Search

Abstract High-yielding traits can potentially improve yield performance under climate change. However, data for these traits are limited to specific field sites. Despite this limitation, field-scale calibrated crop models for high-yielding traits are being applied over large scales using gridded weather and soil datasets. This study investigates the implications of this practice. The SIMPLACE modeling platform was applied using field, 1 km, 25 km, and 50 km input data resolution and sources, with 1881 combinations of three traits [radiation use efficiency (RUE), light extinction coefficient (K), and fruiting efficiency (FE)] for the period 2001–2010 across Germany. Simulations at the grid level were aggregated to the administrative units, enabling the quantification of the aggregation effect. The simulated yield increased by between 1.4 and 3.1 t ha− 1 with a maximum RUE trait value, compared to a control cultivar. No significant yield improvement (

Abstract Extreme climate events can have a significant negative impact on maize productivity, resulting in food scarcity and socioeconomic losses. Thus, quantifying their effect is needed for developing future adaptation and mitigation strategies, especially for countries relying on maize as a staple crop, such as South Africa. While several studies have analyzed the impact of climate extremes on maize yields in South Africa, little is known on the quantitative contribution of combined extreme events to maize yield variability and the causality link of extreme events. This study uses existing stress indices to investigate temporal and spatial patterns of heatwaves, drought, and extreme precipitation during maize growing season between 1986/87 and 2015/16 for South Africa provinces and at national level and quantifies their contribution to yield variability. A causal discovery algorithm was applied to investigate the causal relationship among extreme events. At the province and national levels, heatwaves and extreme precipitation showed no significant trend. However, drought severity increased in several provinces. The modified Combined Stress Index (CSIm) model showed that the maize yield nationwide was associated with drought events (explaining 25% of maize yield variability). Heatwaves has significant influence on maize yield variability (35%) in Free State. In North West province, the maize yield variability (46%) was sensitive to the combination of drought and extreme precipitation. The causal analysis suggests that the occurrence of heatwaves intensified drought, while a causal link between heatwaves and extreme precipitation was not detected. The presented findings provide a deeper insight into the sensitivity of yield data to climate extremes and serve as a basis for future studies on maize yield anomalies.

Abstract This study investigates the main drivers of uncertainties in simulated irrigated maize yield under historical conditions as well as scenarios of increased temperatures and altered irrigation water availability. Using APSIM, MONICA, and SIMPLACE crop models, we quantified the relative contributions of three irrigation water allocation strategies, three sowing dates, and three maize cultivars to the uncertainty in simulated yields. The water allocation strategies were derived from historical records of farmer’s allocation patterns in drip-irrigation scheme of the Genil-Cabra region, Spain (2014–2017). By considering combinations of allocation strategies, the adjusted R 2 values (showing the degree of agreement between simulated and observed yields) increased by 29% compared to unrealistic assumptions of considering only near optimal or deficit irrigation scheduling. The factor decomposition analysis based on historic climate showed that irrigation strategies was the main driver of uncertainty in simulated yields (66%). However, under temperature increase scenarios, the contribution of crop model and cultivar choice to uncertainty in simulated yields were as important as irrigation strategy. This was partially due to different model structure in processes related to the temperature responses. Our study calls for including information on irrigation strategies conducted by farmers to reduce the uncertainty in simulated yields at field scale.

Abstract Crop yield forecasting depends on many interactive factors, including crop genotype, weather, soil, and management practices. This study analyzes the performance of machine learning and deep learning methods for winter wheat yield prediction using an extensive dataset of weather, soil, and crop phenology variables in 271 counties across Germany from 1999 to 2019. We proposed a Convolutional Neural Network (CNN) model, which uses a 1-dimensional convolution operation to capture the time dependencies of environmental variables. We used eight supervised machine learning models as baselines and evaluated their predictive performance using RMSE, MAE, and correlation coefficient metrics to benchmark the yield prediction results. Our findings suggested that nonlinear models such as the proposed CNN, Deep Neural Network (DNN), and XGBoost were more effective in understanding the relationship between the crop yield and input data compared to the linear models. Our proposed CNN model outperformed all other baseline models used for winter wheat yield prediction (7 to 14% lower RMSE, 3 to 15% lower MAE, and 4 to 50% higher correlation coefficient than the best performing baseline across test data). We aggregated soil moisture and meteorological features at the weekly resolution to address the seasonality of the data. We also moved beyond prediction and interpreted the outputs of our proposed CNN model using SHAP and force plots which provided key insights in explaining the yield prediction results (importance of variables by time). We found DUL, wind speed at week ten, and radiation amount at week seven as the most critical features in winter wheat yield prediction.