Scholarlink

Purpose <p>Predicting the availability of cadmium (Cd) in soil is crucial for understanding its potential crop uptake, which has significant implications for food safety and soil management. It is also important to predict and identify the factors modulating the transport of Cd within soil-crop systems. This study aimed to use Cd speciation and soil properties to predict soil available Cd using machine learning (ML) models.</p> Method <p>Paired soil-rice Cd datasets related to soil properties, soil Cd, and its speciation were extracted from the Web of Science search database. Ten machine learning models were developed and evaluated to predict the soil available Cd. To achieve an optimal model, GridSearchCV and 5-fold cross-validation were used. We conducted SHAP analysis to identify important features and the effects of the descriptors on the model prediction.</p> Results <p>After a comprehensive evaluation, the hyperparameter-tuned XGBoost model showed the highest prediction accuracy (R<sup>2</sup> = 0.79). The feature importance analysis revealed that soil total Cd and pH highly influenced Cd availability. Overall, soil properties had more influence than Cd speciation in predictive models of Cd availability. Consequently, a cubic polynomial regression model revealed a moderate, non-linear relationship between soil available Cd and rice grain Cd (R<sup>2</sup> = 0.503) and soil available Cd and wheat grain Cd (R<sup>2</sup> = 0.487) indicating crop specific accumulation patterns.</p> Conclusion <p>Overall, the optimized XGBoost model showed the highest prediction accuracy. Our results highlight that Cd speciation can offer new insights for predicting soil available Cd concentration using ML algorithms. Soil contamination regulations should primarily focus on controlling the concentration of soil total Cd to control Cd availability in agricultural soils.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Prediction of soil available cadmium by machine learning based on cadmium speciation and soil properties: insights from rice and wheat systems

  • 期刊论文

  • 发表日期：

    2025年11月22日

查看全文
  • Yohannes Kebede,
  • Louis A. Tremblay,
  • Yuanwang Liu,
  • Haoran Mi,
  • Xueqi Guo,
  • Okbagaber Andom,
  • G. S. K. Zackariah,
  • K. J. Maneh,
  • Zhaojun Li

摘要

Purpose

Predicting the availability of cadmium (Cd) in soil is crucial for understanding its potential crop uptake, which has significant implications for food safety and soil management. It is also important to predict and identify the factors modulating the transport of Cd within soil-crop systems. This study aimed to use Cd speciation and soil properties to predict soil available Cd using machine learning (ML) models.

Method

Paired soil-rice Cd datasets related to soil properties, soil Cd, and its speciation were extracted from the Web of Science search database. Ten machine learning models were developed and evaluated to predict the soil available Cd. To achieve an optimal model, GridSearchCV and 5-fold cross-validation were used. We conducted SHAP analysis to identify important features and the effects of the descriptors on the model prediction.

Results

After a comprehensive evaluation, the hyperparameter-tuned XGBoost model showed the highest prediction accuracy (R2 = 0.79). The feature importance analysis revealed that soil total Cd and pH highly influenced Cd availability. Overall, soil properties had more influence than Cd speciation in predictive models of Cd availability. Consequently, a cubic polynomial regression model revealed a moderate, non-linear relationship between soil available Cd and rice grain Cd (R2 = 0.503) and soil available Cd and wheat grain Cd (R2 = 0.487) indicating crop specific accumulation patterns.

Conclusion

Overall, the optimized XGBoost model showed the highest prediction accuracy. Our results highlight that Cd speciation can offer new insights for predicting soil available Cd concentration using ML algorithms. Soil contamination regulations should primarily focus on controlling the concentration of soil total Cd to control Cd availability in agricultural soils.