Spatiotemporal Changes in Yangtze Estuary River Islands Revealed by Landsat Imagery

Abstract

As fluvial deposition features, river islands originate from persistently exposed sandbars. Their morphological evolution responds to hydrological dynamics, sediment budgets, and human modifications of river systems. This study conducts a quantitative analysis of the spatiotemporal evolution of four river islands in China’s Yangtze River Estuary (YRE), utilizing multitemporal Landsat imagery (MSS, TM, ETM+, and OLI) at five-year intervals from 1974 to 2024. This analysis employed thresholding, binarization, image registration, cropping, and cluster analysis. Hydrological data (runoff and sediment flux) from Datong Station were concurrently evaluated to explore the driving factors of evolution. The findings suggested the following: (1) MSS/TM/ETM+/OLI images were effective for accurately extracting river island information, and the results were consistent with the accuracy verification. (2) The cumulative area and growth rate of the river islands have exhibited an upward trend over time, with Jiuduansha growing the fastest. (3) Runoff and sediment discharge are the primary natural controls on morphological evolution, with a weak positive correlation (R = 0.293) and a strong negative correlation (R = −0.915) with the area of river islands, respectively. Anthropogenic drivers such as land reclamation, sediment enhancement projects, and the Three Gorges Dam are equally critical.

1. Introduction

River islands are stable landforms that gradually develop from underwater sandbars under favorable river current conditions, ultimately emerging above the water surface as depositional bodies [1]. As products of specific water-sediment conditions, the size, shape, erosion-deposition dynamics, and stability of river islands change in response to fluvial hydrodynamics and sediment deposition processes. Based on their formation processes, river islands can be categorized into three types: primary river islands, dynamic river islands, and accumulative river islands [2]. Typical accumulative islands include Yangzhong Island in the Nanjing section of the Yangtze River, as well as Chongming Island, Changxing Island, and Hengsha Island at YRE. The evolution of river islands provides significant insights into hydrological processes and sedimentary conditions at the scale of river reaches; moreover, it is widely applied in global river geomorphology and sediment assessments, serving as an essential component of fluvial morphological dynamics [3,4,5]. Therefore, exploring the evolutionary patterns of river islands has significant implications for river stabilization, navigation, water-land interface ecology, and species habitats [6,7,8].

The Yangtze River, as the longest river in China, holds significant strategic importance in facilitating the high-quality development of the Yangtze River Economic Belt. The morphological evolution of its river islands is crucial for maintaining island stability and enhancing the functionality of the Golden Waterway. Since the completion of the Three Gorges Dam in June 2003, the sediment flux of the Yangtze River into the sea has dropped sharply from 490 million tons per year to 130 million tons per year, which has led to significant changes in the hydromorphology, ecological environment and other aspects in the middle and lower reaches of the Yangtze River [9,10]. Numerous researchers have investigated river islands in the middle and lower reaches of the Yangtze River both before and after the dam’s construction, drawing on extensive hydrological observation data and multitemporal remote sensing data [11]. Gao conducted a quantitative analysis of eight river islands in the Ma-Wu-Tong segment of the lower Yangtze River using MSS/TM/ETM+ remote sensing imagery, finding that their area initially expanded before subsequently contracting [1]. Jiang et al. [12] studied 64 river islands each exceeding 1 km² in the main course of the Yangtze River, noting that their area generally expanded between 1984 and 2019. Shi et al. [13] analyzed the development of river islands in the Nanjing segment of the lower Yangtze River, demonstrating that sediment discharge is the primary factor influencing their evolution. Additionally, fluctuations in the Yangtze River’s water levels during flood periods are a key factor affecting the temporal dynamics of the river islands [14].

Remote sensing technology, with its capabilities of rapid, efficient, and large-scale coverage, serves as a robust instrument for the dynamic monitoring of river island areas. Its macroscopic characteristics facilitate more convenient extraction of river island boundaries [15]. Some studies have attempted to use Landsat data to monitor the evolution of river islands [1,13]. Leveraging the comprehensive time series of Landsat imagery available on the Google Earth Engine platform, Cao et al. conducted a mapping analysis of the monthly changes in the coastline and tidal flats of the Zhoushan Archipelago from 1985 to 2017 [16]. Similarly, Xu et al. [17] utilized Landsat data and applied water indices and threshold segmentation methods to generate annual land–water maps of Changxing Island and Hengsha Island in YRE from 1987 to 2016.

This study focuses on four major river islands in the Yangtze Estuary, specifically Chongming Island, Changxing Island, Hengsha Island, and Jiuduansha. Utilizing Landsat series datasets including MSS (Multispectral Scanner), TM (Thematic Mapper), ETM+ (Enhanced Thematic Mapper), and OLI (Operational Land Imager) imagery acquired at 5-year intervals from 1974 to 2024 and integrating runoff and sediment flux records from Datong Hydrological Station spanning 1950–2024. The aims of this research are threefold: (1) Extract river island information using the improved Modified Normalized Difference Water Index (MNDWI) threshold method, and quantify the spatiotemporal variation characteristics of the area and shape of the four islands from 1974 to 2024; (2) Systematically monitor and analyze the long-term evolution processes and patterns of these islands, and clarify evolutionary differences across different stages; (3) Construct a formal attribution framework of “natural factor quantification—anthropogenic activity calibration—dual-system coupling verification”, clarify the driving mechanisms of natural and anthropogenic factors (including runoff, sediment flux, land reclamation, sediment enhancement projects, and the Three Gorges Dam) on river island evolution, and minimize the interference of confounders such as tide stage and seasonal hydrology in method design. This study is expected to provide scientific support for the rational utilization of YRE river island resources, ecological protection, and coastal zone management.

2. Study Area and Data

2.1. Study Area

The Yangtze River, the third longest river in the world, ranks third globally in terms of runoff and fourth in sediment discharge. Originating from the Tibetan Plateau at an elevation of approximately 6000 m, it has a drainage basin area of 1.8 × 10⁶ km² and a total length of about 6300 km. It empties into the East China Sea east of Chongming Island in Shanghai. The estuary is divided into northern and southern branches by Chongming Island. In the central part of the estuary, Changxing Island further splits the southern branch into northern and southern navigational channels. Near Hengsha Island, Human infrastructure and navigational channels subdivide the southern channel into two, forming a three-tiered branching system with four outlets draining into the sea.

The four islands selected for this study, from north to south, are Chongming Island, Changxing Island, Hengsha Island, and Jiuduansha (Figure 1). Chongming Island (121°09′30″~121°54′00″ E, 31°27′00″~31°51′15″ N) is located in YRE, bordered by the East China Sea to the east and in close proximity to Shanghai, China’s largest city. As China’s third largest island [18], it forms Chongming District together with Changxing Island and Hengsha Island. Changxing Island and Hengsha Island are, respectively, the second and third largest islands in Shanghai. Changxing Island lies between Chongming Island and Shanghai, facing the Huangpu River estuary, while Hengsha Island is situated east of Changxing Island, positioned opposite Chongming Island. Jiuduansha is located between the northern and southern channels of YRE, consisting of Jiangyan Nansha, Shangsha, Zhongsha, and Xiasha [19]. The study areas have a subtropical monsoon climate with distinct seasons, an average annual temperature of approximately 15 °C, and an average annual precipitation of around 1100 mm. This climatic condition supports natural ecosystems and is suitable for agriculture and other land uses. Forecasting changes in these four islands can effectively reflect variations in the area and shape of the islands in YRE, providing a scientific basis for Yangtze River management.

2.2. Data

2.2.1. Remote Sensing Data

To investigate the spatiotemporal characteristics of these river islands, this study collected Landsat series imagery from 1974 to 2024, including data from the Multi-Spectral Scanner (MSS), Thematic Mapper (TM), Enhanced Thematic Mapper Plus (ETM+), and Operational Land Imager (OLI) (

Table 1. Details of Landsat remote sensing data used in the study.

Date	Satellite	Sensor	Resolution (m)
13 February 1974	Landsat-1	MSS	60
4 March 1979	Landsat-2	MSS	60
23 April 1984	Landsat-5	TM	30
11 August 1989	Landsat-5	TM	30
2 March 1994	Landsat-5	TM	30
1 April 1999	Landsat-5	TM	30
19 July 2004	Landsat-5	TM	30
28 April 2009	Landsat-5	TM	30
30 December 2014	Landsat-7	ETM	30
18 January 2019	Landsat-8	OLI_TIRS	30
20 March 2024	Landsat-8	OLI_TIRS	30

Notes: MSS: Multispectral Scanner; TM: Thematic Mapper; ETM: Thematic Mapper; OLI: Operational Land Imager.

). Priority was given to imagery acquired during the dry season of YRE (November–February of the following year), a period characterized by stable runoff and minimal tidal fluctuations. When dry-season imagery was unavailable, alternatives from March to October with hydrological characteristics similar to the dry season were used. Concurrently, measured tidal level data—sourced from the Wusongkou and Nanmengang Hydrological Stations—was retrieved for each image’s acquisition time, with only imagery featuring tidal levels within “multi-year mean low tide level ± 0.5 m” included. For imagery with deviated tidal levels, high-resolution historical imagery from Google Earth was used to calibrate land–water boundaries. All imagery was normalized to the YRE’s multi-year mean low tide level (based on the Wusong Zero Datum) to ensure consistent benchmarks for area extraction. The final dataset comprises 2 MSS images, 6 TM images, 1 ETM+ image, and 2 OLI images, all sourced from the Geospatial Data Cloud of the Computer Network Information Center, Chinese Academy of Sciences (http://www.gscloud.cn). Notably, the 2014 Landsat-7 SLC-off (Scan Line Corrector-off) imagery underwent processing using the “local linear regression strip repair algorithm”: gaps were filled via 3 × 3 window pixel linear fitting (with a fitting error threshold of 0.02) using contemporaneous strip-free Landsat-8 OLI imagery as reference. Repair effectiveness was quantified using Root Mean Square Error (RMSE = 0.018) and Structural Similarity Index (SSIM = 0.92), verifying acceptable residual uncertainty (RMSE < 0.03, SSIM > 0.9). Landsat imagery corresponding to Path 118 and Row 38 fully covers the study area. Except for the 1974 and 1979 MSS images (60 m spatial resolution), all other images have a 30 m resolution. All data adopt the Universal Transverse Mercator (UTM) Zone 51 N projection and World Geodetic System 1984 (WGS84) coordinate system, with geometric and atmospheric corrections completed during download. Additionally, imagery with cloud cover < 10% was specifically selected to minimize cloud interference in data extraction.

2.2.2. Hydrological Data

Annual runoff and sediment discharge data from the Datong Hydrological Station, spanning 1950 to 2024, were sourced from the Changjiang Water Resources Commission [20] (CWRC) (http://www.cjw.gov.cn/zwzc/zdgk/swgl/cjns (accessed on 20 May 2025)). Located in Chizhou City, Datong Station is 642 km from the YRE and the East China Sea, and it monitors 94% of the downstream area of the Yangtze River Basin. As the nearest hydrological station to the river mouth with long-term observational data, it serves as a critical data source for this study [21]. To enhance the data’s applicability to YRE-related analysis, this study adopted cross-correlation analysis to quantify the time lag of runoff and sediment signals transmitted from Datong Station to YRE, eliminated interferences such as runoff attenuation and sediment deposition in the estuary, and integrated data from estuarine hydrological stations such as Wusongkou and Nanmengang (https://swj.sh.gov.cn (accessed on 1 January 2025)), and tidal observation records to ensure the data could accurately reflect the actual hydrological processes of the YRE. Statistics show that the average annual runoff from 1950 to 2024 is 896.1 billion m³, while the mean annual sediment discharge over the same period is 33.4 million tons. The maximum runoff of 1359 billion m³ was recorded in 1954, and the minimum runoff of 667.1 billion m³ occurred in 2011. In terms of sediment discharge, the highest value of 67.8 million tons was recorded in 1964, and the lowest value of 4.45 million tons was recorded in 2023.

2.3. Image Preprocessing

The ETM+ sensor aboard the Landsat 7 satellite malfunctioned on 31 May 2003, resulting in black stripes covering over 25% of the captured images. The selected Landsat 7 images were repaired using the SLC (Scan Line Corrector)-off model (Figure 2). This paper used the function “tm_destripe” in the ENVI 5.6 (The Environment for Visualizing Images) image processing software package for this. The basis of tm_destripe is the Multiple Image Fixed Window Regression Analysis Model, which can be used to revise stripes [13]. The UTM (Universal Transverse Mercator)-WGS (World Geodetic System) 84 projection of the original images was converted to Lambert Azimuthal Equal Area projection. To address geometric distortions introduced by satellites and sensors during image acquisition, geometric registration was performed on the images. Subsequently, the study area was delineated using the ROI (Region of Interest).

3. Methods

3.1. River Island Area Extraction Methods

This study extracts river islands using the threshold method, leveraging brightness values derived from differences in the reflectivity of various land features in remote sensing images. However, the spectral reflectance characteristics of water in the Yangtze River are complex due to influences from substances such as sediment and chlorophyll, making it difficult to precisely distinguish between water and land using direct reflectance differences alone. Both the Normalized Difference Water Index (NDWI) and Modified Normalized Difference Water Index (MNDWI) effectively identify water I confirmbodies and tidal flats, with simple principles, convenient operation, and highly reliable results [22,23]. Compared to NDWI, MNDWI enhances the reflectance contrast between the shortwave infrared and green bands, generating stronger tonal differentiation that better distinguishes water bodies from building shadows, making it more suitable for river island extraction [23,24,25,26]. Therefore, this study adopts MNDWI for threshold segmentation, with MNDWI calculation formula as follows:

$$MNDWI={\displaystyle \frac{R_G-R_{SWIB}}{R_G+R_{SWIB}}}$$

(1)

where R_G and R_SWIB represent the reflectance values of the green and shortwave infrared bands, respectively.

Given spectral variations across Landsat sensors (MSS/TM/ETM+/OLI), band assignments were standardized to ensure consistency (

Table 2. Band assignments for MNDWI calculation by sensor.

Sensor (Landsat Satellite)	Green Band (RG)	Shortwave Infrared Band (RSWIR)
MSS (Landsat-1/2)	Band 2	Band 4
TM (Landsat-5)	Band 2	Band 5
ETM+ (Landsat-7)	Band 2	Band 5
OLI (Landsat-8)	Band 3	Band 6

), and thresholds were calibrated individually for each sensor type (not a global fixed value) to account for differences in spatial resolution and spectral response:

Through the interpretation of images and prior knowledge, the optimal threshold for accurately distinguishing water bodies from land in the image can be identified (Figure 3). The verification of this threshold yields favorable results (Figure 4). The same method is also applied to other images within the study area, resulting in the acquisition of respective thresholds for each image (

Table 3. Water-land Demarcation Threshold of River Island.

Years	1974	1979	1984	1989	1994	1999	2004	2009	2014	2019	2024
Sensors	MSS	MSS	TM	TM	TM	TM	TM	TM	ETM+	OLI	OLI
Threshold	0.254	0.504	0.359	0.383	0.335	0.263	0.371	0.264	0.347	0.204	0.271

).

After setting the threshold for the single-band image, a binarization approach was applied to separate river islands from the surrounding Yangtze River water. Pixels with brightness values meeting or exceeding this threshold were categorized as river island areas, labeled with “1”, while all other pixels, corresponding to water bodies, were labeled with “0”. The binarization classification rule is defined as follows:

$$\beta =\left\{\begin{array}{c}1 \alpha \ge \gamma \hfill \\ 0 \alpha \le \gamma \hfill \end{array}\right.$$

(2)

where β is the image after binarization, α is the gray value of every pixel of the image, and γ is the threshold value between the land and water values.

After image binarization, cluster analysis is further conducted on the images, with unsupervised classification implemented using the K-means clustering method. The classified images lack spatial continuity; the “Clump Classes” function in ENVI was used to cluster and merge adjacent similar regions, supplemented by the application of mathematical morphological operators to eliminate noise.

The effectiveness of the extraction result map was verified through confusion matrix analysis. For each study year, 1000 validation samples (500 land samples and 500 water samples) were randomly selected from high-resolution data. The results (

Table 4. Accuracy Metrics for representative years.

Year	Sensors	OA (%)	Kappa	PA (Land,%)	UA (Land, %)
1974	MSS	92.8	92.8	92.8	91.8
1979	MSS	93.2	93.1	93.2	92.4
1984	TM	93.6	93.6	93.6	93.0
1989	TM	94.0	94.0	94.0	93.6
1994	TM	95.0	94.8	95.0	94.2
1999	TM	95.2	95.1	95.2	94.6
2004	TM	94.8	94.5	94.8	94.0
2009	TM	95.0	94.9	95.0	94.4
2014	ETM+	94.4	94.1	94.4	93.5
2019	OLI	95.8	95.6	95.8	95.0
2024	OLI	96.0	95.9	96.0	95.3

) show that the overall accuracy of all years exceeded 92%, and the Kappa coefficients were all >0.85, confirming that the extraction results are highly consistent with the true values.

Post-refined binarized images were vectorized in ENVI 5.6 to generate polygon boundaries of river islands, with topological errors such as self-interactions corrected using the “Clean” tool. Island areas were computed using the geometric measurement tool in ArcGIS 10.8 (Redlands, CA, USA), with units converted to km² (rounded to two decimal places.

3.2. Methodology for Analyzing the Trend of Area Change in River Islands

The morphological characteristics and surface area data of the selected river islands from 1974 to 2024 were evaluated through systematic analysis of remote sensing imagery. To clarify trends in area variation, both the individual area anomalies of each river island and the cumulative area anomaly of the four islands were calculated using the following formulas:

$${∆A}_{ik}=A_{ik}-\overline{A_k}$$

(3)

where k stands for a specific island or group of islands, i for a given year; ${∆A}_{ik}$ is the area anomaly of the island(s) in that year; $A_{ik}$ represents the actual area of the island(s) in the same year; and $\overline{A_k}$ refers to the 50-year average area of the specified island(s).

Using data starting from 1974, the formula for calculating the average rate of change is derived as follows:

$${\eta }_{12}={\displaystyle \frac{A_2-A_1}{5\times A_1}}\times 1000\mathrm{‰}$$

(4)

where η₁₂ represents the annual average rate of area change over a specified period, A₁ denotes the initial area of the island, and A₂ signifies its subsequent area. “5” corresponds to 5-year interval. A positive η₁₂ indicates an expansion in area; a value of zero signifies no change; and a negative value denotes a reduction. The 50-year average annual rate of area change is calculated as the mean of η₁₂ across the ten distinct time intervals.

3.3. Analytical Approaches to Assessing Influential Variables

The Pearson correlation coefficient was used to explore the relationship between river island areas and key influencing factors such as annual runoff and sediment transport. The formula applied for this analysis is as follows:

$$R_{xy} ={\displaystyle \frac{{\sum }_{i=1}^n\left[\left(x_i-\overline{x}\right)\left(y_i-\overline{y}\right)\right]}{\sqrt{{{\sum }_{i=1}^n\left(x_i-\overline{x}\right)}^2}{{\sum }_{i=1}^n\left(y_i-\overline{y}\right)}^2}}$$

(5)

4. Results

4.1. Extraction of the River Island Area

To accurately obtain the spatiotemporal distribution characteristics of the four typical river islands (Chongming Island, Changxing Island, Hengsha Island, and Jiuduansha) in the Yangtze River Estuary (YRE) from 1974 to 2024, this study utilized the multi-source Landsat imagery (MSS/TM/ETM+/OLI) selected in Section 2.2 as the data foundation, and relied on the technical workflow of “MNDWI threshold segmentation—binarization modeling—K-means clustering optimization” established in Section 3.1 to conduct extraction. Additionally, multi-dimensional quality control and verification were implemented to ensure the reliability and accuracy of the extraction results, providing high-quality data support for subsequent analysis of evolutionary patterns.

4.1.1. Extraction Technical Workflow and Key Parameters

• Preprocessing and Index Calculation

In the image preprocessing stage (detailed in Section 2.3), geometric precision correction (error < 1 pixel), projection transformation (from UTM Zone 51 N to Lambert Azimuthal Equal Area), and strip repair for Landsat-7 ETM+ SLC-off imagery (RMSE = 0.018, SSIM = 0.92) have been completed. The focus of this stage is to calculate the MNDWI. Based on the sensor band matching scheme determined in Section 3.1 (Table 2), MNDWI values were computed using Formula (1) to enhance the spectral contrast between water bodies and land, laying the foundation for subsequent segmentation.

2.

Threshold Segmentation and Binarization

Considering differences in spatial resolution (60 m for MSS, 30 m for others) and spectral response among different sensors, the “image-by-image calibration” strategy proposed in Section 3.1 was adopted to determine the water-land segmentation threshold. Combined with high-resolution reference images (Google Earth historical imagery, 1:10,000 topographic maps), 200 typical water and land sample points were randomly selected in each island area. The optimal threshold was determined based on the statistical distribution of MNDWI values of the samples (Table 3). Using this threshold, the MNDWI imagery was binarized via Formula (2): pixels with MNDWI values ≥ threshold were labeled as “1” (island area), and those with values < threshold were labeled as “0” (water body), initially realizing water-land separation.

3.

Clustering Optimization and Noise Removal

Isolated noise pixels (such as cloud shadows and water suspended solids) may exist in the binarized imagery. The K-means clustering algorithm (K = 2, representing land and water) was used to optimize the results: a 5 × 5 pixel window was employed to merge spatially continuous regions of the same type. Subsequently, the “Clump Classes” function in ENVI and mathematical morphological operations (dilation-erosion) were applied to eliminate isolated noise patches with an area < 0.01 km², correct the boundary of temporarily submerged areas caused by tidal fluctuations, and finally generate a complete vector mask of the river islands.

4.1.2. Multi-Dimensional Verification of Extraction Results

To systematically verify the reliability of the extraction results, verification was conducted based on three-tier objectives of “accuracy—consistency—anti-interference”, as detailed below:

• Quantitative Accuracy Verification (Based on Confusion Matrix)

Taking high-resolution reference data as the true value, 1000 validation samples (500 land samples and 500 water samples) were randomly selected from the extraction results of each island. The Overall Accuracy (OA), Kappa coefficient, Producer’s Accuracy (PA, for land), and User’s Accuracy (UA, for land) were calculated using the confusion matrix. The selection of reference data followed the principles of “temporal matching and full regional coverage”: 2000–2024, 1 m resolution Google Earth historical imagery (accessible via the Google Earth Pro client: https://www.google.com/earth/versions/#earth-pro (accessed on 1 January 2025)) was used, covering all four river islands; 1974–1999, 5 m resolution aerial photographs sourced from the Shanghai Municipal Bureau of Planning and Natural Resources were adopted, with the original data retrievable through the bureau’s official data service channel (http://ghzyj.sh.gov.cn (accessed on 1 February 2025)) and covering Chongming Island and Changxing Island; 1984–2009, 10 m resolution 1:10,000 topographic maps provided by the Chinese Academy of Sciences were utilized, and relevant data are available through the Science Data Bank of the Chinese Academy of Sciences (https://www.scidb.cn (accessed on 1 February 2025)) to supplement tidal flat verification for Jiuduansha and Hengsha Island. The time difference between all reference data and extracted imagery was <1 year.

The results showed that OA of all years exceeded 92%, with OLI imagery (2019–2024) reaching 95.8–96.0%; Kappa coefficients were all >0.85, meeting the “substantial agreement” standard (Landis-Koch scale); PA for land ranged from 92.8% (1974 MSS) to 96.0% (2024 OLI), and UA for land ranged from 91.8% to 95.3%, indicating low omission and commission errors. The slightly lower accuracy of MSS imagery was due to its 60 m resolution, which struggled to distinguish narrow tidal flats (such as small sandbars in northern Jiuduansha), but it still met long-term evolutionary analysis requirements (OA > 90%, Kappa > 0.8).

2.

Cross-Sensor Consistency Verification

To ensure comparability of extraction results across sensors, spatial consistency analysis was conducted for three core sensor transition nodes (1979–1984: MSS → TM; 2009–2014: TM → ETM+; 2014–2019: ETM+ → OLI). 500 sampling points were evenly set along island boundaries to calculate mask Intersection over Union (IoU) and shoreline offset distance (

Table 5. Cross-Sensor consistency metrics for adjacent time pairs.

Consecutive Time Pair	Sensor Combination	IoU of Island Masks	Mean Shoreline Offset (m)	Max Shoreline Offset (m)	Dominant Cause of Offset
1979–1984	MSS vs.TM	0.89	12.3	18.5	Natural siltation (Chongming North Branch)
2009–2014	TM vs. ETM+	0.92	10.5	15.7	Artificial reclamation (Hengsha East Tidal Flat)
2014–2019	ETM+ vs. OLI	0.95	7.8	12.4	Vegetation—induced sedimentation (Jiuduansha)

).

All transition nodes had IoU > 0.89, with average shoreline offset < 15 m (smaller than sensor resolution: 30–60 m). This verified the effectiveness of the “sensor-by-sensor threshold calibration” strategy in Section 3.1, confirming cross-sensor result consistency for long-term series comparison.

3.

Seasonal Interference Verification

To address extraction biases from tidal fluctuations and seasonal hydrology, further verification was conducted on the basis of “prioritizing dry-season imagery (November–February)” (Section 2.2). For years with wet-season imagery (such as the TM imagery from August 1989 and the TM imagery from July 2004), wet-season results were tide-level calibrated (adjusted to multi-year mean low tide level ± 0.5 m) and compared with dry-season results to quantify seasonal differences and develop differentiated calibration strategies (

Table 6. Seasonal Area Difference in River Islands (1974–2024).

Island Name	Avg. Seasonal Difference (%)	Max. Seasonal Different (%)	Dominant Different Area	Calibration Strategy	Post-Calibration Difference (%)
Chongming Island	1.2	2.0	Dongtan Tidal Flat	Single-period dry-season imagery comparison + tide-level correction	0.8
Changxing Island	0.8	1.5	Qingcaosha Reservoir Periphery	Dyke boundary fixation + local tidal flat calibration	0.5
Hengsha Island	1.5	2.1	East Reclamation Area	Dyke boundary fixation + multi-period imagery superposition	1.2
Jiuduansha	2.0	2.8 (1998 flood)	Zhongsha/Xia Tidal Flats	Multi-period dry-season imagery fusion + flood-period correction	1.5

).

Ultimately, all islands had seasonal area difference < 2.1%, confirming extraction results reflect long-term stable shorelines, unaffected by short-term seasonal hydrology.

4.1.3. Characteristics of Extraction Results in Typical Years

The extraction results of the four islands from 1974 to 2024 (Figure 5) show significant differences in morphological evolution: Chongming Island maintained a narrow, elongated shape with continuous extension at both ends, where scattered sandbars (Yonglongsha, Xinglongsha, Xincunsha) in 1974 gradually connected to the main island through siltation, forming a continuous northern land belt by 2024, and the shoreline of Dongtan in the eastern part showed “step-like” expansion before stabilizing; Changxing Island underwent slow morphological changes, retaining a strip-like shape, with expansion concentrated around the Qingcaosha Reservoir in the north and the Changxing Submerged Dyke in the southeast, and after 2000, artificial reclamation straightened its naturally tortuous shoreline; Hengsha Island evolved from a regular triangle in 1974 to an irregular polygon in 2024, with natural siltation in the northwest being dominant in the early stage, while the expansion direction shifted to eastern artificial reclamation after 2000, and its shape changed from natural irregularity to geometric regularity; Jiuduansha gradually integrated from scattered sandbars in 1974 into an elliptical whole, though the 1998 Yangtze flood (with a peak water level of 12.63 m) temporarily fragmented its shape, and after 1999, it entered a period of rapid siltation—with particularly significant growth between 2014 and 2024—forming a stable east–west oriented main structure. Calculations via the ArcGIS geometric measurement tool showed that the total area of the four islands increased from 1428.32 km² in 1974 to 1805.97 km² in 2024, with a cumulative growth of 377.65 km².

4.2. Analysis on the Overlay Evolution of River Islands

In this study, remote sensing images collected at 5-year intervals from 1974 to 2024 were overlaid to systematically analyze the changes and expansion trends of river islands in YRE over the past 50 years (Figure 6 and Figure 7).

4.2.1. Chongming Island

The expansion of Chongming Island is primarily characterized by the North Branch prograding toward the mainland and Dongtan extending seaward, with the northern and northeastern parts hosting the densest and largest color-coded patches (Figure 6a). Taking the white base map of 1974 as the “initial outline,” there were multiple predominantly isolated sandbars in the northern and eastern areas; over time, the color-coded patches in the North Branch area gradually extended outward from small patches close to the base map, advancing continuously in a belt-like manner to connect the scattered sandbars (Yonglongsha, Xinglongsha, and Xincunsha) to the main island, eventually eliminating the intervening water areas and forming a continuous northern land belt. The Dongtan area, meanwhile, expanded outward in a fan shape: from 1974 to 1990, its color-coded patches advanced rapidly outward, filling the intervening waterway between Tuanjiesha and the main island to link the two, with edges expanding in a “stepped” pattern; from 1990 to 2010, the expansion of these patches slowed, their edges becoming smoother, yet the area continued to extend seaward, with the expansion direction proceeding eastward around Tuanjiesha as the axis; after 2010, the Dongtan area showed minimal color changes and largely maintained its existing shape.

4.2.2. Changxing Island

The expansion of Changxing Island is mainly concentrated on the northern and southeastern sides of the island, as shown in Figure 6b. From 1974 to 1999, the color coverage was extensive, with the color edges showing tortuous, blurred, and gradual transition characteristics; the expansion was dominated by large-scale northward extension, concentrated around Zhongyangsha, and was particularly intense between 1989 and 1994. After 2000, the color edges became straight and regular, coinciding with the alignment of dikes and embankments, and the expansion mainly focused on the reclamation areas near Qingcaosha Reservoir on the northern side and Changxing Submerged Dyke on the southeastern side.

4.2.3. Hengsha Island

The expansion of Hengsha Island is primarily driven by natural silting in the northwest in the early stage and artificial reclamation eastward in the later period. From 1974 to 1999, the color-coded areas of Hengsha Island were mainly distributed on the northwest side, with blurred boundaries, irregular shapes, and intricate interweaving of various colors; the expansion was slow and scattered, extending in a fan-like pattern. From 2000 to 2014, the expansion accelerated, and the color-coded areas shifted from scattered distribution to large-scale, concentrated extension toward the eastern ocean, with their morphology changing from naturally irregular to regular geometric shapes. From 2015 to 2024, the expansion direction was completely locked to the east, where the reclamation area became the sole core of expansion; the color-coded areas formed a closure in the east, with a small portion expanding toward the south (Figure 6c).

4.2.4. Jiuduansha

The expansion pattern of Jiuduansha has evolved from an overall fragmented and disordered state to a regular ellipse (Figure 7). From 1974 to 1979, Jiuduansha transformed from scattered sandbars into a pattern where the Middle and Lower Shoals merged into one while the Upper Shoal remained independent. Its shape transitioned from fragmentation to partial connection, though the edges of the color-coded areas remained blurred and tortuous (Figure 7a). Figure 7b demonstrates the morphological evolution of Jiuduansha from 1984 to 1994, showing its transformation from three separate sandbars into an initially interconnected landmass comprising Upper, Middle, and Lower sections, with concurrent northward expansion. Its shape shifted from scattered islets to an integrated whole, the edges of the color-coded areas became relatively regular, and the area increased to 81.15 km². In July–August 1998, a major flood hit the Yangtze River Basin. Jiuduansha was affected, with the peak flood level reaching 12.63 m, and the morphology of its sandbars was reshaped [20]. The 1999 image (taken 8 months after the flood receded) showed fragmented sandbars of Jiuduansha; by then, Jiuduansha had an irregular distribution and fragmented edges, consisting of the Upper, Middle and Lower Shoals. From 1999 to 2024, the expansion accelerated: the Middle and Lower Shoals merged into one and extended northward; the edges of the color-coded areas changed from fragmented to smooth, and the shape transformed from a transitional connection to an overall ellipse. Particularly between 2014 and 2024, the expansion was intense, with the area increasing to 110.46 km² (Figure 7c).

4.3. River Island Area Evolution Trend

Geometric analysis of remote sensing imagery using ArcGIS enabled the calculation of each island’s area across different time periods (Figure 8). Over the 50-year period, the average areas of the four islands, ranked from largest to smallest, were: Chongming Island, Changxing Island, Hengsha Island, and Jiuduansha. Chongming Island, the largest, had an average area of 1297.77 km² and showed a general growth trend with an annual average increase of 10.41 km², the fastest growth rate among the four. Changxing Island, with an annual average area of 108.94 km², experienced significant early growth followed by fluctuations, but still maintained an overall annual average growth of 0.88 km². Hengsha Island, averaging 84.08 km² annually, grew slowly in the early years but accelerated after 2000, with an annual average increase of 2.53 km², the fastest growth rate. Jiuduansha, the smallest, had an annual average area of 58.57 km² and an annual growth rate of 1.69 km².

From 1974 to 2024, the four islands in YRE had an average total area of 1549.36 km², with an annual expansion of 15.51 km² on average. The overall trend was characterized by initial slow growth followed by accelerated expansion (Figure 9): before 1999, the total area increased modestly at an average annual rate of 14.32 km²; after 1999, however, it expanded more rapidly, with an average annual growth of 16.71 km².

Derived from Equation (4), the annual average rate of area change across the six specified time periods is shown in Figure 10. Jiuduansha exhibited the most significant change, with an annual increase of 81.5‰. In contrast, Chongming Island—the largest of the four—had a growth rate of 8.8‰ per year. By area change rate, the islands rank from highest to lowest as: Jiuduansha, Hengsha Island, Changxing Island, and Chongming Island. Moreover, the annual average rate of area change for each island has been tested at a 95% confidence level and shows a high degree of correlation (Figure 11).

5. Discussion

5.1. Natural Factors Driving Changes in River Islands

This study analyzed the relationship between variations in river island area and concurrent hydrological-sediment factors (Figure 12 and Figure 13), using 1974–2024 annual runoff and sediment discharge data from the Datong Hydrological Station. As a key gauge of upstream water-sediment dynamics, the station monitors 94% of the Yangtze River Basin’s downstream area. Results revealed a weak positive correlation between river island area and annual runoff (Pearson’s r = 0.293, p < 0.05), indicating the island area exhibits only a slight upward trend with increased runoff. This weak association arises because runoff primarily regulates short-term hydrodynamic conditions: higher runoff may intensify erosion in some island segments but also promotes localized sediment redistribution, ultimately leading to minor net changes in island area. In contrast, a strong negative correlation was observed between river island area and annual sediment discharge (r = −0.915, p < 0.05), meaning river islands tend to expand as Datong Station sediment discharge decreases.

River island expansion does not rely solely on upstream sediment flux but is closely tied to the estuary’s overall sediment supply and spatial distribution. While reduced upstream sediment input—reflected by declining Datong Station sediment discharge—would theoretically limit island growth, local estuarine processes have effectively offset this deficit. For nearshore tidal sediment transport, the trumpet-shaped North Branch of YRE allows tidal dynamics to carry large volumes of coastal sediment to island-adjacent shoals, serving as an important supplementary sediment source for northern islands such as the North Branch of Chongming Island [27,28]. For wind-wave redistribution, storm events and wind-wave interactions stir up sediment deposited on tidal flats and transport it to island margins, providing material support for expansion [29]. For estuarine flocculation deposition, freshwater-saltwater mixing in the estuary’s maximum turbidity zone triggers flocculation of fine-grained sediments, forming high-concentration deposition zones that directly supply sediment for island growth [30]. For vegetation-induced sedimentation, tidal flat vegetation such as Spartina alterniflora reduces water flow velocity, with a sediment-trapping efficiency over nine times that of unvegetated areas, significantly accelerating intertidal material accumulation [31,32]. Collectively, these processes show that sediment discharge influences river island evolution by regulating the “total volume and spatial distribution of sediment required for island development.” Even as Datong Station sediment discharge decreases, the estuary’s redistribution of existing sediment still sustains island expansion, confirming sediment discharge’s core role among natural driving factors.

Considering the typical time for sediment to travel from Datong Station to YRE, undergo flocculation, and deposit on island surfaces, this study tested 1–3 year lag periods. Results showed the weak positive correlation between runoff and island area remained stable across all lags (r = 0.278–0.295, p < 0.05), with no significant time lag. This is because runoff primarily affects short-term hydrodynamic conditions like erosion intensity, and its impact on island area is quickly reflected through minor fluctuations. In contrast, the negative correlation between sediment discharge and island area was strongest at a 2-year lag (r = −0.932, p < 0.05), slightly stronger than the concurrent correlation (r = −0.915). This 2-year lag aligns with hydrological timelines: upstream sediment (monitored by Datong Station) takes approximately 2 years to reach the estuary, undergo flocculation and redistribution, and finally deposit on island surfaces to support expansion.

Long-term 1974–2024 data from Datong Station (Figure 14) further contextualizes the relationship between hydrological-sediment factors and island evolution. The station’s multi-year average annual runoff was 8903.63 × 10⁸ m³, showing a fluctuating downward trend (annual reduction: 15.012 × 10⁸ m³) with a 2020 maximum of 11,810 × 10⁸ m³ and 2023 minimum of 6720 × 10⁸ m³. Annual sediment discharge fluctuated more significantly, with a multi-year average of 2.75 × 10⁸ tons (annual decrease: 0.0716 × 10⁸ tons), a 1981 peak of 5.37 × 10⁸ tons, and 2023 trough of 0.445 × 10⁸ tons. A critical water-sediment turning point occurred in 2003, after the Three Gorges Dam’s completion: sediment discharge plummeted from an average of 4.1 × 10⁸ tons/year (1974–2002) to 1.2 × 10⁸ tons/year (2003–2024). Notably, despite a ~100 million ton reduction in the watershed’s annual sediment discharge into the sea, Yangtze River Estuary river islands have continued to expand—highlighting the resilience of the estuary’s diversified sediment supply mechanisms. Additionally, long-term sea-level rise indirectly alters sediment transport pathways by modifying tidal dynamics and sedimentation datums, further regulating sediment supply for island growth [33]. In summary, upstream water-sediment fluxes establish the macro foundation for island evolution, while the estuary’s unique sediment redistribution processes are key to sustaining expansion amid reduced upstream sediment input.

5.2. Human Factors

With the acceleration of urbanization and industrialization in the Yangtze River Basin, coastal regions especially Shanghai have increasingly adopted “marine expansion” as a core strategy to alleviate land scarcity driven by population growth and urban development [34]. Among various human activities, land reclamation stands out as the most direct driver of river island area expansion (Figure 15), though its quantitative impacts on each island vary by project scope and implementation timeline. Chongming Island, sediment accumulation in its northern branch has long compromised drainage capacity; targeted reclamation of Yonglongsha, Xinglongsha, and Xincunsha was implemented in 1992, 2001, and 2010, respectively [35]. Post-project monitoring shows these three reclamation zones contributed 18.7 km², 12.3 km², and 9.5 km² to Chongming Island’s total area increment, accounting for 34% of the island’s total expansion between 1990 and 2020. Changxing Island, the construction of Qingcaosha Reservoir (initiated in 2008, completed in 2010) and southeast tidal flat reclamation (2012–2016) added 15.2 km² and 8.9 km² to its area, respectively—these two projects alone accounted for 62% of the island’s growth during 2008–2016. Hengsha Island’s Dongtan reclamation, aligned with Shanghai’s strategic development plan [36], was implemented in three phases (2005–2008, 2010–2013, 2018–2020), with each phase contributing 7.8 km², 10.1 km², and 12.4 km²; by 2020, reclaimed areas had become the dominant source of the island’s expansion, representing 85% of its total area growth since 2000. Jiuduansha, the deep-water channel project (Phase I: 1998–2002; Phase II: 2006–2010) and Spartina alterniflora planting (large-scale introduction in 2001) have synergistically promoted sedimentation [19]. Quantitative analysis shows the deep-water channel’s siltation-promoting structures increased local sediment deposition rates by 1.2–1.8 cm/year, while Spartina alterniflora -covered zones (expanding from 0.3 km² in 2001 to 12.7 km² in 2020) trapped an additional 21.5 km² of sediment over 20 years—jointly accounting for 73% of Jiuduansha’s area increment during this period.

Beyond reclamation, large-scale hydraulic engineering and excessive sand mining have exerted profound, yet often under-quantified, impacts on river island evolution. The Three Gorges Dam, as a pivotal upstream infrastructure, intercepts a substantial portion of sediment from the upper Yangtze River. According to sediment monitoring data from the Changjiang Water Resources Commission [20], the dam retained 82–87% of upstream sediment during 2003–2024, with an average annual sediment interception volume of 3.2 × 10⁸ tons. A comparative analysis of sediment transport at Datong Station and sediment deposition in the Three Gorges Reservoir from 2003 to 2024 (Figure 16) further confirms this interception effect: reservoir sediment deposition exceeded Datong Station’s sediment transport volume during 2006–2012 and 2018, directly reflecting the dam’s role in reducing downstream sediment supply. This interception has significantly reduced sediment discharge at Datong Station: from 1974 to 2002 (pre-dam), the station’s multi-year average sediment discharge was 4.1 × 10⁸ tons; by 2024, this figure had dropped to 1.08 × 10⁸ tons, representing a 69% reduction relative to the pre-dam average (consistent with the 4.1 × 10⁸ tons baseline, rather than the 1974–2024 multi-year average of 2.75 × 10⁸ tons which includes post-dam data). This sediment deficit has disrupted the natural alluvial balance of river islands: for example, Chongming Island’s northern branch shoreline historically accreting at 100–200 m/year has shifted to an erosion rate of 3–5 m/year since 2010. Concurrently, the dam’s “flood storage and dry-season discharge” operation modifies runoff patterns: during flood seasons (June–August), reduced peak flows lower sediment transport capacity, intensifying erosion in the YRE’s main channel (with a measured erosion depth of 0.8–1.2 m/year in Changxing Island’s southern waters); during dry seasons (December–February), increased downstream discharge temporarily alleviates saltwater intrusion but causes the Chongming Dongtan wetland to shrink by 0.3–0.5 km²/year.

Excessive sand mining in the Yangtze River channel and its affiliated lakes (Dongting Lake, Poyang Lake) has further exacerbated sediment scarcity. As shown in Figure 17, which compares annual sand mining volumes in the basin with Datong Station’s sediment discharge from 2021 to 2024, annual sand mining volumes during this period ranged from 1.5 × 10⁸ to 1.9 × 10⁸ tons, consistently surpassing Datong Station’s annual sediment discharge (1.02 × 10⁸–1.08 × 10⁸ tons during the same period). This imbalance has led to “sandy degradation” of shoals surrounding Hengsha Island: the median grain size of surface sediment in near-island shoals increased from 0.05 mm (2000) to 0.12 mm (2024), reducing the sediment-stabilizing capacity of wetland vegetation and increasing shoreline erosion risk by 40–60%. Additionally, the South-to-North Water Diversion Project (East Route, Phase III completed in 2013) diverts an average of 1×10⁹ m³/year of water from the lower Yangtze, reducing estuarine runoff by 8–12% during dry seasons. This runoff reduction has prolonged saltwater intrusion events: since 2013, the Yangtze Estuary has experienced 3–5 additional days of saltwater intrusion annually (with chloride concentrations exceeding 250 mg/L), threatening the water supply of coastal cities and indirectly affecting intertidal sedimentation by altering freshwater-saltwater mixing dynamics [37,38].

In recent years, Shanghai has begun to balance development and ecological protection. While continuing to implement targeted reclamation projects such as the 2021–2025 Lingang New Area expansion, the city has established nature reserves and natural parks to protect key coastal wetlands. For instance, the Chongming Dongtan National Nature Reserve (expanded to 326 km² in 2019) has restricted reclamation within its boundaries, slowing the island’s eastern expansion rate from 1.2 km²/year (2000–2010) to 0.4 km²/year (2019–2024). Similarly, Jiuduansha’s Spartina alterniflora control project (initiated in 2015) has reduced the invasive plant’s coverage by 45% (from 12.7 km² in 2020 to 6.9 km² in 2024), restoring native wetland ecosystems while adjusting sediment trapping rates to a more ecologically sustainable level. These measures reflect a shift toward “ecologically constrained development”, aiming to reconcile human needs with the long-term stability of river island systems.

6. Conclusions

Based on Landsat series images (MSS/TM/ETM+/OLI) from 1974 to 2024 and hydrological data from the Datong Hydrological Station, this study systematically extracts spatiotemporal evolution information of four river islands in YRE, namely Chongming Island, Changxing Island, Hengsha Island, and Jiuduansha, by employing techniques including the improved modified normalized difference water index (MNDWI) threshold method, binarization modeling, and K-means clustering analysis. Integrating the influences of natural hydrological factors and human activities, the study reveals the evolutionary patterns and driving mechanisms of these islands, thereby providing references for the rationality of human economic activities. The research findings are summarized as follows:

(1)

The Landsat satellite series images, including MSS, TM, ETM+, and OLI, demonstrate strong adaptability in studying surface features and extracting information on land–water areas such as river islands. They enable the efficient acquisition of spatiotemporal evolution characteristics of these islands, and the ETM+ images reconstructed using the SLC-off model also fully meet the requirements of scientific research. The integration of the improved MNDWI threshold method with remote sensing technology enables the effective identification of river island area changes, serving as a reliable technical tool for long-term evolution monitoring. Specifically, it supports Yangtze River waterway management by tracking island shoreline migration and waterway siltation trends, while also issuing early warnings to mitigate shipping losses. Additionally, this approach alleviates “land shortage” pressures in coastal cities such as Shanghai: by efficiently quantifying the newly added area of islands, it lays a solid foundation for the planning of reclamation projects.

(2)

From 1974 to 2024, the average total area of the four river islands reached 1549.36 km², with an annual average growth of 15.51 km², showing an overall characteristic of “slow growth in the early stage and accelerated growth in the later stage”. Ranked by average area, the order is as follows: Chongming Island (1297.77 km²) > Changxing Island (108.94 km²) > Hengsha Island (84.8 km²) > Jiuduansha (58.57 km²). In terms of growth rate, Jiuduansha ranks the fastest (84.5‰ per year) while Chongming Island ranks the slowest (8.7‰ per year). Moreover, there are significant differences in the morphological evolution of each island: Chongming Island maintains a narrow and elongated shape and extends at both ends; Jiuduansha has integrated from scattered sandbars into an elliptical shape; Changxing Island presents a strip-like morphology with slow changes; and Hengsha Island has transformed from a regular triangle to an irregular polygon and expanded eastward. These evolutionary characteristics point out directions for “one island, one strategy” economic development: Jiuduansha, with its high growth rate and primitive ecological status, is suitable for developing a low-disturbance ecological economy; relying on its low growth rate, large area, and the advantage of the Dongtan International Important Wetland, Chongming Island can build a high-value-added eco-tourism IP; Changxing Island and Hengsha Island, due to their moderate growth rates and proximity to Shanghai’s main urban area, can undertake supporting coastal industries. Specifically, Changxing Island can construct a ship parts industrial park, and Hengsha Island can layout a cold chain logistics center, both of which are in line with Shanghai’s “marine expansion” strategic needs.

(3)

Despite the significant reduction in upstream sediment discharge caused by the Three Gorges Project, all four river islands have not ceased expanding, and their expansion drivers exhibit differences. The growth of Chongming Island is jointly driven by the natural sedimentation of Yangtze River, the reclamation of Yonglongsha and Xinglongsha sandbars, and the Chongming Dongtan reclamation project. Changxing Island presents a pattern of “southern erosion and northern accretion”, where the sediment deposition on its northwest side mainly relies on the direct promotion of the Qingcaosha Reservoir reclamation project for area growth. Jiuduansha achieves stable growth through artificial sedimentation promotion projects in deep-water channels and ecological measures such as the introduction of Spartina alterniflora. These findings provide guidance for the optimization of large-scale projects: in reclamation projects, the “artificial reclamation + breakwater + sedimentation promotion dam” model of Hengsha Island can be replicated in land-scarce areas such as the Shanghai Lingang New Area; in water conservancy projects, to address the salinization problem faced by hundreds of mu of cultivated land on Chongming Island due to reduced sediment transport from upstream, “flood season ecological sediment transport” dispatch can be implemented to reduce the area of salinized cultivated land and ensure the safety of agricultural production.

(4)

Correlation analysis shows that the area of the river islands has a weak positive correlation with annual runoff (R = 0.293) and a strong negative correlation with annual sediment discharge (R = −0.915): from the natural perspective, runoff fluctuations and tidal sediment transport directly induce the deposition or erosion of the islands; from the human perspective, the Three Gorges Project intercepts upstream sediment, the South-to-North Water Diversion Project reduces estuarine runoff, excessive sand mining alters the sand quality of shoals, and reclamation projects fix newly added land—all of which jointly change the sediment transport pathways and the evolutionary process of the islands. Among these, the average annual runoff of the Datong Hydrological Station from 1950 to 2024 was 896.1 × 10⁹ m³ (with a maximum of 1181 × 10⁹ m³ in 2020 and a minimum of 667.1 × 10⁹ m³ in 2011), and the annual average water transfer volume of the South-to-North Water Diversion Project is 1 billion m³, which further intensifies the coordinated pressure on water resources and ecology in YRE. Based on this evolutionary mechanism, key support can be provided for the economic security of YRE: in the field of flood control and disaster reduction, large runoff during the flood season will exacerbate shoreline erosion in the northern branch of Chongming Island, and the combined reinforcement measure of “concrete embankment + reed revetment” can be adopted to prevent flood backflow from submerging tens of thousands of mu of farmland along the coast; in the field of water resource allocation, to address the saltwater intrusion problem caused by the South-to-North Water Diversion Project, the “ecological flow” at the Datong Station can be reserved to shorten the duration of saltwater intrusion, reduce water treatment costs, and ensure the safety of urban water supply and the stable operation of the economy.