{"title":"Universal models for comprehensive prediction of soil quality responses to complex amendments in acidic soil","authors":"Pengshun Wang , Prakash Lakshmanan , Qichao Zhu , Siwen Zhang , Donghao Xu , Shuihan Yuan , Fusuo Zhang","doi":"10.1016/j.geoderma.2025.117495","DOIUrl":null,"url":null,"abstract":"<div><div>Soil acidification poses a serious threat to sustainable use of arable land. Applying soil amendments are effective measures to counteract the widely occurred soil acidification. However, a method to accurately predict the amendment effects on soil acidification is lacking. In this study, a total of 41 soil culture treatments, covering organic, inorganic, and their combinations, were conducted to compile a comprehensive dataset, which was further used to establish a model to predict the performance of amendments. The random forest (RF) and multiple linear regression (MLR) were adopted to model soil quality changes due to varying amendments application. 30 % percent of the culture treatment dataset and field observations were used to validate the model performance. The results demonstrate that MLR models are less robust in predicting the change in soil indicators, with the R<sup>2</sup> varying 0.6∼0.82. For some soil indicators, such as exchangeable acid (Ex-Acid), exchangeable calcium and cation exchange capacity (CEC), due to weak adherence to key assumptions such as linearity, homoscedasticity, and normality, which likely impaired their predictive reliability. Such limitations could reduce the model’s fitting accuracy and predictive stability for certain soil properties. The RF model is excellent at reconstructing changes in all soil chemical properties, with R<sup>2</sup> greater than 0.80. This includes soil pH, Ex-Acid, exchangeable calcium, and exchangeable magnesium, except for changes in CEC, which rarely changed after amendments application. Validation of model predictions through multi-site field observations further confirmed the robust predictions for multiple types of amendments. In particular, the prediction results of RF for Ex-acid are better than those of MLR. The overall outcomes suggest that RF model demonstrated greater reliability and adaptability, highlighting its practical value for guiding amendment selection in acid soil management.</div></div>","PeriodicalId":12511,"journal":{"name":"Geoderma","volume":"461 ","pages":"Article 117495"},"PeriodicalIF":6.6000,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Geoderma","FirstCategoryId":"97","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0016706125003362","RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"SOIL SCIENCE","Score":null,"Total":0}
引用次数: 0
Abstract
Soil acidification poses a serious threat to sustainable use of arable land. Applying soil amendments are effective measures to counteract the widely occurred soil acidification. However, a method to accurately predict the amendment effects on soil acidification is lacking. In this study, a total of 41 soil culture treatments, covering organic, inorganic, and their combinations, were conducted to compile a comprehensive dataset, which was further used to establish a model to predict the performance of amendments. The random forest (RF) and multiple linear regression (MLR) were adopted to model soil quality changes due to varying amendments application. 30 % percent of the culture treatment dataset and field observations were used to validate the model performance. The results demonstrate that MLR models are less robust in predicting the change in soil indicators, with the R2 varying 0.6∼0.82. For some soil indicators, such as exchangeable acid (Ex-Acid), exchangeable calcium and cation exchange capacity (CEC), due to weak adherence to key assumptions such as linearity, homoscedasticity, and normality, which likely impaired their predictive reliability. Such limitations could reduce the model’s fitting accuracy and predictive stability for certain soil properties. The RF model is excellent at reconstructing changes in all soil chemical properties, with R2 greater than 0.80. This includes soil pH, Ex-Acid, exchangeable calcium, and exchangeable magnesium, except for changes in CEC, which rarely changed after amendments application. Validation of model predictions through multi-site field observations further confirmed the robust predictions for multiple types of amendments. In particular, the prediction results of RF for Ex-acid are better than those of MLR. The overall outcomes suggest that RF model demonstrated greater reliability and adaptability, highlighting its practical value for guiding amendment selection in acid soil management.
期刊介绍:
Geoderma - the global journal of soil science - welcomes authors, readers and soil research from all parts of the world, encourages worldwide soil studies, and embraces all aspects of soil science and its associated pedagogy. The journal particularly welcomes interdisciplinary work focusing on dynamic soil processes and functions across space and time.