Tropical and Extratropical Cyclone Damages under Climate Change

Matthew Ranson, Carolyn Kousky, Matthias Ruth,

Lesley Jantarasami, Allison Crimmins, and Lisa Tarquinio

Appendix

1  Sensitivity Analysis

Table A1 in this appendix presents a sensitivity analyses designed to explore the robustness of our results. The table shows the estimated distribution of the predicted effect of temperature on storm losses, under several different alternative methodological approaches. The first column shows our preferred results, reproduced here from Table 3 in the main paper.

The second column shows similar results, using data from peer-reviewed studies only. The effects in the North Atlantic are slightly higher, and the effects in the Western North Pacific are somewhat lower, but neither difference is substantial.

The third column shows results based on a weighting scheme in which all estimates from a particular study are given equal weights (instead of weighting by baseline damages for each geography within each study). The estimated treatment effects are nearly identical, indicating that the weighting scheme has little influence on our results.

Finally, the fourth column shows results based only on studies and estimates for which the change in temperature was explicitly reported in the original study. The results are slightly lower, but the difference is not meaningful.

Table A1: Sensitivity Analysis:
Distribution of the Predicted Effect of Temperature on Storm Losses
Mean and Std. Deviation of Treatment Effects, All ΔT
Results from Table 3 in Main Text / Peer-Reviewed Studies Only / Equal Within-Study Weights / Temperature Explicit in Study
Tropical Cyclones
North Atlantic / 0.154 (0.182) / 0.166 (0.137) / 0.152 (0.183) / 0.130 (0.076)
Observations / 107 / 38 / 107 / 43
Studies / 11 / 6 / 11 / 5
Western North Pacific / 0.063 (0.168) / 0.037 (0.233) / 0.067 (0.169) / 0.051 (0.160)
Observations / 29 / 13 / 29 / 26
Studies / 4 / 2 / 4 / 4
Other/Multiple / 0.144 (0.167) / 0.144 (0.167) / 0.130 (0.195) / 0.144 (0.167)
Observations / 42 / 42 / 42 / 42
Studies / 2 / 2 / 2 / 2
Wind Storms
Europe / 0.078 (0.063) / 0.074 (0.072) / 0.077 (0.069) / 0.033 (na)
Observations / 300 / 298 / 300 / 1
Studies / 8 / 6 / 8 / 1
Notes: The table shows variation in the predicted effect of changes in surface air temperature on storm losses, based on different studies, methodologies, and geographies. Specifically, the table shows the estimated mean and standard deviation parameters (μ and σ) for the lognormal distribution in Equation (3). Note that the numbers in parenthesis in the table represent standard deviations of the distribution of coefficients (not the standard error of the mean). Unless otherwise specified, all estimates are weighted by absolute baseline loss within each study, with each study receiving equal weight. Columns (1), (2), (3), and (4) show results for different samples and weights.

2