a, Maps shows the spatial distribution of high and low elevation PIN codes, according to three different statistics summarizing PIN-code-level elevation. For each statistic, we define whether a PIN code is high or low elevation based on whether its value of the statistic is above or below the median value across all PIN codes. The three statistics, indicated in the facet titles, are ‘Elevation’ (i.e., simply the mean elevation across pixels within each PIN code), ‘% < 5m’ (i.e., the proportion of pixels in each PIN code that are below 5 m above sea level), and ‘HAND’ (i.e., the mean height above nearest drainage across pixels within each PIN-code). Maps are from ref. 59. b, Points show the percent increase in five-week mortality due to a single day with total rainfall indicated in the facet title (relative to a day with zero rainfall), differentiated by elevation according to the three measures in (a) (n=19880 throughout this figure). A version with confidence intervals around these point estimates is in Supplementary Fig. 28. c, Points show the percent increase in five-week mortality due to a single hour with total rainfall indicated in the facet title (relative to an hour with zero rainfall), differentiated by elevation according to the three measures in (a). A version with confidence intervals around these point estimates is in Supplementary Fig. 29. d, Points show the percent increase in five-week mortality due to a single hour with total rainfall indicated by the facet title (relative to an hour with zero rainfall), differentiated both by tide height during which the rainfall occurs and elevation according to the three measures in (a). “High tide” refers to a 95th percentile tide height (416 cm) and “Medium tide” refers to a median tide height (262 cm). A version with confidence intervals around these point estimates is in Supplementary Fig. 30.