流域重金属污染是全世界广泛关注的环境问题，我国的重金属污染形势尤其严峻。2011年发布的《湘江流域重金属污染治理实施方案》是我国第一个获国务院批准的流域重金属污染治理试点方案。为了解决湘江流域重金属污染问题，该方案投资了595亿元人民币，主要通过控制工业污染源的方式来削减重金属排放量，从而降低重金属的生态环境风险。然而，流域环境系统是一个高度复杂的系统，流域环境污染往往受到众多的人类活动和自然过程的影响，呈现出点面源混合，大气、土壤、水体相互影响的特点，具有高度的时空异质性。流域水环境模型通过模拟陆地和水体中污染物的迁移转化过程，反映污染物的时空变化规律，是流域环境管理中的重要工具。但是，目前流域水环境模型领域的研究主要集中在氮、磷等营养物质，而对于生态系统和人体健康造成巨大威胁的有害重金属关注较少。现有的流域重金属迁移转化研究工作也主要停留在野外观测和简单的经验统计模型的层面上。因此，构建流域尺度重金属迁移转化机理过程模型，开展重金属流域环境行为模拟研究，明晰流域中重金属的主要源汇与分布，识别流域中重金属的高风险地区和高风险时段，对于流域环境保护、风险评估与风险管理具有十分重要的意义。

本研究首次在经典水文模型SWAT的基础上构建了流域重金属迁移转化模型SWAT-HM，利用SWAT-HM模型模拟了湖南浏阳河上游流域重金属Zn和Cd的时空动态变化过程，得到了不同形态重金属在流域不同环境介质（土壤、泥沙和底泥）中的暴露浓度以及不同环境介质间的迁移量，并运用于流域重金属生态环境风险评估和流域重金属输入输出通量估计中。主要的研究工作和结论如下：

（1）耦合流域水文模型SWAT，构建流域尺度重金属移转化模型SWAT-HM。首先综述了流域环境系统中重金属的形态分布与转化以及主要的迁移途径，提出了重金属多形态转化、多途径迁移的模型框架，首次在经典水文模型SWAT的基础上构建了流域重金属多形态多途径迁移转化模型SWAT-HM，实现了流域尺度长时间的重金属迁移转化连续模拟。该模型主要考虑了重金属两相三态间的吸附反应和老化反应。迁移模块包括陆面过程和河道过程：陆面过程主要考虑重金属随地表径流、壤中流、下渗流、侵蚀泥沙以及植物吸收的迁移过程；河道过程主要考虑重金属沉降、再悬浮、扩散和掩埋等过程。

（2）通过模型率定与验证，测试评估SWAT-HM可靠性。运用SWAT-HM模型在湖南浏阳七宝山矿区所影响的流域进行了适用性评价，模拟了研究区水文过程、泥沙过程和重金属过程。水文泥沙模拟结果在月、日尺度都与实测值符合较好，重金属模拟结果在日值尺度上与监测值趋势一致，数值上也非常吻合。流量的日模拟在率定期和验证期都满足R² > 0.75，NS > 0.7和PBIAS < ±10%。泥沙的日模拟在率定期和验证期都满足R² > 0.40，NS > 0.45和PBIAS < ±20%。根据前人总结的模型评估准则，流量和泥沙的模拟可以分别判定为“好”和“令人满意的”。对于Zn的模拟，85.9%和94.5%的溶解态模拟浓度分别在实测浓度的5倍线和10倍线内；对于Cd的模拟，86.1%和94.2%的溶解态模拟浓度分别在实测浓度的5倍线和10倍线内。同时，SWAT-HM模拟出了重金属经过矿区附近迅速升高，随后逐渐减少的特征。流量、泥沙和重金属的模拟精度依次递减，这是由于重金属的迁移以径流和泥沙为载体，误差的累积导致了精度的依次递减。但是，考虑到流域尺度污染物模拟的复杂性和不确定性，现有结果表明SWAT-HM仍较好地模拟了浏阳河上游流域重金属Zn和Cd的迁移转化特征，可用于后续的重金属污染时空分布分析。

（3）模拟浏阳河上游流域重金属流域行为，分析了研究区重金属Zn和Cd的动态时空分布。从重金属河道迁移量的时间动态来看，年尺度上，流域出口双江口Zn的年平均输出通量为90.8 t/yr，其中溶解态和颗粒态分别占30.7%和69.3%。Cd的年平均输出通量为1.1 t/yr，其中溶解态和颗粒态分别占59.1%和40.9%。从月尺度上，溶解态和颗粒态Zn、Cd最高输出量都出现在6月，最低输出量出现在1月。丰水期（4-6月）溶解态Zn、颗粒态Zn、溶解态Cd和颗粒态Cd输出量分别占总输出量的42.7%、77.7%、45.6%和78.9%。从日尺度上看，溶解态和颗粒态重金属输出量的大幅提升几乎全部出现在强降水后流量增加时，呈现出雨季多旱季少的特点，颗粒态重金属输出量主要由几次暴雨洪水过程贡献，说明了强降雨是形成重金属非点源污染的主要驱动力。从重金属陆相迁移入河量的空间分布来看，无论何种途径（地表径流、壤中流、侵蚀泥沙）的重金属迁移量，七宝山矿区重金属迁移量显著高于周边地区，证明七宝山矿区为重金属污染的最主要来源。从重金属随不同迁移途径的迁移量对比来看，随地表径流和侵蚀泥沙的迁移量日变化较大，随壤中流的迁移量比较稳定。

（4）通过构造不同的气候变化情景，分析不同气象因子对于不同形态重金属输出量的可能影响。气象敏感性分析结果表明，降水的变化对溶解态和颗粒态的影响都显著（p < 0.05），温度的变化对于溶解态重金属影响不显著（p = 0.93），但温度变化对颗粒态重金属有显著影响（p < 0.05）。降雨增加20%会导致溶解态Zn和颗粒态Zn分别增加11.7%和66.7%，温度升高2℃则会导致颗粒态Zn减少4.5%。降雨的变化对重金属输出量的影响比温度明显，且颗粒态重金属对于降雨和温度变化的响应比溶解态重金属敏感。这意味着气候变化导致的气温整体上升也许对重金属行为影响不大，但其所导致的强降水对重金属大范围迁移扩散的影响值得关注。

（5）利用SWAT-HM作为重金属暴露评估工具，评估研究区的生态环境风险。以湖南浏阳河上游流域为例，将SWAT-HM的模拟结果作为重金属暴露评估结果。选择我国地表水环境质量标准，利用商值法评估研究区重金属Zn和Cd环境污染风险；选择已有的Zn和Cd的急性/慢性水质基准，评估研究区Zn和Cd的水生生态风险。从环境质量评估的角度来看，Cd污染程度大于Zn污染，Zn和Cd浓度在每个断面都具有一定的时间波动性，但总体上呈现出上游和下游低，中游矿区附近显著升高的特点。当宝山河与大溪河在44号子流域汇合后，由于稀释作用和水沙交互作用，Zn和Cd浓度都显著降低，均好于III类水标准。从急性水生生态风险来看，Zn的急性生态风险高于Cd，可以发现宝山河经矿区后模拟的溶解态Zn浓度高频率地超过其急性水质基准，而Cd仅在有限几天有超过急性水质基准的现象。进一步分析发现Cd超标主要出现在枯水季，降雨较多的4-9月份超标次数明显减少。从慢性水生生态风险来看，Cd的生态风险显著提高，宝山河部分河段由于模拟的金属Cd浓度在6年内均超过其慢性水质基准，且当宝山河与大溪河汇流后，河水中的Cd浓度仍然高频地超过慢性水质基准值。综上所述，Zn的急性生态风险高于Cd，而Cd的慢性风险高于Zn。研究区重金属Zn和Cd的生态高风险区（短期和长期风险）主要位于宝山河子流域内，Cd的慢性风险区延续到宝山河与大溪河汇合后的下游河段，枯水期的风险高于丰水期。

（6）评估研究区重金属输入输出通量及其不确定性，为流域重金属环境风险管理提供决策依据。本研究将流域尺度重金属模型SWAT-HM应用于中国浏阳河流域上游，并评估了陆地（土壤）和河道（河床底泥）中的重金属Zn平衡，使用SUFI-2和GLUE的组合方法，基于可行解概念，量化了研究流域中主要的输入和输出Zn通量及其不确定性。在陆相过程中，大气沉降的Zn输入和土壤侵蚀的Zn输出被识别为最大的通量。在河道阶段中，底泥层被确定为重要的Zn汇。据此，本研究提出三点环境风险管理策略：第一，基于暴雨事件的土壤侵蚀控制是一种有效的污染控制措施；第二，河道底泥的重金属污染修复，特别是污染严重的河段，应与其它的岸上修复措施同时进行；第三，大气沉降是重要的重金属异地输入源，说明区域的联防联控对于削减流域重金属长期风险也非常重要。

Heavy metal pollution in soil and water is a severe and urgent environmental problem around the world, especially in developing countries such as China. To reduce the heavy metal pollution in the Xiang River Basin, the Xiang River Basin Control Plan for Heavy Metal Pollution, authorized by the State Council in 2011, was the first basin-scale pilot program for the treatment of heavy metal pollution at the watershed scale. This plan which costs 59.5 billion Chinese yuan, mainly cuts heavy metal emissions by controlling industrial pollution sources. However, the river basin environmental system is a highly complex system, where environmental pollution in the river basin often presents a pattern of interaction among the air, soil, and water. For example, the metals move through the terrestrial and aquatic ecosystem via complex networks of flowpaths across all spatial scales. Thus, the concentrations of heavy metals in rivers are affected by both human activities and natural processes, and exhibits significant spatial and temporal heterogeneity. Watershed-scale water quality model is increasingly recognized as an effective and efficient tool to assess the governing networks of surface and subsurface water pathways and associated pollutant fluxes within watersheds. Such models enable us to identify and quantify the sources, sinks, pathways, and fluxes of heavy metals at the watershed scale. However, the current researches of the watershed-scale environmental models mainly focus on nutrients such as nitrogen and phosphorus, and less attention was paid to the toxic heavy metals that pose a huge threat to the ecosystem and human health. Furthermore, the existing researches on heavy metal transport and transformation in river basins are mainly based on field observations and simple empirical statistical models. A major hindrance to such studies is the paucity of the modeling tools, as we have highlighted throughout this thesis. Therefore, a heavy metal model SWAT-HM (Soil and Water Assessment Tool – Heavy Metal) coupled with hydrological model SWAT (Soil and Water Assessment Tool) was developed to help us better understand the environmental behavior and clarify the main sources, sinks, and distribution of heavy metals. It also works on identifying the hot-spots and hot- moments of metal risk, which is of great significance for environment protection, risk assessment, and risk management in metal-polluted regions.

This study focuses on the development of the watershed-scale metal model SWAT-HM and its application in simulating the fate and transport of Zn and Cd in the upper Liuyang River (ULR) watershed in Hunan Province, China. After calibrating and validating the model, the concentrations of Zn and Cd in different environmental media (soil, water, and sediment) and the fluxes between different environmental media were obtained. The results were used to assess the ecological risk and estimate the input and output fluxes of heavy metals in the ULR watershed.

The main results and conclusions are as follows:

(1). Develop a watershed-scale heavy metal model coupled with SWAT to simulate the fate and transport of metals at the watershed scale. This study firstly summarized our current understanding of the influencing factors of heavy metal distribution in the watershed environmental system and its associated input and output pathways. Based on the worldwide applied hydrological model SWAT, a watershed heavy metal transport and transformation model SWAT-HM was then developed. The SWAT-HM accounts for sorption and slow reactions among three metal species; the heavy metals in the upland are allowed to transport vertically through percolation, plant uptake and evaporation-induced water rising as well as horizontally through soil erosion and surface/subsurface runoff; the heavy metals in the water body, in contrast, are modeled to undergo settling, resuspension, diffusion, and burial processes.

(2). Calibrate and validate the SWAT-HM model and simulate the behavior of Zn and Cd in the ULR watershed. To evaluate the reliability of SWAT-HM, the SWAT model was calibrated based on the measured streamflow, sediment, and metal (Zn and Cd) data. Regarding the daily streamflow, R² reached 0.89 and 0.82, NS equaled 0.87 and 0.80, and PBIAS had a value of 5.2% and ?4.7%, for the calibration and validation period, respectively. The SWAT-HM model performance for flow simulations can be judged as “good”. Regarding the daily sediments, R² had a value of 0.81 and 0.55, NS equaled 0.78 and 0.46, and PBIAS had a value of ?7.4% and ?16.3% for the calibration and validation period, respectively. The model performance for sediment simulation was judged as “satisfactory”. For the metal simulation, notwithstanding the high variation in metal concentrations in the river reaches to nearly 3 orders, 85.9% (Zn), 86.1% (Cd) 94.5% (Zn), and 94.2% (Cd) of the simulated concentrations remained within 5-fold and 10-fold of the observed concentrations. The overall good agreement between the simulated and observed concentrations indicated the successful application of the SWAT-HM in the study area.

(3). Simulate the behavior of heavy metals in ULR watershed, and analyze the dynamic spatiotemporal distribution of Zn and Cd in the study area. Regarding the temporal dynamic of heavy metal in the river channel, on an annual scale, the annual average output flux of Zn at the watershed outlet is 90.8 t/yr, of which the dissolved phase and the adsorbed phase account for 30.7% and 69.3%, respectively. The average annual output flux of Cd is 1.1 t/yr, of which the dissolved phase and the adsorbed phase account for 59.1% and 40.9%, respectively. On the monthly scale, the highest output of dissolved and adsorbed phases of Zn and Cd all appeared in June, and the lowest output appeared in January. During the flood season (April-June), the output of dissolved phase Zn, adsorbed phase Zn, dissolved phase Cd and adsorbed phase Cd accounted for 42.7%, 77.7%, 45.6%, and 78.9%, respectively. On a daily scale, the substantial increases in the output of the dissolved and adsorbed metals both occur after heavy precipitation. The output of adsorbed metals is mainly contributed by several storms and floods. It shows that heavy rainfall is the main driving force for the formation of Non-point source metal pollution. Regarding the spatial distribution of heavy metal in the upland, regardless of the route of heavy metal transport (surface runoff, lateral flow, and eroded soil), the heavy metal transport in Qibaoshan mining area is significantly higher than that in the surrounding areas, which proves that Qibaoshan mining area is the most important source of heavy metal pollution. From the comparison of three kinds of transport pathways with surface runoff, soil midstream, and erosion and sedimentation, the daily transport of surface runoff and soil erosion varies greatly, and the transport with the lateral flow is relatively stable.

(4) Construct climate change scenarios to assess the sensitivity of heavy metal migration to the changes in rainfall and temperature. To explore the response of heavy metal behaviors to the climate change, nine different climate change scenarios, each adjusting either the precipitation or the temperature, were used to drive the SWAT-HM model to obtain the sensitivity of Zn output to the changes of precipitation and temperature. The results showed that the Zn was more sensitive to precipitation than temperature, and that particulate Zn responded more strongly than dissolved Zn. This means that the temperature rise caused by climate change may have little effect on the migration of heavy metals, but the impact of the heavy precipitation on the large-scale migration of heavy metals is worthy of attention.

(5) Using SWAT-HM as a heavy metal exposure assessment tool to assess the ecological and environmental risks of Zn and Cd in the ULR watershed. To evaluate the environmental and ecological risks, the consecutively simulated Zn and Cd concentrations in the river channels were used as a surrogate of the Predicted Exposure Concentrations (PEC) of Zn and Cd. Such PECs were compared to their corresponding Predicted No-Effect Concentrations (PNECs) to characterize the aquatic environmental and ecological risk within the ULR watershed. From the perspective of the environmental pollution risk, the risk of Cd was greater than that of Zn. Both Zn and Cd showed the same pattern: a significant increase of risk in the middle reaches of the mining area, and a low risk in the upstream and downstream. Due to the effect of dilution and water-sediment interaction, After the confluence of Baoshan River and Daxi River in the subbasin No.44, the Zn and Cd concentrations reduced significantly, which were below the Class III water quality standard. From the perspective of the acute aquatic ecological risk, the acute ecological risk of Zn is higher than that of Cd. The simulated dissolved Zn concentration in the channels of Baoshan River (No.49, 48, 47, 45, 42, and 40) exceeded its acute water quality criteria frequently; whereas the Cd concentration exceeded the acute water quality criteria for only a few days. Further analysis found that Cd exceeded the criteria mainly in the dry season, and the number of exceedances decreased significantly in the rainy season. From the perspective of the chronic aquatic ecological risks, the risk of Cd increased significantly. The simulated Cd concentration in several channels of the Baoshan River (No. 48, 47, 45, 42, and 40) exceeded its chronic water quality criteria within 6 years. Even after the confluence of the Baoshan River and the Daxi River, the Cd concentration in the river still exceeded the chronic water quality criteria with high frequency.

(6) Assess the input and output fluxes of heavy metals and their uncertainty in the study area, and provide a basis for decision-making under uncertainty for environmental management. A better understanding of the metal budget in a watershed system is critical to evaluate the current pollution status and develop effective management strategies. Based on the concept of “behavioral simulations”, an uncertainty analysis method that couples GLUE and SUFI-2 was proposed to quantify the input and output Zn fluxes and their associated uncertainty. In the upland phase, Zn input from atmospheric deposition and Zn output through soil erosion were identified as the largest fluxes. In the channel phase, the bed sediment was identified as the most important Zn sink. These results suggest three counteract strategies for environmental risk control. First, soil erosion control could be an imperative and effective remediation measure; second, restoration of the metal-polluted bed sediment, especially those of the highly polluted river channels, should be simultaneously implemented with other mitigation measures; third, atmospheric deposition accounted for the most of the input fluxes, indicating that a joint prevention and control of regional air pollution is also important to reduce the long-term risk of heavy metals in the ULR watershed.