Modelling bioaccumulation of heavy metals in soil-crop ecosystems and identifying its controlling factors using machine learning

Highlights

• Three methods were used to model accumulation of heavy metals in soil-crop system.

• Random forest performed best on modelling heavy metals transport in soil-crop system.

• The model performed best on Zn (R² = 0.84) followed by Cu, Cr, Ni, Hg, Cd, As, Pb.

• Plant type was the primary control of heavy metals transport in soil-crop system.

Abstract

The prediction and identification of the factors controlling heavy metal transfer in soil-crop ecosystems are of critical importance. In this study, random forest (RF), gradient boosted machine (GBM), and generalised linear (GLM) models were compared after being used to model and identify prior factors that affect the transfer of heavy metals (HMs) in soil-crop systems in the Yangtze River Delta, China, based on 13 covariates with 1822 pairs of soil-crop samples. The mean bioaccumulation factors (BAFs) for all crops followed the order Cd > Zn > As > Cu > Ni > Hg > Cr > Pb. The RF model showed the best prediction ability for the BAFs of HMs in soil-crop ecosystems, followed by GBM and GLM. The R2 values of the RF models for the BAFs of Zn, Cu, Cr, Ni, Hg, Cd, As, and Pb were 0.84, 0.66, 0.59, 0.58, 0.58, 0.51, 0.30, and 0.17, respectively. The primary controlling factor in soil-to-crop transfer of all HMs under study was plant type, followed by soil heavy metal content and soil organic materials. The model used herein could be used to assist the prediction of heavy metal contents in crops based on heavy metal contents in soil and other covariates, and can significantly reduce the cost, labour, and time requirements involved with laboratory analysis. It can also be used to quantify the importance of variables and identify potential control factors in heavy metal bioaccumulation in soil-crop ecosystems.

Introduction

Soil plays a critical role in ecosystems and is the basic resource for food production, however, soil is threatened by many pollutants, among which heavy metals (HMs) are of great concern (Hu and Cheng, 2013; Zhao et al., 2014; Shao et al., 2018; Hu et al., 2019; Peng et al., 2019). In contrast to organic matter, HMs can reside in soils for long periods of time (Hu et al., 2017a; Jia et al., 2019). The wide distribution of HM-polluted soils has resulted in profound environmental and health issues (Brus et al., 2009; Hu et al., 2017b; Hu et al., 2020). HMs in soils can be slowly but consistently accumulated in the human body via different pathways such as inhalation, ingestion, and dermal contact (De Miguel et al., 2007; Wei and Yang, 2010; Hui et al., 2016). Moreover, the intake of HMs through soil-crop ecosystems has proven to be an important pathway for human exposure to HMs (Liu et al., 2007; Qian et al., 2010; Liu et al., 2013).

In recent four decades, China has experienced rapid economic growth and has undergone a significant transformation from the traditional agriculture-based economy to a manufacturing-based economy (Cheng and Hu, 2010). Industrial development and fast urban expansion have resulted in sharp increases in the amount of sewage sludge produced by industrial, transportation-related, and municipal activities nationwide (Cheng and Hu, 2012). The Yangtze River Delta (YRD), located in Eastern China, is the most developed area in the country and yields 23.6% of the national gross domestic product (Yan et al., 2018). With a population of more than 150 million, the YRD is one of the most densely populated regions in China (Shao et al., 2018; Hu et al., 2017b). It is also an important production area for many important agricultural products such as rice, tea, vegetables, and certain fruits. Many studies have revealed significant accumulation of HMs in soil and crops in the YRD region (Chen and Pu, 2007; Hu et al., 2017c; Fei et al., 2019; Xia et al., 2019). The accumulation of HMs in soil and their subsequent bioaccumulation through the food chain poses great potential health risks to citizens in the YRD. Therefore, it is of great importance to explore HM bioaccumulation in soil-crop systems and identify its controlling factors.

Bioaccumulation of HMs in soil by plants is mainly governed by the uptake mechanism, the physical and chemical properties of the soil, the chemical characteristics of the HMs, physiological characteristics of the plants, and other environmental factors (Peijnenburg et al., 2007). Previous studies analysing the factors affecting the bioaccumulation of HMs in soil-plant systems have mainly focused on these aforementioned factors and have employed traditional statistical analyses (Jin et al., 2005; Liu et al., 2017; Liu et al., 2018). These models only consider quantitative variables, focusing mostly on critical soil properties such as soil pH, soil organic matter content (SOM), and soil HM content. However, they cannot be used to quantitatively analyse the effect of qualitative variables such as soil type and soil parent materials which have also been shown to exert substantial effects on HM accumulation in soil-crop systems. Few studies have taken different factors into consideration or quantitatively analysed the detailed effects of each potential factor and the spatial characteristics of bioaccumulation of different HMs. Further, few researchers have succeeded in building a comprehensive model that is able to predict bioaccumulation factors (BAFs) of HMs in soil-crop systems. Moreover, the methods used in previous studies were unable to identify the importance of the different variables in the modelling process, which is vital for decision-makers in identifying the essential factors controlling HM pollution control and remediation in soils. Aiming to fill this research gap, in this study, we used and compared three kinds of widely used machine learning methods, namely the gradient-boosted model (GBM) (Friedman, 2002), random forest (RF) (Liaw and Wiener., 2002), and the classic linear statistic model referred to as the generalised linear model (GLM) (Nelder and Wedderburn., 1972), to predict the BAFs of different HMs and define the quantitative effects of different factors on the bioaccumulation of eight different HMs. Therefore, there were three main aims of this study. The first was to investigate the HM content and the BAFs of each HM in soil-plant ecosystems. The second objective was to build and compare the models constructed using the GBM, RF, and GLM methods to be used for predicting the BAFs of HMs in soil-plant ecosystems, and to identify the optimal method. The third aim was to identify potential controlling factors in the transfer of HMs in soil-crop ecosystems. This study could contribute to the regulation of HM contamination in soil-crop ecosystems and toward guaranteeing the safety of agricultural products.