A method for identifying specific ecological impact ranges for large water diversion projects

Abstract

This invention provides a method for identifying the specific ecological impact range of large-scale water transfer projects. This method aims to address the problems of existing technologies in assessing the ecological impact of large-scale water transfer projects, which struggle to accurately separate the impact of artificial water conveyance from the natural water system background, and whose assessment results are static and subjective. The identification method includes: data fusion and environmental variable construction, identification of key hydrological driving factors, preliminary ecological impact potential modeling, model calibration and impact decoupling, impact zoning, and seasonal dynamic evaluation. Using the method described in this invention, the specific ecological impact range of large-scale water transfer projects can be quantitatively identified, and the assessment results can provide a scientific basis for water resource optimization and project benefit evaluation.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of geographic information science and ecohydrology, and specifically relates to a method for identifying the specific ecological impact range of large-scale water diversion projects. Background Technology

[0002] Rivers, lakes, and wetlands form the lifeblood of regional ecosystems. Their structure and function directly determine the transport of water, nutrients, and energy, serving as a crucial foundation for maintaining biodiversity and ecosystem services. Large-scale inter-basin water transfer projects, through the construction of artificial water conveyance systems such as canals and pipelines, fundamentally reshape the hydrological network pattern of the receiving area, triggering a series of profound eco-hydrological effects. Therefore, scientifically and accurately assessing the ecological impacts of such large-scale projects is of paramount importance for optimizing project scheduling, formulating ecological compensation policies, and achieving sustainable water resource utilization.

[0003] Currently, existing technical methods for assessing the ecological impact of water networks or water diversion projects mainly fall into the following categories and have their inherent limitations:

[0004] The first category is the analysis method based on fixed geometric buffer zones. This method establishes a buffer zone of fixed width along a natural river channel or artificial irrigation canal and defines it as the ecological impact zone. For example, a 500-meter or 1000-meter area is designated on each side of a river as the riparian zone. The advantage of this method is its simplicity, speed, and ease of implementation. However, its fundamental flaw lies in its oversimplification and static nature, ignoring the actual spatial heterogeneity of the impact area. In reality, the impact of water networks exhibits irregular patterns due to topographic relief, soil permeability, groundwater depth, and seasonal water flow variations. A fixed, homogeneous buffer zone obviously cannot accurately reflect this complex reality, leading to assessment results that are either too large or omit important impact areas, resulting in severely insufficient accuracy.

[0005] The second category is hydrological and hydrodynamic modeling methods based on physical processes. These methods establish complex mathematical models (such as SWAT, MODFLOW, and MIKE SHE) to simulate dynamic changes in surface runoff, river confluence, soil water movement, and groundwater levels, thereby extrapolating the impact of hydrological changes on the surrounding environment. The advantage of these methods lies in their explicit physical mechanisms, theoretically enabling a more realistic depiction of hydrological processes. However, their application at the regional scale faces significant challenges: First, model construction requires massive amounts of high-precision input data (such as detailed river cross-sections, hydrogeological parameters, and soil physical parameters), which are costly and difficult to acquire; second, model parameter calibration is complex and highly uncertain; and finally, the computational demands are enormous, requiring substantial computing resources and hindering rapid, dynamic evaluation, especially when dealing with sudden ecological relocation or comparing multiple scenarios, where their timeliness is severely limited.

[0006] The third category is direct comparative analysis based on remote sensing imagery. This method compares remote sensing images from two or more periods before and after the implementation of a water diversion project, analyzing changes in indicators such as vegetation indices (e.g., NDVI, EVI) and water indices (e.g., NDWI) to qualitatively determine the response area of ​​the ecosystem. This method is intuitive, has a wide coverage, and can reflect the macro-trend of ecological change, but its main drawback is the difficulty in attributing the impact. Changes in regional ecosystems are the result of multiple factors working together, including not only water diversion projects but also variations in rainfall during the same period, temperature fluctuations, land use / cover changes (e.g., urbanization, reforestation), and adjustments to agricultural management practices. Direct comparative methods struggle to isolate the specific contributions of water diversion projects from this complex background "noise," thus failing to accurately define the true spatial boundaries of the ecological impacts caused by artificial water transfer.

[0007] In summary, existing technologies for assessing the ecological impact of large-scale artificial water networks generally suffer from one or more of the following pain points: (1) contradiction between static and dynamic: the methods are either too static and cannot reflect the seasonal dynamics of the impact; (2) trade-off between simplification and complexity: the methods are either too simplified, leading to distortion, or too complex, leading to application difficulties; (3) confusion of impact sources: most importantly, existing methods generally cannot effectively "decouple" or "separate" the impact of artificial water conveyance networks from the inherent natural river systems in the region and the impact of concurrent climate fluctuations.

[0008] Therefore, there is an urgent need for a new technical method that can leverage the advantages of remote sensing big data while overcoming the limitations of traditional methods, to achieve objective, quantitative, and dynamic identification and assessment of the specific ecological impact range of large-scale water diversion projects. Summary of the Invention

[0009] The purpose of this invention is to overcome the shortcomings of existing technologies in assessing the ecological impact of large-scale inter-basin water transfer projects, and to provide a method for identifying the specific ecological impact range of large-scale water transfer projects. Specifically, this invention aims to solve the following core technical problems:

[0010] (1) How to overcome the static and subjective nature of the traditional fixed-width buffer method and realize a dynamic and refined description of the ecological impact range.

[0011] (2) How to avoid the dependence of complex physical models on massive parameters and huge computational overhead, and propose a data-driven, efficient and highly operable regional scale assessment scheme.

[0012] (3) How to quantitatively and objectively “decouple” or “separate” the specific ecological impacts generated by artificial water conveyance networks from the complex background of multiple factors such as natural water systems, artificial water networks and climate fluctuations, so as to achieve accurate attribution of impact sources and accurate definition of impact scope.

[0013] The objective of this invention is achieved through the following technical solution:

[0014] This invention provides a method for identifying specific ecological impact zones of large-scale water diversion projects. This method integrates multi-source geospatial data and, through an innovative modeling and calibration process, achieves accurate identification of the ecological impact zones of artificial water networks. Specifically, it includes the following steps:

[0015] Step 1: Data Fusion and Environmental Variable Construction

[0016] The water receiving area of ​​a large-scale water diversion project was selected as the study area. Multi-source geospatial data covering the study area were acquired and preprocessed. The data included surface water data, land use / cover data, topographic data, and artificial water network data. A set of environmental variables was calculated based on the data.

[0017] The environmental variables include water network structure indicators, hydrological proximity indicators, and key environmental factors.

[0018] Step 2, Identification of key hydrological driving factors

[0019] A machine learning model was used, with land use type as the target variable and environmental variables calculated in step 1 as feature variables for training. Based on the feature importance evaluation results of the model, N key hydrological driving factors were selected.

[0020] Step 3, Preliminary ecological impact potential modeling

[0021] Based on the value ranges of the N key hydrological driving factors selected in step 2, the data are discretized and binned to construct a joint conditional probability model based on an N-dimensional joint hydrological environmental condition space. This model is then used to calculate the conditional probability of a specific target ecosystem under each joint hydrological environmental condition. The conditional probability is then assigned back to the corresponding spatial pixels to generate a preliminary probability map covering the entire study area and reflecting the potential for comprehensive ecological impact.

[0022] Step 4, Model Calibration and Influence Decoupling

[0023] The area defined by the artificial water network data obtained in step 1 is set as the strong influence anchor point area. Based on the anchor point area, the preliminary probability map generated in step 3 is calibrated by a function to amplify the probability value of areas with similar hydrological characteristics to the anchor point area, thereby generating a final probability map that highlights the specific ecological impact of the large-scale water diversion project.

[0024] Step 5, Delineation of the affected area

[0025] The final probability map generated in step 4 is binarized by setting an optimal probability threshold, thereby delineating the specific ecological impact zone of the large-scale water diversion project.

[0026] Step 6, Seasonal Dynamic Evaluation

[0027] By using surface water data from different seasons as input, steps 1 to 5 are repeated to obtain the specific ecological impact zone of the large-scale water diversion project under different seasons.

[0028] Furthermore, in step 1, the multi-source geospatial data also includes soil property data and climate data; the key environmental factors include soil moisture, soil texture, maximum plant root depth, elevation, and average temperature.

[0029] Furthermore, in step 2, the machine learning model is a random forest classifier.

[0030] Furthermore, in step 2, the key hydrological driving factors include the distance DR to the nearest river or artificial canal, the water network density WD, and the water network connectivity index CI.

[0031] Furthermore, in step 4, the strong influence anchor point area is the boundary range of a large backbone irrigation area or an ecological water replenishment beneficiary area confirmed by the authorities.

[0032] Furthermore, in step 4, the function calibration is implemented through a logistic function, the specific form of which is:

[0033]

[0034] In the formula, This is the calibrated probability value. The initial probability value is given, and k and P0 are preset calibration parameters.

[0035] Furthermore, in step 5, the optimal probability threshold is determined by the following method: selecting multiple candidate probability thresholds, performing binarization processing on the final probability map to delineate candidate influence areas; spatially overlaying each candidate influence area with independent ground verification data, calculating the spatial consistency between the two (including the weighted sum of hit rate and coverage), and selecting the candidate threshold with the highest comprehensive consistency as the optimal probability threshold.

[0036] The advantages of this invention compared to the prior art are as follows:

[0037] 1. This invention introduces a model calibration step based on known artificial irrigation areas or ecological water replenishment areas as "anchor points," which can quantitatively separate specific ecological impact areas caused by large-scale water diversion projects from the mixed influences of natural water systems, artificial water networks, and climate background.

[0038] 2. The entire methodology is data-driven, relying on machine learning models and probability statistics. It eliminates the subjectivity of manually setting buffer widths or model parameters in traditional methods, making the evaluation results more objective and repeatable, and presenting them in the form of continuous probability values.

[0039] 3. The framework of the identification method described in this invention has good scalability and can be easily applied to any other area with a large artificial water network. It only requires replacing the corresponding area data and has broad application value.

[0040] 4. The refined and dynamic ecological impact zone distribution map provided by this invention can directly serve water resource management departments, providing strong spatial decision support for optimizing ecological water replenishment plans, scientifically evaluating the ecological benefits of water diversion projects, and formulating ecological compensation standards. Attached Figure Description

[0041] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0042] Figure 1 This is an overall technical flowchart of the method for identifying the specific ecological impact range of a large-scale water diversion project as described in Example 1;

[0043] Figure 2 This is a schematic diagram of the feature importance analysis results of a random forest model used to identify key hydrological driving factors in a typical study area in an application example.

[0044] Figure 3 This is a schematic diagram of the preliminary ecological impact potential map generated based on key hydrological driving factors in an application example.

[0045] Figure 4 This is a schematic diagram of the final probability map generated after calibration and decoupling in an application example, highlighting the ecological impact of artificial water networks;

[0046] Figure 5 This is a spatial distribution map of the final artificial water network ecological impact zone delineated using the optimal threshold in an application example. Detailed Implementation

[0047] The embodiments described are provided to better illustrate the present invention, but are not intended to limit the scope of the invention to the embodiments described. Therefore, non-essential improvements and adjustments made to the embodiments by those skilled in the art based on the above description are still within the scope of protection of the present invention.

[0048] The endpoints and any values ​​of the ranges disclosed herein are not limited to the precise ranges or values, and these ranges or values ​​should be understood to include values ​​close to these ranges or values. For numerical ranges, the endpoint values ​​of the various ranges, the endpoint values ​​of the various ranges and individual point values, and individual point values ​​can be combined with each other to obtain one or more new numerical ranges, which should be considered as specifically disclosed herein.

[0049] The present invention will be described in detail below through embodiments. It should be understood that the following embodiments are only used to exemplify and further explain and illustrate the content of the present invention, and are not intended to limit the present invention.

[0050] Example 1

[0051] like Figure 1 As shown, this embodiment provides a method for identifying the specific ecological impact range of a large-scale water diversion project, including the following steps:

[0052] Step 1: Multi-source data preprocessing and extraction and calculation of environmental variables

[0053] The water-receiving area of ​​a large-scale water diversion project was selected as the study area. Multi-source geospatial data covering the study area were acquired and preprocessed, and then uniformly resampled to the same spatial resolution (e.g., 30 meters) and geographic coordinate system. The data included surface water data, land use / cover data, topographic data, soil property data, climate data, and artificial water network data. Specifically, this included:

[0054] a. Surface water data: For example, monthly or annual surface water distribution raster data from products such as JRC Global Surface Water.

[0055] b. Land use / cover data: For example, annual land use classification data from GlobeLand30 or similar products, including categories such as arable land, forest land, grassland, water bodies, and construction land.

[0056] c. Topographic data: Digital Elevation Model (DEM), used to extract topographic factors such as elevation, slope, and aspect.

[0057] d. Soil property data: For example, data on soil texture (such as clay, silt, and sand content), soil organic matter, and effective soil layer thickness from products such as HWSD or SoilGrids.

[0058] e. Climate data: For example, rasterized monthly or annual precipitation, temperature, and other data interpolated from CRU, WorldClim, or weather stations.

[0059] f. Artificial water network data: As key prior knowledge, obtain vector boundary data of the main canals, branch canals, water outlets of large-scale water diversion projects and major irrigation areas within the study area.

[0060] Based on the above data, a comprehensive set of environmental variables was calculated and generated for each pixel within the study area; these environmental variables aim to comprehensively characterize local hydrothermal, topographical, and soil conditions that influence ecosystem distribution, specifically including:

[0061] a. Hydrological proximity index: Calculate the Euclidean distance (DR) of each cell to the nearest natural river or man-made canal.

[0062] b. Water network structure indicators: These are calculated within the preset analysis units. Among them, water density (WD) is obtained by calculating the ratio of water area to the total area of ​​the unit; water network connectivity index (CI) is obtained by calculating the ratio of the number of independent water patches to the total number of water pixels, and is used to characterize the degree of water network fragmentation.

[0063] c. Other key environmental factors: soil moisture (SM), soil clay content (CC), maximum rooting depth (RD), digital elevation (DE), and average temperature (AT).

[0064] Step 2: Identification of key hydrological driving factors based on machine learning models

[0065] Within the study area, a stratified random sampling method was used to extract a sufficient number of sample points from the land use / cover data, and the land use type corresponding to each sample point was extracted as the target variable (Y), and all environmental variable values ​​calculated in step 1 were used as feature variables (X).

[0066] A Random Forest classifier was selected, and the model was trained using the aforementioned sample set. This model aims to learn and establish a non-linear mapping relationship between environmental variables and land use type distributions. The model's classification accuracy (e.g., overall precision, Kappa coefficient, precision and recall for each category) was evaluated through cross-validation or by reserving a test set.

[0067] The contribution of each environmental variable to land use classification is quantified using the built-in Mean Decrease in Impurity or Mean Decrease in Accuracy assessment function of the random forest model. The top N variables (e.g., N = 3 or 4) that are directly related to hydrological processes are selected as key hydrological drivers, ranked from highest to lowest importance score.

[0068] Studies have shown that water network density (WD), distance from the river (DR), and water network connectivity index (CI) are often the core hydrological factors that determine the regional ecological pattern.

[0069] Step 3: Preliminary assessment of ecological impact potential based on joint conditional probabilities

[0070] The continuous value ranges of the three key hydrological driving factors (DR, WD, and CI) selected in step 2 are discretized into equal-frequency bins to construct an N-dimensional joint hydrological environmental condition space composed of discretized intervals.

[0071] Traverse all pixels within the study area and count those falling within each joint hydrological environmental condition interval I. ijk The number of pixels within a specific target ecosystem (such as woodland or grassland). and the total number of pixels within that interval. Based on this, the conditional probability of this ecosystem occurring under these hydrological conditions can be calculated:

[0072]

[0073] The calculated conditional probability values ​​are assigned back to all their corresponding spatial pixels, thereby generating a preliminary ecological impact potential map covering the entire study area, reflecting the comprehensive potential or possibility of different regions to support specific ecosystems under the combined effects of natural and artificial water networks.

[0074] Step 4: Model calibration and influence decoupling based on known artificial influence areas.

[0075] Using the artificial water network data obtained in step 1, the areas within the study area identified as dominated by artificial water conveyance are defined as "strong influence anchor points." These areas are the most clearly defined and strongest reference standards for artificial influence, such as the boundary range of large backbone irrigation areas, or the officially confirmed ecological water replenishment beneficiary areas on both sides of the main canal of large water diversion projects.

[0076] Within the strong-influence anchor point area, the preliminary probability values ​​of all pixels obtained in step 3 are extracted. Theoretically, the probability of these areas being affected by the artificial water network should approach 1. Based on this assumption, a calibration function is established to map the preliminary probability values ​​to the calibrated probability values. The specific form of this function is:

[0077]

[0078] In the formula, This is the calibrated probability value. The initial probability values ​​are given, and k and P0 are preset calibration parameters. This function aims to amplify the probability values ​​of areas with similar hydrological characteristics to the anchor point area. Specifically, it aims to amplify the probability values ​​of areas with similar hydrological characteristics to the anchor point area (i.e., similar combinations of DR, WD, and CI).

[0079] The calibration function was applied to the preliminary probability map of the entire study area. After this calibration process, the probability values ​​of areas with hydrological characteristics highly similar to those of the strong influence anchor point area were significantly increased, while the probability values ​​of areas with significantly different hydrological characteristics were suppressed or remained unchanged. This process effectively achieved the goal of "amplifying" and "separating" the contribution of the artificial water network from the mixed signal, generating a final probability map that highlights the ecological impact of the artificial water network.

[0080] Step 5: Final delineation and dynamic evaluation of the ecological impact zone of the artificial water network.

[0081] A series of candidate probability thresholds (e.g., 0.70, 0.75, 0.80, 0.85, 0.90) are selected, and the final probability map generated in step 4 is binarized and segmented to obtain candidate impact areas under different thresholds. These candidate impact areas are then spatially overlaid with independent validation data (such as official records of the ecological water replenishment range of large-scale water diversion projects, specific ecological restoration locations mentioned in news reports, and areas with significantly improved vegetation growth) to calculate their degree of fit. The candidate threshold with the highest degree of fit is selected as the optimal threshold, and the final ecological impact area of ​​the artificial water network is delineated accordingly.

[0082] To assess the seasonal differences in impacts, surface water data from different seasons (such as spring, summer, autumn, and winter) can be used as inputs, and steps 1 through 5 can be repeated. This yields maps of the extent, intensity, and spatial distribution of the ecological impact zone of the artificial water network in different seasons, enabling a dynamic assessment of the ecological impacts.

[0083] Application examples:

[0084] This application example uses the key water-receiving area of ​​a large-scale water conservancy project—the southern Hebei region—as a case study to elaborate on the specific implementation process of a method for identifying the specific ecological impact range of a large-scale water diversion project. After the project's implementation, this region developed a complex water system pattern where natural river networks and artificial water conveyance canals coexist.

[0085] The identification method includes the following specific steps:

[0086] Step 1: Multi-source heterogeneous data fusion and construction of environmental variable system

[0087] S1, Data Acquisition and Preprocessing: Collect the following data covering the study area:

[0088] In Step 1, the data acquisition and preprocessing stage, multi-source spatial datasets covering the study area were collected and integrated. Specifically, surface water distribution was obtained from the JRC Global Surface Water v1.4 monthly raster image extracted from the Google Earth Engine (GEE) platform; land use / cover classification was obtained from the publicly released 30-meter resolution annual land use dataset of China; and topographic data was obtained from the SRTM 30-meter digital elevation model (DEM). Soil properties, such as texture and organic matter content, were obtained from the Harmonized World Soil Database (HWSD) v1.2; and the maximum plant root depth estimate was obtained from the publicly released global plant root depth dataset (such as data products based on the research results of Fan, Y. et al.). Climate variables, including monthly average temperature and precipitation, were extracted from the CRUTS v4.06 dataset. Authoritative vector data of a large-scale water conservancy project, including the main canal, water diversion points, and boundaries of large irrigation areas, were obtained from the official water management department. All raster data were reprojected to the WGS 84 / UTMZone 50N coordinate system and uniformly resampled to a spatial resolution of 30 meters to ensure spatial alignment and analytical consistency.

[0089] 2. Environmental Variable Calculation: Using geographic information system software such as ArcGIS Pro or QGIS, calculate the following environmental variables for each 30m × 30m pixel within the study area:

[0090] a. Hydrological Proximity Index (DR): Using the Euclidean distance tool, calculate the distance of each cell to the nearest river or man-made canal.

[0091] b. Water network structure index (WD, CI): Taking township administrative divisions as the analysis unit, calculate the water network density (WD) and water network connectivity index (CI) within each unit.

[0092]

[0093]

[0094] In the formula, The water network density of the analysis unit u; This represents the total area of ​​pixels within the unit that are identified as water bodies. This represents the total area of ​​the analysis unit. CI is the water network connectivity index for analyzing unit u. The lower the CI value, the fewer, larger, and more continuous the water patches, and the better the connectivity of the water network; conversely, the higher the CI value, the more fragmented the water network.

[0095] c. Other key environmental factors: The elevation (DE), soil clay content (CC), mean temperature (AT), soil moisture (SM), and maximum plant root depth (RD) for each pixel were directly extracted from the preprocessed dataset.

[0096] Step 2: Identification of key hydrological driving factors based on machine learning models

[0097] Within the study area, stratified random sampling was conducted according to the area proportion of land use type, resulting in a total of 100,000 sample points. The land use type of each point was extracted as the label (Y), and the corresponding values ​​of eight environmental variables (DR, WD, CI, SM, CC, RD, DE, and AT) were extracted as features (X).

[0098] A Random Forest classifier was built using Python's scikit-learn library, with the number of trees (n_estimators) set to 100. The sample set was divided into training and test sets at a ratio of 70% and 30%, respectively. The model was trained on the training set and validated on the test set, achieving an overall classification accuracy of 91.7% and a Kappa coefficient of 0.86, indicating that the model has high predictive ability.

[0099] Reference Figure 2 The trained model is used to calculate the feature importance of each environmental variable. Figure 2The graph shows the relative contribution of each factor to land use type classification. As can be seen from the graph, hydrological factors such as water network density (WD), soil moisture (SM), and distance from the river (DR) rank among the most important.

[0100] To simplify subsequent probabilistic modeling and highlight core hydrological impacts, this application example selects three key hydrological drivers that are most representative and have clear physical meaning: distance from the river (DR), water network density (WD), and water network connectivity index (CI).

[0101] Step 3: Preliminary ecological impact potential modeling based on joint conditional probabilities

[0102] The value ranges of the three key hydrological driving factors (DR, WD, and CI) selected were discretized by equal frequency and divided into 10 levels (bins) to form a 10×10×10 three-dimensional joint hydrological environmental condition space.

[0103] Taking the "forest" ecosystem as an example, the calculation process of its conditional probability is as follows: First, define L as the land use type, The three-dimensional environmental conditions are defined by the i-th distance from the river (DR) interval, the j-th water network density (WD) interval, and the k-th water network connectivity index (CI) interval. Then, all pixels within the study area are traversed to calculate the environmental conditions falling within each three-dimensional environmental interval. Count of woodland pixels (L=woodland| ) and the total number of pixels N in that interval ijk Its conditional probability It is calculated by the following formula:

[0104]

[0105] In the formula, Count(pixel | L=woodland∩pixel ∈ I) ijk This indicates that the land use type is forest land and its combination of environmental factors falls within the interval. The total number of pixels in the array. Count(pixel | pixel ∈ I) ijk This indicates that the combination of environmental factors falls within the interval. The total number of all pixels within.

[0106] Calculate each interval The conditional probability value is assigned to all pixels falling within that interval. For example, if the probability of woodland appearing is calculated to be 0.45 under the environmental conditions of DR interval i=2, WD interval j=5, and CI interval k=3, then the initial probability value of all pixels in the study area that meet this combination of environmental conditions is assigned to 0.45.

[0107] Reference Figure 3 The calculated conditional probability value for each interval is assigned back to all pixels corresponding to that interval, generating a preliminary ecological impact potential map covering the entire study area. Figure 3 The diagram visually demonstrates the combined potential for forest development in different regions under the combined influence of natural and artificial water networks. It shows that both natural rivers and artificial canals have a high probability of development, but their effects are intertwined and indistinguishable.

[0108] Step 4: Model calibration and influence decoupling based on known artificial influence areas.

[0109] Load the vector boundary data of the large backbone irrigation area obtained in step 1, and define its internal region as the "strong influence anchor point area". These irrigation areas have long been supplied with water by a large water conservancy project, and their ecosystem status is dominated by artificial water transfer.

[0110] Within the anchor point area, preliminary probability values ​​for all pixels are extracted, assuming the "true" probability of human influence in these areas should be 1. By fitting the relationship between these preliminary probability values ​​and the target value of 1, a nonlinear calibration function is established using the logistic function.

[0111]

[0112] In the formula, These are the calibrated probability values; is the initial probability value obtained in step 3; k is the steepness parameter of the function, which controls the strength of probability stretching, and is a preset value greater than 0. For example, in this application example, k can be 10; P0 is the center point translation parameter of the function, that is, the inflection point position of the probability curve, and is a preset value between 0 and 1. For example, in this application example, P0 can be 0.5.

[0113] Reference Figure 4 This calibration function is then applied to the preliminary probability map of the entire region. Figure 4 and Figure 3The comparison shows that, after calibration, the probability values ​​of areas similar to the hydrological characteristics of the irrigation district (i.e., specific combinations of DR, WD, and CI) are significantly amplified, while those of other areas are suppressed. In particular, the probability values ​​are significantly enhanced along narrow strips of the main canals of large-scale water conservancy projects, while the probability values ​​are relatively weakened along some natural rivers far from the artificial water system. This process successfully highlights the influence of the artificial water network from the mixed signal, generating the final probability map.

[0114] Step 5: Final delineation and dynamic evaluation of the ecological impact zone of the artificial water network.

[0115] Select a series of candidate probability thresholds T (e.g., T ∈ {0.70, 0.75, 0.80, 0.85, 0.90}), and then... Figure 4 The final probability map shown is binarized to obtain candidate influence regions A under different thresholds. candidate (T). In this application example, the official record used for verification is a list of names of the beneficiary areas for ecological water replenishment in the central route of a large-scale water conservancy project. This list includes the names of administrative divisions (such as counties, districts, and townships) or specific geographical units (such as specific river names) that are affected by ecological water replenishment. To quantitatively evaluate the quality of the delineation results under different thresholds, this application example uses a comprehensive evaluation method based on hit rate and coverage. The specific steps are as follows:

[0116] 1) Associating Spatial Units: Associating each region name in the official name list with its corresponding geospatial vector boundary to obtain a set of reference verification units. .

[0117] 2) Calculate the overlap area: For each candidate threshold T, calculate its defined candidate influence area. In each reference verification unit u (where u ∈ Area (within) ∩ u).

[0118] 3) Define the evaluation index: Establish a comprehensive evaluation index Score(T) to evaluate the quality of threshold T, as shown in the following formula:

[0119]

[0120] In the formula, Hit rate refers to the area affected by the candidate. The proportion of covered reference validation units to the total number of reference validation units. This metric measures the breadth of model recognition. Average coverage refers to the candidate impact area. The average area coverage across all hit reference validation cells. This metric measures the depth or strength of model recognition. α and β are weighting coefficients, and α + β = 1, used to balance the importance of breadth and depth. In this application example, α = 0.5 and β = 0.5 can be chosen.

[0121] Calculate the comprehensive evaluation score Score(T) for each candidate threshold T. In this application example, the calculation results show that when the threshold T is set to 0.80, the value of Score(0.80) reaches its maximum, indicating that the defined influence area achieves the best balance between recognition breadth (hitting as many areas of official records as possible) and recognition depth (having sufficient coverage within the hit areas). Therefore, 0.80 is determined to be the optimal threshold.

[0122] Reference Figure 5 A threshold of 0.80 was used to segment the final probability map, resulting in a binarized final distribution map of the ecological impact zone of the artificial water network with clear boundaries. Figure 5 The study clearly demonstrates the specific spatial extent of the significant ecological impact of this large-scale water conservancy project in the region during the four seasons: spring (March-May), summer (June-August), autumn (September-November), and winter (December-February). Comparison reveals that the impact is most widespread in summer and least so in winter. However, compared to before the project's implementation, both the scope and intensity of the winter impact have significantly increased, demonstrating the crucial role of artificial water transfer in maintaining ecosystem stability during the dry season.

[0123] Through this application example, the present invention successfully and quantitatively identified the specific ecological impact range of a large-scale water conservancy project from the background of a complex natural-artificial composite water system, and revealed its seasonal dynamic characteristics, providing a scientific basis for the ecological benefit assessment and optimized scheduling of the project.

[0124] Finally, it should be noted that the above is only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention (such as the application of various formulas, the order of steps, etc.) without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for identifying the specific ecological impact range of a large-scale water diversion project, characterized in that, The method includes the following steps: Step 1: Data Fusion and Environmental Variable Construction The water receiving area of ​​a large-scale water diversion project was selected as the study area. Multi-source geospatial data covering the study area were acquired and preprocessed. The data included surface water data, land use / cover data, topographic data, and artificial water network data. A set of environmental variables was calculated based on the data. The environmental variables include water network structure indicators, hydrological proximity indicators, and key environmental factors; Step 2, Identification of key hydrological driving factors A machine learning model is used, with land use type as the target variable and environmental variables calculated in step 1 as feature variables for training. Based on the feature importance evaluation results of the model, N key hydrological driving factors are selected. The machine learning model is a random forest classifier. Step 3, Preliminary ecological impact potential modeling Based on the value range of the N key hydrological driving factors selected in step 2, the data is discretized and binned to construct a joint conditional probability model based on an N-dimensional joint hydrological environmental condition space. The model is then used to calculate the conditional probability of a specific target ecosystem under each joint hydrological environmental condition. The conditional probability is then assigned back to the corresponding spatial pixels to generate a preliminary probability map covering the entire study area and reflecting the potential for comprehensive ecological impact. Step 4, Model Calibration and Influence Decoupling The area defined by the artificial water network data obtained in step 1 is set as the strong influence anchor point area. Based on the anchor point area, the preliminary probability map generated in step 3 is calibrated by function to amplify the probability value of the area with similar hydrological characteristics to the anchor point area, thereby generating a final probability map that highlights the specific ecological impact of the large-scale water diversion project. The strong impact anchor point area refers to the boundary range of a large backbone irrigation area or an ecological water replenishment beneficiary area confirmed by the authorities. The function calibration is implemented through a logistic function, the specific form of which is: In the formula, This is the calibrated probability value. Here are the initial probability values, and k and P0 are the preset calibration parameters; Step 5, Delineation of the affected area The final probability map generated in step 4 is binarized by setting an optimal probability threshold, thereby delineating the specific ecological impact zone of the large-scale water diversion project. Step 6, Seasonal Dynamic Evaluation By using surface water data from different seasons as input, steps 1 to 5 are repeated to obtain the specific ecological impact zone of the large-scale water diversion project under different seasons.

2. The identification method according to claim 1, characterized in that, In step 1, the multi-source geospatial data also includes soil property data and climate data; the key environmental factors include soil moisture, soil texture, maximum plant root depth, elevation, and average temperature.

3. The identification method according to claim 1, characterized in that, In step 2, the key hydrological driving factors include the distance DR to the nearest river or artificial canal, the water network density WD, and the water network connectivity index CI.

4. The identification method according to claim 1, characterized in that, In step 5, the optimal probability threshold is determined by the following method: selecting multiple candidate probability thresholds, binarizing the final probability map to delineate candidate influence areas; spatially overlaying each candidate influence area with independent ground verification data, calculating the spatial consistency between the two, and selecting the candidate threshold with the highest comprehensive consistency as the optimal probability threshold.

Images