Prediction of Metal Phytotoxicity and Threshold Identification in Calcareous Purple Soils: An Integrated Framework Using XGBoost–SHAP and PLS-SEM

Peng, Y.; Grigorieva, I.Y.

Авторы: Peng Yizhou, Iya Yu Grigorieva
Журнал: JOURNAL OF ENVIRONMENTAL CHEMICAL ENGINEERING
Том: 14
Номер: 2
Год издания: 2026
Номер статьи: 122029
DOI: 10.1016/j.jece.2026.122029
Аннотация: Current understanding of carbonate-mediated metal bioavailability relies on idealized amendments and lacks integrated threshold quantification along natural geochemical gradients, leaving speciation–toxicity linkages across carbonate contrasts under-quantified. In this study, a hybrid framework was developed to assess phytotoxicity mitigation in calcareous purple soils with contrasting CaCO₃ (Soil I: 29%; Soil II: 5%) using Brassica napus and Sorghum bicolor, dose–response modeling was used to derive threshold metrics, while XGBoost–SHAP and PLS-SEM were employed to link toxicity to the standardized labile exposure (F₀+F₁), quantify the non-linear contribution of CaCO₃, and evaluate the pathway “CaCO₃ → Labile → RGR. Results demonstrate that elevated CaCO₃ attenuated metal lability and delayed toxicity onset: at 2000 mg·kg⁻¹ Pb, toxicity manifested in 18.5% of samples in Soil I versus 30.5% in Soil II. Consequently, thresholds shifted rightward, with root IC₅₀ increasing for Pb (299 → 460 mg·kg⁻¹, +54%) and As (82 → 125 mg·kg⁻¹, +52%). Under high-exposure scenarios, Soil I conferred growth advantages (Pb ΔRGR +9.41%; As +4.23%), whereas Cd displayed hormetic behavior with attenuated mitigation at higher doses. XGBoost–SHAP identified labile fraction and dosage as dominant toxicity drivers. PLS-SEM revealed a hierarchical mitigation effect with path coefficients of −0.86 (Pb), −0.45 (Cd), and −0.25 (As). The term “Carbonate Leverage” summarizes the mitigation ranking (Pb > Cd > As) and the corresponding effective dose window in this dataset. The derived thresholds and effect sizes provide quantitative references for risk stratification under carbonate-contrasting purple soils, highlighting in-situ CaCO₃ as a modulator.
Добавил в систему: Пэн Ичжоу

	ИСТИНА	Войти в систему Регистрация
	ИПМех РАН
	Главная Поиск Статистика О проекте Помощь

ИСТИНА

ИПМех РАН

Prediction of Metal Phytotoxicity and Threshold Identification in Calcareous Purple Soils: An Integrated Framework Using XGBoost–SHAP and PLS-SEMстатья

Прикрепленные файлы