Silver lining to a climate crisis in multiple prospects for alleviating crop waterlogging under future climates

The study addresses the underexplored impact of soil waterlogging on global crop production under climate change. While most prior assessments emphasize drought, heat, or gradual warming, flooding and waterlogging increasingly threaten agricultural systems and food security. The authors aim to: (1) improve process-based crop modeling of waterlogging effects on photosynthesis and phenology, (2) quantify how climate change alters the frequency and severity of waterlogging across global barley production zones, (3) identify common, phenology-based waterlogging stress patterns via clustering to enable transferability of adaptation strategies, and (4) evaluate how management (sowing time) and genetics (waterlogging-tolerant, spring vs. winter genotypes) mitigate waterlogging-induced yield penalties under future climates (SSP585 for 2040 and 2080). This work is important for designing credible, scalable adaptations that reduce risk in a future with more frequent and compound extremes.

The paper notes that approximately 27% of cultivated land experiences flooding annually, with substantial economic damage. Despite this, global assessments of waterlogging impacts on crops are limited. Most climate impact studies have focused on drought and heat stress, or on gradual climate changes. Existing G×E×M studies often lack scalability and transferability. Prior modeling typically treats waterlogging as a function of soil saturation without adequately accounting for effects on phenology. There is a recognized need to integrate nonlinear, process-based responses in models to guide robust adaptation. The authors build on prior controlled-environment and field experiments on barley waterlogging responses and on advances in process-based modeling (e.g., APSIM), while addressing gaps in representing phenological delays/truncation and in generalizing stress typologies across environments.

Model development and calibration: The authors extended APSIM-Barley (v7.9) to include new waterlogging stress response functions affecting both photosynthesis (oxdef_photo) and phenology (oxdef_pheno), each computed as a function of the fraction of roots waterlogged and crop stage. Stress multipliers range from 1 (no stress) to 0 (full stress). Stage-dependent responses include phenological delays under pre-flowering waterlogging and grain-filling truncation when stress occurs post-flowering. A genotype-specific limiter (y_oxdef_lim_photo) allows representation of adaptive recovery linked to waterlogging tolerance (e.g., via aerenchyma formation). These processes were implemented in APSIM and optimized via multi-objective minimization of squared residuals. Experimental datasets: The modified model was calibrated/evaluated using measured barley data from five two-year experiments across Australia, Argentina, Canada, China, and Ireland, encompassing greenhouse, controlled-environment, and field studies with waterlogging imposed at tillering or via controlled flooding/irrigation. Yield units were standardized to kg ha−1 and yield loss computed relative to controls. Sites and simulation setup: Factorial simulations covered major barley production regions across 13 countries, prioritizing representative soils from the FAO/ISRIC Digital Soil Map. Soil properties (texture, bulk density, pH, SOC) were sourced from ISRIC; groundwater depths from global datasets. APSIM’s SWIM3 soil water module was used. Management included sowing density 180 plants m−2, depth 20 mm, row spacing 200 mm. Nitrogen was maintained above 200 kg N ha−1 in the top 0–30 cm to avoid N limitation and isolate waterlogging effects. Initial plant-available water at sowing was set to 15 mm, and soil conditions were reset annually. Genotypes and sowing: Two lifecycle types were simulated per site: short-season spring and long-season winter genotypes, each with relatively early and late sowing windows per local practice. Phenology parameters (vern_sens, photop_sens, tt_end_of_juvenile) were tuned to match local flowering/maturity timings. Waterlogging tolerance genotypes were also created in silico, parameterized from empirical studies to reflect tolerance to hypoxia/anoxia. Climate data and downscaling: Historical daily TMAX/TMIN, precipitation, and radiation (1985–2016) came from NASA/POWER (1° resolution). Future scenarios used CMIP6/AR6 projections from 27 GCMs under SSP585 for 2030–2059 (2040s) and 2070–2099 (2080s). Monthly GCM outputs were bias-corrected and downscaled to sites using NWAI-WG with inverse distance weighted interpolation and a stochastic weather generator to produce daily series while preserving inter-variable dependencies. CO2 concentrations followed SSP585 trajectories estimated via nonlinear regression across GCMs. Stress typology clustering: Seasonal time series of oxdef_photo were discretized by six phenological phases: early juvenile (JV1), late juvenile (JV2), floral initiation to heading (FIN), flowering to early grain filling (FWR/GF1), early grain filling (GF1), and late grain filling (GF2). For each season-site-genotype-management, waterlogged days (oxdef_photo<1) and mean oxdef_photo were computed by phase. Unsupervised k-means clustering (k=4) grouped trajectories into common seasonal stress patterns for spring (SW0–SW3) and winter (WW0–WW3) barley, minimizing within-cluster variance. Variance explained was assessed for k=4 vs. k=5. Yield impact quantification: For each year and site, yields from default APSIM (without phenology-aware WL stress) were compared with the improved APSIM to compute waterlogging-induced yield penalties (%). Additional analyses assessed changes under future climates, sowing times, and waterlogging-tolerant genotypes, including focus on wettest years (growing-season rainfall >90th percentile). Model performance was evaluated via R2 and RMSE against observed yield losses.

- Model improvement: Incorporating phenology- and photosynthesis-based waterlogging stress substantially improved APSIM performance, increasing R2 from 0.46 to 0.70 and reducing RMSE for waterlogging yield loss from 0.30 to 0.11 across 36 genotype–environment–treatment observations. - Waterlogging under future climates: Severe waterlogging risk increases by 2–10% under future climate (across sites, sowing dates, and GCMs). Despite this, CO2 fertilization and alleviation of cold stress at higher latitudes lead to modest yield increases in many regions when water is not limiting. - Yield penalties due to waterlogging: Accounting for waterlogging reduces simulated future yields by 8–18% in the 2040s and 17–26% in the 2080s compared with simulations that ignore physiological WL effects. Historically, waterlogging penalties averaged 3–11%, rising to 6–14% (2040s) and 10–20% (2080s), varying by genotype and management. - Spring vs. winter barley: Global average WL yield penalties for winter barley were 11% (historical), 14% (2040s), 20% (2080s), with median penalties 130–591 kg ha−1. For spring barley, penalties were 3% (historical), 6% (2040s), 10% (2080s), with median 50–91 kg ha−1. Winter types suffer more due to longer WL duration, even though cereals are most sensitive during reproductive stages. - Sowing-time effects: Spring barley yields increased under future climates for many regions and both sowing times, with relatively limited effect of sowing time on WL severity pattern. For winter barley, earlier sowing reduced frequencies of both mild and severe early-season WL (WW1, WW3), whereas later sowing increased early-onset WL risks. - Stress pattern clustering: Four clusters explained 71% (spring) and 80% (winter) of variance (k=4), with k=5 adding marginal gain. Winter genotypes predominantly experienced early-season WL (juvenile phase, WW3), while spring genotypes experienced late-season WL (SW2–SW3). Under future climates, frequencies of WW1 (7%→17% under early sowing) and WW3 (3%→8% under late sowing) increased, especially in France, UK, Russia, and China. - Adaptation benefits: Combining altered sowing time with waterlogging-tolerant genetics reduced WL penalties by about 18% on average. Waterlogging-tolerant genotypes increased yields by ~14% (early sowing) and ~18% (late sowing) in the 2040s (s.d. 23% and 34%), with similar gains in the 2080s; gains were larger for winter types (≈480–620 kg ha−1) than spring types (≈194–213 kg ha−1). Benefits were greatest in longer, cooler temperate regions (UK, France, Russia, China) and smaller in drier, high ETo/P regions (e.g., Australia). Notably, tolerant genotypes did not reduce yields in drier years and reduced downside risk. - Serendipitous stability of stress patterns: Temporal WL stress patterns relative to phenology are broadly similar between historical and future climates, enabling development and transfer of today’s adaptations to tomorrow’s conditions (e.g., late-season WL tolerance for spring types; early-season WL tolerance for winter types).

The improved, process-based representation of waterlogging in APSIM, including phenology effects, enabled robust quantification of WL impacts and identification of seasonally distinct stress patterns across diverse environments. The findings address the research goals by demonstrating that: (1) WL-induced yield penalties will generally increase under future climates, especially for long-season winter barley, due to higher frequency and duration of early-season WL; (2) despite greater WL risk, overall yields can improve in many regions from CO2 fertilization and milder cold stress; (3) adaptation via earlier sowing (especially for winter types) and adoption of WL-tolerant genotypes substantially mitigates penalties; and (4) clustering of stress trajectories provides a transferable framework to match adaptations to the most probable stress patterns per environment and genotype lifecycle. The similarity between historical and future WL stress patterns is crucial because it allows breeding and agronomy programs to target relevant stress timing (early vs. late) using existing environments as analogues, reducing the need for synthetic stress environments. Regionally, higher WL risks and benefits from tolerance are concentrated in temperate, wetter areas (Europe, parts of China), while drier regions may see less WL risk and smaller gains from WL tolerance. The results underscore the need for contextualized, system-level adaptation portfolios that integrate management timing and genetics, with particular attention to vulnerable smallholder systems where flooding frequency is projected to increase.

This study advances crop-climate impact assessment by embedding phenology-aware waterlogging processes into APSIM and by introducing a generalizable clustering method to categorize seasonal stress patterns across environments, managements, and genotypes. Globally, WL-induced yield penalties in barley are projected to rise from historical levels (3–11%) to 6–14% by the 2040s and 10–20% by the 2080s, with winter types more affected than spring types. However, adaptation strategies—especially earlier sowing of winter barley and adoption of WL-tolerant genotypes—can reduce penalties by roughly 18%, with the largest benefits in long-season temperate regions. Key contributions include: (1) improved process representation of WL impacts on photosynthesis and phenology; (2) a stress-typology framework that explains most variance in seasonal WL patterns and enables transferability of adaptations; and (3) a contextualized adaptation roadmap showing where WL tolerance and sowing shifts will be most impactful. Future research should expand to multi-model crop ensembles, incorporate nutrient dynamics and biotic pressures, refine local genotype parameters for WL tolerance, and explore synergistic adaptation stacking across crops and regions to enhance robustness under increasing climate variability.

- Modeling framework: Only one crop model (APSIM) was used; multi-model ensembles could provide more robust central estimates and uncertainty bounds. - Climate data uncertainty: Precipitation projections are particularly uncertain. Downscaling using NWAI-WG and IDW may introduce biases; gridded datasets can produce more frequent, lower-intensity rainfall events affecting runoff and evaporation. - Scope of stresses: Nitrogen stress and biotic factors (diseases, pests) were not simulated; interactions between WL and nutrient dynamics (e.g., N uptake inhibition) were excluded, potentially underestimating WL impacts. - Parameterization: WL tolerance and phenology parameters were derived from prior studies; site-specific genotype parameters could improve fidelity. - Generalizability: Results focus on barley; while the framework is transferable, application to other crops requires calibration/validation. - Variability and interpretability: Using 27 GCMs increases projection spread, elevating uncertainty and complicating interpretation of local outcomes.