Results from the model runs are analyzed along the following dimensions: (1) global damage costs by scenario; (2) the damage cost components; (3) regional impacts; (4) sensitivity analysis with respect to protection and (5) sensitivity analysis with respect to discounting.
Global damage costs by socio-economic and sea-level rise scenarios
While the choice of socio-economic scenario has an influence on the global damage costs from sea-level rise for the time period analysed for this study, the damage costs vary more over the choice of sea-level rise scenario (Fig. 2).
The damage costs for a 1 m rise are between 4.8 and 5.2 times as high as the damage costs for the 0.5 m sea-level rise, depending on the scenario (except for the 1995 control scenario, where the increase in costs is only 4 times). The increase in costs from 1 m to 2 m is only 2.0 times the damage cost of the 1 m sea-level rise scenario. The assumed bilinear protection costs between the scenario with 0.5 m rise and 1 m rise explains these different increases in damage costs with respect to sea-level rise. While the increase in damage costs from the 1 m to 2 m sea-level rise scenario is almost a factor of two in each of the socio-economic scenarios, the difference between 0.5 m and 1 m sea-level rise does depend somewhat on the socio-economic scenario. In all cases (except the 1995 control scenario) the increase of the total damage is lower than the assumed tenfold increase in protection costs. The overall difference between the SRES scenarios is small.
While the damages from sea-level rise are substantial, they are small compared to the total economy, provided that coastal protection is built. This remains true for the largest 2-m rise scenario. Note that the global total of Fig. 2 hides considerable differences between countries. This issue is discussed in more detail below.
In order to understand the reasons for the differences between the scenarios, a closer look at the four damage cost components is needed.
Disaggregating damage costs by socio-economic and sea-level rise scenarios
Figure 3 shows the damage cost components as calculated by FUND and their share of the total damage cost for the 0.5 m sea-level rise scenario under the assumption that dikes are built, i.e. that people attempt to protect against rising sea levels following current practise against coastal flooding in much of the world (e.g., East Asia and Europe). Note that the results change dramatically if it is assumed that people do not protect; this scenario is analysed in a later section.
Ignoring the control scenario for a moment, three conclusions can be drawn. First, damage costs from dryland loss and migration are a fraction of the costs of protection in every scenario (dryland costs being about one fifth and migration being one tenth of protection costs). Protection costs on the other hand are the most important component for every scenario. This underlines the significance of protection (and adaptation in general). Second, protection costs are less affected by the choice of socio-economic scenario than dryland loss and migration costs. The biggest difference between scenarios for dryland loss and migration costs is a factor of 1.8, for protection costs it is 1.5. Damage costs from wetland loss are even less sensitive to the choice of scenario, with a maximum difference of factor 1.3. Wetland costs are the second most significant damage component in all scenarios. Third, for every cost component except wetland loss, the highest cost scenario is A2, followed by B2, B1 and A1 (the lowest). For wetland costs, the order is reversed, because wetland cost differences between scenarios are mainly driven by the differences in valuation between socio-economic scenarios: higher per capita income place a higher value on wetland loss and therefore produce higher wetland costs. With the other damage costs, higher per capita income mainly leads to more protection, which explains why the effect of higher per capita income is positive in those cases.
Figure 4 presents the disaggregation into damage components for the 1 m sea-level rise scenario. Wetland costs are the only ones that react roughly linearly to the doubling of sea-level rise, they are around two times as high as for the 0.5 m sea-level rise in all scenarios. Protection costs increase between 4.2 to 6.6 times compared to the lower sea-level rise scenario, while dryland loss and migration costs increase by an order of magnitude (factors between 10.7 and 11.4) compared to the lower sea-level rise scenario. Due to the increase in adaptation costs (i.e. the bilinear nature of protection costs), adaptation is significantly more costly in the 1 m sea-level rise scenario and the cost-benefit analysis finds that the optimal length of coast to protect is lower than in the 0.5 m scenario (e.g. it is about 40% lower averaged over time for the A1 scenario, 46% lower for A2, 42% lower for B1 and 45% lower for B2), which leads to a situation where total damage is more evenly divided between the four damage cost components.
While the step from 0.5 m to 1 m sea-level rise changed the distribution of costs between the four components significantly, the step to the 2 m scenario has no such surprises. As can be seen in Fig. 5, all costs roughly double compared to the 1 m scenario. This is not surprising, since the model does not have a change in cost assumptions build into this step.
Regional distribution of damage costs
Sea-level rise damages are not evenly distributed over the world. Figure 6 compares the two scenarios that show the largest difference in total damage cost due to sea-level rise across all regions. While the distribution of damage costs is not the same for the two scenarios, the same regions bear the majority of damage costs in both scenarios. This should not be a surprise as relative exposure to sea-level rise is the main variable that drives relative damages and for example, East Asia and South Asia have large, densely-populated coastal lowlands irrespective of the scenario considered.
The three regions that are widely thought to be the most vulnerable to sea-level rise, i.e. the Pacific, Indian Ocean and Caribbean islands bear only a tiny share of the total global damage. At the same time these damage costs for the small island states are enormous in relation to the size of their economy (Nicholls and Tol 2006). Together with deltaic areas, they will find it most difficult to raise the finances necessary to implement protection.
Figure 7 shows damage costs as percent of GDP for the ten countries with the highest relative impact in 2100 for the A1 scenario with a 1 m sea-level rise. The economies of these ten countries are relatively small. Consequently damages to those countries do not constitute any significant part of the global total or even regional total impact from sea-level rise.
The level of protection, that is the length of coastline that is protected using dikes, is normally determined endogenously by a cost-benefit analysis in FUND. For the first time with a FUND analysis, another set of runs where no protection against sea-level rise is allowed were also conducted. Comparing these two sets of runs with and without protection is insightful for three reasons. First, it shows the huge benefits of protection to sea-level rise in terms of the damages avoided. Second, there might be countries that do not have the means to protect their coastline up to the optimal level that would follow from the cost-benefit analysis. This is especially relevant for large rises in sea level as considered in this analysis (Nicholls et al. 2008). Third, sea-level rise impacts are often presented without considering coastal protection (e.g., Dasgupta et al. 2009). This allows for a comparison between such studies and FUND.
Figure 8 clearly shows the importance of protection, in particular for the 0.5 m sea-level rise scenario. Total damages are between 3.4 and 3.7 times higher when no protection is build for that scenario, depending on the socio-economic scenario. For 1 m and 2 m sea-level rise the damages in the no-protection scenario are only around 1.4 times as high compared to a protection scenario. Since protection costs are assumed to be ten times higher than in the 0.5 m case, this is hardly surprising, as higher protection costs will lead to a lower protection level and therefore to a smaller change if that protection cannot be build. A look at the control Scenario C1995 is particularly interesting, as population and economic indicators are held constant at 1995 levels, while sea-level rise is assumed to occur. Especially in the two scenarios with the higher protection costs (1 m and 2 m), the importance of the significant economic development assumed in all the SRES scenarios can be seen. In both cases, there is little coastal protection in today’s socio-economic situation. The lesson to be learned from this is twofold: (1) protection can significantly lower total damages, but (2) only when economic growth enables this sometimes costly investment in protection to occur. Hence protection and economic growth are coupled, which has often been ignored in earlier analyses where socio-economic conditions are held constant as sea-level rises.
Some of the results for no protection scenarios are peculiar at first sight. For example, the Maldives are estimated to be completely inundated in 2085 for the 1-m rise scenario, which raises the value of its dryland for the time step 2080-4 to very large values. After 2085, the value is zero. This cannot be regarded as a satisfactory valuation from an economic point of view: Such non-marginal damages are outside of the realm of economic valuation. The Maldives disappear much earlier (2050) for the 2 m sea-level rise scenarios without protection, so that the costs of the 2 m scenario fall below that of the 1 m between 2050 and 2085.
Figure 9 displays the benefit gained from protection for specific countries. It shows that protection is a lot more important for some countries than for others, which reflects differences in the efficacy of coastal protection. In densely populated and rich countries, dike building has a high return in that a small expense prevents substantial damage. If people are dispersed and poor, the pay-off to coastal protection is much smaller.
Finally, we investigate how sensitive our cost estimates are to the assumptions made about the increase in costs when a threshold of sea-level rise is breached. The cost of building protection against sea-level rise is assumed to increase by an order of magnitude if sea-levels would rise by more than 1 cm per year for our central results. Figure 10 presents results of a sensitivity analysis that alters the assumed cost increase for a case of rapid sea-level rise. We keep the same threshold level of 1 cm per year, but change the parameter for the cost increase from the central value of 10 (cost increase of an order of magnitude) to a factor of 5 and a factor of 20. For sea-level rise slower than 1 cm per year this does not change predicted costs, as one would expect. For sea-level rise faster than 1 cm per year we find that overall damages increase when protection costs are higher and decrease when they are lower. The increase in overall damages is much more modest than the assumed increase in protection cost: For a protection cost increase of factor 20 instead of factor 10 we find an overall damage increase of 16%, while for a protection cost increase of factor 5 instead of factor 10 overall damage decreases by 22%. This is because protection levels change too, going up (down) if costs go down (up). This suggests that while different assumptions about protection costs matter for overall damage estimates, the general pattern of the results do not alter dramatically and are fairly robust to precise assumptions about protection costs.
Cost-benefit analysis of long term problems have proven to be highly sensitive to the discount rate used to compute net present value estimates. Debates about the appropriate discount rate to be used in climate change economics have been controversial and long standing. Figure 11 presents a sensitivity analysis of our results with respect to the discount rate. In particular, we altered the pure rate of time preference from its central value of 1% to compute estimates that use a pure rate of time preference of 0.1% and 3%, a range that spans commonly used discount rates in the literature on climate change. We find that the choice of discount rate is highly significant for the aggregate monetary estimate of impacts. With a low pure rate of time preference of 0.1% aggregate impact estimates are more than double the estimates that are computed with a high pure rate of time preference of 3%.
Note that we only altered the discount rate used to compute the net present value of the four damage components considered by FUND. We did not alter the discount rate used to compute the optimal protection level. The debate about discounting in climate change economics centers on the question how society should value future impacts from climate change and we investigate the sensitivity of our results to this question by altering the discount rate used to compute the net present value of the four damage components. The discount rate used to compute optimal protection levels on the other hand is not subject to these normative questions, it simply reflect how economic agents will respond to rising sea levels.