What will the climate be like in a hundred years’ time? The answer to this question is highly uncertain, and will depend on a number of socio-economic as well as natural processes, which describe the links between human activity, emissions of greenhouse gases, and warming of the atmosphere. The existing policy discussion in important forums, such as the IPCC and Stern reports (see this Vox column), is largely based on the uncertainty about the biogeophysical and biogeochemical systems, as are analyses such as that of Wigley and Raper (2001). In a recent paper, we include such uncertainty – but highlight uncertainty about the drivers of climate change in the socioeconomic system (CEPR Discussion Paper, 7024).
Clearly, how the climate evolves depends crucially on the policy measures that states can agree upon to reduce greenhouse gas emissions. Our paper, however, abstracts from mitigation and concentrates on evaluating various sources of uncertainty about future climate in a business-as-usual setting – i.e., in the absence of any further policies to stop climate change.
We make use of the RICE model (Regional Dynamic Integrated model of Climate and the Economy), developed by William Nordhaus and others at Yale University. This model quantifies not only how human activity affects the climate but also feedbacks in the other direction – how a warming climate will alter the economy’s productive capacity, as well as many other dimensions of human welfare. Many models used to judge the future climate describe the emissions caused by the economy simply by forecasting how GDP will grow over time. In RICE, population growth, productivity increases, energy use, and capital accumulation are all modelled explicitly, which makes it particularly useful for our purposes. While RICE was originally designed to propose an optimal climate policy, we program it so as to generate business-as-usual outcomes.
The time horizon in RICE is 400 years, and a time period is taken to be ten years. Here is how climate change occurs in the model. In each of eight regions of the world, population and productivity grow according to given processes. This stimulates investments in higher capital stocks and raises consumption of fossil fuels. A long-run supply function for fossil fuels implies that global carbon prices rise as energy becomes more scarce, which somewhat dampens energy use. However, more fossil fuel consumption emits carbon dioxide (CO2), which, along with emissions caused by changes in land use, raises the atmospheric concentrations of greenhouse gases, even though some CO2 is taken up by the biosphere and the oceans (via the carbon cycle). Rising concentrations lead to higher radiative forcing that warms the world and the damages from global warming feed back to the economic system.
The equations in the original RICE model have a set of fixed parameters. We introduce uncertainty by assigning probability distributions to about 80 of these parameters, and then perform a Monte Carlo simulation. This means that we solve the model a large number (10001) of times, each time drawing values for all uncertain parameters from their probability distributions. We use a variety of sources to determine the probability distributions we attach to the various model parameters.
As an example of the socio-economic processes involved, we calculate means and standard deviations for regional population growth based on the forecasts available in the UN World Population Prospects database. The distributions of actual population trajectories used in the simulations are presented in Figure 1.
Figure 1 Fan charts describing population forecasts 150 years ahead
Among the natural processes involved, the most uncertain parameter is the model’s climate sensitivity – a measure of the long-run rise in global mean temperature following a doubling of atmospheric CO2. For this parameter we build on Roe and Baker’s (2007) analysis of feedback uncertainty to generate a skewed distribution similar in shape to that obtained by Murphy et al. (2004). Figure 2 shows the realised distribution of this parameter in our simulations.
Figure 2 Histogram of realised values of the climate sensitivity parameter
Each run of the model gives a future path for the variables of interest, including climate variables such as CO2 emissions, atmospheric concentrations, and global warming. Figure 3 presents a fan chart for global mean temperature over the coming 150 years (left panel), and a snapshot of the distribution at a point in time 100 years ahead (right panel). As the histogram shows, in all of our (10001) simulations, global temperatures go up by more than 2 °C, which the European Union has deemed to be an upper limit for manageable climate change (as referred to in Article 2 of the UNFCCC).
While there are a few simulated futures with a warming of “just” 2-3 °C, anyone who pays close attention to this optimistic tail should seriously consider also the pessimistic tail of the distribution for climate change. As the histogram shows, the highest temperature realisations by 2105 involve a rise around 7 °C. The effects of such temperature changes are very hard to predict, but may include sea levels high enough to threaten major cities as London, Shanghai, or New York, and substantial risks of large-scale shifts in the Earth system, such as collapses of the Gulf Stream or the West Antarctic ice sheet.
Figure 3 Illustration of the uncertainty about future temperature increase
In 2105, the 99% confidence interval for temperature is almost 4 °C wide. This range of warming for the next century is of the same magnitude as the range reported elsewhere, but derived with very different methods.
Sources of uncertainty
What causes variability of global temperature in our simulations? The climate sensitivity parameter is the most important source of uncertainty. This reflects the fact that climate sensitivity is the last link (in the RICE model) in the chain from human activity to global warming. But other reasons are more squarely rooted in the human system. One is lower than expected improvements of energy efficiency in regions with high production and dirty technologies: chiefly the US and China. Another root is higher than expected economic growth in very populous regions, in particular the current low-income countries that host about half of the world population. To put it bluntly, futures in which today’s unfortunate manage to permanently break out of poverty (without large improvements in energy-saving technologies) have substantially higher global warming. Ironically, resolution of one of today’s most pressing global problems aggravates another one.
Figure 4 illustrates how certain socio-economic variables make global warming deviate from the value one would expect given the (uncertain) climate sensitivity parameter. A striking example is the difference between the observations labelled 2 and 3 – the bottom right panel shows that climate sensitivity takes a value close to 4 in both cases, while temperature in a 100 years time differs by more than three degrees. The other panels reveal that observation 3 entails a future with low GDP growth in the two most populous regions in the world and very high improvements in Chinese energy efficiency. Observation 2, by contrast, corresponds to higher GDP growth in China as well as in low-income countries and – most importantly – a very low improvement in Chinese energy efficiency. Similarly, observations 1 and 4 are futures with a three-degree temperature difference, but this time the most important explanation is that low-income countries grow much faster in the future underlying observation 1.
Figure 4 Illustrations of socioeconomic drivers of climate change uncertainty
These futures indicate that alternative socioeconomic developments contribute a great deal to the uncertainty about future global warming. Figure 5 illustrates this further, showing another Monte Carlo simulation with 10001 draws, where climate sensitivity is held constant at its mean value. The 99% confidence interval is nearly 2 °C wide, i.e. half of the corresponding interval in Figure 3.
Figure 5 Uncertainty about future temperature increase for a given value of climate sensitivity
The upshot of our analysis is simple – without swift policy action, global warming will be a major problem even under very optimistic circumstances.
Murphy, J. et al. “Quantification of Modelling Uncertainties in a Large Ensemble of Climate Change Simulations”, Nature 430: 768–772, 2004
Roe, G. H. and Baker, M. B., “Why Is Climate Sensitivity So Unpredictable” , Science 318: 629–632, 2007
Sir N. Stern (ed), The Stern Review on the Economics of Climate Change, Cambridge Univ. Press, Cambridge, 2006.
Wigley, T. M. L. and Raper, S. C. B., “Interpretation of High Projections for Global-Mean Warming” , Science 293: 451–454, 2001