区县尺度土壤重金属时空分布的驱动机制与预测

Most research studies concerning heavy metals in regional soils focus on the source apportionment, pollution evaluation and spatial prediction of the total amount of soil heavy metals, and the driving mechanism of the spatio-temporal evolution process, whereas serious issues of spatio-temporal prediction need further study. Therefore, this research is concerned with the prediction of spatio-temporal patterns of soil heavy metals that take human activities and landscape heterogeneity as the two key drivers, and the annual increment of soil heavy metals as the core object. The study areas will be located in typical industrial zones and mining smelting areas. Firstly, the principal component analyses with a multiple linear regression model will be used to quantify the source apportionment of soil heavy metals increment in the study areas. Based on the results of source apportionment, the corresponding socio-economic factors will be selected as the agent driving factors of various identified sources. The relationship between the driving forces and the total increment of heavy metal elements will be quantified using the multiple linear regression or back propagation neural network methods. Secondly, the random forest and the probability density weighting methods will be applied to establish the quantitative relationship between the increment of soil heavy metals and landscape factors. Finally, according to the socio-economic development parameters and the spatial distribution of the landscape for the forecasting period, the spatio-temporal distribution of soil heavy metals will be predicted using Bayesian Maximum Entropy methods based on previously sampling data. Our research results will help us reveal the driving mechanism of the spatio-temporal distribution of regional soil heavy metals, and will provide the scientific basis for the treatment of heavy metal pollution in soils. Furthermore, the idea and methods of this research proposal will extend the current theories and methods of spatio-temporal prediction of geographical attributes.

现有区域土壤重金属研究侧重于对其总量的源解析、评估及其空间预测，而其时空演变的驱动机制及时空预测有待深入研究。鉴于此，本项目选取工业区和矿业开采冶炼区两种典型区域，以影响土壤重金属累积及空间分异的人类活动和景观异质性两个关键驱动要素为线索，以土壤重金属年度增量为核心：首先，利用PCA/MLR模型对土壤重金属增量进行定量源解析，选取社会经济要素作为各污染源的代理驱动因子，基于MLR或BP神经网络方法，建立驱动因子与土壤重金属元素年度总体增量的定量关系；其次，用随机森林和概率密度加权法，建立土壤重金属增量与景观因子之间的定量关系；最后，根据待预测时期社会经济发展指标和景观要素分布，联合前期采样数据，利用BME法，对土壤重金属含量进行时空预测。研究结果将有助于揭示区域土壤重金属时空演变的驱动机制，为土壤重金属污染的治理提供科学依据，也是对现有的地理属性时空预测理论和方法的拓展。

当前的区域土壤重金属相关研究在区域土壤重金属的时空演变过程及驱动机制方面仍缺乏深入探索，难以把握其未来发展状况。而掌握这些信息，是提出预防、缓解和适应土壤环境变化对策的基础。鉴于上述背景，本基金选择湖北省大冶市为研究区，主要的研究内容包括：（1）基于研究区多年度的土壤样品数据，揭示土壤重金属时空分布状况；（2）对区域土壤重金属污染进行评价，通过定量源解析方法，找到污染源并评估其贡献率。明确造成区域土壤重金属不断累积的环境因子，并探明其影响机制；（3）研究土壤重金属时空预测方法，预测未来土壤重金属元素的空间分布。.主要的研究结果包括：（1）大冶市土壤中Cd、Cu、Pb和Zn超标严重。矿业活动是造成区域土壤重金属污染的主要来源；（2）提出的土壤重金属景观源解析模型，进一步识别出4个具体的污染源景观，并量化了这些景观对每种重金属元素累积的贡献率；（3）基于多种模型，探索了区域土壤重金属时空变异的驱动机制，其中，GWLR模型能够定量衡量土壤重金属富集与人类活动空间非平稳关系的。GWR模型的结果表明，相较于地形和土壤属性，位置因素对土壤重金属累积的空间分异影响较大。随机森林模型的结果表明，人类活动对于研究区土壤铜含量变化的影响要显著高于自然环境本身，但由人为因素导致的土壤铜污染在逐年减缓；（4）对未来土壤重金属时空分布进行了预测，结果表明，未来Co、Cr、Ni、Mn和Zn分布相似，高浓度地区主要为东南部欧家山、中部铜绿山以及北部黄金山一带。Cu污染仍集中在中部铜绿山地区。Pb污染则主要集中在东南部欧家山地区。Cd污染主要集中在中部铜绿山和东南部欧家山地区。.本研究的意义在于：明确了区域土壤重金属时空演变过程及其影响因子和驱动机制，并在景观层面上揭示了造成土壤重金属累积的污染源，使土壤重金属污染的源头阻控和治理依据更明确；同时，预测了未来区域土壤重金属分布状况，有利于降低由土壤中重金属带来的生态和人口健康风险。