Climate change and variability threatens the sustainability of future food productions, especially in semi-arid regions where water resources are limited, and irrigated agriculture is widespread. Increasing temperatures will exacerbate evaporative losses and increase plant water needs. Consequently, higher irrigation intensities would be a logical measure to mitigate climate change impacts in these regions. Using an ensemble of well-parameterized crop model simulations, we show that this mitigation measure is oversimplified and that besides water resources availability, strong temperature increases play a crucial role in crop developments and resulting plant water needs. Our analysis encompasses agricultural areas of the Lower Chenab Canal System in Pakistan (15 000 km2), which is part of the Indus River irrigation system, the largest irrigation system in the world; and covers economically important crop growing areas (e.g., of cotton, rice and maize crops). Climate models project an above average increase in temperature over the study region, and the agro-hydrological and biophysical crops models respond with a strong decline of up to -24% (±12%) in future crop productions. Our modeling results further suggest that evaporative and irrigation demands do not align with increasing future temperature trends. The resulting decline in crop productions is consistent among model projections despite an intensification of irrigation measures and the positive effect of future CO2 enrichments. Overall, our study emphasizes the role of elevated temperature stress, its effects on agricultural production as well as water demand, and its implications for climate change adaption strategies to mitigate adverse impacts in an intensively irrigated region.

There is still limited understanding of how waters mix, where waters come from and for how long they reside in tropical catchments. In this study, we used a tracer-aided model (TAM) and a gamma convolution integral model (GM) to assess runoff generation, mixing processes, water ages and transit times (TT) in the pristine humid tropical rainforest Quebrada Grande catchment in central Costa Rica. Models are based on a four-year data record (2016 to 2019) of continuous hydrometric and stable isotope observations. Both models agreed on a young water component of fewer than 95 days in age for 75% of the study period. The streamflow water ages ranged from around two months for wetter years (2017) and up to 9.5 months for drier (2019) years with a better agreement between the GM estimated TTs and TAM water ages for younger waters. Such short TTs and water ages result from high annual rainfall volumes even during drier years with 4,300 mm of annual precipitation (2019) indicating consistent quick near-surface runoff generation with limited mixing of waters and a supra-regional groundwater flow of likely unmeasured older waters. The TAM in addition to the GM allowed simulating streamflow (KGE > 0.78), suggesting an average groundwater contribution of less than 40% to streamflow. The model parameter uncertainty was constrained in calibration using stable water isotopes (δ2H), justifying the higher TAM model parameterization. We conclude that the multi-model analysis provided consistent water age estimates of a young water dominated catchment. This study represents an outlier compared to the globally predominant old water paradox, exhibiting a tropical rainforest catchment with higher new water fractions than older water.