Geochemical-integrated machine learning approach predicts the distribution of cadmium speciation in European and Chinese topsoils
Naichi Zhang, Chen Lv, Yan Li, Panos Panagos, Cristiano Ballabio, Jun Man, Xueyuan Gu, Fang‐Jie Zhao, Peng Wang, Xingmei Liu, Yifan Qian, Peixin Cui, Tongliang Wu, Meiying Huang, Cun Liu, Yujun Wang
Abstract
Evaluating heavy metals bioavailable is crucial for comprehensive soil contamination assessment but challenging at large scales due to complex and resource-intensive analytical procedures, and the amount of dissolved metal in soils represents the relative solubility and potential mobility of cadmium, which is a key factor determining bioavailability. Here, we developed a geochemical-integrated machine learning framework using multi-source data to predict cadmium speciation distribution in European and Chinese non-industrial topsoils. Average total cadmium content in Chinese topsoils (0.41 mg kg−1) was ~10.8% higher than the Europe, while average dissolved cadmium content (113.2 μg L−1) was ~16.8% higher. Mechanistic interpretation revealed that lower pH, soil organic matter, and amorphous ferrihydrite contents mainly attributed to the higher bioavailability in China. The framework, coupled with knowledge transfer bridging the knowledge gap between geochemical processes and crop uptake, would facilitate the informed decision-making and targeted remediation measures for sustainable agricultural practices and long-term environmental health. Lower pH, soil organic matter, and amorphous ferrihydrite contents attributed to the higher cadmium bioavailability in Chinese topsoils than that of in Europe, according to a geochemical-integrated machine learning framework by incorporating multi-source data.