The impacts of climate change on crop yields have been extensively studied worldwide, yet the mechanisms underlying changes in yields are not fully understood.

The gap

Consensus across the literature

Supporting evidence — 5 representative gaps

• Deep learning model anticipates climate change induced reduction in major commodity crop yields for Canada in 2050 (2026) · doi

  DeepS3 does not explicitly model future changes in agricultural technology, crop genetics, management practices, or socio- economic conditions. While important factors to consider, spatially these variables do not explicit crop-specific projections of exist at the national scale considered here. Consequently, an implicit assumption is that the relationships learned between climate variables and crop yield during the historical training period remain stationary when applied to future climates and regions. Further, our projections should be interpreted as climate-driven changes in land suitability under a fixed technological and management baseline. This assumption may lead to increased uncertainty when extrapolating to novel climate regimes or emerging technological frontiers, but making incorrect assumptions about the technological and management states may exacerbate the problem. Regardless, this limitation should be considered when using these results for long-term planning. Technological innovation, improved crop genetics, or adaptive management practices could partially offset projected declines. Conversely, maladaptation, soil degradation, or increasing climate extremes not captured by seasonal averages could amplify losses, implying that projected impacts may also be underestimated in some regions. It is possible that the reliability of certain 2100 projections is partially maintained by the observation that several high- importance predictors remain closer to the historical training envelope. Across the analyzed crops, terrain slope, soil pH, and diurnal temperature (July–Sept) consistently ranked as the most influential predictors according to SHAP importance. The fact that some high importance variables are not subject to the same degree of extrapolation as early-season predictors may suggest that suitability shifts in some regions are still conditionally informed by multivariate relationships identified during training. Nevertheless, the inherent limitations of linear extrapolation in novel regimes mean that end-of-century maps should be interpreted with caution. They are best viewed as showing potential trends if today’s farming methods don’t change, rather than being definitive predictions of future production. Model evaluation in this study is based on spatial cross- validation within the historical record. Although this provides an out-of-sample test across held-out districts, we did not conduct an independent benchmarking comparison against process-based crop models (e.g., DSSAT/APSIM) or external field-trial datasets at national scale. Such benchmarking would require harmonized assumptions regarding cultivar choice and management practices and region-specific calibration data that are not uniformly available across Canada. Accordingly, our projections should be interpreted as scenario-based, data-driven estimates of climate-driven suitability change under a fixed management baseline.

  Keywords: management crop climate projections technological future practices variables historical training regions interpreted driven suitability importance
• Deep learning model anticipates climate change induced reduction in major commodity crop yields for Canada in 2050 (2026) · doi

  In light of our DeepS3-based crop yield projections, there are several strategies that could be explored to mitigate the impact of climate change on Canada’s agricultural landscape (see Figure 3).

  Keywords: light deeps based crop yield projections there several strategies explored mitigate impact climate change canada
• Will temperature and rainfall changes prevent yield progress in Europe? (2022) · doi

  However, an understanding of the effects of changes in temperature and rainfall throughout the crop cycle on historical yield progress is lacking in Europe (EU).

  Keywords: understanding effects changes temperature rainfall throughout crop cycle historical yield progress lacking europe
• Design and Implementation of a Novel Smart Agriculture’s Irrigation System Integrating IoT Technologies (2026) · doi

  Future research will focus on long-term experimental validation over several seasons and crop diversities to find out statistical relevance and generalizability.

  Keywords: future focus long term experimental validation several seasons crop diversities find statistical relevance generalizability
• Impact of Climate Change on Crop Pests and Diseases: Ensemble Modeling of Time-Varying Weather Effects (2023) · doi

  The impacts of climate change on crop yields have been extensively studied worldwide, yet the mechanisms underlying changes in yields are not fully understood.

  Keywords: yields impacts climate change crop extensively studied worldwide mechanisms underlying changes fully understood