河口咸潮入侵是困扰沿海地区社会经济发展的全球性环境问题之一，对沿海城市用水安全与河口生态系统健康造成了严重威胁。近年来，受气候变化与高强度人类活动的共同影响，珠江口咸潮入侵情势非常严峻，珠江流域入海径流减少是导致珠江口咸潮入侵加剧的重要原因。珠江流域来水主要受降雨控制，土地利用条件影响产汇流过程；同时，流域用水主要受社会经济活动影响。流域来水与用水变化同时影响径流过程，从而影响河口水盐条件。目前，流域来水与用水变化对珠江口咸潮入侵情势的影响尚未明晰，其变化下未来咸潮入侵情势的演变及带来的损失仍待进一步研究。解决以上关键问题将服务于国家陆海统筹重大战略，为粤港澳大湾区高质量发展规划与生态安全保障提供重要支撑。

本文聚焦上述两方面关键问题，以珠江口及受高强度人类活动影响的东江流域为研究对象，首先建立基于FVCOM平台的珠江口三维水盐动力模型，模拟了2019-2020年珠江口咸潮入侵的情势。通过耦合东江流域水文模型（SWAT）与河口水盐模型（FVCOM），揭示了流域降雨与土地利用变化对入海径流及河口咸潮入侵情势的影响，预测了未来降水与土地利用变化下珠江口的咸潮入侵情势。然后，运用机器学习算法（k近邻算法），分析了东江流域社会经济系统不同行业用水的变化对珠江流域径流量的影响，利用可解释的机器学习方法（SHAP分析），识别了影响径流量及咸潮入侵情势的关键因素，揭示了降雨、农业用水和工业用水单因素变化以及多因素变化共同作用情景下珠江口咸潮入侵的情势。最后，综合考虑未来流域降水、土地利用与社会经济用水变化，通过情景模拟，预测了未来珠江口咸潮入侵情势的变化；并采用多区域投入产出模型，核算了由于取水口盐度超标造成的直接和间接经济损失，评估了未来情景下河口咸潮入侵的社会经济影响。主要研究结果如下所示：

（1）在丰水期，珠江八大口门的咸潮入侵问题并不明显，虎门中下游盐度相对较高。在枯水期，各口门均遭遇了不同程度的咸潮入侵影响，在虎跳门、崖门以及虎门受到的影响较大。尤其是虎门，2019-2020年两年间取水口附近选取点位的盐度超标天数达到了370天，多于模拟时间的一半。枯水期虎门取水口附近盐度一直处在超标水平，显著高于其他口门取水口附近盐度。

（2）揭示了虎门、虎跳门以及崖门的盐度垂向分布随时间的变化规律。枯水期虎门及虎跳门存在明显的盐度垂向分层现象，呈现一定的盐度梯度，说明虎门以及虎跳门附近咸水和淡水混合不显著。同时，流场影响盐度输移过程，崖门附近枯水期表层与底层发生了流向交替的现象，造成了表层与底层的水体在垂向上发生了混合，因而盐度分层不明显。

（3）模拟了东江流域降水和土地利用变化情景下，2032-2035年珠江口咸潮入侵情势。未来情景下，枯水季虎门盐度分布存在显著的空间差异，下游点位枯水期盐度最高可达17.07‰，上游点位最高可达3.29‰。2032-2035年6月至9月虎门上游基本不存在盐度超标问题，而11月至来年2月盐度基本处于超标状态，枯水期咸潮防控压力依然很大；由于口门处不同点位盐度差异显著，因此通过合理规划取水口位置可一定程度缓解咸潮入侵的影响。

（4）降水、工业用水以及农业用水单因素变化对于咸潮入侵的影响呈现明显的季节性。当降水减少时，盐度在全年范围内均有不同程度的增加，尤其是在枯水期，降水的减少导致的盐度增加较为明显。农业用水的增加会导致盐度降低，尤其是在丰水期更为显著。工业用水的增加会导致盐度在丰水期增加，在枯水期反而有一定程度的下降。在降水、工业用水以及农业用水三种因素其中的两种共同作用时，降水的减少以及农业用水的增加会导致在枯水期盐度增加，而在丰水期盐度减少，这是两种因素起主导作用的时间不同而形成。而降水减少与工业用水增加时，在大部分时间里都使得盐度增加，仅在秋季盐度会下降。而当工业用水与农业用水同时增加时，盐度几乎不发生变化，说明这两种因素同时作用时，其对于盐度的影响相互抵消。

（5）综合考虑未来流域降水、土地利用与社会经济用水变化，预测了珠江口咸潮入侵情势，模拟结果发现丰水期在行业用水增加的情景下，虎门、蕉门上游及中游的盐度均高于对照情景，这表明行业用水增加会导致河口盐度上升。工业用水增加更易引发或加剧咸潮入侵，其对下游盐水入侵的影响大于农业用水增加的影响。结合多区域投入产出模型，分析咸潮入侵带来的直接经济损失和间接经济损失。情景模拟结果显示，在2032-2035年期间，东江6市因为咸潮而遭受的直接经济损失约为4266.41-4583.3亿元/年，占GDP比例为4.31%-4.77%；由此导致的间接经济损失约为1230.57-1638.39亿元/年。

The intrusion of saline tide into estuaries is one of the global environmental issues troubling the socio-economic development of coastal areas, posing a serious threat to water security and estuarine ecosystem health of coastal cities. In recent years, exacerbated by climate change and intensified human activities, the situation of saline tide intrusion in the Pearl River Estuary has become extremely severe. The water inflow in the Pearl River Basin is primarily controlled by rainfall, while land use also affects the process of water runoff and accumulation. Water use, on the other hand, is mainly influenced by socio-economic activities. As the Pearl River Estuary is situated in the connecting zone between land and sea, changes in water inflow and usage simultaneously affect the runoff process, thereby influencing the water salinity conditions in the estuary. Currently, the effects of changes in water inflow and usage on the intrusion of saline tide in the Pearl River Estuary are not yet clear, and further research is needed to understand the evolution of future saline tide intrusion and the resulting losses. Addressing these key issues will serve the national strategy of integrated land-sea development and provide crucial support for the planning of high-quality development and ecological security in the Greater Bay Area.

This thesis focuses on two key issues, which investigates the Pearl River Estuary and the Dongjiang Basin affected by intense human activities. Firstly, it establishes a three-dimensional hydrodynamic model based on the FVCOM platform to simulate the saline intrusion in the Pearl River Estuary during 2019-2020. By coupling the watershed hydrological model (SWAT) with the estuarine hydrodynamic model (FVCOM), the thesis reveals the impacts of basin rainfall and land use changes on river discharge into the sea and saline intrusion in the estuary. It predicts the future saline intrusion in the Pearl River Estuary under changing precipitation and land use. Next, it employs machine learning algorithms (k-nearest neighbors) to analyze how changes in water use across different industries in the Dongjiang Basin's socio-economic system affect river discharge in the Pearl River Basin. Using interpretable machine learning methods (SHAP analysis), the study identifies key factors influencing river discharge and saline intrusion, revealing the scenarios of saline intrusion in the Pearl River Estuary under single-factor and multi-factor changes in rainfall, agricultural, and industrial water use. Finally, considering future changes in basin precipitation, land use, and socio-economic water use, the thesis uses scenario simulations to predict the changing patterns of saline intrusion in the Pearl River Estuary. It also employs a multi-regional input-output model to calculate the direct and indirect economic losses caused by excessive salinity at the intake, evaluating the socio-economic impacts of estuarine saline intrusion under future scenarios. The main research findings are as follows:

(1) During the wet season, saltwater intrusion in the eight main outlets of the Pearl River Estuary is not significant, with relatively higher salinity in the middle and lower reaches of the Humen outlet. However, during the dry season, all outlets experience varying degrees of saltwater intrusion, with significant impacts observed in the outlets of Huitaomen, Yamen, and Humen. Particularly in Humen, the number of days with salinity exceeding standards at the intake exceeded 370 days between 2019 and 2020, more than half of the simulation period. During the dry season, salinity near the intake consistently exceeded standards and was significantly higher than at other outlets.

(2) The vertical distribution of salinity at Humen, Hutiaomen, and Yamun, and its temporal variation. There is a clear vertical stratification of salinity throughout the month at Humen and Hutiaomen. A certain salinity gradient is observed, indicating insignificant mixing of saline and fresh water in the waters of Humen and Hutiaomen. However, the stratification of flow velocity significantly affects the stratification of salinity transport processes, especially at Yamen during the dry season, where surface and bottom layers undergo alternating flow directions. This phenomenon leads to mixing of surface and bottom layers vertically, resulting in less distinct salinity stratification.

(3) The simulation of precipitation and land use changes in the Dongjiang River Basin indicates the salinity intrusion situation in the Pearl River Estuary from 2032 to 2035. In this future scenario, during the dry season, there are significant spatial differences in salinity distribution at various points. The salinity levels at downstream points during the dry season can reach a maximum of 17.07‰, while upstream points can reach a maximum of 3.29‰. From June to September in the years 2032-2035, there are generally no salinity exceedance issues at the upstream point of Humen, but from November to February of the following year, salinity levels are generally in exceedance, indicating significant challenges in controlling salinity intrusion during the dry season. Due to notable differences in salinity at different locations near the estuary, mitigating the impacts of salinity intrusion can be achieved to some extent through strategic planning of water intake locations.

(4) Single-factor variations in precipitation, industrial water use, and agricultural water use exhibit pronounced seasonal effects on saltwater intrusion. Decreased precipitation leads to increased salinity throughout the year, particularly during the dry season when the decrease in precipitation results in a more noticeable increase in salinity. Increased agricultural water use leads to decreased salinity, especially during the rainy season. Increased industrial water use leads to increased salinity during the rainy season but a slight decrease during the dry season. When precipitation, industrial water use, and agricultural water use are combined, decreased precipitation and increased agricultural water use lead to increased salinity during the dry season and decreased salinity during the rainy season, reflecting the different times when these two factors dominate. Decreased precipitation combined with increased industrial water use generally increases salinity throughout most of the year, except for a decrease during the autumn. When industrial water use and agricultural water use both increase simultaneously, salinity remains almost unchanged, indicating a mutual cancellation of their effects on salinity.

(5) Considering future changes in basin precipitation, land use, and socio-economic water consumption, the prediction of salinity intrusion in the Pearl River Estuary was simulated. The results of the simulation indicate that during high-flow periods with increased industrial water usage, salinity levels in upstream and midstream areas such as Humen and Jiaomen are higher compared to baseline scenarios, suggesting that increased industrial water usage could lead to higher river mouth salinity levels. The increase in industrial water usage is more likely to trigger or exacerbate salinity intrusion, with a greater impact on downstream salinity intrusion compared to increased agricultural water usage. Using a multi-regional input-output model, direct and indirect economic losses resulting from salinity intrusion were analyzed. Simulation results indicate that during the period from 2032 to 2035, the six cities in the Dongjiang Basin suffered direct economic losses ranging from approximately 426.41 billion to 458.33 billion RMB per year due to salinity intrusion, representing 4.31% to 4.77% of the local GDP. These economic losses also led to indirect economic losses of approximately 123.57 billion to 163.84 billion RMB per year.