For purposes of presentation, the climate (Section 4.1) and biophysical (Section 4.3) results focus on the Zambezi River Basin (ZRB). Economic analysis is conducted at the country level for each of the three cases.
We find notable differences in climate outcomes between the unconstrained emissions (UE) and the mitigation scenario (L1S) for the ZRB as a whole by 2050. Schlosser and Strzepek (2015) provide a detailed description of the results obtained for sub-regions of the ZRB. Here, we focus on precipitation in the planting season and temperature in the summer (warmest months) as these are the most important for dry land agriculture. Frequency distributions of planting season (September–November, SON) precipitation change are shown in Fig. 2 (left panel). In the UE scenario, the mode of the distribution of precipitation shifts towards drier outcomes, although both substantial increases and decreases in precipitation are possible. Under L1S climate policy, the range of outcomes is reduced. Notably, the most extreme drying outcomes (decreases of − 0.5 mm/day and higher) are removed. Nevertheless, even under L1S, precipitation remains likely to decrease, with 42% of the L1S distribution at or below a decrease of − 0.2 mm/day in planting season (SON) precipitation.
For surface-air temperature (Fig. 2, right panel), the L1S scenario reduces the modal value of the summer (December–February, DJF) temperature-change distribution by at least 1 °C (less warming), with 56% of the distribution within a 1–1.5 °C increase in summer temperature — contrasted by nearly 56% of the UE distribution spanning summer temperature increases exceeding 2.5 °C (relative to an end of twentieth century average). Similar to what we show for precipitation, strong mitigation eliminates the occurrences of the most extreme temperature increases and shifts the distribution leftward. Specifically, the upper half of the UE distribution of change (exceeding 2.5 °C warming) is excluded in the L1S range of occurrences. Additionally, we find that the minimum warming in the distributions is less affected, indicating that even in the very aggressive mitigation scenario considered for this study, some degree of climate warming, combined with changes in precipitation, is likely unavoidable.
Global fossil fuel prices
The prices of fossil fuels are determined by the supply and demand for fossil fuels, considering interactions with alternative fuels that can act as substitutes. As fossil fuel resources deplete, the cost of production for additional resources tends to rise tempered by gains in extractive technologies. Technological progress also has implications on the demand side through improvements in energy efficiency. Figure 3 shows indices of world producer prices for coal, oil, and natural gas. The indices are constructed as a ratio of the price in the climate policy scenario (L1S) relative to the price in the unconstrained emissions scenario (UE).Footnote 5 For illustrative purposes, the figure extends to 2100 even though the focus of our analysis is to 2050.
The L1S climate policy lowers producer prices relative to UE due to a reduction in oil demand and competition from biofuels. Substantial margins between the cost of production and sale prices result in large price reductions as oil producers try to minimize oil demand reductions. By about 2050, oil prices that producers receive are lower by 60% in the L1S scenario in comparison to UE. This difference grows to almost 80% by the end of the century.
Coal producers also face price decreases under the climate policy, but most of them are already producing close to their marginal costs; therefore, they are not able to reduce their margins. Instead, carbon policies drastically reduce demand for (and production of) coal with some revival when carbon capture and storage (CCS) technology become economic. By 2050 and 2100, coal prices that producers receive are lower by about 10% and 25%, respectively, in L1S compared with UE (Paltsev 2012).
Natural gas price dynamics are more complicated. There are three segments. In the first segment, L1S prices are higher than in the UE scenario due to a switch from coal to natural gas. In the second, tighter emissions targets make natural gas less attractive as it still emits carbon, and there is a need to move to even lower carbon-emitting technologies such as wind, solar, and bioenergy. In the third, two factors serve to push up the price of natural gas relative to UE. The first factor is the large shares of renewables entering the power generation mix. Natural gas producer prices rise (relative to UE) because natural gas serves as a means to resolve intermittency issues. This is the smaller of the two factors. Natural gas is also used with CCS in the second half of the century, which is the larger factor.
The key result, which is qualitatively very similar to the result obtained by Bauer et al. (2016),Footnote 6 is that the price of the most traded fossil fuel, oil, is significantly lower under the L1S policy regime compared with UE. Global emissions policies also cause changes in the prices of other commodities, such as agriculture. However, the relative price shifts are much less pronounced.
Runoff supplies rivers and reservoirs with surface water. Increases in runoff imply that more surface water is available for various uses, including irrigation and hydropower production, although this availability also depends on water resource management as well as available storage and diversion. Decreases in runoff are likely to result in less water available for irrigation and hydropower production while rapid increases in runoff correlate with flood events. For this reason, runoff is a convenient indicator of how changes in climate translate into changes in water availability, which then affects economic outcomes.
The percent change in mean annual runoff, aggregated from 29 to five major basins, is shown in Fig. 4. The graphs within the figure contain the percentage change on the horizontal axis and the estimated kernel density (a measure of likelihood) on the vertical axis. Below each graph, the baseline mean annual runoff is shown as a proxy for the hydrologic significance of each basin. Also, the 10th and 90th percentiles of the modeled historical period—1951 to 1990—are shown in order to illustrate the magnitude of the climate scenario results as compared to the inter-annual variability that has been observed over the 40-year baseline period. The percentage shifts in runoff shown in the graphs represent 10-year means (2041–2050).
For Cahora Basa, the mode from UE and L1S both rest at about zero change and the extremes reach about ± 20%with very little difference between the two policies. The results are similar for the Shire River, although the UE case projects a slightly wetter future, and the overall expectation for both scenarios is also wetter with the mode at about + 5%. In the other three major basins, a decrease in runoff is more likely, at about − 5% for the mode value, which is about the same for both policy cases. However, the differences between the shapes of the two policy distributions are more prominent for these three major basins. In all cases, the UE distribution is noticeably wider, with more extreme tails, reaching − 50% for both the Kafue and Upper Zambezi. These two major basins also show a slightly lower expected runoff for the UE than the L1S scenario.
These changes in runoff, along with changes in irrigation demands, are used to model water resource allocations in order to understand how these changes in surface water supply translate to changes in water availability for the various users. We find that the water sector in Malawi is the least sensitive to climate change. Alternatively, Zambia is predicted to experience losses in terms of hydropower generation, caused mostly by expected decreases in runoff in the west, as well as upstream irrigation demands. The impacts on hydropower generation in Mozambique are mild due to a portfolio type effect stemming from a large contributing area. However, large-scale, high-damage flood events are projected to happen more often, especially under UE policy. This result is consistent in qualitative terms with Shongwe et al. (2009) and Hewitson and Crane (2006).
We elect to focus on the distribution of the average level of GDP over the period 2046–2050 for each of the three case countries. These outcomes are illustrated in Fig. 5, where panels a, b, and c correspond to outcomes for Mozambique, Malawi, and Zambia, respectively. For each country, we show three distributions of GDP outcomes corresponding to (i) the unconstrained emissions case, (ii) level 1 stabilization with world prices maintained at levels from the unconstrained emissions case, and (iii) level 1 stabilization with corresponding world prices. The horizontal axis shows the GDP level (average 2046–2050) as compared to a no climate change baseline. The vertical axis presents a measure of likelihood (kernel density estimates) for the associated GDP outcome under climate change.
The sources of GDP losses in the UE scenario relative to the no climate change baseline are discussed in detail in Arndt and Thurlow (2015) for Mozambique and Arndt et al. (2014) for Malawi. Our focus is on comparing the UE with the L1S global policy scenario. Relative to UE, L1S mitigation policy generates two distinct benefits. First, the distribution of economic outcomes shifts notably to the right (favorably) due uniquely to reduced disruption from climate change as a consequence of reduced temperature rise and a reduced likelihood of strong movements in precipitation (shown by the distribution corresponding to L1S climate with UE world prices). The extent of the shift varies by case country. It is most pronounced in Mozambique driven principally by a substantial reduction in flood probabilities (Arndt and Thurlow 2015).
Corresponding to the reduced dispersion of climate outcomes shown in Fig. 1 and biophysical outcomes as in Fig. 4, economic outcomes also tend to be less dispersed under effective mitigation. This is particularly true for the low-income economies, Mozambique and Malawi, where climate-sensitive sectors such as agriculture play a larger role in GDP. Overall, the distributions of economic outcomes in 2050 exhibit higher mean and reduced variance purely as a consequence of less pronounced changes in climate.
Second, like nearly all low-income and most middle-income developing countries, the case countries considered are substantial net importers of fossil fuels, particularly oil and derived products.Footnote 7 The reduced producer prices particularly for oil in the L1S scenario compared with UE scenario (see Fig. 3) generate a gain in terms of trade, which in turn confers substantial benefits in terms of economic growth (and welfare).Footnote 8 In countries participating in the mitigation regime, these terms of trade shifts would be accompanied by the costs of transitioning to less carbon-intensive energy sources. For simplicity and in order to delineate a best-case scenario for our case countries, we present economic outcomes whereby effective global mitigation occurs but the case countries do not participate. Hence, while avoiding mitigation costs, the case countries are able to import fossil fuels at a substantially lower cost.
The combined effect of these two benefits is to shift the distribution of economic outcomes to the right with the mean GDP outcome improving by about two to six percentage points relative to the unconstrained emissions case. For Malawi and Mozambique, mean and mode outcomes are superior to the no climate change baseline. For Zambia (a lower middle-income country), the mean of the distribution of the GDP outcomes is only slightly worse than the no climate change baseline with about one fourth of the distribution lying above the no climate change baseline.