Method and system for detecting the effect of enhancing the water-holding function of ecological restoration

Abstract

The present application relates to the technical field of environmental monitoring and soil analysis, and specifically relates to a detection method and system for improving the water holding function of ecological restoration, which aims to accurately measure the soil saturated hydraulic conductivity and effective water content at the regional scale, to adaptively estimate the observation error of each remote sensing pixel by collecting multi-temporal and spatial remote sensing data and ground plot monitoring data in the ecological restoration area and the control area, to obtain preliminary simulation results by running the water holding function simulation module, to dynamically correct the model state variables by using the ensemble Kalman filter algorithm to fuse the remote sensing soil moisture data at the monthly or quarterly cycle, to optimize the key parameter field of the model based on the corrected state sequence, and finally to generate the spatialized quantitative evaluation results and visual report of the water holding function improvement amount by comparative analysis. The present application realizes the indirect, large-scale and accurate measurement of the key hydraulic parameters of soil, and is mainly used for the accurate and spatialized detection and evaluation of the water holding function improvement effect after the implementation of the ecological restoration project.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of environmental monitoring and soil analysis, in particular, the present application relates to a detection method and system for improving the water holding function of ecological restoration, especially a detection method and system for quantitatively evaluating the effect of ecological restoration project on improving the water holding function of soil. BACKGROUND

[0002] Developing ecological restoration (such as planting trees and grass) is a key means to improve the water conservation capacity of the region. However, how to scientifically, accurately and widely detect and evaluate the effect of water holding function after restoration has always been a difficulty in the industry. At present, the following problems exist:

[0003] 1. Relies on traditional sample survey, which is difficult to represent the whole region: Although the method of measuring soil moisture by laying sample plots has high precision, it is high in cost and low in efficiency, and the "point" data is difficult to reflect the overall effect of the "surface" of the region, which cannot meet the needs of large-scale engineering effect monitoring.

[0004] 2. Relies on pure remote sensing or model simulation, and the precision and reliability are insufficient: pure remote sensing index (such as EVI, NDVI) cannot directly quantify the water holding capacity of deep soil, and is easily disturbed by weather, season and other factors. When using general ecological model (such as InVEST) for evaluation, the internal key parameters (such as soil permeability and vegetation water consumption coefficient) usually use regional default values, which cannot accurately capture the subtle but key ecological changes caused by soil structure improvement and vegetation community changes after the implementation of restoration project, resulting in significant deviation of the evaluation results.

[0005] 3. Difficulty in fusing "point" and "surface" data: the existing technology lacks effective means to deeply fuse high-precision ground sample "point" data and regional coverage remote sensing "surface" data, resulting in that the data advantages cannot be complementary, and the evaluation results are either lack of precision or representativeness.

[0006] Therefore, there is an urgent need for a new detection technology and system that can fuse multi-source data and accurately reflect the actual effect of ecological restoration project. SUMMARY

[0007] An object of the present application is to solve at least the above problems and provide at least the advantages to be explained later.

[0008] In view of the deficiencies of the existing detection means, the present application aims to provide a detection method and system for ecological restoration water-holding function improvement effect, and the core of the present application is to solve the problem that the key hydraulic parameters (such as saturated hydraulic conductivity) and state variables (such as effective water content) of soil are difficult to be directly and accurately obtained at a large scale, and an advanced soil investigation method for dynamically and spatially explicitly analyzing the soil water-holding characteristics is provided by fusing remote sensing observation and ground investigation.

[0009] In order to achieve these objects and other advantages according to the present application, a detection method for ecological restoration water-holding function improvement effect is provided, comprising the following steps:

[0010] S1, collecting multi-temporal and spatial remote sensing data and ground plot monitoring data of the ecological restoration area and the control area, evaluating the uniformity of the ground surface characteristics in each remote sensing pixel based on the ground plot monitoring data, and adaptively assigning an observation error value to each pixel;

[0011] S2, running a water-holding function simulation module, inputting the multi-temporal and spatial remote sensing data as driving data, and obtaining a preliminary simulation result of the water source conservation amount of the ecological restoration area;

[0012] S3, taking month or quarter as a cycle, taking the soil moisture data inversed by remote sensing as an observation signal, using the ensemble Kalman filtering algorithm, and combining the observation error value to dynamically correct the soil effective water content state variable in the water-holding function simulation module, and obtaining a corrected state sequence;

[0013] S4, after completing the state dynamic correction of a complete growing season or year, adjusting the soil saturated hydraulic conductivity parameter field in the water-holding function simulation module based on the corrected state sequence through a parameter adjustment algorithm, so that the simulation behavior of the water-holding function simulation module matches the corrected state sequence;

[0014] S5, feeding back the adjusted soil saturated hydraulic conductivity parameter field to the water-holding function simulation module, re-running the water-holding function simulation module, and obtaining a spatial distribution map of the water source conservation amount;

[0015] S6, comparing the water source conservation amount results before and after restoration or between the restoration area and the control area, calculating the water-holding function improvement amount, including the spatial distribution, total amount and average improvement intensity of the improvement, and automatically generating a visual detection report.

[0016] Preferably, the step S1 of evaluating the uniformity of the ground surface characteristics in each remote sensing pixel and adaptively assigning an observation error value to each pixel specifically comprises: calculating a ground surface uniformity index corresponding to each remote sensing pixel and a standard deviation of the ground sample monitoring data in each remote sensing pixel based on the ground sample monitoring data, and calculating the observation error value for each remote sensing pixel according to the ground surface uniformity index and the standard deviation of the ground sample monitoring data, wherein the observation error value is positively correlated with the standard deviation of the ground sample monitoring data and negatively correlated with the ground surface uniformity index.

[0017] Preferably, the step S3 of dynamically correcting the soil effective water content state variable in the water holding function simulation module by using the ensemble Kalman filter algorithm and combining the observation error value specifically comprises:

[0018] The observation error value assigned to each remote sensing pixel in the step S1 is taken as the input of the observation error covariance matrix in the ensemble Kalman filter algorithm;

[0019] In each assimilation cycle of a month or a quarter, the following operations are performed:

[0020] The soil effective water content simulated by the water holding function simulation module is taken as the background field;

[0021] The soil moisture data retrieved by remote sensing is taken as the observation field;

[0022] The background field and the observation field are fused and calculated by the ensemble Kalman filter algorithm to generate an analysis field;

[0023] The analysis field is used to replace the soil effective water content state variable in the water holding function simulation module to complete the dynamic correction of the cycle;

[0024] The dynamic correction of all cycles is sequentially performed to form a corrected state sequence.

[0025] Preferably, the step S4 of adjusting the soil saturated hydraulic conductivity parameter field in the water holding function simulation module by the parameter adjustment algorithm specifically comprises:

[0026] The corrected state sequence is taken as the target true value of parameter optimization;

[0027] A target function is constructed, which is used to measure the difference between the state variable sequence simulated by the water holding function simulation module and the corrected state sequence;

[0028] The SCE-UA optimization algorithm is adopted to retrieve and determine the optimal value of the soil saturated hydraulic conductivity parameter field in the water holding function simulation module by minimizing the target function.

[0029] Preferably, the step S5 specifically comprises:

[0030] updating the soil saturated hydraulic conductivity parameter field in the water retention function simulation module with the adjusted soil saturated hydraulic conductivity parameter field in step S4;

[0031] re-running the water retention function simulation module with the multi-temporal and multi-spatial remote sensing data in step S1 as driving data;

[0032] obtaining a spatial distribution map of the water conservation capacity of the ecological restoration area, which is the output of the spatial representativeness of the ground plot monitoring data, the state dynamic correction, and the parameter feedback optimization results.

[0033] Preferably, step S6 specifically comprises:

[0034] Based on the spatial distribution map of the water conservation capacity obtained in step S5, the water conservation capacity result of the ecological restoration area after restoration is spatially overlaid with the water conservation capacity result of the same area before restoration or with the water conservation capacity result of the control area without restoration;

[0035] Through spatial overlay analysis, the water retention function improvement amount is calculated pixel by pixel to generate a spatial distribution map of the water retention function improvement amount;

[0036] Based on the spatial distribution map of the water retention function improvement amount, the total amount of water retention function improvement and the average improvement intensity of the entire ecological restoration area are statistically calculated;

[0037] The spatial distribution map of the water retention function improvement amount, the total amount of water retention function improvement, the average improvement intensity, and related spatial statistical information are automatically integrated to generate a visual inspection report containing charts and maps.

[0038] Preferably, in step S1, the following steps are further included:

[0039] Based on the digital elevation model data and land use type data of the ecological restoration area, the ground plot monitoring data is divided into different ecological hydrological response units;

[0040] For each ecological hydrological response unit, a scale conversion model is established respectively, and the scale conversion model is realized through a geographic weighted regression algorithm, wherein the regression independent variables include the vegetation index and the land surface temperature data of the remote sensing pixel corresponding to the ground plot monitoring data, and the regression dependent variable is the ground plot monitoring data;

[0041] Based on the scale conversion model, the point-distributed ground plot monitoring data is upscaled to planar data matching the remote sensing pixel, and an upscaled conversion uncertainty index of each remote sensing pixel is generated;

[0042] The upscaled planar data and the upscaled conversion uncertainty index are used to evaluate the homogeneity of the ground surface characteristics in each remote sensing pixel.

[0043] Preferably, between steps S3 and S4, the following step is also included:

[0044] Based on the corrected state sequence, the simulation performance index of the water holding function simulation module at different time scales is calculated. According to the simulation performance index, the convergence threshold and iteration step size of the parameter adjustment algorithm are dynamically adjusted. The simulation performance index includes the consistency between short-term simulation deviation and long-term simulation trend.

[0045] Preferably, step S1 further includes the following steps: identifying ecological restoration measures implemented at different times within the ecological restoration area and determining the implementation time nodes of each measure; dividing the entire monitoring period into multiple consecutive restoration stages based on the implementation time nodes; and constructing a stage-specific initial parameter set for the water-holding function simulation module for each restoration stage. When running the water-holding function simulation module, the corresponding initial parameter set is called according to different restoration stages.

[0046] This invention provides a detection system for improving the water retention function of ecological restoration, used to perform the aforementioned detection method, comprising:

[0047] The multi-source data acquisition module is used to acquire multi-temporal and spatial remote sensing data of the ecological restoration area and the control area, as well as ground quadrat monitoring data;

[0048] The water retention function simulation module is used to obtain preliminary simulation results of the water conservation capacity of the ecological restoration area based on multi-temporal and spatial remote sensing data.

[0049] The intelligent data assimilation engine connects the multi-source data acquisition module and the water-holding function simulation module. The intelligent data assimilation engine includes:

[0050] An adaptive error estimation unit is used to evaluate the uniformity of surface characteristics in each remote sensing pixel based on ground quadrat monitoring data, and adaptively assign an observation error value to each pixel.

[0051] The state dynamic correction unit is used to dynamically correct the soil effective water content state variable in the water holding function simulation module by using remote sensing inverted soil moisture data as the observation signal on a monthly or quarterly basis, employing an ensemble Kalman filter algorithm and combining the observation error value, to obtain the corrected state sequence.

[0052] The parameter feedback optimization unit is used to adjust the soil saturated hydraulic conductivity parameter field in the water holding function simulation module based on the corrected state sequence after completing a full growing season or year of dynamic state correction.

[0053] The effect evaluation and output module connects the water-holding function simulation module and the intelligent data assimilation engine. It is used to feed back the adjusted soil saturated hydraulic conductivity parameter field to the water-holding function simulation module, rerun the water-holding function simulation module, obtain the spatial distribution map of water conservation, and calculate the improvement in water-holding function by comparing the water conservation results before and after the restoration or between the restoration area and the control area. This includes the spatial distribution, total amount and average improvement intensity of the improvement, and automatically generates a visual detection report.

[0054] The present invention has at least the following beneficial effects:

[0055] First, high detection accuracy: Through the dual guarantee of state correction and parameter optimization, the uncertainty of the model is significantly reduced, so that the detection results can truly reflect the subtle effects of the repair project.

[0056] Second, point-to-area integration, complementary advantages: It creatively solved the problem of integrating ground "point" data and remote sensing "area" data, ensuring regional representativeness while incorporating the high precision of ground true values.

[0057] Third, strong adaptability: The unique adaptive error estimation mechanism enables the system to maintain reliable detection accuracy even in complex repair areas with high environmental heterogeneity.

[0058] Fourth, automation and operationalization: The system has a clear process and can be automated, providing continuous, objective, and quantitative data support for the acceptance of ecological restoration projects, long-term effectiveness monitoring, and management decisions.

[0059] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0060] Figure 1 This is a schematic diagram of the detection process for one of the technical solutions of the present invention. Detailed Implementation

[0061] The present invention will now be described in further detail with reference to examples, so that those skilled in the art can implement it based on the description.

[0062] It should be noted that, unless otherwise specified, the experimental methods described in the following implementation plan are all conventional methods, and the reagents and materials described are all commercially available unless otherwise specified.

[0063] like Figure 1 As shown, this invention provides a method for detecting the effect of improving water retention function in ecological restoration, comprising the following steps:

[0064] S1. Multi-source Data Collection and Adaptive Observation Error Estimation. Collect multi-temporal and spatial remote sensing data and ground quadrat monitoring data for the ecological restoration area and control area. Based on the ground quadrat monitoring data, assess the homogeneity of surface characteristics within each remote sensing pixel and adaptively assign an observation error value to each pixel. This step establishes an adaptive observation error estimation mechanism by quantifying the spatial heterogeneity and measurement dispersion within pixels, providing accurate error information for subsequent data assimilation and overcoming the limitations of traditional methods that employ uniform error assumptions. The layout of ground quadrat monitoring data should follow the principle of spatial representativeness, and stratified random sampling methods can be used to ensure coverage of different land use types, topographic locations, and types of ecological restoration measures within the ecological restoration area. The quadrat layout density should match the scale of the remote sensing pixels; in principle, there should be at least one quadrat within each representative eco-hydrological response unit.

[0065] S2. Preliminary Simulation of Water Retention Function. The water retention function simulation module is run, inputting multi-temporal and spatial remote sensing data as driving data to obtain preliminary simulation results of the water conservation capacity of the ecological restoration area. This step establishes a benchmark simulation of water retention function based on physical mechanisms, providing an initial reference for subsequent state correction and parameter optimization.

[0066] S3. Dynamic State Correction. On a monthly or quarterly basis, remotely sensed soil moisture data (i.e., remotely sensed observations of surface soil effective water content) is used as the observation signal. An ensemble Kalman filter (EnKF) algorithm, combined with observation error values, is employed to dynamically correct the state variables of soil effective water content (i.e., root zone soil moisture simulated by the model) in the water-holding capacity simulation module, resulting in a corrected state sequence. This step continuously corrects the model state variables by periodically fusing model forecasts and remote sensing observations, ensuring consistency between the simulated state and actual observations, and improving short-term simulation accuracy. Remotely sensed soil moisture data can be obtained from C-band SAR data from the European Space Agency's Sentinel-1 satellite, using change detection or IEM model inversion; or from L-band radiometer products from the NASASMAP (Active and Passive Soil Moisture) satellite. In practical applications, published and validated global or regional soil moisture remote sensing products can be used directly, such as the NASA-USDA (National Aeronautics and Space Administration-United States Department of Agriculture) enhanced SMAP (Soil Moisture Active Passive Detection Satellite) global soil moisture dataset.

[0067] S4. Parameter Feedback Optimization. After completing the dynamic state correction for a full growing season or year, based on the corrected state sequence, the soil saturated hydraulic conductivity parameter field in the water-holding function simulation module is adjusted using a parameter adjustment algorithm to match the simulation behavior of the water-holding function simulation module with the corrected state sequence. This step improves the model's long-term predictive ability and physical consistency by optimizing key hydrological parameters. The definition of a full growing season adopts an objective remote sensing monitoring method, specifically: acquiring the time-series normalized vegetation index (NDVI) data of the ecological restoration area and plotting its annual variation curve; setting the starting threshold of the growing season as an NDVI value that is a certain percentage (e.g., 20%) above the annual baseline value, and setting the ending threshold of the growing season to the same value; the starting date of the growing season is the date when the NDVI curve rises from a low value to the starting threshold, and the ending date of the growing season is the date when the NDVI curve falls from a high value to the ending threshold; the period between these two dates constitutes a full growing season. This method overcomes the problem of mismatch between the actual phenological period of vegetation that may be caused by relying solely on the natural calendar or fixed months.

[0068] S5. Optimize Result Output. The adjusted soil saturated hydraulic conductivity parameter field is fed back to the water-holding capacity simulation module, and the water-holding capacity simulation module is run again to obtain a spatial distribution map of water conservation capacity. This step integrates all previous optimization results to generate a highly reliable spatial distribution map that has been validated by multi-source data and optimized by the model, providing an accurate data foundation for effect evaluation.

[0069] S6. Effect Evaluation and Report Generation. By comparing the water conservation capacity results before and after restoration, or between the restored area and the control area, the improvement in water holding capacity is calculated, including the spatial distribution, total amount, and average improvement intensity, and a visual monitoring report is automatically generated. This step achieves a spatial explicit and quantitative assessment of the improvement in water holding capacity, providing an objective and quantitative scientific basis for the acceptance and management decisions of ecological restoration projects.

[0070] This invention addresses the mismatch between ground-based quadrat monitoring data and remote sensing pixel scale through adaptive observation error estimation, thereby improving the accuracy of data fusion. An ensemble Kalman filter algorithm is employed to dynamically correct the model's state variables, ensuring consistency between simulated conditions and actual observations. Parameter feedback optimization adjusts key parameter fields, enabling the model to more accurately reflect changes in the underlying surface after ecological restoration. Finally, a spatialized assessment of the improvement in water retention capacity is generated, achieving precise quantification of the ecological restoration effect.

[0071] Through a dual mechanism of state correction and parameter optimization, model uncertainty is significantly reduced, enabling detection results to accurately reflect the subtle effects of restoration projects. The system solves the challenge of fusing ground-based point data with remote sensing area data, ensuring both regional representativeness and high accuracy by incorporating true ground values. An adaptive error estimation mechanism ensures reliable detection accuracy even in complex restoration areas with high environmental heterogeneity. The process is clear and can be automated, providing continuous, objective, and quantitative data support for the acceptance of ecological restoration projects, long-term effectiveness monitoring, and management decisions.

[0072] Specifically, this invention provides a method for detecting the improvement effect of water retention function in ecological restoration. The core algorithm of the water retention function simulation module is constructed based on the water conservation module of the mature InVEST (Integrated Valuation of Ecosystem Services and Trade-offs) model. This module is based on the principle of water balance and is suitable for large-scale ecosystem service assessment.

[0073] Construction and operation of the water-holding function simulation module: Before running the water-holding function simulation module, the following multi-temporal and spatiotemporal remote sensing data and geospatial data need to be prepared and input. The typical sources and processing methods of these data are shown below:

[0074] Table 1. Overview of Input Data for the Water Retention Function Simulation Module

[0075]

[0076] The water retention simulation module calculates the water source conservation capacity WC for each grid x using the following steps. x (Unit: mm):

[0077] 1. Calculate the water production Y x Based on the Budyko hydrothermal coupling equilibrium assumption, the annual water production of each grid is calculated.

[0078] , where P x The multi-year average precipitation (mm) for grid x; AET x The annual actual evapotranspiration (mm) of grid x is calculated using the following formula:

[0079] , where R x The Budyko dryness index, ω x Dimensionless parameters for characterizing plant physiological characteristics.

[0080] 2. Calculate the saturated hydraulic conductivity K of the soil. soil-x Estimated based on soil texture data (content of sand, silt, and clay).

[0081] , where φ sand φ silt The values ​​represent the percentage of sand and silt in the soil, respectively. This soil transformation function is estimated based on soil texture data (sand, silt, and clay content). The formula is derived from classic research in the field of soil physics, specifically using the empirical model proposed by Saxton et al. (1986) (soil scientist K.E. Saxton and his collaborators).

[0082] 3. Calculate the Terrain Index (TI) x Extracted based on digital elevation model.

[0083] , where α x The slope is the area of ​​the catchment area on the uphill section of a unit contour line; x Let x be the slope of the grid.

[0084] 4. Calculate the water conservation capacity WC x Based on the above parameters, the final water conservation capacity is obtained using the following formula.

[0085] Among them, Vel j Let Vel be the velocity coefficient for the j-th land use type. j The values ​​can be set according to the recommended values ​​for different land use types in the reference "InVEST Model User Guide", or calibrated using runoff observation data from the study area. The constants (249, 0.9, 3, 300) in the formula are the default parameters of the water conservation module of the InVEST model. In this invention, these parameters are considered as key parameters that can be calibrated using local data. During implementation, measured hydrological data (such as runoff data) of the study area should be collected, and these constants should be calibrated using parameter sensitivity analysis and optimization algorithms (such as SCE-UA) to ensure the simulation accuracy of the model in specific ecological restoration areas.

[0086] For ecological restoration areas, in order to accurately assess the effectiveness of forest and grassland restoration measures, forest land and grassland can be further subdivided in the land use classification of the input model according to vegetation cover or canopy closure. For example, forest land can be divided into forested land, sparse forest land, and shrubland land, and grassland can be divided into high, medium, and low cover grassland.

[0087] In another embodiment of the present invention, in step S1, the uniformity of surface characteristics of each remote sensing pixel is evaluated, and an observation error value is adaptively assigned to each pixel through the following steps:

[0088] Step S1.1: Calculate the surface homogeneity index. Based on the ground quadrat monitoring data, calculate the surface homogeneity index HI corresponding to each remote sensing pixel. Specifically, this includes: extracting all ground quadrat monitoring data within the spatial range of each remote sensing pixel, calculating the mean μ and standard deviation σ of the monitoring data within that pixel, and calculating the coefficient of variation CV. Calculate the surface homogeneity index HI: The HI value range is [0,1]. The larger the HI value, the more uniform the surface characteristics of the pixel.

[0089] Step S1.2: Calculate the standard deviation of the quadrat data. Calculate the standard deviation σ of the ground quadrat monitoring data within each remote sensing pixel. Specifically, for each remote sensing pixel, obtain n ground quadrat monitoring data x1, x2, ..., xn within its range. n Calculate the standard deviation σ: The larger the σ value, the higher the dispersion of the monitoring data within the pixel.

[0090] Step S1.3: Calculate the observation error value. Based on the surface homogeneity index HI and the standard deviation σ of the ground quadrat monitoring data, calculate the observation error value R for each remote sensing pixel: Where k1 and k2 are preset weighting coefficients, with k1 ranging from 0.5 to 1.5 and k2 ranging from 0.3 to 1.0.

[0091] The above calculations ensure that the observation error value R is positively correlated with the standard deviation σ of the ground quadrat monitoring data and negatively correlated with the surface homogeneity index HI.

[0092] Specifically, in step S1.3, the observation error value R is calculated using a formula. The determination of the weighting coefficients k1 and k2 is crucial, as they quantify the relative contributions of measurement uncertainty and spatial heterogeneity to the total observation error, respectively. The following are three recommended methods for determining these coefficients; one can be selected based on the actual data and accuracy requirements:

[0093] Method 1: Combining empirical assignment with sensitivity analysis

[0094] 1. Initial empirical values: Initial settings are made based on the understanding of prior knowledge about the study area. Generally, since the standard deviation σ of the quadrat data is directly derived from the measured values ​​and has high reliability, the initial value of k1 can be set in the upper middle range (e.g., 1.0-1.2); while the surface homogeneity index HI is an indirectly calculated indicator with relatively high uncertainty, so the initial value of k2 can be set in the lower middle range (e.g., 0.4-0.6).

[0095] 2. Sensitivity Analysis and Optimization: Centering on the initial assignment, the combination of k1 and k2 is systematically changed within a certain range (e.g., k1: 0.5-1.5, k2: 0.3-1.0). The subsequent dynamic state correction (S3) process is then run, aiming to minimize the root mean square error (RMSE) between the analysis field and the independent validation dataset (unassimilated ground measurement data) to select the optimal combination of k1 and k2. This method is both practical and scientifically sound in practice.

[0096] Method 2: Regression method based on historical data training

[0097] 1. Training dataset construction: Collect data from representative periods in historical projects or study areas, including ground quadrat monitoring data, corresponding remote sensing image data, and high-precision reference data (such as dense sampling data or airborne remote sensing data).

[0098] 2. True Value Error Estimation: For each historical remote sensing pixel, calculate the difference between all ground quadrat data within it and the high-precision reference data, and use this as the true observation error R for that pixel. true .

[0099] 3. Model Fitting: Using the standard deviation σ and (1-HI) of the sample data for each pixel as independent variables, and R0 as the formula, the model fitting is performed. true Perform multiple linear regression with the variable as the dependent variable: The coefficients β1 and β2 obtained from the regression can be directly used as the optimal estimates of k1 and k2. This method is the most objective, but it relies on high-quality historical datasets.

[0100] Method 3: Search Method Based on Optimization Algorithms

[0101] 1. Definition of objective function: In a typical correction period, the objective function J(k1, k2) is defined as the difference (e.g., RMSE) between the state sequence obtained after dynamic state correction and the independent verification data of the same period.

[0102] 2. Parameter Space Search: Employing efficient optimization algorithms (such as the Nelder-Mead simplex method or particle swarm optimization), a search is performed within a reasonable range of values ​​for k1 and k2 to find the parameter that minimizes the objective function J. and .

[0103] 3. Determining the optimal solution: The optimal solution obtained from the search... and This method is applied throughout the entire monitoring period. It is highly automated and can adaptively find the optimal parameters, but the computational load is relatively large.

[0104] Independent validation datasets can be obtained through one of the following methods: (1) reserving some ground quadrats within the ecological restoration area that do not participate in the scale transformation and assimilation process, specifically for validation; (2) collecting higher-precision third-party mapping or monitoring data within the study area, such as high-resolution aerial remote sensing soil moisture data; (3) retaining the observation data of the last time period as the validation set in terms of time scale.

[0105] In practical applications, Method 1 is recommended as it achieves a good balance between computational cost and result reliability, and has been validated in multiple eco-hydrological monitoring projects. Once k1 and k2 are determined using any of the above methods, they can be used as fixed parameters for estimating observation errors throughout the monitoring period or in similar ecological zones, thus ensuring the robustness and accuracy of the method presented in this invention.

[0106] This scheme quantifies the degree of spatial heterogeneity within pixels using the surface homogeneity index HI, providing spatial structure information for observation error estimation. The standard deviation σ of the sample plot data reflects the dispersion of monitoring data within pixels, characterizing measurement uncertainty. An adaptive observation error estimation model is established through weighted combination, ensuring that error estimation considers both spatial heterogeneity and measurement uncertainty. This provides error estimates consistent with the characteristics of each pixel for subsequent data assimilation, improving the accuracy and reliability of state correction.

[0107] In specific implementation, the following points should be noted: ground quadrat monitoring data should include key ecological indicators such as soil moisture content and vegetation parameters; the remote sensing pixel scale and quadrat layout density should be matched to ensure statistical validity; weighting coefficients can be determined based on historical data training or set based on expert experience; and observation error values ​​should be standardized to meet the requirements of subsequent assimilation algorithms.

[0108] In another embodiment of the present invention, in step S3, the ensemble Kalman filter algorithm, combined with the observation error value, is used to dynamically correct the soil effective water content state variable in the water-holding function simulation module through the following steps:

[0109] Step S3.1: Input observation error value. The observation error value R assigned to each remote sensing pixel in step S1 is used as the input to the observation error covariance matrix in the ensemble Kalman filter algorithm. Specifically, this includes:

[0110] Read the observation error value R of each pixel calculated in step S1, construct the observation error covariance matrix R, whose diagonal elements are the observation error values ​​R of each pixel, and input the observation error covariance matrix R into the initialization parameters of the ensemble Kalman filter algorithm.

[0111] Step S3.2, Assimilation Cycle Settings. Within each monthly or quarterly assimilation cycle, perform the following operations:

[0112] Step S3.2.1, Background Field Setting. The effective soil water content simulated by the water-holding function simulation module (model simulation value) is used as the background field Xb. Specifically, at the beginning of the assimilation cycle, the simulated value of the effective soil water content output by the water-holding function simulation module is read, and the simulated value of the effective soil water content is used as the background field Xb, representing the state estimate predicted by the model.

[0113] Step S3.2.2, Observation Field Setup. The remote sensing inverted soil moisture data (remote sensing observations) is used as the observation field Yo. Specifically, this includes: acquiring remote sensing inverted soil moisture data corresponding to the assimilation period, performing quality control on the remote sensing inverted soil moisture data, removing outliers, and using the quality-controlled data as the observation field Yo, representing the actual observed state estimate;

[0114] Step S3.2.3, Fusion Calculation. The background field Xb and the observation field Yo are fused using an ensemble Kalman filter algorithm to generate the analysis field Xa. Specifically, this includes: calculating the covariance matrix Pb of the background field Xb, and calculating the Kalman gain matrix K. Calculate and analyze field Xa: Here, H is the observation operator, used to establish the physical transformation relationship from the model state space (soil effective water content) to the observation space (remote sensing-retrieved soil moisture). The observation operator H is implemented through a transformation function, which, based on the vertical distribution characteristics of soil moisture, converts the model-simulated soil effective water content (usually representing the mean value of a certain root layer depth, such as 0-60cm) into a soil moisture estimate that matches the remote sensing observation depth (such as the surface layer 0-5cm). A typical implementation uses an exponential decay function model: , where θ surface θ is the estimated value of surface soil moisture obtained from the conversion. rootzone Here, is the effective water content of the root zone soil simulated by the model, depth is the remote sensing depth (e.g., 5 cm), and λ is the attenuation coefficient, an empirical parameter related to soil texture and vegetation cover. This transformation makes the model state comparable to remote sensing observations in terms of physical dimensions and numerical values.

[0115] Step S3.2.4, State Update. Replace the soil effective water content state variable in the water-holding function simulation module with the analysis field Xa to complete the dynamic correction for this cycle. Specifically, this includes: writing the analysis field Xa back to the state variable storage area of ​​the water-holding function simulation module, updating the initial conditions of the water-holding function simulation module, and providing the corrected state for the next simulation cycle.

[0116] Step S3.3, State Sequence Construction. Dynamic correction is performed sequentially for all cycles to form a corrected state sequence. Specifically, this includes: recording the corrected state values ​​for each assimilation cycle in chronological order, constructing a time-continuous corrected state sequence Xcorr(t), and storing the corrected state sequence for subsequent parameter optimization.

[0117] This scheme quantifies observation uncertainty by inputting the adaptively calculated observation error value into the assimilation algorithm. It ensures the continuous consistency between the model state and the actual observation by periodically performing state correction. It optimally combines the advantages of model prediction and observation information through ensemble Kalman filtering fusion calculation. It provides high-quality training data for parameter optimization through the continuous construction of state sequences. It establishes a dynamic feedback mechanism between the model state and multi-source observation data, thereby improving simulation accuracy.

[0118] In practical implementation, the selection of the assimilation period should take into account both computational efficiency and the requirements of state update frequency. The design of the observation operator should take into account the physical relationship between the model state and the observed variables. The set size of the ensemble Kalman filter should be determined in balance according to computational resources and accuracy requirements. The update of the state variables should ensure compatibility with the numerical solution scheme of the model. The corrected state sequence should contain a sufficient time span to reflect the dynamic characteristics of the system.

[0119] In another embodiment of the present invention, in step S4, the soil saturated hydraulic conductivity parameter field in the water-holding function simulation module is adjusted by the parameter adjustment algorithm through the following steps:

[0120] Step S4.1, Setting the Target Truth. The corrected state sequence is used as the target truth for parameter optimization, specifically including:

[0121] Read the corrected state sequence Xcorr(t) formed in step S3, use the corrected state sequence Xcorr(t) as the reference benchmark for the parameter optimization process, establish a time index, and ensure that the corrected state sequence is aligned with the simulation results in the time dimension.

[0122] Step S4.2, Objective Function Construction. The objective function is constructed to measure the difference between the state variable sequence simulated by the water-holding function simulation module and the corrected state sequence. Specifically, it includes:

[0123] Define the objective function Where Xsim(t,θ) is the sequence of simulated state variables of the water-holding function simulation module under parameter θ, Xcorr(t) is the corrected state sequence, and the summation symbol Σ represents the cumulative calculation over the entire time series.

[0124] Step S4.3: Parameter Optimization Solution. The SCE-UA (Shuffled Complex Evolution University of Arizona) optimization algorithm is used to minimize the objective function, inverting and determining the optimal values ​​of the soil saturated hydraulic conductivity parameter field in the water-holding capacity simulation module. Specifically, this includes:

[0125] Step S4.3.1: Algorithm Initialization. Set the control parameters of the SCE-UA algorithm, including population size, number of complexes, and maximum number of iterations. Define the search space for the soil saturated hydraulic conductivity parameter field θ, set upper and lower limits for the parameters, initialize the population, and randomly generate an initial parameter set within the parameter search space. The control parameters of the SCE-UA algorithm can be set as follows: the population size can be set to 10-20 times the number of parameters to be optimized, the number of complexes is usually 2-5, and the maximum number of iterations can be set to 1000-5000 times depending on the computing resources.

[0126] Step S4.3.2, Iterative Optimization Process. Execute the following iterative steps until the convergence condition is met: For each parameter individual θi in the current population, run the water-holding function simulation module to obtain the simulated state sequence Xsim(t,θi), calculate the objective function value J(θi) corresponding to each parameter individual, sort and partition the population according to the objective function value, perform competitive evolution operation within each complex, generate new parameter individuals, update the population, and retain the better parameter individuals;

[0127] Step S4.3.3: Determining the optimal parameter. When the objective function value is less than a preset threshold or the maximum number of iterations is reached, the optimization process is terminated, and the parameter individual with the smallest objective function value in the population is selected as the optimal parameter θ. * Output the optimal value θ of the soil saturated hydraulic conductivity parameter field. * .

[0128] This scheme ensures that the parameter optimization process is based on high-quality observational constraint data by using the corrected state sequence as the target ground truth. It quantitatively characterizes the difference between the simulated sequence and the corrected sequence through the objective function, providing a clear direction for parameter optimization. It achieves a global search of the parameter space through the SCE-UA algorithm, avoiding getting trapped in local optima. It maintains the spatial correlation between parameters by performing an overall inversion of the soil saturated hydraulic conductivity parameter field. Through parameter optimization, it maximizes the fit between the model's simulated behavior and the corrected state sequence, thereby improving the model's long-term predictive ability.

[0129] In practical implementation, the construction of the objective function should take into account the uncertainty differences of the observation data at different time points. The parameter settings of the SCE-UA algorithm should balance the needs of computational efficiency and optimization accuracy. The determination of the parameter search space should be based on prior knowledge to ensure physical rationality. The setting of convergence conditions should take into account the balance between optimization effect and computational cost. The parameter evolution should be monitored during the optimization process to ensure the stability of the optimization process.

[0130] In another embodiment of the present invention, after optimizing the soil saturated hydraulic conductivity parameter field, the optimized soil saturated hydraulic conductivity parameter field is fed back to the water-holding capacity simulation module and re-run to obtain a spatial distribution map of water conservation capacity. This step is a crucial stage for integrating the results of all previous core processing steps (point-area data fusion, dynamic state correction, and parameter feedback optimization), and its purpose is to generate a highly reliable spatial distribution map of water conservation capacity that has been validated by multi-source data and optimized by the model. Specifically, it includes:

[0131] S5.1 Parameter Field Update and Injection. The optimal soil saturated hydraulic conductivity parameter field θ obtained in step S4 through a parameter adjustment algorithm (such as the SCE-UA algorithm) is updated and injected. * (Unit: mm / h), and fully fed back to the water-holding function simulation module in the form of spatial raster data. Specifically, this includes:

[0132] Read the optimized soil saturated hydraulic conductivity parameter field θ * The raster file must have a spatial extent and cell size that are completely consistent with the computational grid of the water-holding function simulation module. In the water-holding function simulation module, θ is used... * The raster data system systematically replaces the original, unoptimized soil saturated hydraulic conductivity parameter field of the module. This operation is achieved by modifying the module's parameter input interface or directly overwriting its internal parameter array, ensuring that subsequent simulations are based on the new parameter field.

[0133] S5.2, Driving Data Preparation and Model Rerun. Using the same preprocessed multi-temporal remote sensing data as in step S2 as the driving data, rerun the water-holding function simulation module that has undergone parameter updates. Specifically, this includes:

[0134] The driving data includes, but is not limited to: time-series precipitation P (unit: mm), actual evapotranspiration AET (unit: mm), vegetation index data, land use type data, etc. The water-holding capacity simulation module is then activated, utilizing the updated soil saturated hydraulic conductivity parameter field θ. * In addition to the aforementioned driving data, the water conservation capacity of each pixel is recalculated based on its embedded hydrological process algorithm (e.g., an algorithm based on the principle of water balance).

[0135] The precipitation and reference evapotranspiration data were preprocessed using the Kriging spatial interpolation method. This method utilizes the spatial autocorrelation of station data to provide optimal unbiased estimates. Compared to deterministic interpolation methods such as the inverse distance weighting method, it better reflects the spatial distribution structure of meteorological elements and provides an estimate of interpolation errors. After interpolating from the stations to the continuous spatial field, the resolution was further downscaled to match the model's computational grid (e.g., 30 meters) by incorporating digital elevation models (DEMs) as covariates.

[0136] The soil texture data is sourced from global soil databases (such as HWSD), with a relatively coarse original resolution (approximately 1 km). To address the mismatch between the spatial resolution of the original soil texture data and the model's computational grid, a machine learning-based spatial downscaling method is employed to process the soil texture (sand, silt, and clay content) data. The specific steps include: a) extracting topographic attributes as key environmental covariates from a high-resolution digital elevation model (DEM); b) training a regression model using the original coarse-resolution soil data as the response variable and the corresponding location's environmental covariates as features; c) applying the trained model to high-resolution environmental covariate data across the entire study area to predict and generate soil texture distribution maps at the corresponding resolution. This method can capture the nonlinear relationship between soil properties and topography, achieving statistical downscaling from "area" to "point," and significantly improving the spatial detail representation of the soil parameter field.

[0137] The root depth is determined based on typical eco-hydrological parameters under different land use types and vegetation covers. For example, it can be referenced from MODIS global root depth distribution products, or assigned corresponding empirical values ​​based on existing research findings on dominant tree species in the study area (such as deep-rooted trees, shallow-rooted shrubs, or herbs). This method avoids the logical error of simply linking the geological structural indicator of "bedrock depth" with the biological indicator of "root depth," ensuring the scientific nature of the parameter setting.

[0138] S5.3 Spatial Distribution Map Generation and Output. After the water-holding function simulation module finishes running, it outputs a spatial distribution map of the water conservation capacity in the ecological restoration area. This map is expressed in raster form, with each pixel representing the water conservation capacity S (unit: mm) at that location. This spatial distribution map is an integrated representation of the following technical processing chain:

[0139] Point-area data fusion, its underlying driving data fusion incorporates ground quadrat monitoring information upscaled through a scale transformation model, enhancing regional representativeness. Dynamic state correction: the model operates based on a soil effective water content state sequence that is periodically corrected using an ensemble Kalman filter algorithm, more closely resembling the actual state. Parameter feedback optimization: the model's key parameter—the soil saturated hydraulic conductivity parameter field—is the result of inversion optimization constrained by the corrected state sequence, making it more representative of the underlying surface characteristics after ecological restoration.

[0140] Step S5 is not an isolated or simple process of model repetition. It is a crucial step in realizing the value of the dual-loop assimilation mechanism (dynamic state correction loop and parameter feedback optimization loop) proposed in this invention. By feeding back the optimized parameter field to the model and re-running it, the final output spatial distribution map of water conservation not only possesses the spatial continuity advantage provided by model simulation, but more importantly, its reliability is significantly enhanced through deep fusion of multi-source data, dynamic correction of the model state, and feedback optimization of key parameters. This lays a solid data foundation for subsequent accurate and spatially explicit evaluation of the water-holding capacity improvement effect, overcoming the shortcomings of one-time simulation or schemes that only perform state correction in terms of long-term reliability and parameter adaptability.

[0141] In another embodiment of the present invention, after obtaining a spatial distribution map of water conservation capacity in the ecological restoration area, the improvement in water holding capacity is calculated based on the spatial distribution map, and a visual monitoring report is automatically generated. This step aims to provide a spatially explicit quantitative assessment of the water holding capacity improvement effect of the ecological restoration project and to automatically output a comprehensive monitoring report.

[0142] S6.1 Preparation of Data for Effect Comparison. Prepare spatial distribution map data of water conservation capacity for effect comparison, specifically including one or both of the following two scenarios:

[0143] By comparing time series data, a spatial distribution map of water conservation capacity in the same ecological restoration area before restoration was obtained. pre .

[0144] Spatial comparison: Obtain the spatial distribution map S of water conservation capacity in the control area where no remediation was implemented during the same time period. control .

[0145] Simultaneously, the spatial distribution map of water conservation capacity of the ecological restoration area generated in step S5 after restoration or during a specific assessment period is read. post or S treatment All spatial distribution maps must have the same spatial reference and pixel size.

[0146] S6.2 Spatial Overlay Analysis and Calculation of Water Holding Capacity Enhancement. S post (or S)treatment ) and S pre (or S) control Spatial overlay analysis was performed to calculate the water retention capacity improvement on a pixel-by-pixel basis. :

[0147] When comparing time sequence, .

[0148] When spatially contrasted .

[0149] Calculated This creates a new spatial raster layer, namely a spatial distribution map of the water-holding capacity improvement. Each cell in this map... The value (unit: mm) represents the net increase in water retention capacity at that location.

[0150] S6.3, Statistics on Improvement Effect. Based on the spatial distribution map of the improvement in water retention capacity, regional statistics are conducted, and the following core evaluation indicators are calculated:

[0151] Increase the total amount of water retention capacity The sum of ΔS values ​​for all pixels in the entire ecological restoration area is calculated and converted into volume or depth-area equivalent, typically in meters (m). 3 or mm·km 2 The calculation formula can be: ,in For the increase in the i-th pixel, A cell This represents the area of ​​a single pixel.

[0152] Average lifting intensity : Calculate the arithmetic mean of the ΔS values ​​of all pixels within the ecological restoration area, in mm. , where N is the total number of pixels in the ecological restoration area.

[0153] S6.4 Automatic Generation of Visual Inspection Report. The above calculation results are automatically integrated with the spatial distribution map to generate a visual inspection report containing charts and maps. The report content should include at least:

[0154] A spatial distribution map of the water-holding capacity improvement, serving as a core appendix, visually illustrates the spatial pattern and heterogeneity of the improvement effect. A list or chart of key statistical indicators clearly outlines the total improvement in water-holding capacity. and average improvement intensity Spatial statistics, including the number of pixels or area percentage of the increase in intensity across different intensity ranges (histogram or pie chart), are included. Necessary legends, scale bars, and text descriptions ensure the report's professionalism and readability. The report generation process utilizes pre-set templates and automated scripts to ensure rapid and standardized output of assessment results.

[0155] The core of this step lies in conducting a spatially explicit, pixel-by-pixel quantitative effect assessment based on the high-precision, high-reliability spatial distribution map of water conservation capacity generated through all the aforementioned optimization steps (S1 to S5). This method can accurately reveal the spatial heterogeneity of the water retention capacity improvement effect, answering the refined question of where and by how much improvement has been achieved, overcoming the limitations of traditional methods that rely on sporadic point assessments or regional overall average assessments. The feature of automatically generating comprehensive visualization reports greatly enhances the practical value and operational application potential of this method in ecological restoration project acceptance, long-term effectiveness monitoring, and management decision-making.

[0156] In another embodiment of the present invention, step S1 further includes the following specific implementation methods for improving the accuracy and reliability of scale conversion of ground quadrat monitoring data:

[0157] S1.a. Delineation of Eco-hydrological Response Units. Based on digital elevation model data and land use type data of the ecological restoration area, the study area is divided into several relatively homogeneous eco-hydrological response units. Specifically, these include:

[0158] Input digital elevation model data to calculate topographic factors such as slope and aspect. Overlay land use data with topographic factors and use clustering algorithms or decision tree rules to classify areas with similar hydrological response characteristics into the same eco-hydrological response unit.

[0159] S1.b. Construction of the Zonal Scale Transformation Model. For each eco-hydrological response unit defined in step S1.a, a scale transformation model is established. This model is implemented using a geographically weighted regression algorithm, with the specific steps as follows:

[0160] The regression variables were set using vegetation index and land surface temperature data of the remote sensing pixels corresponding to the ground quadrat monitoring data as independent variables. The ground quadrat monitoring data itself at the corresponding location was used as the dependent variable. A geographically weighted regression model was established, constructing a geographically weighted regression model for all quadrat points within each eco-hydrological response unit. The core of this model lies in the fact that its regression coefficients vary with geographic spatial location, enabling it to capture the spatial non-stationarity of the relationship between independent and dependent variables.

[0161] Specifically, in the geographic weighted regression algorithm, a Gaussian function is used as the spatial weight function to characterize the decay of spatial relationships. The weight function is defined as: w ij It is the regression weight of position j to position i, d ijLet be the Euclidean distance between locations i and j, and b be the bandwidth parameter used to control the rate at which the weights decay with distance. The bandwidth b can be determined through cross-validation or the AICc criterion. The Gaussian function provides smooth weight decay, is suitable for most continuous geographic phenomena, and is key to achieving high-precision upscaling transformation.

[0162] S1.c. Area Data Transformation and Uncertainty Quantification. Based on the unit-scale transformation models established in step S1.b, data transformation is performed and the transformation quality is evaluated:

[0163] Upscaling transforms point-distributed ground quadrat monitoring data into an areal data layer matching the scale of remote sensing pixels, using the scaling transformation model of its respective unit. Uncertainty index generation involves simultaneously calculating and outputting the upscaling uncertainty index for each remote sensing pixel during the transformation process. This index can be quantified based on the local prediction standard deviation or residuals of a geographically weighted regression model.

[0164] S1.d. Uniformity assessment data update. The upscaled areal data and upscaling uncertainty index obtained in step S1.c are used to replace the original ground quadrat monitoring data that have not undergone scale transformation, and the uniformity of surface characteristics in each remote sensing pixel is assessed.

[0165] This implementation addresses the common issue of surface heterogeneity in ecological restoration areas by constructing a process of zoning → modeling → transformation → uncertainty assessment. Eco-hydrological response unit division, based on topography and land use, ensures relatively homogeneous hydrological processes and surface characteristics within each unit, laying the foundation for establishing a high-precision transformation model. The application of a geographic weighted regression algorithm, which considers the influence of spatial location on regression relationships, better captures the spatial non-stationarity of the relationship between vegetation indices, surface temperature, and ground monitoring parameters, significantly improving the accuracy of the transformation from "point" to "area" under complex underlying surface conditions. The upscaling transformation uncertainty index quantifies the uncertainty of the transformation process itself for the first time at the scale transformation stage. This index provides a more scientific and refined input for adaptive observation error estimation in subsequent step S1, thereby improving the reliability of the entire data assimilation chain from the source. Through the above refined processing, not only is the scale transformation from ground "point" data to remote sensing "area" data achieved, but more importantly, a quantitative assessment of the transformation quality is provided, forming a dual output of "data + quality index," providing a fundamental guarantee for the reliability of all subsequent processing steps.

[0166] In another embodiment of the present invention, after completing the dynamic state correction in step S3 and before performing the parameter feedback optimization in step S4, the following specific implementation method is also included: dynamically adjusting the parameter optimization process based on simulation performance. This step aims to enhance the adaptability of the method in the long-term dynamic process of ecological restoration by adaptively controlling the subsequent parameter optimization process through multi-timescale model performance evaluation.

[0167] S3a, Calculation of multi-timescale simulation performance indicators. Based on the corrected state sequence X obtained in step S3. corr(t) The original simulation state sequence X of the water-holding function simulation module sim(t) Calculate the simulation performance metrics in the following two dimensions:

[0168] Short-term simulation bias B short : Calculate X within each assimilation period sim(t) With X corr(t) The root mean square of the difference is then used to calculate the average of the root mean squares of all periods over the entire evaluation period. , where t cycle It represents the time range of one assimilation cycle.

[0169] Long-term simulation trend fit C long : Calculate X over the entire growing season or annual assessment period. sim(t) Sequence and X corr(t) The correlation coefficient between sequences. , where T represents the entire evaluation period.

[0170] S3b. Adaptive adjustment of parameter optimization process. Based on the simulation performance indicators calculated in step S3a, dynamically adjust the key control parameters of the parameter adjustment algorithm (taking the SCE-UA algorithm as an example) to be used in step S4:

[0171] Adjustment of convergence threshold ε: If B short A large value indicates significant short-term dynamic bias in the model. Therefore, the convergence threshold ε should be appropriately relaxed (i.e., the value of ε should be increased) to avoid excessive time consumption in the optimization process due to local fitting. If C long A lower value indicates that the model has a poor ability to capture long-term trends. Therefore, the convergence threshold ε should be tightened (i.e., the value of ε should be reduced) in order to seek a more accurate global optimum.

[0172] Adjustment of iteration step size factor α: If B short The value is large and C long The value is acceptable, indicating that the model needs significant parameter adjustments to improve short-term dynamics. Therefore, iterative step size factor α should be appropriately increased to accelerate the search speed. If B short The value is small and C longA low value indicates that the model may need fine-tuning to capture long-term trends. In this case, it is appropriate to reduce the iteration step size factor α to enhance local search capabilities.

[0173] The adjustment rules can be implemented through predefined decision matrices or empirical formulas, ensuring the clarity and repeatability of the adjustment logic.

[0174] This implementation method addresses the technical problem of insufficient adaptability of traditional parameter optimization methods in the long-term dynamic process of ecological restoration through a time-series performance feedback mechanism. Multi-timescale performance evaluation, by simultaneously considering the consistency between short-term simulation deviations and long-term simulation trends, comprehensively diagnoses the model's performance differences across different time dimensions during ecological restoration. Adaptive control of the parameter optimization process dynamically adjusts the convergence threshold and iteration step size of the optimization algorithm based on the model performance diagnosis results, achieving a match between the optimization strategy and the model's current state and avoiding the limitations of fixed parameter optimization settings. Enhancing the applicability to dynamic processes, the introduction of this feedback loop allows the entire detection method to better adapt to the characteristics of the evolution of underlying surface conditions over time caused by ecological restoration, significantly improving the reliability of parameter optimization results in long-term dynamic simulations, thereby enhancing the scientific rigor of the final evaluation of water retention capacity improvement.

[0175] In another embodiment of the present invention, step S1 further includes the following specific implementation methods for handling the temporal dynamic effects of ecological restoration measures:

[0176] S1.e. Identification and Phase Division of Restoration Measures. Identify ecological restoration measures implemented in different time sequences within the ecological restoration area and determine the specific implementation time nodes for each measure. Using these implementation time nodes as boundaries, the entire monitoring period is divided into multiple consecutive restoration phases. For example, if there are two key restoration measure implementation time points t1 and t2 within the monitoring period, the entire period is divided into: Phase 1 (start to t1), Phase 2 (t1 to t2), and Phase 3 (t2 to end).

[0177] S1.f. Construction of Stage-Specific Initial Parameter Sets. For each remediation stage defined in step S1.e, a stage-specific initial parameter set for the water-holding capacity simulation module is constructed. Specifically, this includes:

[0178] For the first repair phase (Phase 1), the initial parameter set θ initphase1 It can be set based on baseline survey data before the repair or the default parameters of the area.

[0179] For subsequent repair phases (such as Phase 2 and Phase 3), the initial parameter set θ initphase2 θ initphase3The results of ecosystem succession in the previous stage and the expected impact of newly implemented restoration measures should be fully considered, and targeted adjustments should be made based on the parameter set of the previous stage.

[0180] S1.g. Dynamic Parameter Calling Mechanism. A dynamic parameter calling mechanism is introduced when running the water-holding function simulation module. This mechanism is based on the current simulation time t. sim During the corresponding repair phase, the relevant phase-specific initial parameter set is automatically invoked and loaded. The specific process is as follows:

[0181] During simulation initialization, the initial parameter set θ of Phase1 is loaded. initphase1 At simulation time t sim When the start time node of the next repair stage is reached (e.g., t1), the current parameter field of the water-holding function simulation module will be automatically switched to θ. initphase2 Similarly, in t sim When a subsequent time node (e.g., t2) is reached, switch to θ. initphase3 .

[0182] This implementation method addresses the deep-seated technical problem of the often overlooked temporal dynamic effects in ecological restoration projects. It constructs a complete time dimension encompassing time-series identification, stage division, parameter differentiation, and dynamic retrieval. By accurately identifying the implementation time points of restoration measures and dividing them into stages, it lays a foundation for refined processing in the time dimension for dynamic assessment. Setting differentiated initial parameters for different restoration stages allows for a more accurate characterization of the ecosystem's evolutionary characteristics and hydrological response under the dominance of different restoration measures at each stage. During the simulation, the parameter set is dynamically switched according to the time stages, achieving precise simulation of the entire ecological restoration process. This overcomes the limitations of traditional methods that treat the restoration area as a static unit and ignore the cumulative, superimposed, and interactive effects of restoration measures over time. Through refined processing in the time dimension, this implementation method achieves a technological leap from static assessment to dynamic process assessment, significantly improving the applicability of the detection method to ecological restoration projects with complex temporal characteristics and the scientific rigor and attribution accuracy of the assessment results.

[0183] This invention also provides a detection system for improving the water retention capacity of ecological restoration by performing the above-described methods. The system relies on a distributed computing environment comprised of data storage and management servers, high-performance computing nodes, and network devices. For example, for an area of ​​1000 km²... 2 For the ecological restoration area, a one-year simulation using 30-meter resolution data is recommended. It is suggested to configure computing nodes with at least 32 CPU cores and 128GB of memory to ensure the process is completed within a reasonable timeframe (e.g., 24 hours). At the software level, the system works collaboratively through the following specifically implemented modules:

[0184] The multi-source data acquisition module is implemented as a service with data access, parsing, and quality management functions. This module automatically downloads or receives raw data from various remote sensing data sources and ground sensor networks listed in Table 1 through scheduled tasks. Its built-in data preprocessing unit calls open-source library functions such as GDAL and Orfeo ToolBox to perform geometric correction, radiometric calibration, atmospheric correction, and format standardization operations on the raw data. Finally, it outputs a spatiotemporally unified raster dataset and attribute table, which are stored in a spatiotemporal database for subsequent modules to use.

[0185] The water retention simulation module is implemented as a parameterized model executor. This module runs by calling the kernel of the InVEST model's water conservation module (or an executable program of a similar model such as SWAT) compiled in C++ or Fortran. In practice, the module reads the driver data path, model parameters, and output settings according to the input parameter configuration file (such as INI format). After startup, the module iteratively executes its embedded hydrological process algorithm at time steps, calculating water conservation capacity pixel by pixel, and writes the simulation results (including preliminary simulation results, intermediate results of state variables, and the final spatial distribution map) to a specified directory in standard raster formats such as GeoTIFF. This module accepts calls and parameter passing through the system command-line interface.

[0186] The intelligent data assimilation engine is the core processing hub of this system. It is specifically implemented as a dedicated algorithm suite written in Python, MATLAB, or C++, containing three sequentially executed processors:

[0187] The adaptive error estimation unit is implemented as a script program, whose input is ground quadrat monitoring data read from the database. The core logic of the program is to perform statistical calculations as described in S1.1 to S1.3, including calculating the surface homogeneity index HI and the standard deviation σ of the quadrat data for each pixel, and applying the formula... The observation error value R for each pixel is calculated. The calculation results are constructed into an observation error covariance matrix R in diagonal matrix form and stored as a file for the state dynamic correction unit to read.

[0188] The dynamic state correction unit is implemented as a standalone computational program integrating an ensemble Kalman filter algorithm. Its operation is triggered by a timed task and periodically (monthly or quarterly) performs the following operations: reading the soil effective water content background field Xb output from the water-holding capacity simulation module; reading quality-controlled remote sensing inversion soil moisture data as the observation field Yo; reading the observation error covariance matrix R; and calling EnKF algorithm library functions to execute the formula. and The calculation generates the analysis field Xa. Finally, through file read / write operations, the initial state file of the water-holding function simulation module is overwritten with the analysis field Xa, completing the update of the state variables. The corrected states of all cycles are recorded in chronological order to form a corrected state sequence file.

[0189] The parameter feedback optimization unit, implemented as an optimization controller, is at its core an executable program encapsulating optimization algorithms such as SCE-UA. This unit starts after completing a full time interval of state correction. It reads the corrected state sequence Xcorr(t) and constructs the objective function. Subsequently, the SCE-UA algorithm library was invoked, and the water-holding function simulation module was run iteratively multiple times (each run receiving a different set of candidate parameter θ values) to find the optimal parameter θ that minimizes J(θ). * Ultimately, θ * Write a new soil saturated hydraulic conductivity parameter field file.

[0190] The effect evaluation and output module is implemented as an application integrating Geographic Information System (GIS) functions and a report generation engine. This module uses spatial analysis libraries such as ArcPy or GDAL to read the spatial distribution maps of water conservation capacity before and after restoration, or between the restored and control areas, and performs raster calculations (such as subtraction) to generate a spatial distribution map of the improvement in water holding capacity. Subsequently, it calls regional statistical functions such as Zonal Statistics to calculate indicators such as the total improvement and average improvement intensity. Finally, based on a preset Jinja2 or similar template, the module automatically combines the resulting maps, statistical charts, and text descriptions into a visual inspection report in PDF or HTML format. During report generation, statistical charts can be generated using Python's Matplotlib or Seaborn libraries, and spatial distribution maps can be processed using GDAL or ArcPy libraries and embedded into the report template. Finally, a template engine such as Jinja2 automatically integrates the charts, maps, and text descriptions into a report in PDF or HTML format.

[0191] System Workflow: During runtime, data is transferred between system modules through predefined, structured data interface files (such as GeoTIFF raster data and CSV tabular data). The main control script (such as a Python script or Shell script) is responsible for calling each executable program or service in sequence according to the workflow, and determining the success or failure of execution by checking the return value of each step and generating logs. When a step fails, the system can automatically retry or terminate the process and issue an alert to ensure the robustness of the workflow.

[0192] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. Other modifications can be easily made by those skilled in the art. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and examples shown and described herein.

Claims

1. A method for detecting the effect of improving water retention function in ecological restoration, characterized in that, Includes the following steps: S1. Collect multi-temporal and spatial remote sensing data of the ecological restoration area and control area, as well as ground quadrat monitoring data. Based on the ground quadrat monitoring data, evaluate the uniformity of surface characteristics in each remote sensing pixel and adaptively assign an observation error value to each pixel. S2. Run the water retention function simulation module, input multi-temporal and spatial remote sensing data as driving data, and obtain preliminary simulation results of water conservation capacity in the ecological restoration area; S3. Using the soil moisture data retrieved from remote sensing as the observation signal on a monthly or quarterly basis, the ensemble Kalman filter algorithm is used, combined with the observation error value, to dynamically correct the soil effective water content state variable in the water holding function simulation module, and obtain the corrected state sequence. S4. After completing the dynamic state correction for a full growing season or year, based on the corrected state sequence, adjust the soil saturated hydraulic conductivity parameter field in the water-holding function simulation module through the parameter adjustment algorithm so that the simulation behavior of the water-holding function simulation module matches the corrected state sequence. S5. Feed the adjusted soil saturated hydraulic conductivity parameter field back to the water holding function simulation module, rerun the water holding function simulation module, and obtain the spatial distribution map of water conservation. S6. By comparing the water conservation results before and after restoration or between the restored area and the control area, calculate the improvement in water holding capacity, including the spatial distribution, total amount and average improvement intensity, and automatically generate a visual detection report. Step S1 further includes the following steps: Based on digital elevation model data and land use type data of the ecological restoration area, the ground quadrat monitoring data are divided into different eco-hydrological response units; For each eco-hydrological response unit, a scale transformation model is established. The scale transformation model is implemented through a geographic weighted regression algorithm. The independent variables of the regression include the vegetation index and surface temperature data of the remote sensing pixels corresponding to the ground quadrat monitoring data, and the dependent variable of the regression is the ground quadrat monitoring data. Based on the scaling transformation model, the point-distributed ground quadrat monitoring data is upscaled and transformed into areal data that matches the remote sensing pixels, while generating an upscaling transformation uncertainty index for each remote sensing pixel. Using upscaled areal data and upscaled uncertainty index, the uniformity of surface characteristics in each remote sensing pixel is evaluated.

2. The method for detecting the effect of improving water retention function in ecological restoration as described in claim 1, characterized in that, Step S1, which evaluates the uniformity of surface characteristics in each remote sensing pixel and adaptively assigns an observation error value to each pixel, specifically includes: based on ground quadrat monitoring data, calculating the surface uniformity index corresponding to each remote sensing pixel and the standard deviation of the ground quadrat monitoring data within each remote sensing pixel; and calculating the observation error value for each remote sensing pixel based on the surface uniformity index and the standard deviation of the ground quadrat monitoring data. The observation error value is positively correlated with the standard deviation of the ground quadrat monitoring data and negatively correlated with the surface uniformity index.

3. The method for detecting the effect of improving water retention function in ecological restoration as described in claim 1, characterized in that, Step S3 employs an ensemble Kalman filter algorithm, combined with observation error values, to dynamically correct the soil effective water content state variable in the water-holding capacity simulation module. Specifically, this includes: The observation error value assigned to each remote sensing pixel in step S1 is used as the input to the observation error covariance matrix in the ensemble Kalman filter algorithm. Within each monthly or quarterly assimilation cycle, perform the following operations: The effective soil water content simulated by the water-holding function simulation module is used as the background field; Soil moisture data retrieved from remote sensing was used as the observation field; The background field and the observation field are fused and calculated using an ensemble Kalman filter algorithm to generate the analysis field; Replace the soil effective water content state variable in the water-holding function simulation module with the analysis field to complete the dynamic correction for this cycle; The dynamic corrections for all cycles are performed sequentially to form a corrected state sequence.

4. The method for detecting the effect of improving water retention function in ecological restoration as described in claim 1, characterized in that, Step S4, which involves adjusting the soil saturated hydraulic conductivity parameter field in the water-holding capacity simulation module using a parameter adjustment algorithm, specifically includes: The corrected state sequence is used as the target true value for parameter optimization; Construct an objective function to measure the difference between the state variable sequence simulated by the water-holding function simulation module and the corrected state sequence; The SCE-UA optimization algorithm is used to minimize the objective function and invert and determine the optimal values ​​of the soil saturated hydraulic conductivity parameter field in the water-holding function simulation module.

5. The method for detecting the effect of improving water retention function in ecological restoration as described in claim 1, characterized in that, Step S5 specifically includes: Update the soil saturated hydraulic conductivity parameter field adjusted in step S4 to the water holding function simulation module, replacing its original soil saturated hydraulic conductivity parameter field. Using the multi-temporal remote sensing data from step S1 as driving data, the water holding function simulation module with updated parameters is rerun. A spatial distribution map of water conservation capacity in the ecological restoration area is obtained. This spatial distribution map is an output that integrates the spatial representativeness of ground quadrat monitoring data, dynamic state correction, and parameter feedback optimization results.

6. The method for detecting the effect of improving water retention function in ecological restoration as described in claim 1, characterized in that, Step S6 specifically includes: Based on the spatial distribution map of water conservation obtained in step S5, the water conservation results after restoration in the ecological restoration area are spatially overlaid with the water conservation results before restoration in the same area or with the water conservation results of the control area without restoration. Through spatial overlay analysis, the improvement in water retention capacity is calculated pixel by pixel, and a spatial distribution map of the improvement in water retention capacity is generated. Based on the spatial distribution map of the improvement in water retention capacity, the total improvement in water retention capacity and the average improvement intensity of the entire ecological restoration area were statistically calculated. The system automatically integrates spatial distribution maps of water retention capacity improvement, total water retention capacity improvement, average improvement intensity, and related spatial statistics to generate a visual inspection report containing charts and maps.

7. The method for detecting the effect of improving water retention function in ecological restoration as described in claim 1, characterized in that, Between steps S3 and S4, the following steps are also included: Based on the corrected state sequence, the simulation performance index of the water holding function simulation module at different time scales is calculated. According to the simulation performance index, the convergence threshold and iteration step size of the parameter adjustment algorithm are dynamically adjusted. The simulation performance index includes the consistency between short-term simulation deviation and long-term simulation trend.

8. The method for detecting the effect of improving water retention function in ecological restoration as described in claim 1, characterized in that, Step S1 also includes the following steps: identifying ecological restoration measures implemented at different times within the ecological restoration area and determining the implementation time nodes of each measure. Using the implementation time nodes as boundaries, the entire monitoring period is divided into multiple continuous restoration stages. For each restoration stage, a stage-specific initial parameter set for the water-holding function simulation module is constructed. When running the water-holding function simulation module, the corresponding initial parameter set is called according to different restoration stages.

9. A detection system for improving the water retention function of ecological restoration, used to perform the detection method according to any one of claims 1-8, characterized in that, include: The multi-source data acquisition module is used to acquire multi-temporal and spatial remote sensing data of the ecological restoration area and the control area, as well as ground quadrat monitoring data; The water retention function simulation module is used to obtain preliminary simulation results of the water conservation capacity of the ecological restoration area based on multi-temporal and spatial remote sensing data. The intelligent data assimilation engine connects the multi-source data acquisition module and the water-holding function simulation module. The intelligent data assimilation engine includes: An adaptive error estimation unit is used to evaluate the uniformity of surface characteristics in each remote sensing pixel based on ground quadrat monitoring data, and adaptively assign an observation error value to each pixel. The state dynamic correction unit is used to dynamically correct the soil effective water content state variable in the water holding function simulation module by using remote sensing inverted soil moisture data as the observation signal on a monthly or quarterly basis, employing an ensemble Kalman filter algorithm and combining the observation error value, to obtain the corrected state sequence. The parameter feedback optimization unit is used to adjust the soil saturated hydraulic conductivity parameter field in the water holding function simulation module based on the corrected state sequence after completing a full growing season or year of dynamic state correction. The effect evaluation and output module connects the water-holding function simulation module and the intelligent data assimilation engine. It is used to feed back the adjusted soil saturated hydraulic conductivity parameter field to the water-holding function simulation module, rerun the water-holding function simulation module, obtain the spatial distribution map of water conservation, and calculate the improvement in water-holding function by comparing the water conservation results before and after the restoration or between the restoration area and the control area. This includes the spatial distribution, total amount and average improvement intensity of the improvement, and automatically generates a visual detection report.

Images