Climate change significantly alters plant phenology, impacting crop yields. Rising temperatures accelerate phenological development, potentially shortening growing seasons. However, farmers adapt by adjusting sowing dates and cultivar choices, influencing crop phenology and yield. This study addresses the underrepresentation of farmers' adaptive behavior in crop-model-based climate change yield assessments. The research question focuses on quantifying the impact of farmers' adaptive management on future global crop yields under climate change. Understanding this interaction is crucial for accurate climate change impact assessments and for developing effective adaptation strategies to ensure global food security. The study's purpose is to integrate farmers' adaptive decision-making into global-scale crop models to project future yields under different climate scenarios. The importance of this work lies in its potential to refine climate change impact assessments and inform policy decisions related to agricultural adaptation and food security.

Previous research has highlighted the impact of climate change on plant phenology, showing earlier and faster phenological progress in both wild and cultivated plants. Faster-growing cycles associated with higher temperatures are considered a main mechanism of climate change impacts on crop yields. However, the phenology of annual crops is also influenced by farmers' decisions on sowing dates and varietal choice. While some studies have incorporated adaptive cultivar selection to maintain the length of growing periods, they often neglect farmers' adjustments of sowing dates or only consider static crop calendars. The current understanding of farmers’ adaptation to climate change and its impact on future crop yields at a global scale is limited by insufficient data on management practices, phenological phases, and crop development parameters. Existing studies have begun exploring decision-making rules simulating dynamic farming practices, finding that climate is the primary driver of farmers' choices, yet these lack global-scale assessments under climate change.

This study combined two rule-based methods to simulate farmers' crop calendar selection worldwide. These methods, primarily driven by crop physiology and long-term changes in monthly temperature, precipitation, and seasonality, were improved and evaluated against observed crop calendars using observation-based climate data. Projections of location-specific sowing and maturity dates were derived for five crops (maize, rice, sorghum, soybean, wheat) and two water management regimes (rainfed and irrigated) for historical (1986–2005) and two future (2060–2079, 2080–2099) climatic periods. Thermal unit requirements (TUreq, °C day) between sowing and maturity dates were computed to represent cultivars adapted to these calendars. The process-based global gridded crop model LPJmL was used to simulate annual crop growing periods and yields from 1986 to 2099 under a reference scenario assuming no management action in response to climate change. Counterfactual scenarios included no adaptation, timely adaptation, and delayed adaptation of sowing dates and cultivars. The sensitivity of yields to individual measures (adapting sowing dates or cultivars) was also tested. Simulations considered single crop cycles per year, excluding multi-cropping systems. Multiple General Circulation Models (GCMs) (HadGEM2-ES, GFDL-ESM2M, IPSL-CMSA-LR, MIROCS) were used to assess robustness against climate projection uncertainties. The rule-based model distinguishes between spring and winter wheat types, enhancing simulation accuracy in warmer climates where winter wheat may not go dormant. The model determines sowing dates based on temperature or precipitation seasonality, while maturity dates are optimized to avoid adverse conditions like heat stress or water limitations. The LPJmL model simulates daily crop phenology and yields, integrating these simulated crop calendars with climate projections and other management inputs (irrigation, fertilization, tillage, and crop residue management) from existing datasets.

Simulations revealed that timely adaptation to future climate significantly shifts temperature-driven sowing dates in extratropical regions (>30°N–S). Spring crops showed advancing sowing dates, while winter wheat sowing dates were postponed to avoid frost damage. Precipitation-driven sowing dates in the tropics showed minimal change. Adapted maturity dates were generally later than in the no-adaptation scenario, except in water-limited regions. Cultivars adapted to future climate generally required accumulating more thermal units to reach maturity. Globally, timely adaptation increased crop yields by 12% (GCM range 9–15%) compared to no adaptation. The largest yield gains were for maize (+17%) and rice (+17%), with the smallest for wheat (+7%). Adaptation benefits were larger under warmer climate scenarios. The positive CO₂ fertilization effect, more pronounced for C₃ crops, was further enhanced by adaptation. Adaptation generally decreased yield variability for some crops but increased it for others, although it never exceeded that of the reference period. Timely adaptation proved more effective than delayed adaptation, particularly for maize, rice, and sorghum. Spatial patterns showed the largest yield gains (>30%) in high-latitude temperate regions, where warming extends the growing season. Significant benefits (10–30%) were also found in tropical regions and major breadbasket regions. The relatively small effectiveness of adaptation in mid-latitude regions was attributed to the less effective adaptation of wheat.

The findings demonstrate that neglecting farmers' adaptive management practices leads to an overestimation of climate change's negative impacts on crop yields. The integration of agro-climatic rule-based approaches for projecting farmers' decisions into future scenarios highlights the sizeable effect of growing-period adaptation. This adaptation not only avoids negative impacts but also enhances yields of crops already benefiting from climate change, supporting increased food supply for a growing population. The study's global-scale assessment, accounting for locally adapted growing seasons, provides a significant improvement over previous approaches relying on static historical management practices. The study underscores the need to integrate dynamic management decisions into global models of agricultural systems, improving the accuracy of future crop yield projections. While some adaptation will be direct (e.g., changing sowing dates), the introduction of new cultivars requires collaboration between farmers, breeders, and markets.

This study demonstrates the significant potential for increasing global crop yields by incorporating farmers' adaptive management strategies into climate change impact assessments. Timely adaptation of sowing dates and cultivars can significantly mitigate the negative effects of climate change and enhance the positive impacts of CO2 fertilization. The findings highlight the importance of considering adaptive behavior in future crop yield projections, suggesting a need for further research integrating biophysical and socioeconomic factors to improve model accuracy and inform sustainable agricultural practices. Future work should focus on improving the representation of extreme climate events and incorporating photoperiod responses into crop models. Additionally, integrating socioeconomic factors like land-use change and technological progress is crucial for a more comprehensive understanding of adaptation potentials.

The study's limitations include the exclusion of multi-cropping systems and crop rotations, simplifying the complexity of real-world agricultural practices. The model assumes a single crop cycle per year and does not consider the potential impact of intermediate phenological phases, like flowering, on yield. The LPJmL model's representation of extreme climate impacts, such as heat stress, frost damage, and waterlogging, could also be improved. Furthermore, using air temperature instead of canopy temperature might introduce biases, particularly under water stress. The lack of photoperiod responses in the model might have led to an oversensitivity of crop growing periods to temperature increases. While the model incorporates many aspects of farmers' adaptation, other factors influencing their decisions are not considered.