When drought strikes a field of maize or wheat, the plants do far more than simply slow down. Beneath the soil surface, they begin rerouting the sugars produced by photosynthesis, shifting carbon away from leaves and toward roots in a desperate bid to find more water. This hidden reallocation of resources, known as photosynthate allocation, has long been one of the weakest links in the crop models that scientists rely on to forecast harvests and soil carbon storage. Now, a team of researchers at the Institute of Atmospheric Physics of the Chinese Academy of Sciences has rebuilt this core component of the Agro-C model, and the results suggest that getting this one physiological detail right can dramatically sharpen predictions of how crops fare when water runs short.
The study, published in Geoscientific Model Development by Wei Xu, Wen Zhang, Yongqiang Yu, and Qing Zhang, tackles a stubborn limitation shared by most crop simulation frameworks. Models such as APSIM, DSSAT, CropSyst, and AquaCrop typically describe carbon allocation using fixed coefficients or rules tied only to developmental stage. In APSIM, for example, carbon flows to leaves during early vegetative growth and then progressively shifts to stems, reproductive organs, and grains, regardless of what the soil is doing. In the CASA and DNDC models, the root-to-shoot ratio varies only by vegetation type. These static schemes capture the broad arc of a growing season, but they miss the physiological plasticity that real plants display when soil moisture fluctuates from week to week.
The consequences of that simplification are not trivial. Water availability regulates photosynthesis, biomass production, and the distribution of assimilated carbon among plant organs, and it does so differently at each stage of development. Under moderate drought, crops often send more carbon below ground, thickening their root systems to improve water uptake. Under severe deficit, however, carbon assimilation itself collapses, suppressing both aboveground and belowground biomass and ultimately limiting yield. Wheat, a C3 crop, tends to enhance root allocation under mild drought but inhibits root development under severe stress, while maize, a C4 crop, follows its own distinct pattern. A model blind to these dynamics cannot hope to capture the carbon trade-offs that determine both food production and the organic matter returned to the soil.
The team’s solution rests on two interlocking innovations. The first is a water-driven, stage-dependent allocation scheme that dynamically adjusts the fraction of photosynthate directed to roots and to leaves as soil moisture changes. The second is a process-based root water uptake module that replaces the crude supply-demand water stress factors used elsewhere. Rather than simply comparing potential transpiration with actual uptake, the new module computes the water potential gradient between soil and root xylem using the van Genuchten water retention function, then multiplies that gradient by the root hydraulic conductivity, which itself depends on root dry mass, specific root surface area, and stage-specific conductance values drawn from the literature. This means the model’s sense of drought is grounded in the actual physics of water moving through the soil-plant-atmosphere continuum.
The researchers also confronted a subtler symptom of drought: leaf aging. Water stress accelerates leaf senescence, abscission, and mortality, shrinking the photosynthetic engine precisely when the plant needs it most. To represent this, the team introduced a dynamically adjusted yellow-to-green leaf ratio, calibrated against the water deficit at each developmental stage. The crop development index, which runs from zero at emergence to two at maturity, was divided into fine intervals, and within each interval the adjustments to root allocation, leaf allocation, and leaf senescence were expressed as empirical functions of the water balance difference between crop demand and actual uptake. The result is a model in which carbon allocation responds continuously to both the calendar of development and the state of the soil.
Building and testing such a model demanded an unusually rich dataset. The team compiled 608 observations of aboveground and belowground biomass for maize and wheat from 21 published studies, spanning 198 observations at maturity and 411 at earlier stages. Twelve studies, covering 214 maize and 164 wheat records across multiple growth stages, served for calibration, while nine studies with 101 maize and 129 wheat records were reserved for validation. Each record had to include irrigation amounts or soil moisture conditions, fertilizer inputs, and detailed site information, with meteorological data drawn from the ERA5 reanalysis. This breadth allowed the calibration to capture allocation behavior across the full spectrum from severe drought to waterlogging.
The validation results are striking. For maize, simulation accuracy for aboveground biomass improved from a coefficient of determination of 0.70 with a root mean squared error of 153.15 grams of carbon per square meter to 0.95 with an error of just 66.03 grams. Belowground biomass for maize improved from 0.49 to 0.61, with the error nearly halved from 20.55 to 11.18 grams of carbon per square meter. Wheat showed equally dramatic gains: aboveground biomass rose from a coefficient of 0.31 to 0.74, and belowground biomass from 0.42 to 0.48, with the pre-optimized model faltering precisely in the water-stress treatments that dominate the validation set. Across all soil moisture regimes, the optimized model achieved coefficients between 0.77 and 0.95 for maize aboveground biomass and 0.80 to 0.89 for wheat, with belowground values reaching 0.62 to 0.83.
Perhaps the most scientifically interesting findings concern how differently the two crops spend their carbon under identical stress. During the reproductive transition, maize under optimal moisture invests heavily in leaves, sustaining leaf area and prolonging photosynthesis to support pollination and kernel formation. Wheat, by contrast, steadily diverts carbon away from leaves toward the developing spike, relying on remobilization of stem reserves during grain filling while its root system senesces rapidly. At maturity, maize root allocation actually declines under extreme drought because premature root senescence sets in, whereas wheat accelerates the transfer of assimilates from roots to grains. The study also found that excessive moisture is not harmless: waterlogged soils raised the yellow-to-green leaf ratio at maize maturity and weakened root activity through poor aeration, disrupting allocation in ways that impair yield.
These improvements matter beyond the field trial. Agro-C is a coupled crop-soil carbon model in which simulated crop residues and root biomass feed directly into the soil carbon module as carbon inputs. By sharpening the estimate of belowground biomass, the optimized model reduces a major source of uncertainty in projections of soil organic carbon storage, a critical lever in climate mitigation accounting. The authors note that their framework offers a more mechanistic alternative to the threshold-based functions of AquaCrop or the empirical allocation rules of APSIM, positioning carbon allocation as a living, moisture-sensitive process rather than a lookup table.
Limitations remain, and the authors are candid about them. Nitrogen interactions with water are not yet explicitly coupled in the allocation scheme, root architecture and turnover are simplified, and the model does not distinguish among drip, sprinkler, and flood irrigation strategies. Belowground simulations still lag aboveground ones, particularly during late reproductive stages when root biomass is small and even minor deviations depress the coefficient of determination. Yet even there, root mean squared errors of roughly 8 grams of carbon per square meter indicate the model remains practically reliable. As droughts intensify and freshwater resources tighten worldwide, tools that can faithfully simulate how crops negotiate the trade-off between feeding their leaves and feeding their roots will become indispensable for planning irrigation, forecasting yields, and managing the carbon that agriculture returns to the earth.
Subject of Research: Dynamic water-stress-driven photosynthate allocation modeling in the Agro-C crop model for maize and wheat
Article Title: Improving photosynthate allocation dynamic simulations of crops under water stress conditions
Article References: Xu, W., Zhang, W., Yu, Y., & Zhang, Q. (2026). Improving photosynthate allocation dynamic simulations of crops under water stress conditions. Geoscientific Model Development, 19(20), 9973-9991. https://doi.org/10.5194/gmd-19-9973-2026
Image Credits: AI Generated
Keywords: Agro-C model, photosynthate allocation, water stress, drought, maize, wheat, root water uptake, carbon allocation, soil organic carbon, crop modeling, leaf senescence, Geoscientific Model Development
News Source: Alan Morgan. (October 19, 2026). Smarter Crop Model Tracks How Drought Shifts Carbon Between Roots and Leaves. Scienmag.
