Warming of hot extremes alleviated by expanding irrigation

The study investigates how expansion of irrigation has influenced historical changes in hot temperature extremes relative to other anthropogenic forcings, particularly greenhouse gas-driven warming. Although irrigation’s local climatic imprint has been observed and modeled, its contribution has not been systematically included in assessments of observed climate change drivers. Given the large increase in irrigated area during the 20th century, the authors aim to disentangle irrigation’s effect on hot extremes from the broader warming signal using observations and earth system model simulations, with special attention to strongly irrigated regions such as South Asia.

Prior work has detected irrigation imprints on local climate using in-situ temperature and energy flux observations and remotely sensed soil moisture. Modeling studies show irrigation can delay the Indian Summer Monsoon onset and alter precipitation patterns regionally and remotely. While annual mean temperature effects are limited, emerging evidence indicates irrigation substantially affects temperature extremes, notably cooling the hottest days. Despite this, land management (including irrigation) has often been omitted from historical climate change attribution. The irrigated area expanded dramatically in the 20th century, motivating a formal assessment of its role in observed changes in hot extremes.

Data: Monthly gridded temperatures from CRU TS v4.02 (0.5°) for 1900–2017 were used to compute the average daily maximum temperature during the hottest month (TXm), averaged over 1901–1930 and 1981–2010. Irrigation extent was from the Historical Irrigation Dataset (HID), converted to irrigated cell fraction and remapped to the model grid. Observational attribution method: A spatial window searching algorithm was applied to CRU and CESM outputs to reconstruct local irrigation-induced contributions to ΔTXm. For each eligible pixel, multiple linear regression across an 11×11 grid-cell window related total temperature change to changes in irrigated fraction and spatial predictors (latitude, longitude, elevation). The irrigation effect at the center pixel was obtained by multiplying the regression coefficient for irrigation by the local change in irrigated fraction. The method is designed to isolate local biogeophysical cooling from irrigation mainly via increased evaporative fraction. Climate model experiments: The Community Earth System Model (CESM v1.2 with CLM4.0 including an interactive irrigation module) was used to perform four ensembles (five members each): CTL (all forcings except irrigation, 1976–2010), IRR (as CTL with irrigation on, 1976–2010), CTL_20C (all forcings except irrigation, 1896–1930), and IRR_20C (as CTL_20C with irrigation on). Simulations used prescribed transient GHGs, SSTs, sea ice, and MODIS-based phenology. Irrigation areas were prescribed for years 1915 and 2000 to represent early and late 20th-century extent; irrigation demand and timing were computed internally based on soil moisture stress during crop growing seasons. Event attribution metric: Probability ratio (PR) = P_new / P_ref was computed for exceedance of the local early-20th-century 99th percentile of daily maximum temperature (TX) and higher percentiles. PRs were computed for all land and subsets (irrigated lands >10% irrigated fraction; South Asia defined as Pakistan, India, Nepal, Bangladesh). Effects of global warming (all forcings except irrigation), irrigation expansion alone, and their combination were quantified by comparing appropriate ensembles. Human exposure: Using HYDE (rural) and GPW (total) population density datasets (year 2000) remapped to the CESM grid, the number of people less exposed to hot extremes was estimated where irrigation led to PR < 0.5 for the 99th percentile TX, yielding lower and upper bounds depending on exposure assumptions.

• Observations show that TXm warmed less over regions with substantial irrigation expansion (>25% of grid cell). Pixels with increases in irrigated fraction above ~35% often experienced cooling trends. The effect is strongest over South Asia (Pakistan, India, Nepal, Bangladesh).
• The window-search reconstruction indicates irrigation expansion exerts a cooling influence on TXm that grows with irrigation extent; removing the reconstructed irrigation signal largely eliminates the negative correlation between observed warming and irrigation extent.
• CESM simulations corroborate substantial irrigation-induced cooling during hot extremes, consistent with satellite land surface temperature comparisons. While model magnitude differs, spatial patterns and sign agree with observations.
• Global warming increases the likelihood of hot extremes nearly everywhere (PR typically 2–3 for the 99th percentile TX), with larger PRs for more extreme percentiles (>99.5%, >99.75%, >99.9%).
• Irrigation expansion reduces the likelihood of hot extremes predominantly over irrigation hotspots. Over South Asia, irrigation locally reduced the likelihood (PR) of exceeding early-20th-century 99th percentile TX by factors of roughly 2–8 (i.e., making hot extremes 2–8 times less likely than without irrigation).
• Combining effects shows irrigation can partially or wholly cancel, and locally even reverse, the anthropogenic warming of hot extremes in some regions. Over irrigated lands and across South Asia, the net change in hot extreme likelihood is near zero.
• The physical basis includes increased water application during hot days and reduced land–atmosphere coupling strength, lowering temperature variability sensitivity in irrigated areas.
• Population exposure: Between approximately 0.79 and 1.29 billion people around the year 2000 were less exposed to hot extremes due to irrigation-induced cooling. Irrigated crops (>40% of global yields) likewise benefited from capped temperature extremes.
• Historical irrigation area expanded massively (e.g., ~0.63 to 3.06 million km² from 1900 to 2005), enabling the observed masking of hot extreme warming in irrigation hotspots.

The findings demonstrate that irrigation expansion has substantially masked the anthropogenic increase in hot extremes in intensively irrigated regions, particularly South Asia, thereby addressing the research question of irrigation’s role relative to other forcings. By isolating irrigation effects in observations and through targeted CESM experiments, the study shows that local land management can rival or offset greenhouse-gas-driven changes in extreme heat frequency at regional scales. This improves understanding of historical drivers of observed temperature changes and underscores the need to incorporate land management processes in climate attribution and projections. The results are consistent with other land-use findings (e.g., deforestation increasing hot extremes in mid-latitudes), indicating multiple land surface changes contribute to regional extremes. The masking effect raises implications for risk assessment: regions benefiting from irrigation-induced cooling may face rapid increases in hot extreme frequency if irrigation practices change, stagnate, or become more efficient (reducing evaporative cooling), or if water availability constraints reduce irrigation. The study emphasizes that current climate models often omit irrigation, potentially biasing projections of future heat extremes in densely populated hotspots.

Irrigation expansion during the 20th century has regionally counteracted the anthropogenic increase in hot extremes, with observations and CESM simulations jointly indicating that in irrigation hotspots—especially South Asia—the likelihood of extreme hot days was reduced sufficiently to yield little net change despite global warming. This work highlights irrigation’s unintended yet significant benefit in reducing human exposure to extreme heat and protecting crop yields in affected regions. Given uncertainties about future irrigation extent and efficiency, and the tendency of current Earth system models to neglect irrigation, the authors recommend: (1) inclusion of transient irrigation processes and areas in historical and future simulations, (2) testing climate responses under scenarios of changing irrigation extent and water-use efficiency, and (3) multi-model attribution frameworks that separate contributions from various land cover and management changes to extreme temperatures.

• Model design assumes fixed irrigation extent for early and present-day periods; transient irrigation area changes are not simulated.
• Only a generic C3 crop is represented; crop diversity and phenology differences are not captured.
• Irrigation water sources (e.g., groundwater vs. surface water) and techniques (sprinkler, drip, flood, ponding) are not differentiated; CESM applies water to the soil surface from runoff, ignoring local water availability constraints.
• Paddy water ponding is disabled, likely underestimating irrigation cooling impacts in regions with paddy fields (South, East, Southeast Asia).
• Land cover other than irrigated croplands is fixed to year 2000 for biogeophysical effects, omitting other land management/land cover change effects; only biogeochemical effects are represented via time-varying GHGs.
• The attribution window-search method focuses on local biogeophysical effects and may not capture nonlocal indirect effects (e.g., cloud/precipitation changes); it may be conservative due to predictor dependencies and spatial smoothing.
• Observational constraints: daily extremes are not directly available globally in long-term datasets; results for daily extremes rely on modeling. The study emphasizes forced responses averaged across ensemble members; local observed changes also reflect natural variability.
• Observation-based estimates show sensitivity to dataset versions and interpolation, indicating uncertainty in magnitude estimates.