, an error in the amplitude of the forcing of a factor of two would simply induce a change by a factor of two in the regression coefficient, if the response of the system could be considered as linear (e.g. Tett et al. 1999).

Using an EBM to obtain the fingerprints of the externally forced signals and various reconstructions of Northern Hemisphere mean temperature, Hegerl et al. (2003) were able to detect clearly the influence of volcanic forcing on temperature changes during the past millennium. Furthermore, in their model, the time-scale of the response is similar to the one deduced from the reconstructions. The influence of anthropogenic forcings during the late 20th century is also detected, in agreement with studies focusing on the 20th century. Solar forcing has been marginally detected only over some periods and in some records, thus they conclude that the role of this forcing has probably been weak for the multi-decadal variability of the temperature averaged over the Northern Hemisphere.

It is possible to combine model results and data to constrain the magnitude of the changes during the past millennium as well as the value of some important characteristics of the climate system such as the climate sensitivity. Using a coupled physical-biogeochemical climate model, driven by different reconstructions of past changes in solar irradiance and volcanic forcing, Gerber et al. (2003) compared the simulated CO2 and temperature evolution with available observations. In their analysis, for large changes in solar irradiance and hemispheric mean temperature, they find large discrepancies between observed CO2 concentration and that simulated by the model. They argue that, between 1100 and 1700, changes in multi-decadal Northern Hemispheric mean surface temperature larger than 1°C are inconsistent with CO2 records and thus not realistic. Following a similar objective to Gerber et al. (2003), Hegerl et al. (2006) recently performed a large ensemble of simulations with an EBM over the past millennium varying two important model parameters, namely the equilibrium climate sensitivity to a doubling of CO2 and the effective ocean diffusivity, and using different scaling of the reconstructions of past radiative forcing. By comparing the results of this ensemble with several reconstructions of surface temperature in the Northern Hemisphere at (nearly) hemispheric scale, using a technique based on a Bayesian approach, they derive a probability density function (PDF) for climate sensitivity. Such methods, however, could be strongly dependent on prior assumptions made in the process of the PDF estimation (e.g. Frame et al. 2005) and additional work is still required before being able to constrain efficiently the model parameters using all the available information from the data.

0 0

Post a comment