AI-Driven Soil Pollution Assessment in Cotton Fields
Heavy metal pollution in agricultural soils poses a major risk to environmental and human health, especially in intensive crop production areas such as cotton fields. This study explores the integration of multi-source data—remote sensing, natural environmental indicators, and human activity metrics—with machine learning techniques to predict the levels and distribution of soil available heavy metals (AHM) in northern Xinjiang, China. The aim is to enhance prediction accuracy, identify key influencing factors, and support sustainable agricultural practices through advanced data fusion and modeling strategies.
Multi-Source Data Fusion for Soil Pollution Prediction
The fusion of multi-source datasets offers a more comprehensive view of soil conditions. This research highlights the effectiveness of integrating remote sensing, environmental data, and anthropogenic activity metrics to create a unified dataset. By identifying and selecting the most significant features through importance analysis, the fused dataset enhanced the performance of machine learning models in predicting AHM content, outperforming models trained on isolated data sources.
Machine Learning Algorithms in Soil Heavy Metal Modeling
Four machine learning algorithms—LightGBM, Random Forest (RF), Partial Least Squares Regression (PLSR), and XGBoost—were utilized to construct heavy metal pollution prediction models. The models trained on the fused important variables consistently outperformed those trained on full or remote-sensing-only datasets. This comparative analysis underscores the value of model selection and hybrid data use in improving predictive accuracy for environmental pollutants.
Spatial Distribution and Severity of Heavy Metal Pollution
The study mapped the spatial variability of nickel, chromium, and lead across the cotton fields. Results showed that nickel pollution was severe in the central region, chromium was concentrated in the east, and lead was spread across both central and eastern areas. These insights are crucial for targeted soil remediation and pollution control, especially in areas undergoing intensive cotton cultivation.
Human Activities as Dominant Influencers of AHM Levels
Among the influencing variables, fertilizers and plastic film usage emerged as dominant contributors to AHM levels. Their correlation with increased heavy metal concentrations suggests that agricultural practices directly affect soil health. This finding emphasizes the urgent need for sustainable farming inputs and waste management practices to minimize long-term environmental damage.
Implications for Sustainable Agriculture and Environmental Monitoring
The proposed data fusion and ML-based prediction framework serves as a scientific basis for real-time soil pollution monitoring and proactive decision-making. The methodology not only improves prediction accuracy but also provides actionable insights into managing pollution sources. This contributes significantly to sustainable agricultural development and soil health preservation in ecologically sensitive cotton-growing regions.
Technology Scientists Awards
#SoilPollution
#CottonFarming
#MachineLearning
#SustainableAgriculture
#HeavyMetals
#RemoteSensing
#EnvironmentalMonitoring
#AgriTech
#LightGBM
#XGBoost
#RandomForest
#PLSR
#NickelPollution
#LeadContamination
#ChromiumLevels
#SoilHealth
#DataFusion
#XinjiangAgriculture
#SoilScience
#SmartFarming
Comments
Post a Comment