Fig. 4: Critical role of anthropogenic soil warming in the surge of soil-based compound dry-hot extremes (CDHEs).

a Observed regional mean annual frequency of soil-based CDHEs over 1980–2017 in China with anthropogenically-induced changes in soil temperature and moisture removed. b Same as (a), but for the annual coverage area of soil-based CDHEs. A grid cell is defined as part of the affected area if it experiences at least three soil-based CDHE days during a given warm season. c Regional mean annual frequency of soil-based CDHEs over 1980–2017 in China driven by anthropogenically-induced changes in soil temperature (red line), anthropogenically-induced changes in soil moisture (blue line), and nonlinear interactions between them (purple line), respectively. d Same as (c), but for the annual coverage area of soil-based CDHEs. The dashed line represents smoothed variations derived from LOESS local regression with a default span of 0.75. The linear regression slope (b), its 95% confidence interval (values in parenthesis), and p-value (p) are shown. The statistical trend analyses are conducted using the two-sided Mann–Kendall trend test. Since the effects of anthropogenic soil temperature and moisture changes on soil-based CDHE changes are represented by the estimated differences between different datasets (adjusted vs. unadjusted datasets), some times series of CDHE frequency and coverage area contain negative values. Nevertheless, our analysis focuses on long-term trends. Note that linear trends are computed using the non-parametric Theil–Sen trend estimation method, and thus the sums of these four components differ slightly from the trends shown in Fig. 2b, c.