Leveraging Spectral Neighborhood Information for Corn Yield Prediction with Spatial-Lagged Machine Learning Modeling: Can Neighborhood Information Outperform Vegetation Indices?

AI, Journal Year: 2025, Volume and Issue: 6(3), P. 58 - 58

Published: March 13, 2025

Accurate and reliable crop yield prediction is essential for optimizing agricultural management, resource allocation, decision-making, while also supporting farmers stakeholders in adapting to climate change increasing global demand. This study introduces an innovative approach by incorporating spatially lagged spectral data (SLSD) through the spatial-lagged machine learning (SLML) model, enhanced version of spatial lag X (SLX) model. The research aims show that SLSD improves compared traditional vegetation index (VI)-based methods. Conducted on a 19-hectare cornfield at ARS Grassland, Soil, Water Research Laboratory during 2023 growing season, this used five-band multispectral image 8581 measurements ranging from 1.69 15.86 Mg/Ha. Four predictor sets were evaluated: Set 1 (spectral bands), 2 bands + neighborhood data), 3 VIs), 4 top VIs data). These evaluated using SLX model four decision-tree-based SLML models (RF, XGB, ET, GBR), with performance assessed R2 RMSE. Results showed (Set 2) outperformed VI-based approaches 3), emphasizing importance context. models, particularly RF, performed best 4–8 neighbors, excessive neighbors slightly reduced accuracy. In 3, improved predictions, but smaller subset (10–15 indices) was sufficient optimal prediction. slight gains over Sets XGB RF achieving highest values. Key predictors included (e.g., Green_lag, NIR_lag, RedEdge_lag) CREI, GCI, NCPI, ARI, CCCI), highlighting value integrating corn underscores context lays foundation future across diverse settings, focusing size, data, refining dependencies localized search algorithms.

Language: Английский

Estimation of Silage Maize Plant Moisture Content Based on UAV Multispectral Data and Ensemble Learning Methods

Agriculture, Journal Year: 2025, Volume and Issue: 15(7), P. 746 - 746

Published: March 31, 2025

Plant moisture content (PMC) serves as a crucial indicator of crop water status, directly affecting agricultural productivity, product quality, and the effectiveness precision irrigation. Conventional methods for PMC assessment predominantly rely on destructive sampling techniques, which are labor-intensive impede real-time monitoring. This study investigates silage maize cultivated in Hexi region China, leveraging multispectral data acquired via an unmanned aerial vehicle (UAV) to estimate across different phenological stages. A stacked ensemble learning framework was developed, integrating Back Propagation Neural Network (BPNN), Random Forest Regression (RFR), Support Vector (SVR), with Partial Least Squares (PLSR) employed feature fusion. The findings indicate that incorporating vegetation indices into spectral variables significantly improved prediction performance. standalone models demonstrated coefficient determination (R2) values ranging from 0.43 0.69, root mean square error (RMSE) spanning 0.61% 1.43%. In contrast, model exhibited superior accuracy, achieving R2 between 0.61 0.87 RMSE 0.54% 1.38%. methodology offers scalable, non-invasive alternative estimation, facilitating data-driven irrigation optimization regions facing scarcity.

Language: Английский

Predicting Sugarcane Yield Through Temporal Analysis of Satellite Imagery During the Growth Phase

Agronomy, Journal Year: 2025, Volume and Issue: 15(4), P. 793 - 793

Published: March 24, 2025

This research investigates how to estimate sugarcane (Saccharum officinarum L.) yield at harvest by using an average satellite image time-series collected during the growth phase. study aims evaluate effectiveness of various modeling approaches, including a heteroskedastic gamma regression model, Random Forest, and Artificial Neural Networks, in predicting based on satellite-derived vegetation indices environmental variables. Key covariates analyzed include varieties, production cycles, accumulated precipitation phase, mean GNDVI index. The analysis was conducted two locations over consecutive growing seasons. emphasizes integration data with advanced statistical machine learning techniques enhance prediction agricultural systems, specifically focusing cultivation. results indicate that model outperformed other methods explaining variability, particularly commercial fields, achieving Coefficient Determination (R2) 0.89. These findings highlight potential these models support informed decision-making optimize practices, providing valuable insights for precision farming. Overall, this represent initial step toward developing more robust yield. Future work will involve incorporating additional variables better assess impacts stresses, such as high temperatures water deficits, crop’s agronomic performance.

Language: Английский