Daily reference crop evapotranspiration (ET₀) is crucial for precision irrigation management, yet traditional prediction methods struggle to capture its dynamic variations due to the complexity and nonlinearity of meteorological conditions. To address this, we propose an Improved Informer model to enhance ET₀ prediction accuracy, providing a scientific basis for agricultural water management. Using meteorological and soil data from the Yingde region, we employed the Maximal Information Coefficient (MIC) to identify key influencing factors and integrated Residual Cycle Forecasting (RCF), Star Aggregate Redistribute (STAR), and Fully Adaptive Normalization (FAN) techniques into the Informer model. MIC analysis identified total shortwave radiation, sunshine duration, maximum temperature at 2 m, soil temperature at 28–100 cm depth, and surface pressure as optimal features. Under the five-feature scenario (S3), the improved model achieved superior performance compared to Long Short-Term Memory (LSTM) and the original Informer models, with MAE reduced to 0.065 (LSTM: 0.637, Informer: 0.171) and MSE to 0.007 (LSTM: 0.678, Informer: 0.060). The inference time was also reduced by 31%, highlighting the enhanced computational efficiency. The Improved Informer model effectively captures the periodic and nonlinear characteristics of ET₀, offering a novel solution for precision irrigation management with significant practical implications.

Rice blast, caused by Magnaporthe oryzae, is a devastating disease that jeopardizes global rice production and food security. Precision agriculture demands timely and accurate monitoring tools to enable targeted intervention. This study introduces a novel deep learning framework that fuses time-series hyperspectral imaging with an advanced Autoformer model (AutoMSD) to dynamically track rice blast progression. The proposed AutoMSD model integrates multi-scale convolution and adaptive sequence decomposition, effectively decoding complex spatio-temporal patterns associated with disease development. When deployed on a 7-day hyperspectral dataset, AutoMSD achieved 86.67% prediction accuracy using only 3 days of historical data, surpassing conventional approaches. This accuracy at an early infection stage underscores the model’s strong potential for practical field deployment. Our work provides a scalable and robust decision-support tool that paves the way for site-specific disease management, reduced pesticide usage, and enhanced sustainability in rice cultivation systems.

Rice blast caused by Pyricularia oryzae progresses from symptom-free leaves to extensive tissue death, while neighboring severity grades often show similar spectra. This study developed XGBSVMRF-SA to integrate regional spectral-angle (SA) features with heterogeneous stacking of XGBoost 3.1.2, support vector machine, and random forest class probabilities. The analysis included 833 spectra from 40 recorded pot groups covering six severity grades (0–5). The 2151-band spectra underwent Savitzky–Golay smoothing and second-derivative preprocessing. Twelve near-infrared and shortwave-infrared SA features and 18 class probabilities formed a 30-dimensional logistic-regression input. Group-balanced grade centers and grouped out-of-fold predictions were generated within training partitions, without resampling or threshold adjustment. In matched grouped five-fold validation, XGBoost achieved 90.88% accuracy and 90.31% macro F1. Heterogeneous stacking increased these values to 91.34% and 90.61%; adding SA further increased them to 92.07% and 91.46%. The complete framework gained 1.19 and 1.16 percentage points over XGBoost, including SA-specific increments of 0.72 and 0.85 percentage points over the matched stack. Local perturbations of spectral-region boundaries produced a macro F1 span of 0.14 percentage points. These results support regional SA as a supplementary spectral-geometric input alongside class probabilities in heterogeneous stacking for quantitative rice blast grading.