Fig. 4: Attribution of changes in CDHW across different income groups.

Observed and simulated frequencies of CDHWs for a low-income and b high-income countries from 1981 to 2020. The solid black line in (a, b) indicates the annual frequency of CDHWs based on the ERA5 reanalysis dataset (OBS), and solid orange and green lines represent the ensemble mean results based on CMIP6 climate model simulations with ALL (ANT + NAT) and NAT forcings, respectively (Supplementary Table 1). The dashed lines represent the linear trend, and the shadings show the 5 to 95% ranges based on the ensemble simulations of ALL and NAT, respectively. c The best estimates of the scaling factors for each income group and (d) attributable increasing trends (unit: %/year) generated from a two-signal [ANT (ALL-NAT) and NAT] analysis of changes in the frequency of CDHWs spanning from 1981 to 2020. The error bars in (c, d) illustrate the uncertainty range of 5 to 95% for the respective estimates. The frequency is calculated for each grid cell and then weighted by the area of grid cells for different income groups with consideration of the weights assigned to grid areas. The frequency is calculated for each year as the area-weighted average of grid cells in different income groups.