Warming of hot extremes alleviated by expanding irrigation

Abstract

Irrigation affects climate conditions - and especially hot extremes - in various regions across the globe. Yet how these climatic effects compare to other anthropogenic forcings is largely unknown. Here we provide observational and model evidence that expanding irrigation has dampened historical anthropogenic warming during hot days, with particularly strong effects over South Asia. We show that irrigation expansion can explain the negative correlation between global observed changes in daytime summer temperatures and present-day irrigation extent. While global warming increases the likelihood of hot extremes almost globally, irrigation can regionally cancel or even reverse the effects of all other forcings combined. Around one billion people (0.79-1.29) currently benefit from this dampened increase in hot extremes because irrigation massively expanded throughout the 20[Formula: see text] century. Our results therefore highlight that irrigation substantially reduced human exposure to warming of hot extremes but question whether this benefit will continue towards the future.

Fig. 1. Observed warming rates affected by irrigation.

Boxplots of the total ($\mathrm{\Delta }\mathrm{TXm}$, a, b) and irrigation-induced ($\mathrm{\Delta }{\mathrm{TXm}}_{\mathrm{irr}}$, c, d) change in average daily maximum temperature during the hottest month of the year for global land (a, c) and South Asia (b, d) between 1901 and 1930 and 1981 and 2010. Results were binned by the change in grid cell fraction equipped for irrigation (hereafter referred to as irrigated fraction $f_{\mathrm{irr}}$) between both periods. Cell counts per bin for $\mathrm{\Delta }{\mathrm{TXm}}_{\mathrm{irr}}$ are indicated in e, f. Blue bars represent results from the CRU observational data, orange bars show results from the CESM global climate simulations. Boxplots indicate the spatial distribution (centre line: median; box limits: upper and lower quartiles; whiskers and outliers: not shown) and are only plotted for bins containing $\ge 5$ pixels. Note that most grid cells with $f_{\mathrm{irr}}>0.5$ for the global land are situated in South Asia.

Fig. 2. Change in probability of hot extremes from expanding irrigation and other forcings.

Ensemble-mean likelihood of exceeding 99th percentile of daytime temperature ($\mathrm{TX}$) as simulated by CESM, considering all forcings except irrigation (a), irrigation expansion only (b), and all forcings including irrigation expansion (c). Probability ratios (PRs) are shown for the present-day (1981–2010) relative to the early 20th century reference period (1901–1930), except for b where the reference is a counter-factual present-day world without irrigation.

Fig. 3. Regional masking of trends in hot extremes due to irrigation.

Median probability ratio (PR) for individual daytime temperature (TX) percentiles considering all forcings except irrigation (red), irrigation only (dark blue), and all forcings including irrigation (light blue) for all land (a), all irrigated land (b), and South Asia (c). Irrigated land is defined here as all pixels with $>$10% irrigated crop fraction, and corresponds to $~$5% of all land area. South Asia is defined as Pakistan, India, Nepal and Bangladesh, and represents $~$3% of all land. Note that the bars are non-additive because of differences in the reference ensemble.