Online climate change projections report 3.2.5 Earth system uncertainties
We sample uncertainties in ocean, sulphur cycle and terrestrial processes by running three additional perturbed physics ensembles, each consisting of 16 perturbed variants of HadCM3. Each of these ensembles is driven from 1860–2100 by the same time series of forcing agents used in PPE_A1B. In each of these ensembles parameters in the module targeted for perturbation are varied within ranges obtained by consultation with experts, while parameters in other modules are held fixed at values used in the standard model variant. In all cases parameter combinations were determined using a Latin Hypercube sampling design (McKay et al. 1979).
Ocean transport
The first ensemble addresses uncertainties in ocean transport, building on preliminary simulations reported by Collins et al (2007). The ensemble members sample perturbations to parameters controlling various aspects of the resolved and subgrid-scale transports of heat, salt and momentum in both the horizontal and vertical. In these simulations, future global mean temperature rise shows a limited dependence on these ocean parameters (Figure 3.2), much smaller than the uncertainties arising from atmospheric processes.
Sulphur cycle
The second ensemble samples uncertainties in atmospheric sulphur cycle processes, represented in HadCM3 using the module described by Jones et al. (2001). It simulates sulphate aerosol concentrations from prescribed emission fields of anthropogenic sulphur dioxide (SO2), natural dimethyl sulphide and tropospheric sulphur arising from quasi-regular volcanic eruptions. Three modes of aerosol are represented, comprising sulphate dissolved in cloud droplets plus two free particle modes. The model simulates production of sulphate by oxidation of SO2, transport within the atmosphere, rain out and transfers between the different aerosol modes. The atmospheric sulphur burden affects radiation via the direct (cooling) influence of scattering and absorption of incoming solar radiation, and through increases in cloud albedo resulting from the action of sulphate aerosols as cloud condensation nuclei (the first indirect effect). As mentioned earlier, the second indirect effect, in which reductions in cloud droplet size reduce precipitation efficiency and increase cloud lifetime, is not included since the calculation of precipitation in HadCM3 does not allow for any dependence on cloud droplet number concentrations. The 16 member ensemble of HadCM3 simulations samples simultaneous perturbations to parameters controlling key aspects of the processes outlined above, including emissions of aerosol precursors. All ensemble members used the settings for atmosphere and ocean module parameters employed in the standard variant of HadCM3. This ensemble simulates a wide range of atmospheric sulphur burdens (although perturbations to some of the atmosphere module parameters in PPE_A1B and PPE_A1B_NOGHG also have a significant impact on these). The impact of sulphur cycle perturbations on global mean temperature changes is modest compared with that in PPE_A1B (Figure 3.2).
Terrestrial ecosystem
Uncertainties in terrestrial ecosystem processes are sampled in a third ensemble in which the TRIFFID dynamic vegetation module of Cox (2001) is added to HadCM3, to form an Earth System model HadCM3C. TRIFFID simulates soil carbon, and the growth and replacement of five functional types of vegetation (broadleaf tree, needleleaf tree, C3 grass, C4 grass and shrubs). The functional types vary according to the net available carbon and competition between plant types, parameterised using empirical relationships. Soil carbon can be increased by litterfall and is returned to the atmosphere by microbial respiration, which depends on temperature and soil moisture. CO2 fluxes at the land–atmosphere interface are determined by photosynthesis and plant and microbial respiration. In order to simulate carbon fluxes at the ocean–atmosphere interface, an ocean carbon cycle module (Cox et al. 2001) is also included. This simulates exchange of gaseous CO2 with the atmosphere, the transport of dissolved inorganic carbon and cycling of carbon by marine biota via a nutrient–phytoplankton–zooplankton–detritus ecosystem module (Palmer and Totterdell 2001) that accounts for the effects of light penetration, alkalinity and nutrient availability on biological carbon uptake. In previous carbon cycle experiments using HadCM3 (e.g. Cox et al. 2000; Jones et al. 2003), the horizontal resolution of the ocean module was reduced; however, here we maintain the standard resolution of 1.25 x 1.25 degrees in order to ensure that our carbon cycle simulations are physically consistent with the other coupled ocean–atmosphere ensembles included in our methodology.
A 16-member ensemble was produced, sampling simultaneous perturbations to TRIFFID parameters controlling soil carbon and the five vegetation functional types. A further ensemble member with standard TRIFFID settings was also run. Parameters in the ocean carbon cycle module were held fixed at standard values in these simulations, because resource and time limitations made it impractical to perform the ensemble of long preliminary integrations (e.g. Cox et al. 2001) which would have been required to achieve equilibrium in ocean–atmosphere carbon fluxes. The specification of forcing agents was as in PPE_A1B, except that CO2 was input as a time series of emissions rather than concentrations, in order to allow carbon cycle feedbacks to operate. This ensemble simulates a substantial range of future changes in CO2 concentration (669-1130 parts per million at the year 2100, for example), and therefore of global mean surface temperature (Figure 3.2), comparable to the spread found by sampling physical surface and atmospheric processes in PPE_A1B. Uncertainties in the ocean carbon sink are not sampled in our simulations (as explained above), however the spread of responses obtained is similar to that found in a previous multi-model ensemble of carbon cycle simulations carried out in the Coupled Climate Carbon Cycle Intercomparison Project (C4MIP) by Friedlingstein et al (2006). The C4MIP ensemble sampled variations in both terrestrial and ocean carbon cycle processes and found that climate-induced changes in carbon storage were explained mainly by the former.
In addition to their impacts on global mean surface temperature (Figure 3.2), the ocean, sulphur cycle and terrestrial ecosystem PPEs all show some statistically significant impacts on patterns of regional change in some parts of the world. For example, the sulphur cycle PPE shows a significant spread in temperature changes in the Arctic Ocean and over interior regions of the northern Eurasian landmass (because surface albedo feedbacks amplify the effects of perturbations to the response of surface temperature), and in precipitation changes over tropical regions of the central and western Pacific Ocean (due to the strong coupling with sea surface temperature changes in these regions). The ocean PPE shows similar impacts over the Arctic and tropical Pacific Oceans, while the terrestrial carbon cycle PPE shows a large spread of precipitation changes over Amazonia, due to the regional influence of ecosystem-atmosphere interactions (Betts et al. 2004). However the impacts on changes over the UK (beyond those directly explained by variations in the global mean warming) turn out to be relatively minor.
- Last updated: Sunday, 11 March 2012