这是用户在 2025-1-3 12:11 为 https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2023WR034738 保存的双语快照页面，由 沉浸式翻译 提供双语支持。了解如何保存？
Water Resources Research
水资源研究
Volume 59, Issue 12 e2023WR034738
Research Article  研究论文
Open Access  开放获取

A Process-Based Model to Track Water Pollutant Generation at High Resolution and Its Pathway to Discharge
基于过程的水污染物生成高分辨率追踪模型及其排放途径

Yujie Zhuang, 

Yujie Zhuang

State Key Laboratory of Pollution Control and Resource Reuse, School of the Environment, Nanjing University, Nanjing, P. R. China

Contribution: Conceptualization, Methodology, Software, Validation, Formal analysis, ​Investigation, Data curation, Writing - original draft, Writing - review & editing, Visualization, Project administration

Search for more papers by this author
Xin Liu, 

Corresponding Author

Xin Liu

State Key Laboratory of Pollution Control and Resource Reuse, School of the Environment, Nanjing University, Nanjing, P. R. China

Correspondence to:

X. Liu and Z. Yuan,

xinliu@nju.edu.cn;

yuanzw@nju.edu.cn

Contribution: Validation, Formal analysis, ​Investigation, Writing - review & editing, Supervision, Funding acquisition

Search for more papers by this author
Zengwei Yuan, 

Corresponding Author

Zengwei Yuan

State Key Laboratory of Pollution Control and Resource Reuse, School of the Environment, Nanjing University, Nanjing, P. R. China

Correspondence to:

X. Liu and Z. Yuan,

xinliu@nju.edu.cn;

yuanzw@nju.edu.cn

Contribution: Conceptualization, Formal analysis, Resources, Writing - review & editing, Supervision, Funding acquisition

Search for more papers by this author
Hu Sheng, 

Hu Sheng

Lishui Institute of Ecology and Environment, Nanjing University, Nanjing, P. R. China

Contribution: Software, Formal analysis, Data curation

Search for more papers by this author
Jianqi Gao, 

Jianqi Gao

State Key Laboratory of Pollution Control and Resource Reuse, School of the Environment, Nanjing University, Nanjing, P. R. China

Contribution: Validation, ​Investigation

Search for more papers by this author
First published: 01 December 2023
https://doi.org/10.1029/2023WR034738 IF: 4.6 Q1
https://doi.org/10.1029/2023WR034738IF：4.6 Q1 译文中：https://doi.org/10.1029/2023WR034738IF：4.6 Q1

首次发布：2023 年 12 月 1 日 https://doi.org/10.1029/2023WR034738 IF: 4.6 Q1

Abstract  摘要

The estimation of water pollutant loads is of crucial importance as it decides data inputs for both subsequent water quality models and watershed pollutant mitigation strategy. However, the generation of water pollutant loads at high resolution and the life-cycle pathway of them from sources, through pipelines and wastewater treatment plants, finally to recipient water bodies remain unclear. This study aims to establish a process-based Water Pollutant Loads Tracking model and applies it to a rapidly urbanizing watershed in Taihu Lake Basin, China at the resolution of 5 m × 5 m in 2017, whose terrain is flat, slowly lowing from west to east. Results show that, of 261.55 tons of the total phosphorus (TP) generation, 71.28% is collected, and 64.32% is treated, with only 26.46% discharged to the water bodies and 1.25% to the target ones. The spatial hotspots of TP generation are mainly concentrated in residential and industrial areas. Direct discharges from point sources and nonpoint runoffs, especially in rainy seasons, are recognized as the main challenges for local water pollution control. The main innovation of this study is to quantify the generation of water pollutant loads at high resolution and to trace the subsequent pathway to collection and discharge, and then to identify the potential substantial gap between generation and discharge, demonstrating the efficacy and broader applicability of this model.
水质污染物负荷的估算至关重要，因为它决定了后续水质模型的数据输入和流域污染物减排策略。然而，高分辨率的水污染物负荷生成以及它们从源头、通过管道和污水处理厂，最终到达受纳水体的生命周期路径仍然不明确。本研究旨在建立一个基于过程的污染物负荷追踪模型，并将其应用于 2017 年中国太湖流域一个快速城市化的流域，分辨率为 5 米×5 米，地形从西向东缓慢降低。结果显示，在 261.55 吨总磷（TP）生成量中，71.28%被收集，64.32%得到处理，仅有 26.46%排放到水体中，1.25%排放到目标水体中。TP 生成的空间热点主要集中在住宅和工业区域。直接从点源排放和非点源径流，尤其是在雨季，被认为是当地水污染控制的主要挑战。 本研究的主要创新在于以高分辨率量化水污染物负荷的产生，并追踪其随后的收集和排放路径，然后识别产生与排放之间的潜在重大差距，证明了该模型的效力和更广泛的应用性。

Key Points  关键点

  • A process-based mass balance model is developed to quantify the generation pattern of water pollutant loads at high resolution
    基于过程的大规模平衡模型被开发出来，以量化高分辨率下水污染物负荷的产生模式

  • The model is a reliable tool for tracing water pollutants from sources to recipient water bodies without any topographic limitation
    该模型是一种可靠的工具，能够在没有任何地形限制的情况下追踪水源到受水体的水污染物

  • Substantial positive gap between generation and discharge indicates the effectiveness of local pollution control measures
    显著的正向代际差距和排放差距表明了地方污染控制措施的有效性

1 Introduction  1 引言

The frequent occurrence of water pollution incidents across the world remains a considerable challenge to approach sustainable development goals (Mekonnen & Hoekstra, 2015; Schwarzenbach et al., 2010). Water quality deterioration causes negative impacts on aquatic ecosystems and human societies (Bartos & Kerkez, 2021; Hamidi et al., 2021; Martín-Hernández et al., 2021), such as harmful algal blooms (Iiames et al., 2021), biodiversity loss (Vörösmarty et al., 2010), threatening drinking water safety (Mortazavi-Naeini et al., 2019), harming human health (Adebayo & Areo, 2021; Carmichael & Boyer, 2016), and eventually fueling regional conflicts (Y. Liu et al., 2021). It has been recognized that anthropogenic factors, such as population density, economic development level, and wastewater treatment infrastructure, are the main causes, especially in urbanizing watersheds with flat terrains (B. Chen et al., 2019; Duan et al., 2009; D. Liu et al., 2022). Regulators initially attempted to restrict discharge concentrations to improve the basins’ water environment, with limited success (B. Wang et al., 2023). The total maximum daily load and other projects have optimized control strategies to confine the discharge loads so that nutrients input from additional anthropogenic sources can match the self-purification capacity of water bodies, and achieved certain results (Stephenson et al., 2021). However, the generation pattern of water pollutant loads at high resolution remains blank, and the detailed pathway between generation and discharge has not been clarified, resulting in the potential substantial gap between the two processes has not been considered (S. Li et al., 2017; J. Zhang et al., 2019). Ultimately, the reliability of the load management task decomposition for each pollution source is compromised (R. Zhou et al., 2023).
全球范围内水污染事件的频繁发生仍然是实现可持续发展目标的一大挑战（Mekonnen & Hoekstra，2015；Schwarzenbach 等，2010）。水质恶化对水生生态系统和人类社会产生负面影响（Bartos & Kerkez，2021；Hamidi 等，2021；Martín-Hernández 等，2021），例如有害藻华（Iiames 等，2021）、生物多样性丧失（Vörösmarty 等，2010）、威胁饮用水安全（Mortazavi-Naeini 等，2019）、损害人类健康（Adebayo & Areo，2021；Carmichael & Boyer，2016），并最终加剧区域冲突（Y. Liu 等，2021）。人们已经认识到，人为因素，如人口密度、经济发展水平、污水处理基础设施等，是主要原因，尤其是在地形平坦的城市化流域（B. Chen 等，2019；Duan 等，2009；D. Liu 等，2022）。监管机构最初试图限制排放浓度以改善流域水环境，但效果有限（B. Wang 等，2023）。 总最大日负荷和其他项目已优化控制策略，以限制排放负荷，使额外人为来源的营养物质输入与水体的自净能力相匹配，并取得了一定成果（Stephenson 等，2021）。然而，高分辨率下水污染物负荷的产生模式仍然空白，生成与排放之间的详细路径尚未明确，导致这两个过程之间可能存在的重大差距尚未被考虑（S. Li 等，2017；J. Zhang 等，2019）。最终，针对每个污染源的负荷管理任务分解的可靠性受损（R. Zhou 等，2023）。

Watershed models such as the Soil and Water Assessment Tool model (Arnold et al., 1998) and the integrated catchment model (Whitehead et al., 1998), commonly used tools in watershed management practices, have been developed to simulate and quantify the impact of human activities and natural processes on water quality degradation (L. Chen et al., 2018). These models can address both point and nonpoint sources of pollution by considering the nutrient cycling within pools and the transformation of different nutrient forms, which are often driven by a combination of physical, chemical, and biological mechanisms (Fu et al., 2019; L. Yuan et al., 2020). However, numerous parameters and complex calculations are involved, requiring tremendous efforts of data input, calibration, and validation (Pang et al., 2020; Ranjan & Mishra, 2022). To alleviate this difficulty, budget models such as substance flow analysis (Wu et al., 2016) and inventory analysis (Xu et al., 2023) are primarily driven by physical mechanisms, reducing modeling costs (Zou et al., 2020). However, they focus on the development and utilization of resources (Shi et al., 2019) and describe how much pollutant, for example, phosphorus (P), is delivered from human activities to water bodies in a specified amount of time (Hobbie et al., 2017). As a result, the pathway of water pollutants from sources to recipient water bodies has been considered as a black box but not characterized from the life-cycle perspective. Not only is it impossible to quantify the generation pattern to support an effective breakdown of load management, it is also impossible to identify direct discharges from point sources, posing a serious threat to water quality and leading to excesses (Jones et al., 2022). In addition, budget models rely on macro statistical data on an annual basis (Dong et al., 2020) and focus on different large spatial scales, mostly global (e.g., with a spatial resolution of 5 arc-minutes, approximately 10 km at the equator) (Mekonnen & Hoekstra, 2018), national (X. Liu et al., 2020), provincial (Wu et al., 2014), or large watershed (Jiang & Yuan, 2015; Yan et al., 2021) and sub-watershed (e.g., with an average spatial resolution of 5 km) (Gao et al., 2013) scales, which are typically not available at high spatio-temporal resolution and lack details in temporal variation (Z. Yuan et al., 2019) and spatial heterogeneity of water pollutants (B. Liu et al., 2013). Accordingly, they cannot be well coupled with water quality models to simulate the in-stream processes, and the water quality response is not available. In contrast, spatial scales of watershed models reach monthly, daily, and even hourly (Yang et al., 2016), and the target watersheds are divided into sub-watersheds or even smaller spatial units (e.g., hydrological response units, landscape units) with the help of digital elevation model (DEM) (Lazar et al., 2010; Ren et al., 2022), which both strongly support the operation of the embedded modules of water quality simulation. These models are therefore widely applicable to mountainous catchments but have limitations in flat urbanizing watersheds due to anthropogenic disturbance of drainage systems (e.g., sluices, underground wastewater collection pipelines) and flat terrains (Datta et al., 2022).
流域模型，如土壤和水评估工具模型（Arnold 等，1998）和综合流域模型（Whitehead 等，1998），是流域管理实践中常用的工具，已被开发用于模拟和量化人类活动及自然过程对水质退化的影响（L. Chen 等，2018）。这些模型可以通过考虑池内的养分循环和不同养分形式的转化来处理污染的点源和非点源，这些转化通常由物理、化学和生物机制的组合驱动（Fu 等，2019；L. Yuan 等，2020）。然而，涉及众多参数和复杂的计算，需要大量数据输入、校准和验证的努力（Pang 等，2020；Ranjan & Mishra，2022）。为了减轻这种困难，预算模型如物质流动分析（Wu 等，2016）和清单分析（Xu 等，2023）主要受物理机制驱动，降低建模成本（Zou 等，2020）。然而，它们侧重于资源和开发利用的发展与利用（Shi 等。，2019)并描述在特定时间内，例如磷（P），从人类活动输送到水体中的污染物量（Hobbie 等人，2017）。因此，从源头到受纳水体的水污染物路径被视为一个黑箱，但并未从生命周期角度进行表征。不仅无法量化生成模式以支持有效的负荷管理分解，而且也无法识别来自点源的直接排放，这对水质构成严重威胁，导致过剩（Jones 等人，2022）。此外，预算模型依赖于年度的宏观统计数据（Dong 等人，2020），并关注不同的大型空间尺度，大多为全球尺度（例如，空间分辨率为 5 弧分，大约在赤道处为 10 公里）（Mekonnen & Hoekstra，2018），国家尺度（X. Liu 等人，2020），省尺度（Wu 等人，2014）或大型流域（Jiang & Yuan，2015；Yan 等人，2021）以及子流域（例如，平均空间分辨率为 5 公里）（Gao 等人。, 2013) 的尺度，通常在时空高分辨率下不可用，且在污染物水质的时间变化（Z. Yuan 等人，2019）和空间异质性（B. Liu 等人，2013）方面缺乏细节。因此，它们不能很好地与水质模型耦合以模拟河流中的过程，水质响应不可用。相比之下，流域模型的时空尺度可达月度、日度甚至小时度（Yang 等人，2016），目标流域在数字高程模型（DEM）（Lazar 等人，2010；Ren 等人，2022）的帮助下被划分为子流域或更小的空间单元（例如，水文响应单元、景观单元），这都强烈支持水质模拟嵌入模块的运行。因此，这些模型在山区流域中得到广泛应用，但由于排水系统（例如，闸门、地下污水收集管道）的人为干扰和平坦地形（Datta 等人，2022），在平坦的城市化流域中存在局限性。

Considering the above deficiencies, a mass balance model based mainly on physical processes, namely Water Pollutant Loads Tracking (WPLT) model, is developed in this study to (a) make up the blank in the generation pattern of water pollutant loads at high spatio-temporal resolution; (b) characterize the life-cycle pathway of water pollutants (including generation, collection, treatment, discharge, etc.) without constraints by terrain; (c) reveal the potential substantial gap between generation and discharge to facilitate the efficient decomposition of the load management task in watersheds; (d) identify point-source direct discharges to provide accurate data inputs for robust simulations of water quality models. This model is universal and can be applicable to all types of terrains. In this study, we specifically focus on highlighting the advantage of this model in flat terrains, considering anthropogenic factors as the primary contributor to water quality deterioration. To address this, the model was then applied to a typical rapidly urbanizing area with flat terrains in Taihu Lake Basin, China, to characterize the total phosphorus (TP) loads from sources to recipient water bodies at monthly and daily scales in 2017 with the resolution of 5 m × 5 m. To our best knowledge, it is the first attempt to clarify the life-cycle pathway of water pollutants, which would contribute to a systemic understanding of water pollutant loads terminology and provide valuable implications for local water pollution control.
考虑到上述不足，本研究开发了一种主要基于物理过程的物质平衡模型，即水污染物负荷追踪（WPLT）模型，旨在（a）填补高时空分辨率下水污染物负荷生成模式中的空白；（b）描述水污染物（包括生成、收集、处理、排放等）的生命周期路径，不受地形限制；（c）揭示生成和排放之间可能存在的重大差距，以促进流域负荷管理任务的分解效率；（d）识别点源直接排放，为水质模型稳健模拟提供准确的数据输入。此模型具有通用性，适用于所有类型的地形。在本研究中，我们特别关注强调该模型在平坦地形中的优势，将人为因素视为水质恶化的主要贡献者。 为解决这一问题，该模型随后被应用于中国太湖流域典型的快速城市化地区平坦地形，以 2017 年月度和日度尺度，以 5 米×5 米的分辨率，对总磷（TP）负荷从源到受纳水体的负荷进行表征。据我们所知，这是首次尝试阐明水污染物的生命周期路径，这将有助于对水污染物负荷术语的系统理解，并为当地水污染控制提供有价值的启示。

2 Materials and Methods
2 材料与方法

2.1 Model Description  2.1 模型描述

The hierarchy of WPLT model is shown in Figure 1, of which water pollution sources are divided into point and nonpoint sources. Since the model mainly focuses on anthropogenic impacts, we assumed some natural sources, such as atmospheric deposition, to be negligible (X. Liu et al., 2016). Point sources include those with easily identifiable origins, such as industrial production (e.g., food processing factories, textile factories) and residential and commercial consumption (e.g., residences, restaurants, and shopping malls). Note that the precision of industrial sources can be improved by being broken down into sub-industries following the national standard Industrial Classification for National Economic Activities (GB/T 4754-2017) and further into the enterprise level with the support of plant-based data. In cases where wastewater pretreatment systems and septic tanks exist, discharges from these facilities are referred to as the generation of water pollutants. Most of the point-source wastewater is collected and transported through municipal pipes to wastewater treatment plants (WWTPs) for pollutant removal before environmental discharge, while a small amount is directly discharged into freshwater bodies in untreated form. The WWTPs not only refer to centralized municipal WWTPs that mainly treat wastewater related to urban households, industries and businesses, but also include those built on site in rural areas to prevent domestic wastewater from polluting nearby freshwater ecosystems and groundwater. Nonpoint sources are those that contribute pollutants from dispersed areas. With the exception of a portion of rainfall collected by rainwater pipes in built-up areas, the major transport mechanisms for nonpoint pollution are surface runoff induced from rainfall in unbuilt areas and artificial drainage activities, such as paddy farming, aquaculture, and animal breeding, while the rest of the nonpoint transportation may be retained by the soil, or evaporated to the air to a small extent. Once pollutants from point and nonpoint sources reach freshwater systems, they transfer from upstream to downstream and only a certain proportion would ultimately enter the target water bodies and cause local pollution problems.
图 1 展示了 WPLT 模型的层次结构，其中水污染源被分为点和非点源。由于该模型主要关注人为影响，我们假设一些自然源，如大气沉降，可以忽略不计（刘晓等，2016）。点源包括那些容易识别来源的，例如工业生产（例如，食品加工厂、纺织厂）和居民及商业消费（例如，住宅、餐馆和购物中心）。请注意，通过按照国家标准《国民经济行业分类》（GB/T 4754-2017）将工业源细分为子行业，并在植物数据支持下进一步细分为企业层面，可以提高工业源的精确度。在存在废水预处理系统和化粪池的情况下，这些设施排放的废水被称为水污染物的产生。 大部分的点源废水被收集并通过市政管道运输到污水处理厂（WWTPs）进行污染物去除，然后再排放到环境中，而一小部分则未经处理直接排放到淡水水体中。WWTPs 不仅指主要处理城市家庭、工业和企业相关废水的集中式市政污水处理厂，还包括在农村地区建设的，以防止生活污水污染附近淡水生态系统和地下水的现场污水处理厂。非点源是指从分散区域贡献污染物的那些。除了收集到建筑区雨水管道的一部分降雨外，非点污染的主要传输机制是来自未建设区域的降雨引起的地表径流和人工排水活动，如水稻种植、水产养殖和动物饲养，其余的非点传输可能被土壤保留，或部分蒸发到空气中。 一旦点源和非点源的污染物进入淡水系统，它们就会从上游转移到下游，只有一部分最终进入目标水体，造成局部污染问题。

Details are in the caption following the image
Figure 1  图 1
Open in figure viewer
打开图库查看器PowerPoint  幻灯片

Conceptual framework of the Water Pollutant Loads Tracking model.
水污染物负荷追踪模型的概念框架。

Physical variables in the WPLT model can be classified into several processes following the life-cycle pathway of water pollutants. Taking TP for example, the starting point of the model is when TP leaves pollution generation units and the end point is when TP enters water bodies. The absorption and circulation of P within each process, such as in plants, is not considered. Variables GkPSs ${G}_{k}^{\text{PSs}}$ and GkNPSs ${G}_{k}^{\text{NPSs}}$ represent the generation of water pollutant loads from point sources and nonpoint sources, respectively; variables CkPSs ${C}_{k}^{\text{PSs}}$ and   和CkNPSs ${C}_{k}^{\text{NPSs}}$ represent the water pollutant loads collected by municipal and rainwater pipes, respectively, while variables
分别代表市政和雨水管道收集的水污染物负荷，而变量U⁢CkPSs $\mathrm{U}{\mathrm{C}}_{k}^{\text{PSs}}$ and   和U⁢CkNPSs $\mathrm{U}{\mathrm{C}}_{k}^{\text{NPSs}}$ represent those uncollected; variable
代表未收集的；变量TkPSs ${T}_{k}^{\text{PSs}}$ represents the water pollutant loads treated by WWTPs; variables
代表由污水处理厂处理的污染物负荷；变量DkWWTPs ${D}_{k}^{\text{WWTPs}}$, DkPSs ${D}_{k}^{\text{PSs}}$, and   ，并且DkNPSs ${D}_{k}^{\text{NPSs}}$ represent the water pollutant loads in treated outflow, and directly discharged from point sources and nonpoint sources, respectively; variables
代表处理后的出水污染物负荷，以及分别直接从点源和非点源排放的变量EkTWB ${E}_{k}^{\text{TWB}}$ and   和EkOWB ${E}_{k}^{\text{OWB}}$ represent the water pollutant loads entering the target and the other water bodies, respectively. Additionally, variable
分别代表进入目标和其它水体中的水污染物负荷。此外，变量LkWWTPs ${L}_{k}^{\text{WWTPs}}$ represents the water pollutant loads that are removed from wastewater outflow and accumulate to waste sludge or go into the air as a gaseous substance. Variable
表示从废水排放中去除的水污染物负荷，积累成废污泥或以气态物质的形式进入大气。变量LkPSs ${L}_{k}^{\text{PSs}}$ signifies the loss of water pollutant loads from point sources in pipe transport. Meanwhile variable
表示管道运输中点源水污染物负荷的损失。同时变量LkNPSs ${L}_{k}^{\text{NPSs}}$ represents the release of water pollutant loads from nonpoint sources into the soil and the air. The letter k indicates the type of water pollutant, such as TP.
表示非点源向土壤和空气中排放的水污染物负荷。字母 k 表示水污染物的类型，如 TP。

The main benefit of the WPLT model is the ability to identify water pollution hotspots by considering various factors, such as the source, process, and spatial and temporal aspects of the pollution, without any topographic limitation (Figure 2), which would help set more precise targets for future water pollution control interventions. The model is suitable for highly urbanized areas, and can also be feasibly adapted to other less developed areas. Another advantage of this model is that the generation and collection of water pollutants from point sources are independently analyzed, which would make them independent of each other and can be used for comparison to check the consistency.
WPLT 模型的主要优势在于能够通过考虑污染的来源、过程以及空间和时间方面等因素来识别水污染热点，没有任何地形限制（图 2），这有助于为未来的水污染控制干预措施设定更精确的目标。该模型适用于高度城市化的地区，并且可以实际地适应其他欠发达地区。该模型的另一个优点是独立分析了来自点源的水污染物的生成和收集，这使得它们相互独立，可以用于比较以检查一致性。

Details are in the caption following the image
Figure 2  图 2
Open in figure viewer
打开图库查看器PowerPoint  幻灯片

Comparison of the Water Pollutant Loads Tracking (WPLT) model with others.
比较水污染物负荷追踪（WPLT）模型与其他模型。

2.1.1 Pollutant Generation
2.1.1 污染物产生

Variable GkPSs ${G}_{k}^{\text{PSs}}$ is calculated as follows:
变量 GkPSs ${G}_{k}^{\text{PSs}}$ 的计算方法如下：
GkPSs=∑uGk,uPS ${G}_{k}^{\text{PSs}}=\sum\limits _{u}{G}_{k,u}^{\text{PS}}$ (1)
Gk,uPS⁢=∫⁢quPS⁡(t)⋅ck,uPS⁡(t)⋅d⁢t ${G}_{k,u}^{\text{PS}}=\int {q}_{u}^{\text{PS}}(t)\cdot {c}_{k,u}^{\text{PS}}(t)\cdot dt$ (2)
where Gk,uPS ${G}_{k,u}^{\text{PS}}$ is water pollutant k generated from point source unit u; quPS⁡(t) ${q}_{u}^{\text{PS}}(t)$ is the wastewater generated from point source unit u at time t;
Gk,uPS ${G}_{k,u}^{\text{PS}}$ 为由点源单元 u 产生的污染物 k； quPS⁡(t) ${q}_{u}^{\text{PS}}(t)$ 为在时间 t 由点源单元 u 产生的废水；ck,uPS⁡(t) ${c}_{k,u}^{\text{PS}}(t)$ is the concentration of water pollutant k at the outlet of point source unit u at time t. To facilitate the summary of the results, the inverse distance weighting (IDW) (Bartier & Keller, 1996), a simple and commonly used spatial interpolation method, is used to perform a point-to-raster operation on
Gk,uPS ${G}_{k,u}^{\text{PS}}$ 为由点源单元 u 产生的污染物 k； quPS⁡(t) ${q}_{u}^{\text{PS}}(t)$ 为点源单元 u 在时间 t 产生的废水； ck,uPS⁡(t) ${c}_{k,u}^{\text{PS}}(t)$ 为点源单元 u 在时间 t 出口处污染物 k 的浓度。为了便于结果总结，采用逆距离加权法（IDW）（Bartier & Keller，1996），这是一种简单且常用的空间插值方法，对点进行栅格转换操作。GkPSs ${G}_{k}^{\text{PSs}}$. It uses the distance between the interpolated point and the sample point as a weight, and samples closer to the interpolated point are assigned more weights (Hou et al., 2017). The operational process of IDW and the calculation formulas of quPS⁡(t) ${q}_{u}^{\text{PS}}(t)$ are detailed in the Supporting Information.
Gk,uPS ${G}_{k,u}^{\text{PS}}$ 为由点源单元 u 产生的污染物 k； quPS⁡(t) ${q}_{u}^{\text{PS}}(t)$ 为点源单元 u 在时间 t 产生的废水； ck,uPS⁡(t) ${c}_{k,u}^{\text{PS}}(t)$ 为点源单元 u 在时间 t 出口处污染物 k 的浓度。为了便于结果总结，采用逆距离加权法（IDW）（Bartier & Keller，1996），这是一种简单且常用的空间插值方法，对 GkPSs ${G}_{k}^{\text{PSs}}$ 进行点至栅格操作。它使用插值点与样本点之间的距离作为权重，距离插值点较近的样本赋予更高的权重（Hou 等人，2017）。IDW 的操作过程和 quPS⁡(t) ${q}_{u}^{\text{PS}}(t)$ 的计算公式详见补充信息。
Variable GkNPSs ${G}_{k}^{\text{NPSs}}$ is calculated as follows:
变量 GkNPSs ${G}_{k}^{\text{NPSs}}$ 的计算方法如下：
GkNPSs=∑i∑gGk,i,gNPS ${G}_{k}^{\text{NPSs}}=\sum\limits _{i}\sum\limits _{g}{G}_{k,i,g}^{\text{NPS}}$ (3)
Gk,i,gNPS⁢=∫⁢qi,gNPS⁡(t)⋅ck,i,gNPS⁡(t)⋅d⁢t ${G}_{k,i,g}^{\text{NPS}}=\int {q}_{i,g}^{\text{NPS}}(t)\cdot {c}_{k,i,g}^{\text{NPS}}(t)\cdot dt$ (4)
where Gk,i,gNPS ${G}_{k,i,g}^{\text{NPS}}$ is water pollutant k generated from nonpoint source of subtype i in grid g; qi,gNPS⁡(t) ${q}_{i,g}^{\text{NPS}}(t)$ is the wastewater generated from nonpoint source of subtype i in grid g at time t; ck,i,gNPS⁡(t) ${c}_{k,i,g}^{\text{NPS}}(t)$ is the concentration of water pollutant k generated from nonpoint source of subtype i in grid g at time t. The calculation formulas of
Gk,i,gNPS ${G}_{k,i,g}^{\text{NPS}}$ 为网格 g 中来自子类型 i 的非点源产生的污染物 k； qi,gNPS⁡(t) ${q}_{i,g}^{\text{NPS}}(t)$ 为网格 g 中在时间 t 由子类型 i 的非点源产生的废水； ck,i,gNPS⁡(t) ${c}_{k,i,g}^{\text{NPS}}(t)$ 为网格 g 中在时间 t 由子类型 i 的非点源产生的污染物 k 的浓度。qi,gNPS⁡(t) ${q}_{i,g}^{\text{NPS}}(t)$ are detailed in the Supporting Information.
Gk,i,gNPS ${G}_{k,i,g}^{\text{NPS}}$ 为网格 g 中来自子类型 i 的非点源产生的污染物 k； qi,gNPS⁡(t) ${q}_{i,g}^{\text{NPS}}(t)$ 为网格 g 中在时间 t 由子类型 i 的非点源产生的废水； ck,i,gNPS⁡(t) ${c}_{k,i,g}^{\text{NPS}}(t)$ 为网格 g 中在时间 t 由子类型 i 的非点源产生的污染物 k 的浓度。 qi,gNPS⁡(t) ${q}_{i,g}^{\text{NPS}}(t)$ 的计算公式详见补充信息。

2.1.2 Pollutant Collection and Treatment
2.1.2 污染物收集与处理

Variables   变量CkPSs ${C}_{k}^{\text{PSs}}$ and   和TkPSs ${T}_{k}^{\text{PSs}}$ are calculated as follows:
变量 CkPSs ${C}_{k}^{\text{PSs}}$ 和 TkPSs ${T}_{k}^{\text{PSs}}$ 的计算方法如下：
CkPSs=TkPSs1−⁢FkMP ${C}_{k}^{\text{PSs}}=\frac{{T}_{k}^{\text{PSs}}}{1-{F}_{k}^{\text{MP}}}$ (5)
TkPSs=∑uTk,uWWTP ${T}_{k}^{\text{PSs}}=\sum\limits _{u}{T}_{k,u}^{\text{WWTP}}$ (6)
Tk,uWWTP⁢=∫⁢qIn,⁢uWWTP⁡(t)⋅cIn,⁢k,uWWTP⁡(t)⋅d⁢t ${T}_{k,u}^{\text{WWTP}}=\int {q}_{\text{In},u}^{\text{WWTP}}(t)\cdot {c}_{\text{In},k,u}^{\text{WWTP}}(t)\cdot dt$ (7)
where   哪里FkMP ${F}_{k}^{\text{MP}}$ is the loss factor of pollutant k from point sources in municipal pipe transport, which is related to pipe materials, residual service life, soil types and so on (DeSilva et al., 2005);
是市政管道运输中污染源 k 的损失系数，与管道材料、剩余使用寿命、土壤类型等有关（DeSilva 等，2005 年）；Tk,uWWTP ${T}_{k,u}^{\text{WWTP}}$ is water pollutant k collected and then treated by WWTP u;
是水污染物 k 被收集后由污水处理厂 u 进行处理；qIn,⁢uWWTP⁡(t) ${q}_{\text{In},u}^{\text{WWTP}}(t)$ is the wastewater inflow to WWTP u at time t;
是时间 t 时污水处理厂 u 的废水进水；cIn,⁢k,uWWTP⁡(t) ${c}_{\text{In},k,u}^{\text{WWTP}}(t)$ is the concentration of water pollutant k in the inflow of WWTP u at time t.
FkMP ${F}_{k}^{\text{MP}}$ 为市政管道运输中点源污染物的损失系数，与管道材料、剩余使用寿命、土壤类型等有关（DeSilva 等，）； Tk,uWWTP ${T}_{k,u}^{\text{WWTP}}$ 为收集并经污水处理厂（WWTP）处理的污水污染物； qIn,⁢uWWTP⁡(t) ${q}_{\text{In},u}^{\text{WWTP}}(t)$ 为时间 t 时污水处理厂（WWTP）的废水进水量； cIn,⁢k,uWWTP⁡(t) ${c}_{\text{In},k,u}^{\text{WWTP}}(t)$ 为时间 t 时污水处理厂（WWTP）进水中水污染物浓度。
Variable CkNPSs ${C}_{k}^{\text{NPSs}}$ is calculated as follows:
变量 CkNPSs ${C}_{k}^{\text{NPSs}}$ 的计算方法如下：
CkNPSs=∑g⁢∈built−upareasGk⁢,gRR ${C}_{k}^{\text{NPSs}}=\sum\limits _{g\in \text{built}-\text{up}\,\text{areas}}{G}_{k,\mathrm{g}}^{\text{RR}}$ (8)
where Gk,gR⁢R ${G}_{k,g}^{\mathrm{R}\mathrm{R}}$ is water pollutant k generated from rainfall runoff in grid g in built-up areas.
Gk,gR⁢R ${G}_{k,g}^{\mathrm{R}\mathrm{R}}$ 是来自建成区网格 g 的降雨径流中产生的污染物 k。

2.1.3 Pollutant Discharge
2.1.3 污染物排放

Variable DkWWTPs ${D}_{k}^{\text{WWTPs}}$ is calculated as follows:
变量 DkWWTPs ${D}_{k}^{\text{WWTPs}}$ 的计算方法如下：
DkWWTPs=∑uDk,uWWTP ${D}_{k}^{\text{WWTPs}}=\sum\limits _{u}{D}_{k,u}^{\text{WWTP}}$ (9)
Dk,uWWTP⁢=∫⁢qOut,⁢uWWTP⁡(t)⋅cOut,⁢k,uWWTP⁡(t)⋅d⁢t ${D}_{k,u}^{\text{WWTP}}=\int {q}_{\text{Out},u}^{\text{WWTP}}(t)\cdot {c}_{\text{Out},k,u}^{\text{WWTP}}(t)\cdot dt$ (10)
where Dk,uWWTP ${D}_{k,u}^{\text{WWTP}}$ is water pollutant k discharged from WWTP u; qOut,⁢uWWTP⁡(t) ${q}_{\text{Out},u}^{\text{WWTP}}(t)$ is the wastewater outflow from WWTP u at time t; cOut,⁢k,uWWTP⁡(t) ${c}_{\text{Out},k,u}^{\text{WWTP}}(t)$ is the concentration of water pollutant k in the outflow of WWTP u at time t.
Dk,uWWTP ${D}_{k,u}^{\text{WWTP}}$ 为来自污水处理厂 u 排放的水污染物 k； qOut,⁢uWWTP⁡(t) ${q}_{\text{Out},u}^{\text{WWTP}}(t)$ 为污水处理厂 u 在时间 t 的废水排放； cOut,⁢k,uWWTP⁡(t) ${c}_{\text{Out},k,u}^{\text{WWTP}}(t)$ 为污水处理厂 u 在时间 t 的废水排放中水污染物 k 的浓度。
Variable DkPSs ${D}_{k}^{\text{PSs}}$ is calculated as follows:
变量 DkPSs ${D}_{k}^{\text{PSs}}$ 的计算方法如下：
DkPSs={(GkPSs−CkPSs)×(1−⁢FkSP),GkPSs≥CkPSs0,GkPSs<CkPSs ${D}_{k}^{\text{PSs}}=\left\{\begin{array}{@{}cc@{}}\left({G}_{k}^{\text{PSs}}-{C}_{k}^{\text{PSs}}\right)\times \left(1-{F}_{k}^{\text{SP}}\right)& ,{G}_{k}^{\text{PSs}}\ge {C}_{k}^{\text{PSs}}\\ 0& ,{G}_{k}^{\text{PSs}}< {C}_{k}^{\text{PSs}}\end{array}\right.$ (11)
where   哪里FkSP ${F}_{k}^{\text{SP}}$ is the loss factor of pollutant k from point sources in subsurface pipe transport, and the factors affecting it are similar to those affecting
是污染物 k 从地下管道输送中的源点损失系数，影响它的因素与影响那些的因素相似FkMP ${F}_{k}^{\text{MP}}$. Two conditions are proposed for variables
两个条件被提出用于变量GkPSs ${G}_{k}^{\text{PSs}}$ and   和CkPSs ${C}_{k}^{\text{PSs}}$: (a) if   （a）如果GkPSs≥CkPSs ${G}_{k}^{\text{PSs}}\ge {C}_{k}^{\text{PSs}}$, it reveals that part of the generated pollutants, which equals to the difference between
，揭示了部分生成的污染物，相当于差异部分GkPSs ${G}_{k}^{\text{PSs}}$ and   和CkPSs ${C}_{k}^{\text{PSs}}$, is not collected by municipal pipes but discharged directly to the water environment; (b) if
未通过市政管道收集，而是直接排放到水环境中；（b）如果GkPSs<CkPSs ${G}_{k}^{\text{PSs}}< {C}_{k}^{\text{PSs}}$, it shows that the generation of pollutants is less than the collection by municipal pipes, which indicates that point-source wastewater may be underestimated or the collection by municipal pipes is mixed with other forms of water.
FkSP ${F}_{k}^{\text{SP}}$ 是地下管道输送中污染源 k 的损失系数，影响它的因素与影响 FkMP ${F}_{k}^{\text{MP}}$ 的因素类似。针对变量 GkPSs ${G}_{k}^{\text{PSs}}$ 和 CkPSs ${C}_{k}^{\text{PSs}}$ 提出两种条件：（a）如果 GkPSs≥CkPSs ${G}_{k}^{\text{PSs}}\ge {C}_{k}^{\text{PSs}}$ ，则表明产生的部分污染物，即 GkPSs ${G}_{k}^{\text{PSs}}$ 与 CkPSs ${C}_{k}^{\text{PSs}}$ 之差，没有被市政管道收集，而是直接排放到水环境中；（b）如果 GkPSs<CkPSs ${G}_{k}^{\text{PSs}}< {C}_{k}^{\text{PSs}}$ ，则表明污染物的产生量小于市政管道的收集量，这表明点源废水可能被低估，或者市政管道的收集与其他形式的水混合。
Variable   变量DkNPSs ${D}_{k}^{\text{NPSs}}$ is calculated as follows:
如下计算：
DkNPSs=∑i∑gGk,i,gNPSs⋅fk,i,gNPS ${D}_{k}^{\text{NPSs}}=\sum\limits _{i}\sum\limits _{g}{G}_{k,i,g}^{\text{NPSs}}\cdot {f}_{k,i,g}^{\text{NPS}}$ (12)
fk,i,gNPS⁢=1−⁢Fk,i,ginterception−Fk,i,ginfiltration−Fk,i,gevaporation−Fk,i,gleakage ${f}_{k,i,g}^{\text{NPS}}=1-{F}_{k,i,g}^{\text{interception}}-{F}_{k,i,g}^{\text{infiltration}}-{F}_{k,i,g}^{\text{evaporation}}-{F}_{k,i,g}^{\text{leakage}}$ (13)
where   哪里fk,i,gNPS ${f}_{k,i,g}^{\text{NPS}}$ is the ratio of water pollutant k from nonpoint source of subtype i in grid g that enters water bodies, and it is determined by the factors of interception (
是网格 g 中来自亚型 i 的非点源污染物 k 进入水体时的比例，它由截留（Fk,i,ginterception ${F}_{k,i,g}^{\text{interception}}$), infiltration (  )，渗透（Fk,i,ginfiltration ${F}_{k,i,g}^{\text{infiltration}}$), evaporation (  )，蒸发（Fk,i,gevaporation ${F}_{k,i,g}^{\text{evaporation}}$), and leakage in rainwater pipe transport (
)，雨水管道输送泄漏（Fk,i,gleakage ${F}_{k,i,g}^{\text{leakage}}$). These factors are related to various parameters. Path curbs and field ridges play a significant role in interception (Z. Yuan et al., 2021). As for infiltration, it is impacted by a number of parameters, such as land use, soil properties, and the moisture content of the soil itself (Possantti et al., 2023). Furthermore, evaporation is influenced by atmospheric temperature and air humidity (Mukundan et al., 2020). Despite efforts to reduce pipeline water loss, leaks and bursts may still happen (Sheng et al., 2020), which are similarly affected by pipe materials, residual service life, and soil types. Interested readers can refer to the relevant literature for details as the specific modeling and calculation of these factors are beyond the scope of this study. When applying the universal calculation formulas above, they can be adapted to fit the specific local circumstances by appropriately simplifying factors that have limited effects on the results.
). 这些因素与各种参数相关。路径路缘和田间脊对拦截作用有显著影响（Z. Yuan 等，2021）。至于渗透，它受到许多参数的影响，如土地利用、土壤性质以及土壤本身的含水量（Possantti 等，2023）。此外，蒸发受大气温度和空气湿度的影响（Mukundan 等，2020）。尽管努力减少管道水损失，泄漏和爆裂仍可能发生（Sheng 等，2020），这些同样受到管道材料、剩余使用寿命和土壤类型的影响。有兴趣的读者可参考相关文献以获取详细信息，因为对这些因素的特定建模和计算超出了本研究范围。在应用上述通用计算公式时，可以通过适当简化对结果影响有限的因素来适应特定地方情况。

2.1.4 Pollutant Entering the Target Water Bodies
2.1.4 污染物进入目标水体

Variable EkTWB ${E}_{k}^{\text{TWB}}$ is calculated as follows:
变量 EkTWB ${E}_{k}^{\text{TWB}}$ 的计算方法如下：
EkTWB=(∑uDk,uWWTP⋅JuWWTP)+(DkPSs+DkNPSs)⋅f ${E}_{k}^{\text{TWB}}=\left(\sum\limits _{u}{D}_{k,u}^{\text{WWTP}}\cdot {J}_{u}^{\text{WWTP}}\right)+\left({D}_{k}^{\text{PSs}}+{D}_{k}^{\text{NPSs}}\right)\cdot f$ (14)
JuWWTP={1,u⁢∈TW0,u⁢∉TW ${J}_{u}^{\text{WWTP}}=\left\{\begin{array}{@{}cc@{}}1& ,u\in \text{TW}\\ 0& ,u\notin \text{TW}\end{array}\right.$ (15)
f=NTWNALL $f=\frac{{N}_{\text{TW}}}{{N}_{\text{ALL}}}$ (16)
where JuWWTP ${J}_{u}^{\text{WWTP}}$ is the discriminative factor for water pollutants from WWTP u to target water bodies (JuWWTP⁢=1 ${J}_{u}^{\text{WWTP}}=1$) and to other water bodies (JuWWTP⁢=0 ${J}_{u}^{\text{WWTP}}=0$); f is the ratio of the number of drainage outlets in the target water bodies (NTW) to that in the whole study area (NALL), which indicates the proportion of direct discharges from point and nonpoint sources that enter the target water bodies.
JuWWTP ${J}_{u}^{\text{WWTP}}$ 是从污水处理厂 u 到目标水体（ JuWWTP⁢=1 ${J}_{u}^{\text{WWTP}}=1$ ）和其他水体（ JuWWTP⁢=0 ${J}_{u}^{\text{WWTP}}=0$ ）的水污染物判别因子；f 是目标水体（N TW ）中排水出口数量与整个研究区域（N ALL ）中排水出口数量的比率，这表示直接排放到目标水体的点源和非点源排放的比例。

2.2 Study Area  2.2 研究区域

Here, this study applied the WPLT model to a rapidly urbanizing watershed in Taihu Lake Basin, China to trace the TP loads in 2017. Taihu Lake has been suffered from eutrophication mainly caused by excessive inputs of pollutants from surrounding areas (Yi et al., 2017). Therefore, Wangyu River, a north–south channel that connects the Yangtze River and Taihu Lake, plays a critical role in diverting water from the Yangtze River into Taihu Lake to curb its pollution. The study area is located on the west bank of Wangyu River in Wuxi City, where two rivers, Jiuli River and Bodugang River, flow through and into Wangyu River. A total of 234.26 km2 of the reaches of Jiuli River and Bodugang River, from upstream sluices to downstream assessment sections, are selected as the study area (Figure S1 in Supporting Information S1). The terrain boundary is set comprehensively based on information about the coverage areas of WWTPs, administrative divisions, and water system distribution. A national water quality monitoring site is placed at the confluence of Jiuli River and Wangyu River, and a provincial water quality monitoring site is placed at the confluence of Bodugang River and Wangyu River, which provide important information for assessing the performance of local water pollution control.
此处，本研究将 WPLT 模型应用于中国太湖流域快速城市化的流域，以追踪 2017 年的 TP 负荷。太湖一直遭受富营养化之苦，主要由周边地区污染物过量输入引起（Yi 等，2017）。因此，王渔河，一条连接长江和太湖的南北通道，在将长江水引入太湖以遏制其污染方面发挥着关键作用。研究区域位于无锡市王渔河西岸，这里有两条河流，即九里河和博徒港河，流经并汇入王渔河。从上游闸门到下游评估段，共选取了九里河和博徒港河的 234.26 公里作为研究区域（支持信息 S1 中的图 S1）。地形边界是基于污水处理厂覆盖区域、行政区划和水系分布信息综合设定的。 全国水质监测站点位于巨流河与王渔河交汇处，省级水质监测站点位于博度岗河与王渔河交汇处，这些站点为评估当地水污染控制效果提供了重要信息。

The number of grids for global studies with a spatial resolution of 5 arc-minutes (Mekonnen & Hoekstra, 2015; Schyns et al., 2019) was counted as 9,931,200 while that with 0.5° (Whelan et al., 2007; J. Zhou et al., 2022) was 259,200. Accordingly, the alternative resolutions of the study area were calculated, namely 5 m × 5 m and 30 m × 30 m, and they were visually compared with the initial resolution of the remote sensing image (0.5 m × 0.5 m) for the presentation of land use in the same plot (Figure S2 in Supporting Information S1). Finally, the resolution of 5 m × 5 m was selected, which can not only ensure high calculation accuracy and accurate location, but also reduce the difficulty and workload of pollution source control.
全球研究空间分辨率为 5 弧分（Mekonnen & Hoekstra，2015；Schyns 等，2019）的网格数量为 9931200，而 0.5°（Whelan 等，2007；J. Zhou 等，2022）的为 259200。据此，计算了研究区域的不同分辨率，即 5 米×5 米和 30 米×30 米，并将它们与遥感图像的初始分辨率（0.5 米×0.5 米）进行视觉比较，以展示同一地块的土地利用情况（支持信息 S1 中的图 S2）。最后，选择了 5 米×5 米的分辨率，这不仅能够确保高计算精度和准确的位置，还能降低污染源控制的难度和工作量。

The proportion of constructed regions in the city area of Wuxi rose from 5.72% in 1998 to 20.56% in 2017, with an increase of 259.57% (WMSB & NBSSOW, 2018). In the study area, the constructed regions, including industrial land, urban residence, commercial land, and road, account for about 25.86%. Meanwhile, the cropland, forest, and water bodies occupy about 31.66%, 20.06%, and 7.82% of the total area, respectively. There are five rainfall observation stations in the study area, which record rainfall data on a daily basis. In addition, the area has three municipal WWTPs (i.e., Meicun, Ehu, and Anzhen) and 239 rural on-site WWTPs. The designed daily wastewater treatment capacity of Meicun, Ehu, and Anzhen WWTPs are 110,000 tons, 10,000 tons, and 20,000 tons, respectively. Meicun WWTP adopts both AAO (anaerobic/anoxic/oxic)—SBR (sequencing batch reactor)—cloth media filter and MBR (membrane bioreactor) technologies, Ehu WWTP uses the improved AAO technology, and the wastewater treatment process of Anzhen WWTP is based on the SBR technology. In rural areas, improved AAO—constructed wetlands technology is mostly applied in on-site WWTPs. Rice–wheat rotation is one of the main agricultural production systems in the study area, of which rice is planted in paddy soils from June to November. The field is maintained by ponding water throughout the rice growth period, and surface runoff can be caused by heavy rainfall events and intentional drainage by farmers around August. Aquaculture in this region is dominated by black carp culture in ponds, with a small portion of other fishes such as silver carp and crustaceans. According to our investigation, these ponds would have drained effluents after fish harvesting, usually at the beginning of each year. There are a few pig and poultry production facilities in the east rural area of this study, all of which have been installed with septic tank systems for liquid manure storage and treatment.
无锡市区建成区比例从 1998 年的 5.72%增长到 2017 年的 20.56%，增长了 259.57%（WMSB & NBSSOW，2018）。研究区域内，建成区包括工业用地、城市住宅、商业用地和道路，占约 25.86%。同时，耕地、森林和水域分别占总面积的 31.66%、20.06%和 7.82%。研究区域内有五个降雨观测站，每日记录降雨数据。此外，该区域有三个市政污水处理厂（即梅村、鹅湖和安镇）和 239 个农村就地污水处理厂。梅村、鹅湖和安镇污水处理厂设计的日污水处理能力分别为 11 万吨、1 万吨和 2 万吨。梅村污水处理厂采用 AAO（厌氧/缺氧/好氧）—SBR（连续批式反应器）—布料介质过滤器和 MBR（膜生物反应器）技术，鹅湖污水处理厂使用改进的 AAO 技术，安镇污水处理厂的水处理工艺基于 SBR 技术。在农村地区，就地污水处理厂主要应用改进的 AAO—人工湿地技术。 水稻-小麦轮作是该研究区域的主要农业生产系统之一，其中水稻从六月到十一月种植在稻田土壤中。在整个水稻生长期内，田间通过蓄水来维护，表面径流可能由强降雨事件和农民在八月份的故意排水引起。该地区的水产养殖以池塘中的黑鱼养殖为主，还有一小部分其他鱼类如银鱼和甲壳类。根据我们的调查，这些池塘在捕鱼后通常会每年年初排放废水。在该研究区域的东部农村地区有一些猪和家禽生产设施，所有这些设施都安装了化粪池系统用于液体粪便的储存和处理。

2.3 Data Sources  2.3 数据来源

The application of the WPLT model was supported by collecting study area data from various sources. Note that the availability and quality of data varied considerably within and between key parameters, and we followed the multi-source data integration method (Z. Yuan et al., 2018) to prepare input data sets. The brief description of raw data is shown in Table 1, and more details are provided in the Supporting Information.
WPLT 模型的运用得到了从多个来源收集研究区域数据的支持。请注意，数据在关键参数内部和之间的可用性和质量差异很大，我们遵循了多源数据集成方法（Z. Yuan 等，2018）来准备输入数据集。原始数据的简要描述如表 1 所示，更多细节请参阅支持信息。

Table 1. Data Requirements of the Water Pollutant Loads Tracking Model
Processes  过程 Data requirements  数据需求
Generation  生成 • Geographic information: Latitude–longitude coordinates of point sources, land use types, soil types, etc.
地理信息：点源经纬度坐标、土地利用类型、土壤类型等。
• Activity data: Industrial output values, population, precipitation, planting area, number of livestock animals, etc.
• 活动数据：工业产出值、人口、降水量、种植面积、牲畜数量等。
• Parameters: Wastewater generation coefficients, TP concentration, etc.
• 参数：废水产生系数、TP 浓度等。
Collection and treatment
收集与处理		 • Geographic information: Coverage areas of WWTPs
• 地理信息：污水处理厂覆盖区域
• Activity data: Inflow volume of WWTPs
• 活动数据：污水处理厂进水量
• Parameters: TP concentration in WWTP inflow
• 参数：污水处理厂进水中的 TP 浓度
Discharge  排出 • Geographic information: None
• 地理信息：无
• Activity data: Outflow volume of WWTPs
• 活动数据：污水处理厂排放量
• Parameters: TP concentration in WWTP outflow, the ratio of TP from nonpoint sources that enters water bodies
• 参数：污水处理厂出水中的 TP 浓度，进入水体中非点源 TP 的比率
Entering the target water bodies
进入目标水体		 • Geographic information: Location of WWTP drainage outlets
• 地理信息：污水处理厂排水出口位置
• Activity data: Numbers of drainage outlets in the target water bodies and whole study area
• 活动数据：目标水体和整个研究区域的排水出口数量
• Parameters: None

At the generation process of point sources, we were able to collect plant-level information for 4,100 enterprises on their names, unified social credit codes, latitude–longitude coordinates, output values in 2017, and industrial category to which they belong. In addition, the amount of water pollutants generated by key enterprises in 2017 under monitoring was used to calculate industry-specific pollutant generation coefficients (Table S2 in Supporting Information S1). As for residential and commercial consumption, three main input data sets with temporal and spatial variations were used. The first is the population map, which was derived from WorldPop (Tatem, 2017) and then corrected with township-level statistics (Table S3 in Supporting Information S1) (ABXW, 2018; WMSB & NBSSOW, 2018; WND, 2018); the second is daily wastewater generation per capita (Table S4 in Supporting Information S1), estimated either from water use statistics (WMSB & NBSSOW, 2018) and the national standard Code for Urban Wastewater and Stormwater Engineering Planning (GB 50318-2017) or from wastewater inputs to WWTPs; the third is the concentrations of water pollutants (Table S5 in Supporting Information S1), which were obtained from the 24 hr field monitoring experiments at the drainage outlets of five representative residential areas. Since the experiments were carried out in the spring, we also calculated seasonal correction factors of pollutant concentrations in other seasons (Z. Liu et al., 2009; Ma et al., 2013).
在点源生成过程中，我们收集了 4100 家企业的植物级信息，包括企业名称、统一社会信用代码、经纬度坐标、2017 年产出值以及所属行业类别。此外，利用 2017 年监测的关键企业产生的水污染物量来计算行业特定的污染物生成系数（见支持信息 S1 中的表 S2）。至于居民和商业消费，使用了三个具有时间和空间变化的输入数据集。 第一个是人口地图，该地图来源于 WorldPop（Tatem，2017），然后通过乡镇级统计数据（S1 支持信息中的表 S3）（ABXW，2018；WMSB & NBSSOW，2018；WND，2018）进行校正；第二个是人均日污水产生量（S1 支持信息中的表 S4），估计值来自用水统计数据（WMSB & NBSSOW，2018）和《城市污水和雨水工程规划国家标准》（GB 50318-2017）或来自污水处理厂（WWTPs）的污水输入；第三个是水污染物浓度（S1 支持信息中的表 S5），这些数据是通过在五个代表性住宅区的排水出口进行的 24 小时现场监测实验获得的。由于实验是在春季进行的，我们还计算了其他季节污染物浓度的季节校正系数（Z. Liu 等人，2009；Ma 等人，2013）。

At the generation process of nonpoint sources, a localized Soil Conservation Service Curve Number (SCS-CN) method was used to calculate the rainfall-runoff, and the daily rainfall data was collected from local rainfall observation stations. The SCS-CN method can relate precipitation, soil, vegetation, and land use relationships with the values of curve numbers to estimate the potential maximum retention, and further the direct runoff (Mishra et al., 2006). By conducting natural and simulated rainfall observation experiments, we measured the initial abstraction of rainfall, and then localized the initial abstraction rate (Ji et al., 2022). In addition, we investigated the artificial drainage activities from agricultural sources. The rice cultivation area (Table S6 in Supporting Information S1) was collected from the local government and the accumulated drainage depth (Table S7 in Supporting Information S1) was examined as the interval of 3.00–8.30 cm. The pollutant concentration (Table S8 in Supporting Information S1) was taken from the existing literature (Z. Liu et al., 2009; Ma et al., 2013; D. Zhang et al., 1997). The similar data acquisition was also applied for aquaculture (Tables S9 and S10 in Supporting Information S1), and the difference in pollutant concentration between inflow and outflow (Table S11 in Supporting Information S1) was from the previous study (Ma et al., 2013). As the pig and poultry production facilities have treated the wastewater and recycled the outflow to nearby farmlands, here we only took the extensive animal raising into account. The stock and slaughter of pig and poultry (Table S12 in Supporting Information S1) were collected from subtown-level statistics, and the annual pollutant generation coefficients (Table S14 in Supporting Information S1) were based on the literature (DSTEMARAC & OLGFNCPS, 2009; Ma et al., 2013). The land use was determined by using GeoEye-1 image of 2017 with a spatial resolution of 0.5 m × 0.5 m (later upscaled to 5 m × 5 m), and in total 10 types of land use were classified according to the national standard Current Land Use Classification (GB/T 21010-2017).
在非点源生成过程中，采用了局部化的土壤保持服务曲线数（SCS-CN）方法来计算降雨径流，并从当地降雨观测站收集了每日降雨数据。SCS-CN 方法可以通过曲线数来关联降水、土壤、植被和土地利用之间的关系，以估算潜在最大滞留量，以及进一步估算直接径流（Mishra 等，2006）。通过进行自然和模拟降雨观测实验，我们测量了降雨的初始蒸发量，然后局部化了初始蒸发率（Ji 等，2022）。此外，我们还调查了农业来源的人工排水活动。水稻种植面积（支持信息 S1 中的表 S6）来自当地政府，累积排水深度（支持信息 S1 中的表 S7）以 3.00-8.30 厘米为间隔进行考察。污染物浓度（支持信息 S1 中的表 S8）取自现有文献（Z. Liu 等，2009；Ma 等，2013；D. Zhang 等，1997）。 相似的数据采集也应用于水产养殖（支持信息 S1 中的表 S9 和 S10），以及进水和出水污染物浓度的差异（支持信息 S1 中的表 S11）来自先前的研究（Ma 等，2013）。由于猪和家禽生产设施已处理废水并将出水回收到附近的农田，因此在此我们只考虑了粗放型畜牧业。猪和家禽的存栏和屠宰量（支持信息 S1 中的表 S12）来自次城镇级别的统计数据，而年污染物产生系数（支持信息 S1 中的表 S14）基于文献（DSTEMARAC & OLGFNCPS，2009；Ma 等，2013）。土地利用是通过 2017 年的 GeoEye-1 图像确定的，分辨率为 0.5 m × 0.5 m（后来扩大到 5 m × 5 m），总共根据国家标准《当前土地利用分类》（GB/T 21010-2017）分类了 10 种土地利用类型。

At the processes of collection, treatment, discharge, and entering the target water bodies, the location and the number of drainage outlets for the three municipal WWTPs and 239 sets of rural on-site WWTPs were obtained, as well as their designed daily wastewater treatment capacities (tons/day). In addition, we collected the actual daily wastewater volume and pollutant concentration in inflow and outflow for all the three municipal WWTPs and 28 sets of rural on-site WWTPs in 2017. The ratio of pollutants from nonpoint sources that enter water bodies was obtained from the literature (Ma et al., 2013; Tan, 2016).
在收集、处理、排放以及进入目标水体过程中，获得了三个城市污水处理厂和 239 套农村就地污水处理设施的排水口位置和数量，以及它们设计的每日废水处理能力（吨/天）。此外，我们还收集了 2017 年三个城市污水处理厂和 28 套农村就地污水处理设施的实际每日废水体积和污染物浓度。非点源污染物进入水体的比例来自文献（Ma 等，2013；Tan，2016）。

3 Results  3 结果

3.1 Characterization of the Life-Cycle Pathway
3.1 生命周期途径表征

In 2017, the total TP generation in the study area was estimated to be 261.55 t (Figure 3), of which residential and commercial consumption contributed nearly three-quarters (74.75%), followed by rainfall runoff (17.88%) and industrial production (5.04%). Due to the small scale of paddy farming, aquaculture, and animal breeding, these agricultural sources contributed little to TP. However, in terms of the volume of wastewater, rainfall-runoff accounted for up to 50.70%, which was higher than industrial production (23.45%) and residential and commercial consumption (20.83%) combined (Figure S6 in Supporting Information S1). The difference in the composition of wastewater volumes and TP loads is mainly caused by the relatively higher TP concentration in household wastewater, ranging from 3.93 to 6.79 mg/L, than rainwater runoffs and most of the pretreated industrial wastewater. The major industries in this area were textile and equipment manufacturing, which produced wastewater with TP concentration of 0.25–0.74 mg/L. In addition, before entering the municipal pipes, industrial wastewater must meet local discharge standards and is supervised by real-time online monitoring system. Household wastewater is usually divided into blackwater, which is discharged from toilets, and greywater from all the other sources, such as kitchen sink, washing machine, shower, etc. The TP concentration of household wastewater in this study is in line with those found in other Asian countries, around 6 mg/L for Indonesia, Vietnam, and Malaysia (Widyarani et al., 2022).
2017 年，研究区域总 TP 产生量估计为 261.55 吨（图 3），其中住宅和商业消费贡献了近四分之三（74.75%），其次是雨水径流（17.88%）和工业生产（5.04%）。由于水稻种植、水产养殖和畜牧业规模较小，这些农业来源对 TP 的贡献很小。然而，从废水体积来看，雨水径流占比高达 50.70%，超过了工业生产（23.45%）和住宅及商业消费（20.83%）的总和（支持信息 S1 中的图 S6）。废水体积和 TP 负荷组成的差异主要是由于家庭污水中 TP 浓度相对较高，介于 3.93 至 6.79 mg/L 之间，高于雨水径流和大部分预处理后的工业废水。该区域主要产业为纺织和设备制造，产生的废水 TP 浓度为 0.25–0.74 mg/L。此外，工业废水在进入市政管道前必须符合当地排放标准，并受到实时在线监控系统监督。 家庭污水通常分为黑水，即从厕所排放的污水，以及来自其他来源的灰水，如厨房水池、洗衣机、淋浴等。本研究中家庭污水的 TP 浓度与其他亚洲国家的研究结果相符，约为印度尼西亚、越南和马来西亚的 6 mg/L（Widyarani 等，2022）。

Details are in the caption following the image
Figure 3  图 3
Open in figure viewer
打开图库查看器PowerPoint  幻灯片

Life-cycle pathway of total phosphorus (TP) from sources to recipient water bodies.
生命周期中总磷（TP）从来源到受纳水体的路径。

Located in the most developed regions of China, the study area is facilitated with a wastewater treatment system that has a wide range of coverage. In 2017, the total TP collected by WWTPs in the study area was about 168.24 t (Figure 3), occupying 64.32% of the total TP generation. The three municipal WWTPs together contributed more than 97% to the total TP collection of point sources, while rural on-site WWTPs accounted for less than 3%. It is mainly because that, though the number of rural on-site WWTPs is about 80 times that of municipal WWTPs, the design capacity of daily wastewater treatment of rural on-site WWTPs (2–120 tons) is far less than the latter (10,000–110,000 tons). In 2017, the average wastewater processing volume of all the rural on-site WWTPs only occupied no more than 0.03% of that of municipal WWTPs. As the increase in rural young population in urban employment, many rural houses covered by rural on-site WWTPs are vacant most time of the year, making it difficult for these facilities to run steadily. Improper and delayed equipment operation and maintenance are also possible reasons. Therefore, through the implementation of the rural revitalization strategy, the local government has made great efforts to improve rural wastewater treatment, for example, by connecting rural branch pipelines to main pipelines using pumping stations and onward to municipal WWTPs.
位于中国最发达地区的研究区域配备了覆盖范围广泛的污水处理系统。2017 年，研究区域内污水处理厂收集的总 TP 约为 168.24 吨（图 3），占 TP 总产量的 64.32%。三个市政污水处理厂共同贡献了超过 97%的点源 TP 收集量，而农村就地污水处理厂仅占不到 3%。这主要是因为，尽管农村就地污水处理厂数量约为市政污水处理厂的 80 倍，但农村就地污水处理厂每日污水处理设计能力（2-120 吨）远低于后者（10,000-110,000 吨）。2017 年，所有农村就地污水处理厂的平均污水处理量仅占市政污水处理厂总量的 0.03%以下。随着农村年轻人口在城市就业的增加，许多被农村就地污水处理厂覆盖的农村房屋在一年中的大部分时间都是空置的，这使得这些设施难以稳定运行。设备操作和维护不当及延迟也是可能的原因。 因此，通过实施乡村振兴战略，地方政府做出了巨大努力来改善农村污水处理，例如，通过使用泵站将农村支管道连接到主管道，并进一步连接到市政污水处理厂。

We estimated the wastewater collection rate for each municipal WWTP, finding that Meicun WWTP has a much higher value (95.07%) than the other two (Anzhen WWTP: 70.24%; Ehu WWTP: 73.13%). All the three values are in accord with the empirical experience of WWTP managers. This is also consistent with the order of urbanization level of the three regions (Meicun WWTP area > Ehu WWTP area > Anzhen WWTP area) (Figure S10 in Supporting Information S1). The total wastewater treatment volume of Meicun WWTP in 2017 was about 7 times and 13 times that of Anzhen WWTP and Ehu WWTP, respectively. Despite the difference in volume, there is no significant variation in the values of TP concentration of inflow to these municipal WWTPs. However, it should be noted that, the terrain of the study area is flat, slowly lowing from west to east. In a few eastern areas, the wastewater in rural branch pipelines cannot rely gravity to flow into the main pipelines. The current practice is to divert the wastewater to nearly ponds for storage and wait for further treatment procedures. This not only makes the municipal pipes useless, but also seriously pollutes the local water bodies.
我们估计了每个市政污水处理厂（WWTP）的废水收集率，发现梅村 WWTP 的值（95.07%）远高于其他两个（安镇 WWTP：70.24%；鹅湖 WWTP：73.13%）。这三个值都与污水处理厂管理人员的经验相符。这也与三个地区的城市化水平顺序一致（梅村 WWTP 区域 > 鹅湖 WWTP 区域 > 安镇 WWTP 区域）（支持信息 S1 中的图 S10）。2017 年梅村 WWTP 的总废水处理量分别是安镇 WWTP 和鹅湖 WWTP 的约 7 倍和 13 倍。尽管处理量有差异，但这些市政污水处理厂进水中的总磷（TP）浓度值没有显著变化。然而，需要注意的是，研究区域的地面是平坦的，从西向东缓慢下降。在少数东部地区，农村支管道中的废水不能依靠重力流入主管道。目前的做法是将废水转移到近池塘进行储存，等待进一步的处理程序。这不仅使市政管道无用，还严重污染了当地水体。

After wastewater treatment and runoff interception loss, the contribution of various sources to the environmental discharge differs greatly from that to the generation, implying a substantial gap between the two processes. In 2017, around 69.20 t of TP was discharged to freshwater bodies (Figure 3), which was about 26.46% of the amount generated. Direct discharge accounted for more than half (58.48%), followed by rainfall runoff (27.79%), and WWTPs (11.54%). The large proportion of direct discharge is mainly because the wastewater in some rural areas cannot be effectively collected by municipal WWTPs, and there may be illegal industrial wastewater discharge to save disposal costs. The contribution of direct discharge to wastewater volume (5.90%) is less than one-eighth of that of WWTPs (48.77%), but its contribution to TP is more than five times that of WWTPs, implying the high efficiency of WWTPs in removing TP. About 160.26 t of TP was removed from the wastewater outflow using chemical or bio-based technologies and then accumulated in the sewage sludge, with an average TP removal efficiency of 95.25%. Also, thanks to the high TP removal efficiency of WWTPs, the contribution of point sources at the discharge process is less significant than that at the generation process. The TP concentrations in treated outflow all meet the national requirement for discharges from municipal WWTPs (GB18918-2002). However, the sampling results of 28 rural on-site WWTPs (0.02–5.80 mg/L) show that a few of them fail to comply with local regulations (0.5 mg/L), possibly due to backward treatment technology and equipment maintenance.
在废水处理和径流拦截损失之后，各种来源对环境排放的贡献与对产生的贡献差异很大，暗示了这两个过程之间存在巨大差距。2017 年，大约有 69.20 吨 TP 排放到淡水水体中（图 3），这大约是产生量的 26.46%。直接排放占超过一半（58.48%），其次是降雨径流（27.79%），然后是污水处理厂（11.54%）。直接排放所占比例较大主要是由于某些农村地区的废水无法被市政污水处理厂有效收集，并且可能存在非法工业废水排放以节省处理成本。直接排放对废水体积的贡献（5.90%）不到污水处理厂（48.77%）的八分之一，但其对 TP 的贡献超过污水处理厂的五倍，暗示污水处理厂在去除 TP 方面的高效率。大约有 160.26 吨 TP 通过化学或基于生物的技术从废水排放中去除，然后积累在污泥中，平均 TP 去除效率为 95.25%。 此外，得益于污水处理厂（WWTPs）的高 TP 去除效率，排放过程中的点源贡献比生成过程中的贡献小得多。处理后的出水 TP 浓度均符合国家关于城市污水处理厂排放的要求（GB18918-2002）。然而，28 个农村就地污水处理厂的采样结果（0.02–5.80 mg/L）显示，其中一些未能符合当地规定（0.5 mg/L），可能由于落后的处理技术和设备维护。

In 2017, it is estimated that 3.27 t of TP finally entered the reaches of Jiuli River and Bodugang River (Figure 3), accounting for about 1.25% of TP generation and 4.72% of TP discharge. The contribution of various sources to the TP entering the target water bodies further changed. WWTPs, especially municipal ones, were the major contributors, representing 62.53% of the TP loads to the target water bodies. The dominance of WWTPs can be attributed to two possible reasons: on one hand, two of the three municipal WWTPs (Anzhen and Ehu WWTPs) have their drainage outlets in the reaches of Jiuli River and Bodugang River; on the other hand, direct discharges from point sources and runoff from nonpoint sources can reach freshwater systems at various spatial locations, which may not necessarily enter the target water bodies and therefore play a far less significant role in the TP loads at the recipient side.
2017 年，估计有 3.27 吨 TP 最终进入嘉陵江和北干河（图 3），占 TP 产生的约 1.25%和排放的 4.72%。各种来源对进入目标水体的 TP 的贡献进一步变化。污水处理厂，尤其是市政污水处理厂，是主要贡献者，代表了目标水体 TP 负荷的 62.53%。污水处理厂的主导地位可以归因于两个可能的原因：一方面，三个市政污水处理厂中的两个（安贞和鹅湖污水处理厂）的排水口位于嘉陵江和北干河的流域内；另一方面，来自点源的直接排放和非点源的径流可以在不同的空间位置到达淡水系统，这些可能不会进入目标水体，因此在接收端的 TP 负荷中发挥的作用远不如前者显著。

3.2 Spatial Interpretation of Generation Pattern
3.2 生成模式的空間解釋

The 5 m × 5 m spatial generation pattern of TP for study area in 2017 is illustrated in Figure 4. The hotspots of TP generation show a C-shaped distribution, with the highest generation densities found in the southwestern parts, which have relatively higher population density and are more urbanized. In general, residential and commercial consumption has established the overall spatial pattern, and industrial production further defined hotspots in certain areas where industries cluster, such as textile and equipment manufacturing. We find that, of all the 4,100 industrial enterprises, one textile enterprise and one equipment manufacturing enterprise respectively contributed to 18.18% and 14.13% of the TP generated by industrial production, while the rest 3,198 industrial enterprises performed very similar (0.00%–2.81%). In addition, 144 industrial enterprises were estimated to contribute 80.01% of the TP generated by industrial production, of which 30 enterprises are in the textile sector (Codes C17 and C18), and 90 are in the equipment manufacturing sector (Codes C33, C34, C35, C38, and C39).
2017 年研究区域 TP 的 5m×5m 空间生成模式如图 4 所示。TP 生成热点呈 C 形分布，西南部生成密度最高，该地区人口密度相对较高且城市化程度更高。总体而言，住宅和商业消费已确立整体空间模式，而工业生产进一步定义了某些产业集群区域的亮点，如纺织和设备制造。我们发现，在所有 4100 家工业企业中，一家纺织企业和一家设备制造企业分别贡献了工业生产产生的 TP 的 18.18%和 14.13%，而其余 3198 家工业企业表现非常相似（0.00%–2.81%）。此外，估计有 144 家工业企业贡献了工业生产产生的 TP 的 80.01%，其中 30 家企业属于纺织行业（代码 C17 和 C18），90 家企业属于设备制造行业（代码 C33、C34、C35、C38 和 C39）。

Details are in the caption following the image
Figure 4  图 4
Open in figure viewer
打开图库查看器PowerPoint  幻灯片

Spatial distribution of total phosphorus (TP) generation from sources of industrial production (IP), residential and commercial consumption (RCC), rainfall runoff (RR), paddy farming (PF), aquaculture (AQ), and animal breeding (AB).
空间分布：来自工业生产（IP）、居民和商业消费（RCC）、降雨径流（RR）、水稻种植（PF）、水产养殖（AQ）和畜牧业（AB）来源的总磷（TP）产生。

Due to the increase of impermeable surface cover, it was estimated that urban residence contributed the greatest of 17.29% to rainfall runoff calculated using the SCS-CN method. It is mainly because that the runoff depths of impermeable areas are about three to four times that of permeable ones, such as green land and cropland. As for artificial drainage activities, we find that the distribution of TP generated from paddy farming is very similar across the arable land, possibly the result of well-guided cultivation techniques. Animal breeding has a limited contribution to TP generation, merely from backyard pig and poultry breeding scattered in a few rural areas. TP generated from aquaculture is more concentrated eastward, where the river network is denser. Jiangsu Province is a pioneer in aquaculture pollution control by recently issuing the Discharge Standard of Water from Aquaculture Ponds (DB32/4043-2021), which requires the TP discharge concentration in Taihu Lake Basin to not exceed 0.4 mg/L. The successful implementation would hopefully improve the environmental performance of aquaculture and prevent pollution of water environment.
由于不透水地表覆盖的增加，估计城市居民对使用 SCS-CN 方法计算的雨水径流贡献最大，达到 17.29%。这主要是因为不透水区域的径流深度大约是透水区域（如绿地和耕地）的三到四倍。至于人工排水活动，我们发现来自稻田耕作的 TP 分布在整个耕地中非常相似，可能是由于良好的栽培技术指导。畜牧业对 TP 产生的贡献有限，仅来自少数农村地区的散养猪和家禽。来自水产养殖的 TP 更集中于东部，那里河流网络更密集。江苏省最近发布了《水产养殖池塘出水排放标准》（DB32/4043-2021），作为水产养殖污染控制的先驱，该标准要求太湖流域的 TP 排放浓度不超过 0.4 mg/L。成功的实施有望提高水产养殖的环境绩效并防止水环境污染。

3.3 Temporal Changes of Generation Pattern
3.3 代际模式的时间变化

Based on information such as the daily inflow and outflow data of WWTPs and the daily rainfall data of observation stations, Figure 5 depicts the temporal distribution of TP at the three processes of generation, discharge, and entering the target water bodies on the monthly and daily scales. In general, the TP of various sources at the generation and discharge processes peaked in September, followed by a small peak in April. The monthly fluctuation of TP entering the target water bodies is more gentle than that at the discharge process.
基于污水处理厂每日进出水数据以及观测站每日降雨数据，图 5 展示了 TP 在生成、排放和进入目标水体三个过程中的时间分布，按月度和日度尺度。总体而言，生成和排放过程中各种来源的 TP 在 9 月份达到峰值，随后在 4 月份出现一个小峰值。TP 进入目标水体的月度波动比排放过程中的波动更为平缓。

Details are in the caption following the image
Figure 5  图 5
Open in figure viewer
打开图库查看器PowerPoint  幻灯片

Temporal distribution of total phosphorus generation, discharge, and entering the target water bodies on the monthly (a–c) and daily (d–f) scales.
时间尺度上总磷生成、排放和进入目标水体的情况，按月（a-c）和按日（d-f）划分。

The peak periods of TP generated from point sources are March, April, May, November, and December, while August and September are its relatively obvious trough periods. According to our estimates, the daily maximum and minimum value of TP generated from point sources occurred in February (1,232.67 kg) and September (159.41 kg). Considering that industrial production is usually in a stable status, the fluctuation of point-source generation is interpreted to be mainly caused by residential and commercial consumption. We also notice that the temporal trend of TP loads and wastewater volume are somewhat opposite (Figure S16 in Supporting Information S1). Households tend to reduce the frequency of bathing, laundry, car washing, and other cleaning activities in cold seasons, and foods that are rich in animal fat and meat such as hot pot and barbecue are more preferred in winter diets (Hu & Lian, 2012). These factors possibly increase the TP concentration in wastewater and lead to the above-mentioned peak periods. Other findings also showed that seasonal changes in concentration have a greater effect on loads than wastewater volume (Peng et al., 2012; L. Wang, 2018). The monthly fluctuation of point-source TP shows a relatively close trend at the processes of discharge and entering the target water bodies, which is less around April while more around September, possibly due to the temperature that affects the wastewater treatment process of municipal WWTPs (C. Li et al., 2019).
高峰期 TP 来自点源生成的是 3 月、4 月、5 月、11 月和 12 月，而 8 月和 9 月是其相对明显的低谷期。根据我们的估计，来自点源的 TP 日最大值和最小值分别出现在 2 月（1,232.67 千克）和 9 月（159.41 千克）。考虑到工业生产通常处于稳定状态，点源生成的波动被解释为主要由居民和商业消费引起。我们还注意到，TP 负荷量和废水量的时间趋势在一定程度上是相反的（支持信息 S1 中的图 S16）。家庭在寒冷季节倾向于减少洗澡、洗衣、洗车等清洁活动的频率，而在冬季饮食中更偏好富含动物脂肪和肉类的食物，如火锅和烧烤（Hu & Lian，2012）。这些因素可能增加废水中 TP 的浓度，导致上述峰值期。其他研究也表明，季节性变化对负荷量的影响大于废水量（Peng 等，2012；L. Wang，2018）。 月均点源 TP 波动在排放和进入目标水体过程中呈现相对接近的趋势，4 月左右较少，9 月左右较多，可能由于影响市政污水处理厂（WWTPs）废水处理过程的温度（C. Li 等人，2019）。

For nonpoint sources, the monthly fluctuation of TP shows a consistent trend at the three processes, that is, around August and September is the top peak period, and April is the second peak period. This fluctuation trend is basically consistent with local rainfall records, of which the rainy season from August to September is characterized with heavy rainfall events with high occurrence frequency. The local maximum daily rainfall in 2017 appeared on 25 September, ranging from 92.0 to 200.3 mm among five stations. Also, small rainfall intervals increase the soil moisture to a relatively saturated level, which in turn would lead to higher runoff (Z. Yuan et al., 2021). Therefore, on the daily scale, the TP loads of nonpoint sources at the three processes which we estimated all reached the maximum values (7,434.26, 2,834.22, and 56.68 kg, respectively). In addition, the artificial drainage of paddy fields was conducted around August and also reinforces the above trend.
对于非点源，TP 的月度波动在三个过程中表现出一致的趋势，即 8 月和 9 月是最高峰值期，4 月是次峰值期。这种波动趋势基本上与当地降雨记录一致，其中 8 月至 9 月的雨季以高发生频率的重降雨事件为特征。2017 年当地最大日降雨量出现在 9 月 25 日，五个站点降雨量在 92.0 至 200.3 毫米之间。此外，小雨间隔增加了土壤湿度至相对饱和水平，这反过来又会导致更高的径流（Z. Yuan 等人，2021）。因此，在日尺度上，我们估计的三个过程中非点源的 TP 负荷均达到了最大值（分别为 7,434.26、2,834.22 和 56.68 千克）。另外，稻田的人工排水在 8 月左右进行，也加强了上述趋势。

4 Discussion  4 讨论

4.1 Validation and Uncertainty Analysis
4.1 验证与不确定性分析

To reduce modeling biases, efforts were made to assure the accuracy of activity data and parameter values in the context of various data sources. For example, the TP concentration generated from residential and commercial consumption in this study (3.93–6.79 mg/L, Table S18 in Supporting Information S1) is within the range of estimates from other studies (1.80–11.81 mg/L) (Hu & Lian, 2012; Jiang et al., 2017; Peng et al., 2012; L. Wang, 2018). The TP intensity generated by rainfall runoff in this study (0.20 t/km2) is consistent with values in the literature (0.12–0.58 t/km2) (Gao et al., 2013; B. Liu et al., 2013; S. Wang et al., 2013; L. Zhuang et al., 2015). Also, the discharge intensity of TP from industrial production (0.01 t/km2) exhibits high similarity with previous estimates (0.01–0.02 t/km2) (Ji, 2019; Kong et al., 2018; Kuang et al., 2015; Y. Li et al., 2018). The different industrial structures in different regions and the TP removal rate of local WWTPs (Hua et al., 2022) may have caused differences in the results.
为减少模型偏差，努力确保了在各种数据源背景下活动数据和参数值的准确性。例如，本研究中由住宅和商业消费产生的 TP 浓度（3.93–6.79 mg/L，见支持信息 S1 中的表 S18）与其他研究的估计值范围一致（1.80–11.81 mg/L）（Hu & Lian，2012；Jiang 等，2017；Peng 等，2012；Wang L.，2018）。本研究中由降雨径流产生的 TP 强度（0.20 t/km 2 ）与文献中的值一致（0.12–0.58 t/km 2 ）（Gao 等，2013；Liu B.等，2013；Wang S.等，2013；Zhuang L.等，2015）。此外，工业生产中 TP 的排放强度（0.01 t/km 2 ）与之前的估计值高度相似（0.01–0.02 t/km 2 ）（Ji，2019；Kong 等，2018；Kuang 等，2015；Li Y.等，2018）。不同地区的不同工业结构和当地城市污水处理厂的 TP 去除率（Hua 等，2022）可能导致了结果的不同。

In addition, we conducted interactive crosschecks between life-cycle processes to further increase robustness. For instance, the total TP generation from various sources was estimated using spatially available data sets, while the total TP collected and then treated by WWTPs was also independently calculated based on continuously monitored data sets from the WWTPs themselves. Collection rate can be derived by comparing the values between both processes, and it matches local understandings of performance of municipal WWTPs (expert consultation). Furthermore, the total TP loads discharged into water bodies are also in good agreement with the TP increments of rivers passing through the study area, which were calculated by multiplying the river flow volume by the TP concentration difference between the downstream observation station and the upstream background (Figure 6).
此外，我们进行了生命周期过程的交互式交叉校验，以进一步提高其鲁棒性。例如，使用空间可用的数据集估计了来自各种来源的总 TP 生成量，而由 WWTPs 收集并处理的总量也是基于 WWTPs 自身持续监测的数据集独立计算的。通过比较两个过程中的值，可以推导出收集率，这与当地对市政 WWTPs 性能的理解相符（专家咨询）。此外，排入水体的总 TP 负荷也与穿过研究区域的河流 TP 增量一致，该增量是通过将河流流量乘以下游观测站与上游背景之间的 TP 浓度差计算得出的（图 6）。

Details are in the caption following the image
Figure 6  图 6
Open in figure viewer
打开图库查看器PowerPoint  幻灯片

Comparison of total phosphorus (TP) discharges from various sources and increments of river passing through the study area on the monthly scale.
比较各来源的总磷（TP）排放量和研究区域河流月度通过量的增量。

Besides the above, Monte Carlo simulation has been applied to quantitatively test the propagation of input uncertainties and variability into the final results. Based on the previous studies, we provided activity data sampled from a continuous uniform distribution while parameters sampled from a triangular distribution, and then conducted the simulation of 10,000 times, finding that the overall uncertainty ranges from −11.84% to +11.68% within 95% confidence intervals (Figure 7). As the parameter values were derived from official automatic monitoring data of key enterprises, the uncertainty of industrial production is small. Uncertainties in residential and commercial consumption and rainfall runoff are controllable because their parameter values were supported by official statistics and on-site sampling data. The parameter values for paddy farming, aquaculture, and animal breeding were taken from interviews and literature, making the uncertainties relatively large. However, due to the small amount of these three sources, they have little influence on the whole analysis but deserve further investigation. It should be noted that the model is primarily driven by physical mechanisms and therefore does not take into account the chemical transformation of water pollutants from nonpoint sources, which may have a slight impact on the estimation of overall loads.
除了上述内容外，蒙特卡洛模拟已被应用于定量测试输入不确定性和变异性对最终结果的影响。基于先前的研究，我们提供了从连续均匀分布中抽取的活动数据，以及从三角分布中抽取的参数，然后进行了 10,000 次模拟，发现整体不确定性在 95%置信区间内为-11.84%至+11.68%（见图 7）。由于参数值来自关键企业的官方自动监测数据，工业生产的不确定性较小。居民和商业消费以及降雨径流的不确定性是可控的，因为它们的参数值得到了官方统计数据和现场采样数据的支持。水稻种植、水产养殖和畜牧业的参数值来自访谈和文献，使得不确定性相对较大。然而，由于这三个来源的数据量较少，它们对整体分析的影响很小，但仍值得进一步研究。 应注意的是，该模型主要受物理机制驱动，因此未考虑来自非点源的水污染物化学转化，这可能会对总体负荷估计产生轻微影响。

Details are in the caption following the image
Figure 7  图 7
Open in figure viewer
打开图库查看器PowerPoint  幻灯片

Uncertainties of total phosphorus (TP) generated from various sources expressed as 95% confidence intervals around central estimates.
不确定的总磷（TP）来源，以中心估计值的 95%置信区间表示。

4.2 Scenario Analysis  4.2 场景分析

The application of the WPLT model in the study area has shown the pathway of TP from generation sources to entering the water bodies and identified the main water pollution hotspots, for example, direct discharge and nonpoint sources. Combined with the method of control variates, a scenario analysis consisting of six scenarios was further conducted to estimate the change of TP loads in response to possible implementation of management measures. The impact on TP loads from climate change and social-economic changes was not considered in the scenario settings. (a) Business as usual is the current baseline scenario without any additional control measures. (b) Scenario A1 upgrades the WWTPs, in which the TP removal rate will be achieved at the highest existing level in practice. (c) A2 is the scenario of banning direct discharge, which enters WWTPs for treatment and is discharged afterward according to the existing treatment level. (d) A3, A3-2, A3-3, and A3-4 are a set of scenarios assuming the amount of wastewater and water pollutant loads generated by point sources can be controlled, with a reduction ratio of 10%, 20%, 30%, and 40%, respectively. (e) We further assumed that the above scenarios can be supplemented to each other and set scenarios of B1, B2, and B3, which are the combination of A1 + A2, A1 + A3, and A2 + A3, respectively, and (f) scenario C, a comprehensive scenario of A1 + A2 + A3.
WPLT 模型在研究区域的运用展示了 TP 从生成源到进入水体的路径，并确定了主要的水污染热点，例如直接排放和非点源。结合控制变量的方法，进一步进行了包括六个情景的场景分析，以估计在可能实施管理措施的情况下 TP 负荷的变化。在情景设置中未考虑气候变化和社会经济变化对 TP 负荷的影响。（a）现状基线情景为“照常进行”，没有任何额外的控制措施。（b）情景 A1 升级了污水处理厂，其实际 TP 去除率将达到最高水平。（c）情景 A2 为禁止直接排放，进入污水处理厂进行处理，然后根据现有处理水平排放。（d）情景 A3、A3-2、A3-3 和 A3-4 是一组假设点源产生的废水和水污染物负荷量可以控制的情景，分别减少 10%、20%、30%和 40%。 (е) 我们进一步假设上述场景可以相互补充，并设置了 B1、B2 和 B3 场景，分别是 A1 + A2、A1 + A3 和 A2 + A3 的组合，以及(f)场景 C，即 A1 + A2 + A3 的综合场景。

As far as A1, A2, and A3 are concerned, A2 has the best reduction effect and are followed by A3 and A1 (Figure 8). This is mainly because the ratio of direct discharge to discharge of WWTPs for TP reached over 500%. As the reduction efficiency of A3, A3-2, A3-3, and A3-4 increases sequentially, we find that the reduction at source is positively correlated with the cut of the discharge to the environment. With the integration of multiple control measures, the three-factor combination scenario performs better than the two-factor combination scenarios, as well as the single-factor scenarios, which shows synergistic effect rather than antagonism of various measures.
关于 A1、A2 和 A3，A2 具有最佳的减排效果，其次是 A3 和 A1（图 8）。这主要是因为 TP 的直接排放与污水处理厂排放的比率超过 500%。随着 A3、A3-2、A3-3 和 A3-4 的减排效率依次提高，我们发现源头减排与对环境排放的削减呈正相关。在多种控制措施整合下，三因素组合场景的性能优于两因素组合场景，以及单因素场景，这表明各种措施之间存在协同效应而非对抗效应。

Details are in the caption following the image
Figure 8  图 8
Open in figure viewer
打开图库查看器PowerPoint  幻灯片

Comparison of total phosphorus discharges reduction efficiency under various scenarios.
比较在各种情景下总磷排放减少效率。

The improvement of the three influencing factors can be specifically started as follows: (a) for municipal WWTPs, tools such as advanced process control (Palatsi et al., 2021), which are more efficient in removing water pollutants, can be used in practice. It should be also noted that sewage sludge can be further treated as secondary raw material and used as a valuable source of nutrients and organic matter in agricultural cultivation (Goel et al., 2021; Tang et al., 2022). (b) Proper planning, construction, and maintenance of sewage pipelines can not only avoid the direct discharge of water pollutants, but also improve the water environment in rural areas. (c) For residential and commercial consumption, a blander diet and using phosphate-free personal/home care products are suggested to reduce pollutant concentration in household wastewater. Industrial TP mainly comes from textile and equipment manufacturing sectors, of which technological upgrades, such as reuse of treated textile wastewater in production processes (Cinperi et al., 2019) and use of low-phosphorus additives in equipment manufacturing processes (Sarin et al., 2002), are strongly needed.
三个影响因素的改进可以具体从以下方面开始：（a）对于城市污水处理厂，可以使用如先进过程控制（Palatsi 等人，2021 年）等工具，这些工具在去除水污染物方面更有效率。还应注意的是，污泥可以进一步作为二次原料处理，并在农业栽培中用作有价值的营养和有机物质来源（Goel 等人，2021 年；Tang 等人，2022 年）。（b）合理规划、建设和维护污水管道不仅可以避免水污染物的直接排放，还可以改善农村地区的环境。（c）对于居民和商业消费，建议采用口味较淡的饮食和使用无磷的个人/家庭护理产品，以降低家庭污水中污染物的浓度。工业 TP 主要来自纺织和设备制造行业，其中技术升级，如在生产过程中重复使用处理过的纺织废水（Cinperi 等人，2019 年）和使用设备制造过程中的低磷添加剂（Sarin 等人，2002 年），是迫切需要的。

4.3 Applicability of the WPLT Model
4.3 WPLT 模型的适用性

This single-year case study, conducted on both monthly and daily scales, demonstrates the effectiveness of the WPLT in identifying temporal hotspots of water pollutants. In addition, the model can be used for long time series studies on an annual basis if required. Moreover, this model can trace water pollutants from sources to recipient water bodies at high spatial resolution regardless of the topographic constraints. It is worth noting that flat terrain areas with fertile soil, mild climate, and convenient communications account for about 9% of China's total area, and are undergoing rapid urbanization. Consequently, the WPLT model presents enormous potential for application beyond these areas.
这项针对单一年份进行的案例研究，在月度和日度尺度上开展，证明了 WPLT 在识别水污染物时间热点方面的有效性。此外，如果需要，该模型还可以用于年度长期时间序列研究。此外，该模型可以在不考虑地形限制的情况下，以高空间分辨率追踪水污染物从源头到受纳水体的路径。值得注意的是，具有肥沃土壤、温和气候和便利交通的平原地区占中国总面积的约 9%，并且正在经历快速城市化。因此，WPLT 模型在这些地区之外的应用潜力巨大。

The main data requirements for the WPLT model are shown in Table 1 and are relatively easy to obtain. The detailed data of the study area did support the application of the WPLT model, but for areas where only limited data may be available, the data requirements of the model can be met in the following ways: (a) with the help of other variables that establish regression relationship with the target variable; (b) localize and correct the data based on international databases; (c) refer to literature data of other similar regions.
表 1 显示了 WPLT 模型的主要数据需求，这些需求相对容易获得。研究区域的具体数据支持了 WPLT 模型的应用，但对于可能只有有限数据的地区，模型的数据需求可以通过以下方式满足：（a）借助与其他变量建立回归关系的变量；（b）基于国际数据库本地化和修正数据；（c）参考其他类似地区的文献数据。

The WPLT model has the following enlightenment for general policy implications around the world: (a) quantifying the generation pattern of water pollutant loads at high spatio-temporal resolution to identify hotspots and raise public awareness regarding main sources of pollution. (b) Characterizing the life-cycle pathway of water pollutants to reveal the potential substantial gap between generation and discharge, enabling the efficient decomposition of the load management task for each generation source. (c) By identifying direct discharges from point sources and improving the accuracy of discharges, the results can serve as an essential input to mathematical models that estimate water quality and make them operate robustly, which would be helpful for making targeted regulatory actions and assessing the water quality response of load management to improve the water quality.
WPLT 模型对全球一般政策启示有以下几点：（a）以高时空分辨率量化水污染物负荷的生成模式，以识别热点并提高公众对污染主要来源的认识。（b）描述水污染物的生命周期路径，揭示生成和排放之间的潜在重大差距，使每个生成源的负荷管理任务分解更有效。（c）通过识别来自点源的直接排放并提高排放的准确性，结果可以作为估计水质和使数学模型稳健运行的基本输入，这有助于制定有针对性的监管措施和评估负荷管理对水质响应的效果，以改善水质。

Studies have shown that the legacy contamination of P, nitrogen (N), and other nutrient elements is the key factor hindering the improvement of water quality (Basu et al., 2022; Jarvie et al., 2013). The residual P can be transformed into a state of bioavailability, thus polluting the water bodies (Doydora et al., 2020; Van Meter et al., 2021). By supplementing the P input from human activities, the WPLT model can assist in the estimation of legacy P. In addition to P, this model can also describe the pathway of N and other water pollutants.
研究表明，P、氮（N）和其他营养元素的残留污染是阻碍水质改善的关键因素（Basu 等，2022；Jarvie 等，2013）。残留的 P 可以转化为生物可利用状态，从而污染水体（Doydora 等，2020；Van Meter 等，2021）。通过补充人类活动带来的 P 输入，WPLT 模型可以帮助估算残留 P。除了 P，该模型还可以描述 N 和其他水污染物的途径。

5 Conclusions  5 结论

The WPLT model established in this work defines the technical term of water pollutant loads from a life-cycle perspective by tracing its pathway from generation sources to recipient water bodies. The model's application in Taihu Lake Basin, China in 2017 shows good estimation accuracy when compared with local empirical experience. Point-source pollution is greatly reduced after WWTP collection and treatment, recognizing the challenges from direct discharge and nonpoint sources, as well as the potential substantial gap between generation and discharge. TP generation has its peak in September and is spatially clustered in urban residence and industrial land. We have further conducted the scenario analysis to evaluate the life-cycle pollution reduction potentials and found that increasing the wastewater collection rate is a relatively feasible and effective way to eliminate direct discharge and water pollutant loads. With no dependence on DEM-based watershed delineation, the WPLT model is useful for urbanizing flat terrains to identify water pollution hotspots at high spatial and temporal resolution. It has been proven to be a reliable approximation to understand water pollutant loads and an effective tool to facilitate setting up more precise and targeted water pollution control strategies in rapidly urbanizing watersheds.
该工作中建立的 WPLT 模型从生命周期角度定义了水污染物负荷的技术术语，通过追踪其从生成源到受纳水体的路径。2017 年在中国太湖流域的应用表明，与当地经验相比，该模型具有良好的估计精度。在 WWTP 收集和处理后，点源污染大大减少，认识到直接排放和非点源带来的挑战，以及生成和排放之间可能存在的巨大差距。TP 生成在 9 月达到峰值，并在城市住宅和工业用地空间上聚集。我们进一步进行了情景分析，以评估生命周期污染减少潜力，发现提高废水收集率是消除直接排放和水污染物负荷相对可行且有效的方法。WPLT 模型不依赖于基于 DEM 的流域划分，对于城市化的平坦地形识别高时空分辨率的水污染热点是有用的。 已被证明是理解水污染物负荷的可靠近似，并且是促进制定更精确和有针对性的水污染控制策略的有效工具，适用于快速城市化的流域。

The results of the WPLT model in the study area were verified and a quantitative uncertainty analysis was conducted, both confirming the stability of this model. Meanwhile, although the conceptual framework of the WPLT model considers the loss of pipeline transport and the loss (including retention, infiltration, and evaporation) of nonpoint sources from generation to recipient water bodies, the calculation of these processes is simplified when this model is applied in the study area where the pipeline construction is relatively new and the proportion of nonpoint sources is relatively small, which may be a limitation of this study. Salient recommendations for future studies are to appropriately reduce these simplifications based on local conditions and include estimates of legacy contamination as that is also a key factor hindering water quality improvement.
研究区域 WPLT 模型的结果得到了验证，并进行了定量不确定性分析，两者均证实了该模型稳定性。同时，尽管 WPLT 模型的概念框架考虑了管道运输的损失以及从源头到受水体的非点源损失（包括滞留、渗透和蒸发），但当该模型应用于管道建设相对较新且非点源比例相对较小的研究区域时，这些过程的计算被简化，这可能是本研究的局限性。未来研究的显著建议是根据当地条件适当减少这些简化，并包括对历史污染的估计，因为这也是阻碍水质改善的关键因素之一。

Acknowledgments  致谢

This study was financially supported by the National Science Fund for Distinguished Young Scholars (Grant 41925004) and the National Natural Science Foundation of China (Grants 41871214 and 41801212).
本研究得到了国家自然科学基金优秀青年科学基金项目（项目编号 41925004）和国家自然科学基金项目（项目编号 41871214 和 41801212）的资助。

    Conflict of Interest  利益冲突

    The authors declare no conflicts of interest relevant to this study.
    作者声明本研究不存在任何利益冲突。

      back
      更多…
      image icon