Ecological pattern regulation and control method, device and system based on ecological hydrological collaboration

Abstract

The invention discloses an ecological pattern regulation and control method, device and system based on ecological hydrological coordination, and belongs to the field of ecological pattern regulation and control. After a remote sensing image and a visible light image of a preset area are obtained, ecological and hydrological indexes can be determined based on the remote sensing image and the visible light image, modeling is performed to obtain a target machine learning model, then boundary benefits of each feature variable in the target learning model are obtained, a feature curve is obtained through fitting, and a critical point is obtained through calculation; according to the critical point and the characteristic curve, the effect type of the target characteristic variable on the ecological pattern in the preset area can be determined in combination with the actual characteristic value, and the characteristic value of the corresponding target characteristic variable can be adjusted based on the effect type. According to the scheme, the synergistic effect of hydrology and ecological processes in an ecological system can be comprehensively considered, the key threshold value influencing the ecological pattern change is dynamically recognized based on the multi-source data and machine learning technology, and accurate regulation and control of the regional ecological pattern are ensured.

Description

Technical Field

[0001] The present application relates to the technical field of ecological pattern regulation, and in particular, to an ecological pattern regulation method, device and system based on ecological and hydrological synergy. Background Art

[0002] With the intensification of global climate change and human activities, terrestrial ecosystems are facing serious environmental pressures, including drought, precipitation changes, and frequent extreme weather events. These changes have a profound impact on the stability and sustainability of regional ecological patterns (such as lake water area, vegetation area, etc.). Especially in arid and semi-arid areas, the complex coupling of hydrological processes (such as precipitation, evaporation, etc.) and ecological processes (such as net primary productivity, normalized difference vegetation index, etc.) is crucial to the regulatory mechanism of regional ecological patterns. Therefore, studying the synergistic relationship between ecological and hydrological processes and determining the critical thresholds of their impact on ecological patterns are of great significance for understanding the dynamic changes of terrestrial ecosystems, predicting environmental responses, and formulating scientific management strategies.

[0003] At present, traditional ecosystem regulation strategies mostly rely on single hydrological or ecological indicators, while ignoring the complex interactions between the two, making it difficult to reveal the complex feedback mechanisms and critical thresholds of the system. In recent years, the development of ecohydrology has provided a new theoretical framework for exploring the coupled regulation of ecological and hydrological processes. However, the current identification methods for ecohydrological synergy thresholds often rely on empirical models or statistical analysis of historical data, which have large uncertainties and are difficult to adapt to the dynamic changes in different ecosystem types and climate backgrounds. In different ecosystem types, the interactions between hydrological and ecological processes show significant differences, which makes it difficult to universalize a unified threshold identification method, thus affecting the regulation of regional ecological patterns. Summary of the invention

[0004] In order to overcome the shortcomings of the prior art, the present application provides an ecological pattern regulation method, device and system based on eco-hydrological synergy to solve the problem that the current identification method for eco-hydrological synergy thresholds often relies on empirical models or statistical analysis of historical data, has large uncertainties, and is difficult to adapt to the dynamic changes in different ecosystem types and climate backgrounds. In different ecosystem types, the interaction between hydrological processes and ecological processes shows significant differences, which makes it difficult to universalize a unified threshold identification method, thus affecting the regulation of regional ecological patterns.

[0005] The technical solution adopted by this application to solve its technical problem is:

[0006] First, a method for regulating ecological pattern based on ecological and hydrological synergy is provided, including:

[0007] Obtain remote sensing images and visible light images of the preset area;

[0008] Acquire ecological data and hydrological data of the preset area based on the remote sensing image and the visible light image, and obtain ecological and hydrological indicators based on the ecological data and hydrological data;

[0009] Modeling the relationship between the target variable and the characteristic variable to obtain a target machine learning model; wherein the target variable is the ecological pattern, and the characteristic variable is the ecological and hydrological indicators;

[0010] Obtaining the boundary benefit of each characteristic variable in the target machine learning model, wherein the boundary benefit is used to characterize the contribution of the characteristic variable to the ecological pattern;

[0011] The characteristic curve is obtained by fitting, and the critical point when the boundary benefit of each target characteristic variable is 0 is calculated, wherein the ordinate of the characteristic curve is the boundary benefit, and the abscissa is the characteristic value of the target characteristic variable;

[0012] Determining the effect type of the target characteristic variable of the preset area on the ecological pattern based on the actual characteristic value, the critical point and the characteristic curve, wherein the effect type includes a positive effect and a negative effect;

[0013] The characteristic value of the corresponding target characteristic variable is adjusted based on the effect type.

[0014] Furthermore, the modeling of the relationship between the target variable and the feature variable to obtain the target machine learning model includes:

[0015] The relationship between the target variable and the feature variable is modeled using a first algorithm and a second algorithm respectively to obtain a first initial machine learning model and a second initial machine learning model, wherein the first algorithm and the second algorithm are different machine learning algorithms;

[0016] Divide all feature variables into N feature subsets, and input the N feature subsets into the first initial machine learning model and the second initial machine learning model respectively to obtain performance indicators corresponding to the feature subsets, wherein each feature subset includes some feature variables, and the performance indicators include the correlation coefficient, root mean square error, and mean absolute percentage error between the true value of the feature variable and the model predicted value;

[0017] Based on the performance indicator, a target feature subset and a target algorithm are determined; and the relationship between the target variable and the feature variable is modeled using the target algorithm and the feature variables in the target feature subset to obtain a target machine learning model.

[0018] Furthermore, the target algorithm and the characteristic variables in the target characteristic subset are used to model the relationship between the target variable and the characteristic variables to obtain the target machine learning model, including:

[0019] Modeling the relationship between the target variable and the feature variable using the target algorithm and the feature variables in the target feature subset to obtain an intermediate machine learning model;

[0020] The target parameters of the intermediate machine learning model are adjusted to obtain a target machine learning model, wherein the target parameters include quantity estimation, learning rate, and maximum depth.

[0021] Furthermore, it also includes:

[0022] Verifying the causal relationship between the feature variables of the target feature subset and the target variable;

[0023] The characteristic variable having a causal relationship with the target variable is used as the target characteristic variable.

[0024] Furthermore, it also includes:

[0025] Obtain the goodness of fit and significance of the characteristic curve obtained by fitting;

[0026] When the goodness of fit and significance meet the preset requirements, it is determined as the target fitting curve; when the goodness of fit and significance do not meet the preset requirements, refitting is performed until the goodness of fit and significance of the characteristic curve obtained by fitting meet the preset requirements.

[0027] Furthermore, the obtaining of ecological data and hydrological data of the preset area based on the remote sensing image and the visible light image includes:

[0028] Preprocessing the remote sensing image to obtain a normalized vegetation index of the preset area, wherein the preprocessing includes clipping, radiation calibration, and atmospheric correction;

[0029] Determining the type of each sub-area in the preset area based on the normalized difference vegetation index;

[0030] The types of sub-areas include built-up areas, water bodies and vegetation.

[0031] Furthermore, the determining the type of each sub-area in the preset area based on the normalized difference vegetation index includes:

[0032] Determine the first sub-region where the normalized difference vegetation index is less than a preset value as a building site or a water body;

[0033] The support vector machine classification algorithm is used to identify the building land in the first sub-area by extracting training samples, and the improved normalized difference water index in the first sub-area throughout the year is calculated to extract the water body.

[0034] Furthermore, the obtaining of ecological data and hydrological data of the preset area based on the remote sensing image and the visible light image includes:

[0035] Determine the second sub-region where the normalized vegetation index is greater than or equal to a preset value as vegetation;

[0036] Performing linear interpolation on the data set of the second sub-area;

[0037] Using a first preset function to process the upward trend of the interpolated data, and using a second preset function to process the downward trend of the interpolated data, so as to obtain a growing season data set that conforms to the development process of the vegetation growing season;

[0038] The type of each vegetation in the second sub-area is determined based on the temporal changes of the growing season dataset.

[0039] In the second aspect, an ecological pattern control device based on ecological and hydrological synergy is provided, comprising:

[0040] An image acquisition module is used to acquire remote sensing images and visible light images of a preset area;

[0041] An index acquisition module, used to acquire ecological data and hydrological data of the preset area based on the remote sensing image and the visible light image, and obtain ecological and hydrological indicators based on the ecological data and hydrological data;

[0042] A model building module, used to model the relationship between the target variable and the characteristic variable to obtain a target machine learning model; wherein the target variable is the ecological pattern, and the characteristic variable is the ecological and hydrological indicators;

[0043] A benefit acquisition module, used to acquire the boundary benefit of each characteristic variable in the target machine learning model, wherein the boundary benefit is used to characterize the contribution of the characteristic variable to the ecological pattern;

[0044] A critical determination module is used to fit a characteristic curve and calculate the critical point when the boundary benefit of each target characteristic variable is 0, wherein the ordinate of the characteristic curve is the boundary benefit and the abscissa is the characteristic value of the target characteristic variable;

[0045] An effect determination module, used to determine the effect type of the target characteristic variable of the preset area on the ecological pattern based on the actual characteristic value, the critical point and the characteristic curve, wherein the effect type includes a positive effect and a negative effect;

[0046] A feature adjustment module is used to adjust the feature value of the corresponding target feature variable based on the effect type.

[0047] The third aspect is to provide an ecological pattern control system based on ecological and hydrological synergy, including:

[0048] at least one processor and at least one memory;

[0049] The memory stores executable instructions of the processor;

[0050] The processor is configured to execute the above-mentioned ecological pattern regulation method based on ecological and hydrological synergy.

[0051] Beneficial effects:

[0052] The technical solution of the present application provides an ecological pattern regulation method, device and system based on ecological and hydrological synergy. After obtaining remote sensing images and visible light images of a preset area, it is possible to determine ecological and hydrological indicators based on remote sensing images and visible light images, and to model the target machine learning model, and then obtain the boundary benefits of each characteristic variable in the target learning model, fit the characteristic curve, and calculate the critical point. According to the critical point and the characteristic curve and combined with the actual characteristic value, the effect type of the target characteristic variable on the ecological pattern in the preset area can be determined, and the characteristic value of the corresponding target characteristic variable can be adjusted based on the effect type. The present application solution can comprehensively consider the synergistic effect of hydrology and ecological processes in the ecosystem, and based on multi-source data and machine learning technology, dynamically identify the key thresholds that affect changes in the ecological pattern, and ensure accurate regulation of the regional ecological pattern. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0054] Figure 1 This is a flow chart of an ecological pattern regulation method based on ecological and hydrological synergy provided in an embodiment of the present application;

[0055] Figure 2 is a flow chart of a method for acquiring a growing season data set provided in an embodiment of the present application;

[0056] Figure 3 It is a flow chart of a construction land identification method provided by an embodiment of the present application;

[0057] Figure 4This is a flow chart of a water extraction method provided in an embodiment of the present application;

[0058] Figure 5 This is a schematic diagram of the structure of an ecological pattern control device based on ecological and hydrological synergy provided in an embodiment of the present application;

[0059] Figure 6 It is a schematic diagram of the structure of an ecological pattern regulation system based on ecological and hydrological synergy provided in an embodiment of the present application. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical scheme and advantages of the present application clearer, the technical scheme of the present application is described in detail below in conjunction with the accompanying drawings and embodiments. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other implementation methods obtained by ordinary technicians in the field without making creative work belong to the scope of protection of the present application.

[0061] Reference Figure 1 The present application embodiment provides an ecological pattern regulation method based on ecological and hydrological synergy, including:

[0062] S11: Acquire remote sensing images and visible light images of a preset area; wherein the preset area is an area to be regulated, illustratively, the remote sensing image is a LANDSAT image, and the visible light image is acquired by a drone.

[0063] S12: Based on the remote sensing image and the visible light image, the ecological data and hydrological data of the preset area are obtained, and the ecological and hydrological indicators are obtained based on the ecological data and hydrological data; the ecological data include but are not limited to: diversity index, uniformity index, spread index, landscape fragmentation index, net primary productivity, vegetation coverage, normalized vegetation index, human disturbance, which are set according to actual needs; hydrological data include precipitation, potential evaporation, groundwater depth, water consumption, which are set according to actual needs; due to different situations in different regions, the indicators may also be different. It is determined specifically according to the sub-region type within the preset region. Specifically as follows:

[0064] The remote sensing image is preprocessed to obtain a normalized vegetation index of the preset area, wherein the preprocessing includes clipping, radiation calibration, and atmospheric correction; based on the normalized vegetation index, the type of each sub-area in the preset area is determined; the type of the sub-area includes building land, water bodies, and vegetation.

[0065] Determine the first sub-region where the normalized difference vegetation index is less than a preset value as a building site or a water body;

[0066] The support vector machine classification algorithm is used to identify the building land in the first sub-area by extracting training samples, and the improved normalized difference water index in the first sub-area throughout the year is calculated to extract the water body.

[0067] In addition, since there are many types of vegetation and different vegetation have different impacts on the ecological pattern of the region, it is necessary to determine the specific vegetation type.

[0068] The second sub-region where the normalized vegetation index is greater than or equal to a preset value is determined as vegetation; the data set of the second sub-region is linearly interpolated; the rising trend of the interpolated data is processed using a first preset function, and the falling trend of the interpolated data is processed using a second preset function to obtain a growing season data set that conforms to the development process of the vegetation growing season; the type of each vegetation in the second sub-region is determined based on the temporal changes of the growing season data set.

[0069] S13: Modeling the relationship between the target variable and the characteristic variable to obtain a target machine learning model; wherein the target variable is the ecological pattern, and the characteristic variable is the ecological and hydrological indicators;

[0070] As a preferred implementation of the present application, the relationship between the target variable and the feature variable is modeled to obtain a target machine learning model, including:

[0071] The first algorithm and the second algorithm are respectively used to model the relationship between the target variable and the feature variable to obtain the first initial machine learning model and the second initial machine learning model, wherein the first algorithm and the second algorithm are different machine learning algorithms; all feature variables are divided into N feature subsets, and the N feature subsets are respectively input into the first initial machine learning model and the second initial machine learning model to obtain the performance indicators corresponding to the feature subsets, wherein each feature subset includes some feature variables, and the performance indicators include the correlation coefficient, root mean square error and mean absolute percentage error between the true value of the feature variable and the model prediction value; wherein any two feature subsets are not exactly the same, but the same feature variables may exist. In addition, the number of feature variables in any two feature subsets may be different. Based on the performance indicators, the target feature subset and the target algorithm are determined; the target algorithm and the feature variables in the target feature subset are used to model the relationship between the target variable and the feature variable to obtain the target machine learning model.

[0072] The target algorithm and the characteristic variables in the target characteristic subset are used to model the relationship between the target variable and the characteristic variables to obtain a target machine learning model, including:

[0073] The target algorithm and the feature variables in the target feature subset are used to model the relationship between the target variable and the feature variable to obtain an intermediate machine learning model; the target parameters of the intermediate machine learning model are adjusted to obtain a target machine learning model, wherein the target parameters include quantity estimation, learning rate and maximum depth.

[0074] In addition, the causal relationship between the characteristic variables of the target characteristic subset and the target variable is verified; the characteristic variables having a causal relationship with the target variable are used as the target characteristic variables. In this way, only the marginal benefits of the target characteristic variables can be calculated in the subsequent marginal benefits calculation.

[0075] It should be noted that this application uses two algorithms for modeling, and the final result is more accurate. In practice, only one algorithm can be used for modeling.

[0076] S14: Obtain the marginal benefit of each characteristic variable in the target machine learning model, wherein the marginal benefit is used to characterize the contribution of the characteristic variable to the ecological pattern. Exemplarily, the SHAP (SHapley Additive ex Planations) method is used to obtain the SHAP values ​​of all characteristic variables to quantify the importance of the characteristic variables.

[0077] S15: fitting to obtain a characteristic curve, and calculating the critical point when the boundary benefit of each target characteristic variable is 0, wherein the ordinate of the characteristic curve is the boundary benefit, and the abscissa is the characteristic value of the target characteristic variable;

[0078] Also includes:

[0079] Obtain the goodness of fit and significance of the characteristic curve obtained by fitting;

[0080] When the goodness of fit and significance meet the preset requirements, it is determined as the target fitting curve; when the goodness of fit and significance do not meet the preset requirements, refitting is performed until the goodness of fit and significance of the characteristic curve obtained by fitting meet the preset requirements.

[0081] S16: Determine the effect type of the target characteristic variable of the preset area on the ecological pattern based on the actual characteristic value, the critical point and the characteristic curve, wherein the effect type includes a positive effect and a negative effect;

[0082] For example, when the actual characteristic value is greater than the critical point, it may mean that the ecological pressure is too great or the resources are over-exploited, which is a positive effect. When it is less than the threshold: If the indicator is lower than the threshold, it indicates that the ecosystem may be in a degraded state and compensatory measures need to be taken, such as increasing soil and water conservation, improving soil quality, and enhancing biodiversity.

[0083] S17: Adjust the characteristic value of the corresponding target characteristic variable based on the effect type. That is, the theoretical characteristic value of the target characteristic variable can be output and displayed based on the effect type, so that relevant personnel can take corresponding intervention measures for the preset area after obtaining the theoretical characteristic value. It is understandable that there are too many ecological and hydrological indicators, and different indicators have different effects on the ecological pattern. Intervention measures need to be combined with specific variables and regional characteristics.

[0084] The ecological pattern regulation method based on eco-hydrological synergy provided in the embodiment of the present application can determine the ecological and hydrological indicators based on the remote sensing images and visible light images after obtaining the remote sensing images and visible light images of the preset area, and model the target machine learning model, and then obtain the boundary benefits of each characteristic variable in the target learning model, fit the characteristic curve, calculate the critical point, and determine the effect type of the target characteristic variable on the ecological pattern in the preset area according to the critical point and the characteristic curve in combination with the actual characteristic value, and adjust the characteristic value of the corresponding target characteristic variable based on the effect type. The scheme of the present application can comprehensively consider the synergy of hydrology and ecological processes in the ecosystem, and dynamically identify the key thresholds that affect the changes in the ecological pattern based on multi-source data and machine learning technology to ensure accurate regulation of the regional ecological pattern. In addition, compared with the traditional regulation that requires manual calculation and determination of thresholds, the present application can automatically determine the critical point based on remote sensing images and visible light images in combination with machine learning models, which greatly reduces the intensity of manual labor.

[0085] In order to more clearly illustrate the present application scheme, a specific scheme is provided below, comprising the following steps:

[0086] Step 1: Data collection: Collect basic information of the study area, including geographic coordinates and land types; use remote sensing technology to obtain LANDSAT images and drone visible light images; obtain ecological data (diversity index, evenness index, spread index, landscape fragmentation index, net primary productivity, vegetation cover, normalized difference vegetation index, human disturbance, etc.) and hydrological data (precipitation, potential evaporation, groundwater depth, water consumption, etc.).

[0087] Step 2: Land use data preprocessing:

[0088] ① Call the software ENVI5.3 to preprocess the acquired original remote sensing images in the software ENVI5.3, including cropping, radiation calibration, atmospheric correction and other steps.

[0089] ② For the obtained NDVI (normalized difference vegetation index) data, first remove the area with NDVI < 0.05 (the area with NDVI < 0.05 is water body and building land, and the classification accuracy of vegetation community is improved after removal), then perform linear interpolation on the obtained data set, call timesat3.3 software, and use the double logic function filtering method in timesat3.3 software to smooth the data set to make it conform to the development process of vegetation growing season, and finally obtain the growing season data set every 15 days. Among them, Figure 2 As shown in the figure, the double logistic function filtering method adjusts the smoothness of the time series through two logistic functions, one function is used to process the upward trend of the data, and the other function is used to process the downward trend of the data. This method can better capture the dynamic changes of the data, reduce the impact of noise and outliers, improve the accuracy of prediction, and is less likely to be disturbed by outliers / extreme values ​​than other filtering methods. In this study, the NDVI dataset with NDVI < 0.05 was removed and imported into timesat3.3 three times to make the vegetation present an irregular state of three peaks and troughs. The double logistic method was used to smooth the dataset to make it consistent with the development process of the vegetation growing season, so as to determine the vegetation growing season, greening period and yellowing period in the region, and the dataset was fitted to obtain the NDVI dataset of the study area in different growing seasons, which is the basic data required for the decision tree.

[0090] ③ Non-vegetation classification: Select the original remote sensing images in October after preprocessing and use the support vector machine classification algorithm in ENVI5.3 software to extract training samples to identify the building land in the study area, such as Figure 3 As shown, the geographical location of the building land is determined based on the collected drone images, and sufficient training samples are extracted. The ENVI5.3 software will extract similar samples in the remote sensing image based on the training samples. The extracted samples are the building land with high accuracy.

[0091] The GEE platform is used to calculate the annual MNDWI (modified normalized difference water index) data to extract water bodies. Figure 4 As shown, call the GEE platform and define a for loop. First, set the required time range, filter time, study area, and cloud cover, traverse the cloud mask function, extract the MNDWI function, take the median synthesis, and crop to the study area size. Secondly, display the true color synthesized study area image, and then display the water body extraction result. Finally, define a function to export to the cloud disk to display the water body extraction result task.

[0092] ④ Vegetation classification, based on the land use classification decision tree method, according to the NDVI time series changes of each vegetation type in Landsat8 (Landsat5, 9) images, by adjusting and optimizing the threshold, a decision tree is established to classify different vegetation types in the study area, and finally the vegetation decision tree recognition rules of long-term and spatial sequence Landsat images are established. The NDVI value reflected by each vegetation community according to the changes in seasonal time is different, which is the basis for classifying all vegetation. The adjustment and optimization of the threshold can be considered that the NDVI values ​​of different vegetation communities at the same time are different. By continuously adjusting the NDVI threshold, the accuracy of vegetation classification is improved, thereby achieving an optimization effect.

[0093] Step 3: Select machine learning for modeling: Use two machine learning algorithms, extreme gradient boosting (XGBoost) and random forest (RF), to model the relationship between the target variable (ecological pattern) and characteristic variables (ecological and hydrological indicators).

[0094] Step 4: Feature selection and model evaluation:

[0095] ① Recursive feature elimination (RFE) is used to identify the optimal feature subsets of XGBoost and RF models, and other feature variables are removed.

[0096] ② The indicators for evaluating the model are the correlation coefficient (R), root mean square error (RMSE) and mean absolute percentage error (MAPE) between the true value of the feature variable and the model prediction value. The 10-fold cross-validation method is used to test the model with the best feature subset, and the mean of the model's R, RMSE and MAPE on the test data set is used as the final model performance indicator. The model with the best performance indicator is retained.

[0097] Step 5: Parameter optimization: XGBoost and RF model parameter tuning includes: "n_estimators" (i.e., number of estimates), "learning_rate" (i.e., learning rate), and "max_depth" (i.e., maximum depth). Among them, "n_estimators" takes values ​​from 30 to 300, with a step size of 10; "learning_rate" takes a discrete value set of 0.01, 0.02, 0.05, 0.1, 0.12, 0.15, and 0.2; and "max_depth" takes an integer ranging from 3 to 8.

[0098] Step 6: Machine learning model interpretation: Use the SHAP (SHapley Additive exPlanations) method to obtain the SHAP values ​​of all feature variables, quantify the importance of feature variables, and calculate the boundary benefits of individual variables on the ecological pattern (boundary benefits: the importance contribution of a specific value of a single variable to the model output).

[0099] Step 7: Test the causal relationship between feature variables and target variables: EconML estimates personalized causal effects from observational or experimental data by embedding machine learning into an interpretable causal model. First, it is assumed that there is a causal relationship between each pair of selected feature variables and the target variable. After verification by EconML, a causal network is generated, a network framework is established, and the validity of each causal relationship is tested using the EconML model. In causal analysis, the average treatment effect (ATE) represents the average difference between the intervention group and the control group, which can identify whether a feature variable has a direct impact on the target variable.

[0100] Step 8: Critical point output:

[0101] ① The critical point is defined as the threshold of the impact of ecological and hydrological indicators on the ecological pattern, that is, the position where the SHAP value is 0, indicating the turning point of the positive or negative effect of the eco-hydrological element on the ecological pattern.

[0102] ② Select the characteristic variable that has a causal relationship with the target variable, use the SHAP value as the vertical axis (y) and the corresponding characteristic value as the horizontal axis (x), draw a graph, and perform curve fitting on their relationship.

[0103] Use goodness of fit (R 2 ) and significance (p value) are used as performance indicators to evaluate the fitting effect. If the performance indicators meet the requirements, it is determined as the characteristic curve. If not, it is refitted.

[0104] Step 9: Control measures:

[0105] ①Maintaining positive effects and blocking negative effects

[0106] When the characteristic value is greater than the threshold value (positive influence):

[0107] Technical goal: Maintain or strengthen positive effects and prevent excessive intervention leading to ecological redundancy.

[0108] Specific measures: Dynamically monitor positive effect indicators (such as biodiversity index and hydrological connectivity); if the indicators do not meet expectations, fine-tune threshold parameters; limit human activities that may weaken positive effects (such as controlling tourism development intensity and optimizing water resource allocation).

[0109] When the characteristic value is less than the threshold value (negative influence):

[0110] Technical goal: Block the spread of negative effects and repair damaged ecological functions.

[0111] Specific measures: Initiate ecological restoration projects (such as replanting vegetation and soil carbon sequestration); urgently shut down pollution sources or over-development projects (such as suspending mining and restricting agricultural water withdrawal).

[0112] ②Two-way dynamic control mechanism

[0113] Dynamically switch the control mode according to the deviation direction between the characteristic value and the threshold:

[0114] Positive deviation (eigenvalue > threshold): lock the current gain state and optimize resource space allocation, and optimize resource allocation through algorithms (such as allocating water resources preferentially to ecologically fragile areas).

[0115] Negative deviation (eigenvalue < threshold): triggering a multi-node linked ecological loss blocking mechanism, combined with IoT devices (such as smart sluice gates and drone monitoring) to implement rapid response.

[0116] ③Threshold elasticity correction strategy

[0117] To avoid rigid regulation due to fixed thresholds, the threshold range can be dynamically modified based on historical data and real-time feedback (e.g. using different thresholds for rainy season / dry season).

[0118] The present invention provides an ecological pattern regulation method based on ecological and hydrological synergy. The method can reveal the complex interaction of ecological and hydrological processes within the ecosystem by comprehensively analyzing ecological indicators and hydrological indicators, effectively overcoming the limitations of traditional single indicator analysis methods, and thus more accurately reflecting the real dynamic changes of the ecosystem.

[0119] The present invention uses machine learning and multi-source data fusion technology, combined with the synergistic change relationship between ecological and hydrological indicators, to identify the key thresholds that affect the ecological pattern. Compared with traditional empirical models, this method can dynamically adjust the recognition threshold, significantly improving the accuracy and applicability of recognition, especially for complex ecological environments such as arid and semi-arid areas.

[0120] By determining the critical threshold of eco-hydrological synergy, the present invention can help managers to timely predict and take scientific intervention measures when the ecosystem faces extreme climate events or environmental pressures to prevent ecosystem imbalance or degradation. In addition, the method is applicable to different types of terrestrial ecosystems, has strong universality and promotion value, and provides an important theoretical basis and decision-making support tool for ecological environmental protection, desertification prevention and control, and water resources management.

[0121] This paper provides a new research method for evaluating and predicting the response of ecosystems to future climate change by revealing the coordinated regulatory mechanism of ecological and hydrological processes on regional ecological patterns. This method not only helps to understand the long-term evolution trend of terrestrial ecosystems, but also provides a scientific basis for the formulation of regional ecological planning and sustainable development policies, effectively promoting the health and sustainable development of regional ecosystems.

[0122] Based on the same inventive concept, Figure 5As shown, the present application also provides an ecological pattern control device 50 based on ecological and hydrological synergy, comprising:

[0123] An image acquisition module 51 is used to acquire remote sensing images and visible light images of a preset area;

[0124] An index acquisition module 52, used to acquire ecological data and hydrological data of the preset area based on the remote sensing image and the visible light image, and obtain ecological and hydrological indicators based on the ecological data and hydrological data;

[0125] The obtaining of ecological data and hydrological data of the preset area based on the remote sensing image and the visible light image includes:

[0126] The remote sensing image is preprocessed to obtain a normalized vegetation index of the preset area, wherein the preprocessing includes clipping, radiation calibration, and atmospheric correction; based on the normalized vegetation index, the type of each sub-area in the preset area is determined; the type of the sub-area includes building land, water bodies, and vegetation.

[0127] The determining the type of each sub-area in the preset area based on the normalized difference vegetation index includes:

[0128] The first sub-region where the normalized vegetation index is less than a preset value is determined as a building land or a water body; a support vector machine classification algorithm is used to identify the building land in the first sub-region by extracting training samples, and an improved normalized difference water index in the first sub-region throughout the year is calculated to extract the water body.

[0129] The second sub-region where the normalized vegetation index is greater than or equal to a preset value is determined as vegetation; the data set of the second sub-region is linearly interpolated; the rising trend of the interpolated data is processed using a first preset function, and the falling trend of the interpolated data is processed using a second preset function to obtain a growing season data set that conforms to the development process of the vegetation growing season; the type of each vegetation in the second sub-region is determined based on the temporal changes of the growing season data set.

[0130] A model building module 53 is used to model the relationship between the target variable and the characteristic variable to obtain a target machine learning model; wherein the target variable is the ecological pattern, and the characteristic variable is the ecological and hydrological indicators;

[0131] The method of modeling the relationship between the target variable and the feature variable to obtain the target machine learning model includes:

[0132] A first algorithm and a second algorithm are respectively used to model the relationship between the target variable and the feature variable to obtain a first initial machine learning model and a second initial machine learning model, wherein the first algorithm and the second algorithm are different machine learning algorithms; all feature variables are divided into N feature subsets, and the N feature subsets are respectively input into the first initial machine learning model and the second initial machine learning model to obtain performance indicators corresponding to the feature subsets, wherein each feature subset includes some feature variables, and for any two of them, the performance indicators include the correlation coefficient, root mean square error and mean absolute percentage error between the true value of the feature variable and the model predicted value; a target feature subset and a target algorithm are determined based on the performance indicators; the target algorithm and the feature variables in the target feature subset are used to model the relationship between the target variable and the feature variable to obtain a target machine learning model.

[0133] Furthermore, the target algorithm and the characteristic variables in the target characteristic subset are used to model the relationship between the target variable and the characteristic variables to obtain a target machine learning model, including:

[0134] The target algorithm and the feature variables in the target feature subset are used to model the relationship between the target variable and the feature variable to obtain an intermediate machine learning model; the target parameters of the intermediate machine learning model are adjusted to obtain a target machine learning model, wherein the target parameters include quantity estimation, learning rate and maximum depth.

[0135] A benefit acquisition module 54 is used to acquire the boundary benefit of each characteristic variable in the target machine learning model, wherein the boundary benefit is used to characterize the contribution of the characteristic variable to the ecological pattern;

[0136] Also includes:

[0137] Verify the causal relationship between the characteristic variables of the target characteristic subset and the target variable; and use the characteristic variables having a causal relationship with the target variable as the target characteristic variables.

[0138] The critical determination module 55 is used to fit the characteristic curve and calculate the critical point when the boundary benefit of each target characteristic variable is 0, wherein the ordinate of the characteristic curve is the boundary benefit and the abscissa is the characteristic value of the target characteristic variable;

[0139] Also includes:

[0140] The goodness of fit and significance of the characteristic curve obtained by fitting are obtained; when the goodness of fit and significance meet the preset requirements, it is determined as the target fitting curve; when the goodness of fit and significance do not meet the preset requirements, refitting is performed until the goodness of fit and significance of the characteristic curve obtained by fitting meet the preset requirements.

[0141] An effect determination module 56, for determining the effect type of the target characteristic variable of the preset area on the ecological pattern based on the actual characteristic value, the critical point and the characteristic curve, wherein the effect type includes a positive effect and a negative effect;

[0142] The feature adjustment module 57 is used to adjust the feature value of the corresponding target feature variable based on the effect type.

[0143] The ecological pattern control device based on ecological and hydrological synergy provided in the embodiment of the present application can determine ecological and hydrological indicators based on remote sensing images and visible light images after obtaining remote sensing images and visible light images of a preset area, and perform modeling to obtain a target machine learning model, and then obtain the boundary benefits of each characteristic variable in the target learning model, fit the characteristic curve, calculate the critical point, and determine the effect type of the target characteristic variable on the ecological pattern in the preset area based on the critical point and the characteristic curve in combination with the actual characteristic value, and adjust the characteristic value of the corresponding target characteristic variable based on the effect type. The present application scheme can comprehensively consider the synergistic effect of hydrology and ecological processes in the ecosystem, and based on multi-source data and machine learning technology, dynamically identify the key thresholds that affect changes in the ecological pattern, and ensure accurate regulation of the regional ecological pattern.

[0144] Based on the same inventive concept, Figure 6 As shown, the present application also provides an ecological pattern control system 60 based on ecological and hydrological synergy, including:

[0145] at least one processor 61 and at least one memory 62;

[0146] The memory stores executable instructions of the processor;

[0147] The processor is configured to execute the ecological pattern regulation method based on ecological and hydrological synergy provided in the above embodiment.

[0148] The ecological pattern control system based on ecological and hydrological synergy provided by the embodiment of the present application stores the executable instructions of the processor in the memory. When the executable instructions are executed, the processor can determine the ecological and hydrological indicators based on the remote sensing images and visible light images after acquiring the remote sensing images and visible light images of the preset area, and perform modeling to obtain the target machine learning model, and then obtain the boundary benefits of each characteristic variable in the target learning model, fit the characteristic curve, calculate the critical point, and determine the effect type of the target characteristic variable on the ecological pattern in the preset area according to the critical point and the characteristic curve in combination with the actual characteristic value, and adjust the characteristic value of the corresponding target characteristic variable based on the effect type. The scheme of the present application can comprehensively consider the synergistic effect of hydrology and ecological processes in the ecosystem, and dynamically identify the key thresholds that affect the changes in the ecological pattern based on multi-source data and machine learning technology to ensure accurate regulation of the regional ecological pattern.

[0149] It should be noted that, in the description of this application, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, the meaning of "plurality" refers to at least two.

[0150] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

Claims

1. A method for regulating ecological pattern based on ecological and hydrological synergy, characterized in that: include: Obtain remote sensing images and visible light images of the preset area; Acquire ecological data and hydrological data of the preset area based on the remote sensing image and the visible light image, and obtain ecological and hydrological indicators based on the ecological data and hydrological data; Modeling the relationship between the target variable and the characteristic variable to obtain a target machine learning model; wherein the target variable is the ecological pattern, and the characteristic variable is the ecological and hydrological indicators; Obtaining the boundary benefit of each characteristic variable in the target machine learning model, wherein the boundary benefit is used to characterize the contribution of the characteristic variable to the ecological pattern; The characteristic curve is obtained by fitting, and the critical point when the boundary benefit of each target characteristic variable is 0 is calculated, wherein the ordinate of the characteristic curve is the boundary benefit, and the abscissa is the characteristic value of the target characteristic variable; Determining the effect type of the target characteristic variable of the preset area on the ecological pattern based on the actual characteristic value, the critical point and the characteristic curve, wherein the effect type includes a positive effect and a negative effect; The characteristic value of the corresponding target characteristic variable is adjusted based on the effect type.

2. The method according to claim 1, characterized in that: The method of modeling the relationship between the target variable and the feature variable to obtain a target machine learning model includes: The relationship between the target variable and the feature variable is modeled using a first algorithm and a second algorithm respectively to obtain a first initial machine learning model and a second initial machine learning model, wherein the first algorithm and the second algorithm are different machine learning algorithms; Divide all feature variables into N feature subsets, and input the N feature subsets into the first initial machine learning model and the second initial machine learning model respectively to obtain performance indicators corresponding to the feature subsets, wherein each feature subset includes some feature variables, and the performance indicators include the correlation coefficient, root mean square error, and mean absolute percentage error between the true value of the feature variable and the model predicted value; Based on the performance indicator, a target feature subset and a target algorithm are determined; and the relationship between the target variable and the feature variable is modeled using the target algorithm and the feature variables in the target feature subset to obtain a target machine learning model.

3. The method according to claim 2, characterized in that: The target algorithm and the characteristic variables in the target characteristic subset are used to model the relationship between the target variable and the characteristic variables to obtain the target machine learning model, including: Modeling the relationship between the target variable and the feature variable using the target algorithm and the feature variables in the target feature subset to obtain an intermediate machine learning model; The target parameters of the intermediate machine learning model are adjusted to obtain a target machine learning model, wherein the target parameters include quantity estimation, learning rate, and maximum depth.

4. The method according to claim 2, characterized in that: Also includes: Verifying the causal relationship between the feature variables of the target feature subset and the target variable; The characteristic variable having a causal relationship with the target variable is used as the target characteristic variable.

5. The method according to claim 1, characterized in that Also includes: Obtain the goodness of fit and significance of the characteristic curve obtained by fitting; When the goodness of fit and significance meet the preset requirements, it is determined as the target fitting curve; when the goodness of fit and significance do not meet the preset requirements, refitting is performed until the goodness of fit and significance of the characteristic curve obtained by fitting meet the preset requirements.

6. The method according to claim 1, characterized in that: The obtaining of ecological data and hydrological data of the preset area based on the remote sensing image and the visible light image includes: Preprocessing the remote sensing image to obtain a normalized vegetation index of the preset area, wherein the preprocessing includes clipping, radiation calibration, and atmospheric correction; Determining the type of each sub-area in the preset area based on the normalized difference vegetation index; The types of sub-areas include built-up areas, water bodies and vegetation.

7. The method according to claim 6, characterized in that: The determining the type of each sub-area in the preset area based on the normalized difference vegetation index includes: Determine the first sub-region where the normalized difference vegetation index is less than a preset value as a building site or a water body; The support vector machine classification algorithm is used to identify the building land in the first sub-area by extracting training samples, and the improved normalized difference water index in the first sub-area throughout the year is calculated to extract the water body.

8. The method according to claim 6, characterized in that: The obtaining of ecological data and hydrological data of the preset area based on the remote sensing image and the visible light image includes: Determine the second sub-region where the normalized vegetation index is greater than or equal to a preset value as vegetation; Performing linear interpolation on the data set of the second sub-area; Using a first preset function to process the upward trend of the interpolated data, and using a second preset function to process the downward trend of the interpolated data, so as to obtain a growing season data set that conforms to the development process of the vegetation growing season; The type of each vegetation in the second sub-area is determined based on the temporal changes of the growing season dataset.

9. An ecological pattern control device based on ecological and hydrological synergy, characterized in that: include: An image acquisition module is used to acquire remote sensing images and visible light images of a preset area; An index acquisition module, used to acquire ecological data and hydrological data of the preset area based on the remote sensing image and the visible light image, and obtain ecological and hydrological indicators based on the ecological data and hydrological data; A model building module, used to model the relationship between the target variable and the characteristic variable to obtain a target machine learning model; wherein the target variable is the ecological pattern, and the characteristic variable is the ecological and hydrological indicators; A benefit acquisition module, used to acquire the boundary benefit of each characteristic variable in the target machine learning model, wherein the boundary benefit is used to characterize the contribution of the characteristic variable to the ecological pattern; A critical determination module is used to fit a characteristic curve and calculate the critical point when the boundary benefit of each target characteristic variable is 0, wherein the ordinate of the characteristic curve is the boundary benefit and the abscissa is the characteristic value of the target characteristic variable; An effect determination module, used to determine the effect type of the target characteristic variable of the preset area on the ecological pattern based on the actual characteristic value, the critical point and the characteristic curve, wherein the effect type includes a positive effect and a negative effect; A feature adjustment module is used to adjust the feature value of the corresponding target feature variable based on the effect type.

10. An ecological pattern control system based on ecological and hydrological synergy, characterized in that: include: at least one processor and at least one memory; The memory stores executable instructions of the processor; The processor is configured to execute the method according to any one of claims 1 to 8.

Images