Crop yield prediction helps to enhance the stability of agricultural product supply and promote sustainable agricultural development, both of which are crucial for food production and security. To develop simple yet highly accurate crop yield prediction models, this study proposed a spring- and summer-maize yield prediction model based on the deep hybrid kernel extreme learning machine (DHKELM) algorithm. In this study, four tree-based feature importance analysis algorithms, including classification and regression tree, gradient boosting decision tree, random forest, and extreme gradient boosting algorithms, were utilized to analyze the importance of the factors affecting the yield of spring and summer maize. Then, based on the analysis of the four algorithms, different combinations of factors were established to obtain the optimal combination of features. Moreover, to improve the prediction accuracy of the machine learning model, this study utilized three optimization algorithms, including the bald eagle search algorithm, chaos game optimization (CGO) algorithm, and carnivorous plant algorithm, to optimize the hyperparameters in the DHKELM algorithm. The results of the study showed that planting density and plant height were important factors affecting maize yield, and net solar radiation (R_n) received during the reproductive period exhibited the highest relative importance. Appropriate feature combinations can effectively improve model prediction accuracy. The optimal feature combination for spring maize included planting density, plant height, R_n, mean temperature (T_mean), minimum temperature (T_min), and cumulative temperature, and the optimal feature combination for summer maize included R_n, plant height, planting density, T_min, and T_mean. Among the three optimization algorithms, the CGO algorithm exhibited the best optimization effect and could significantly improve the prediction accuracy of the DHKELM algorithm. When the optimal combination of features was used as input, the CGO–DHKELM model used for maize yield prediction provided the following values: RMSE=1.488 t/hm², R²=0.862, MAE=1.051 t/hm², and NSE=0.852 for spring maize; RMSE=1.498 t/hm², R²=0.892, MAE=1.055 t/hm², and NSE=0.891 for summer maize. Thus, the findings of the study provide a reference for high-precision prediction of spring and summer maize yields in China.

Evapotranspiration (ET) is a key component of the water cycle, and accurate estimation of reference crop evapotranspiration (ET_o) is essential for irrigation management. To build a precise, lightweight ET_o estimation model, this study takes key meteorological factors as inputs and applies machine learning models and hybrid models Crested Porcupine Optimizer kernel extreme learning machine (CPO-KELM), the Dung Beetle Optimizer Algorithm KELM (DBO-KELM), and Particle Swarm Optimization KELM (PSO-KELM) to estimate ET_o at 38 meteorological stations across China’s seven major geographical regions. The results indicate that maximum temperature (T_max), average temperature (T_ave), and relative humidity (RH) are the primary factors affecting ET_o and were therefore used as model inputs. The standalone kernel extreme learning machine (KELM) model shows acceptable ET_o estimation performance, with R², RMSE, MAE, and NSE ranging from 0.802-0.885, 0.512-0.911, 0.464-0.970, and 0.802-0.885, respectively. Hybrid models outperform the standalone KELM, among which CPO-KELM is the most accurate: its R², RMSE, MAE, and NSE range from 0.881-0.942, 0.413-1.147, 0.284-0.763, and 0.881-0.942. At the regional scale, the CPO-KELM model exhibits its best performance in the Northeast and North China regions, with R², RMSE, MAE, and NSE ranging from 0.923-0.936, 0.413-0.511, 0.284-0.358, and 0.923-0.936, respectively. In contrast, its weakest performance is observed in parts of South China and Northwest China, with R², RMSE, MAE, and NSE ranging from 0.881-0.905, 0.675-1.147, 0.506-0.763, and 0.881-0.905. Compared to standalone KELM, CPO-KELM improves accuracy significantly: R² and NSE rise by 6.4%-9.9%, while MAE drops by 21.3%-38.8%. Thus, the hybrid CPO-KELM model effectively enhances ET_o estimation accuracy across China’s regions. Therefore, the proposed CPO-KELM hybrid model provides a high-accuracy and lightweight alternative for ET_o estimation in data-scarce regions, and offers reliable technical support for intelligent water resources management and irrigation optimization across diverse climatic zones in China.

Accurate detection of tomato leaf diseases is crucial for early prevention and ensuring agricultural production. This study addresses six tomato leaf diseases: bacterial spot, early blight, late blight, leaf mold, septoria leaf spot, and yellow leaf curl virus. A lightweight detection model, YOLO-LGS, is proposed to achieve efficient and automated disease detection. The dataset of tomato leaf diseases was first augmented to enrich the disease features, thereby improving the model’s detection performance. The YOLO-LGS model is built on the YOLOv11 architecture, incorporating lightweight group attention net (LWGANet) to reconstruct the backbone network, replacing the convolutional block with parallel spatial attention mechanism with the grouped channel-wise self-attention (GCSA) mechanism, and introducing separated and enhanced attention module (SEAM) into the detection head to balance performance and efficiency. Experimental results show that the YOLO-LGS model achieves an mAP50 of 0.693 and an F1 score of 0.677, outperforming other YOLO models (YOLOv8s, YOLOv9s, YOLOv10s, and YOLOv11s). Additionally, the model’s parameter size is only 6.333 M, and its GFLOPs is 13.4, representing reductions of 32.739% and 37.089%, respectively, compared to YOLOv11s, significantly lowering computational cost while maintaining detection performance. The results demonstrate the effectiveness of LWGANet, GCSA, and SEAM. The development of the YOLO-LGS model provides an efficient, lightweight solution for tomato leaf disease detection in resource-constrained environments.