A vegetation coverage prediction method and system based on multi-source data fusion

Abstract

The present invention discloses a vegetation coverage prediction method and system based on multi-source data fusion, which relates to the technical field of vegetation coverage prediction and includes the following steps: obtaining historical vegetation coverage data and land use type data, and obtaining a matching time series of land use transformation and vegetation recovery lag; performing frequency band extraction on the matching time series based on Fourier transform to obtain an initial change frequency of the matching degree; performing random perturbation on the initial change frequency of the matching degree based on a preset first constraint condition to obtain a plurality of different matching degree change frequencies; obtaining vegetation coverage degradation impact data and ecological restoration potential data corresponding to different matching degree change frequencies, and constructing an evaluation model; and correcting the vegetation coverage prediction value of the preset first prediction model according to the data output by the evaluation model, thereby solving the problem of inaccurate vegetation coverage prediction caused by not considering the relationship fluctuation between the speed of land use type conversion and the lag degree of vegetation recovery.

Description

Technical Field

[0001] The present invention relates to the technical field of vegetation coverage prediction, and more particularly to a multi-source data fusion vegetation coverage prediction method and system. Background Art

[0002] The coupling relationship between land use and vegetation cover is particularly important in urban development. By combining land use change with vegetation cover change, we can effectively analyze the impacts of land development and urbanization on the ecological environment. Using remote sensing data and GIS technology, combined with land use type and vegetation cover, we can develop models linking land use and vegetation cover change, providing a scientific basis for decision-making in urban planning, ecological restoration, and other areas.

[0003] For example, the invention patent announcement with the announcement number CN114254802B discloses a method for predicting spatiotemporal changes in vegetation cover driven by climate change, including the following steps: step 1: determining the research area and data source for establishing a vegetation prediction model; step 2: using historical measured vegetation cover grid data as the dependent variable, introducing different lags as independent variables, constructing a pixel-by-pixel multiple linear regression model, and comprehensively considering the model's coefficient of determination, the five-year average analysis method, and the results of the propensity rate analysis to screen and construct the best multiple linear regression model; step 3: using meteorological data in the future period as input to calculate the future vegetation cover grid data; step 4: performing downscaling processing based on the future vegetation cover grid data and the historical observed vegetation cover data; and step 5: drawing a future interannual change curve. The present invention constructs a regression model in a pixel-by-pixel manner, obtains a regression model with a higher degree of fit in meteorological data with lower spatiotemporal resolution, and predicts the response of vegetation cover to climate change in the future period.

[0004] The above disclosed technical solutions have at least the following technical problems:

[0005] Traditional vegetation cover prediction methods ignore the mutual influence between land use type transformation and vegetation recovery lag, and cannot accurately capture the complex relationship between the two in the dynamic change process.

[0006] In view of the above problems, the present invention proposes a solution. Summary of the Invention

[0007] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a vegetation coverage prediction method and system based on multi-source data fusion. Through a vegetation coverage prediction method and system based on multi-source data fusion, the problem of inaccurate vegetation coverage prediction caused by the traditional method not considering the fluctuations in the relationship between the speed of land use type conversion and the lag degree of vegetation recovery is solved.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A method for predicting vegetation coverage using multi-source data fusion includes the following steps: obtaining historical vegetation coverage data and land use type data, and obtaining a matching time series of land use transition and vegetation restoration lag; extracting frequency bands from the matching time series based on Fourier transform to obtain an initial matching degree change frequency; randomly perturbing the initial matching degree change frequency based on a preset first constraint condition to obtain a number of different matching degree change frequencies; obtaining vegetation coverage degradation impact data and ecological restoration potential data corresponding to different matching degree change frequencies, and constructing an evaluation model; and revising the vegetation coverage prediction value of the preset first prediction model based on data output by the evaluation model.

[0010] In a preferred embodiment, the method of obtaining historical vegetation cover data and land use type data, and obtaining a matching time series of land use transformation and vegetation restoration lags, is specifically as follows: obtaining historical vegetation cover data and land use type data, and constructing a land use type transfer matrix; calculating the trace of the area change of each land use type in adjacent periods based on the land use type transfer matrix to obtain a land use transformation rate time series; constructing a time-lagged cross-correlation function of the land use area change rate and the vegetation coverage change rate of each land use type; obtaining the time delay when the time-lagged cross-correlation function first reaches a peak, and using the time delay as a lag period; obtaining the length of the lag period and calculating the vegetation restoration lag rate series, and using the cubic spline interpolation method to time-align the land use transformation rate time series and the vegetation restoration lag rate series to obtain a matching time series.

[0011] In a preferred embodiment, the frequency band extraction of the matching degree time series based on Fourier transform is performed to obtain the initial change frequency of the matching degree, specifically: the matching degree time series is Fourier transformed using a sliding window mechanism, and the main frequency band whose amplitude ratio in each window exceeds a preset threshold is extracted; the frequency mean of the main frequency band is calculated and used as the initial change frequency.

[0012] In a preferred embodiment, the initial change frequency of the matching degree is randomly disturbed based on the preset first constraint condition to obtain several different matching degree change frequencies, specifically: the maximum vegetation recovery lag rate, land use transformation rate and average lag period are calculated, and the first constraint condition is constructed; according to the first constraint condition, the disturbance amplitude of the Poisson distribution is obtained; the Poisson distribution is used for random disturbance to obtain several random disturbance terms within the disturbance amplitude; the random disturbance terms are added to the initial change frequency of the matching degree to obtain several different matching degree change frequencies.

[0013] In a preferred embodiment, the vegetation cover degradation impact data includes vegetation cover degradation impact characteristics, and the specific method of obtaining the vegetation cover degradation impact characteristics is as follows: obtain multi-phase remote sensing NDVI data and calculate the vegetation cover change rate of each phase; perform data analysis on the vegetation cover change rate, identify vegetation degradation hotspot areas and calculate the standard deviation and mean of the NDVI change rate of the vegetation degradation hotspot areas; obtain the degradation fluctuation value based on the standard deviation and mean of the NDVI change rate; obtain climate data, calculate the ratio of precipitation in the drought period to the annual average precipitation, and obtain the drought frequency index; calculate the vegetation cover degradation impact characteristics based on the degradation fluctuation value and the drought frequency index.

[0014] In a preferred embodiment, the ecological restoration potential data includes ecological restoration potential characteristics, and the specific method of obtaining the ecological restoration potential characteristics is as follows: obtain the species distribution database, and calculate the overlap between historical vegetation coverage and species suitable habitats to obtain a first potential value; obtain remote sensing precipitation data, and calculate the water collection rate corresponding to different slopes; calculate the mean and variance of the water collection rate, and use the ratio of the mean and variance of the water collection rate as the terrain wetness index; calculate the ecological restoration potential characteristics based on the first potential value and the terrain wetness index.

[0015] The technical effects and advantages of the vegetation coverage prediction method and system based on multi-source data fusion of the present invention are as follows:

[0016] 1. The present invention introduces the land use transformation rate and the vegetation recovery lag rate, uses Fourier transform analysis to match the time series, extracts the frequency of change in the matching degree, and then quantifies and corrects the vegetation recovery lag. This method, based on the evaluation model, combines multi-phase remote sensing data and climate data. It not only takes into account the speed of land use type transformation, but also comprehensively analyzes the lag effect of vegetation recovery, effectively improving the accuracy of vegetation coverage prediction. In addition, by introducing the random perturbation technology of Poisson distribution, a variety of possible change frequencies are provided for the prediction model, which enhances the robustness and adaptability of the model, solves the problem that traditional methods do not consider the relationship fluctuations between the speed of land use type conversion and the lag degree of vegetation recovery, thereby leading to inaccurate vegetation coverage prediction, and improves the accuracy of vegetation coverage prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The figure is a flow chart of a vegetation coverage prediction method based on multi-source data fusion according to the present invention.

[0018] Figure 2 The diagram is a structural diagram of a multi-source data fusion vegetation coverage prediction system of the present invention. DETAILED DESCRIPTION

[0019] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0020] Example 1, Figure 1 The present invention provides a vegetation coverage prediction method based on multi-source data fusion, which includes the following steps:

[0021] S1, obtain historical vegetation cover data and land use type data, and obtain the matching time series of land use transition and vegetation restoration lag;

[0022] In this example, historical vegetation cover data and land use type data are obtained, and the matching time series of land use transition and vegetation restoration lag are obtained, specifically:

[0023] Obtain historical vegetation cover data and land use type data, and construct a land use type transfer matrix;

[0024] Based on the land use type transition matrix, the trace of the area change of each land use type in adjacent periods is calculated to obtain the time series of land use transformation rate;

[0025] Construct the time-lagged cross-correlation function between the land use area change rate and vegetation cover change rate of each land use type;

[0026] Obtain the time delay when the time-lag cross-correlation function first reaches a peak, and use the time delay as the lag period;

[0027] The lag period length was obtained and the vegetation recovery lag rate series was calculated. The land use transition rate time series and the vegetation recovery lag rate series were aligned using the cubic spline interpolation method to obtain a matching time series.

[0028] The specific calculation formula for the matching degree time series between land use transition rate and vegetation restoration lag rate is as follows:

[0029]

[0030] in, is the matching degree at time t, is the trace of the change in area of ​​each land use type in adjacent periods, is the preset weight coefficient, is the vegetation coverage of the i-th land use type, is the average vegetation cover, is the number of land use types.

[0031] It should be noted that the historical vegetation cover data and land use type data are first obtained, and a land use type transfer matrix is ​​constructed. Through this matrix, the trace of the area change of each land use type in adjacent periods can be calculated, and then the land use transformation rate time series can be obtained. Then, based on the time-lagged cross-correlation function between the area change rate of land use type and the vegetation coverage change rate, the time-lagged relationship between the land use area change and the vegetation coverage change of each land use type is calculated, and by analyzing the peak position of the time-lagged cross-correlation function, the time-lagged period, that is, the lag effect of the impact of land use transformation on vegetation restoration, is obtained. Furthermore, by calculating the vegetation recovery lag rate series, the land use transformation rate time series and the vegetation recovery lag rate series are time-aligned in combination with the cubic spline interpolation method to obtain the final matching time series. The specific advantages are as follows:

[0032] First, by temporally aligning the land-use transition rate and the lag rate of vegetation restoration, we can comprehensively and systematically reveal the long-term impact of land-use change on vegetation restoration. In particular, by taking into account the time lag factor, we can effectively avoid the bias caused by traditional methods that ignore the time effect. Second, by smoothing the time series of different periods using the cubic spline interpolation method, we can help eliminate the interference of time series fluctuations on the analysis results, thereby improving the accuracy and reliability of the results. Finally, the analysis method based on the time-lag cross-correlation function can more accurately capture the intrinsic relationship between land-use change and vegetation restoration, thereby providing a scientific basis for the formulation of land-use management, ecological restoration and sustainable development strategies.

[0033] S2, extract the frequency band of the matching time series based on Fourier transform to obtain the initial change frequency of the matching degree;

[0034] In this example, frequency bands are extracted from the matching degree time series based on Fourier transform to obtain the initial change frequency of the matching degree, specifically:

[0035] A sliding window mechanism is used to Fourier transform the matching degree time series, and the main frequency band whose amplitude ratio exceeds the preset threshold in each window is extracted;

[0036] Calculate the frequency mean of the main frequency band and use it as the initial change frequency.

[0037] It should be noted that a sliding window mechanism is used to perform Fourier transform on the matching degree time series, and the main frequency band whose amplitude ratio exceeds the preset threshold is extracted in each window. Through Fourier transform, the time domain signal can be converted to the frequency domain, revealing the changing characteristics of different frequency components in the time series. In each sliding window, the main frequency band whose amplitude ratio exceeds the preset threshold is selected to ensure that the extracted frequency band is closely related to the main change pattern of the time series. Then, the frequency mean of the main frequency band in each window is calculated, and the frequency mean is used as the initial change frequency to further analyze the changing trend of the time series. The specific advantages are as follows:

[0038] Converting the matching degree time series from the time domain to the frequency domain through Fourier transform can reveal potential periodic changes and frequency characteristics, making the analysis results more intuitive and clear; secondly, using the sliding window mechanism to perform local frequency domain analysis on the time series can capture the dynamic changes in the time series and the frequency fluctuations in different time periods, avoiding the defect that the overall analysis may ignore local changes; thirdly, by extracting the frequency bands whose amplitude ratio exceeds the preset threshold, the noise component can be effectively removed to ensure that the extracted frequency bands are closely related to the actual change frequency of the matching degree; finally, calculating the frequency mean of the main frequency band and using it as the initial change frequency helps to accurately grasp the initial change trend of the time series, providing a stable and reliable frequency estimate, and laying a solid foundation for subsequent trend analysis and prediction.

[0039] S3, randomly perturbing the initial change frequency of the matching degree based on the preset first constraint condition to obtain a plurality of different matching degree change frequencies;

[0040] In this example, the initial change frequency of the matching degree is randomly perturbed based on the preset first constraint condition to obtain several different matching degree change frequencies, specifically:

[0041] Calculate the maximum vegetation recovery hysteresis rate, land use transformation rate and average hysteresis period, and establish the first constraint condition;

[0042] According to the first constraint, obtain the disturbance amplitude of the Poisson distribution;

[0043] Use Poisson distribution for random perturbation to obtain several random perturbation terms within the perturbation amplitude;

[0044] The random perturbation term is added to the initial change frequency of the matching degree to obtain several different matching degree change frequencies.

[0045] The specific formula of the first constraint is as follows:

[0046]

[0047] in, is the frequency disturbance, is the initial change frequency of matching degree, is the maximum vegetation recovery hysteresis rate, is the land use transition rate, is the average lag period.

[0048] It should be noted that, according to the first constraint, the disturbance amplitude of the Poisson distribution is obtained. The Poisson distribution is used to simulate the randomness of the disturbance. Its characteristic is that it can reflect the fluctuations within a certain range, thereby providing a probability distribution basis for subsequent random disturbances. Next, the Poisson distribution is used to randomly perturb the initial change frequency of the matching degree to obtain random disturbance terms within several disturbance amplitudes. These random disturbance terms are added to the initial change frequency of the matching degree to generate multiple different matching degree change frequencies. The specific advantages are as follows:

[0049] By calculating the maximum vegetation recovery lag rate, land use transformation rate and average lag period, a reasonable and scientific first constraint condition was constructed to ensure that the actual system change law will not be deviated from during the disturbance process, thus ensuring the accuracy and reliability of the model; secondly, the use of Poisson distribution for disturbance can simulate system changes under different scenarios by introducing randomness, providing diversified frequency changes for analysis and avoiding the limitations of a single model; thirdly, by adding the disturbance term to the initial change frequency of the matching degree, multiple perturbation frequencies can be generated.

[0050] S4, obtain the vegetation cover degradation impact data and ecological restoration potential data corresponding to different matching degree change frequencies, and build an evaluation model;

[0051] In this example, the vegetation cover degradation impact data includes vegetation cover degradation impact characteristics, and the specific method for obtaining the vegetation cover degradation impact characteristics is as follows:

[0052] Obtain multi-temporal remote sensing NDVI data and calculate the vegetation cover change rate for each phase;

[0053] Analyze the vegetation cover change rate data, identify vegetation degradation hotspots, and calculate the standard deviation and mean of the NDVI change rate in vegetation degradation hotspots;

[0054] Based on the standard deviation and mean of the NDVI change rate, the degradation fluctuation value is obtained;

[0055] Obtain climate data, calculate the ratio of precipitation during the drought period to the annual average precipitation, and obtain the drought frequency index;

[0056] The impact characteristics of vegetation cover degradation were calculated based on the degradation fluctuation value and drought frequency index.

[0057] The specific calculation formula for the impact characteristics of vegetation cover degradation is as follows:

[0058]

[0059] in, The impact characteristics of vegetation cover degradation are: is the degradation fluctuation value, is the drought frequency index, is the vegetation cover change rate, is the preset weight coefficient.

[0060] In this example, the ecological restoration potential data includes ecological restoration potential characteristics, and the specific method for obtaining the ecological restoration potential characteristics is as follows:

[0061] Obtain the species distribution database and calculate the overlap between historical vegetation coverage and species suitable habitat to obtain the first potential value;

[0062] Obtain remote sensing precipitation data and calculate the water collection rate corresponding to different slopes;

[0063] The mean and variance of the water collection rate were calculated, and the ratio of the mean and variance of the water collection rate was used as the terrain wetness index;

[0064] The ecological restoration potential characteristics were calculated based on the first potential value and terrain wetness index.

[0065] The calculation formula for the ecological restoration potential characteristic is as follows:

[0066]

[0067] in, The ecological restoration potential characteristics are is the terrain wetness index, is the seed dispersal distance, is the maximum propagation distance, and are the preset weight coefficients respectively.

[0068] In this example, we build an evaluation model, specifically:

[0069] The impact characteristics of vegetation cover degradation and ecological restoration potential were input into the random forest model;

[0070] The random forest model was trained using a cross-validation strategy to obtain the frequency evaluation value of the matching degree change.

[0071] It is important to note that the input variables used are the vegetation cover degradation impact characteristics and the ecological restoration potential characteristics, respectively reflecting the vegetation degradation during land use transition and its corresponding ecological restoration capacity. These characteristics include, but are not limited to, data on land use change, vegetation cover, and soil quality, and can comprehensively capture the dynamic characteristics of ecosystem changes. These characteristics are then modeled using a random forest model. Random forest is an ensemble learning algorithm that uses a combination of multiple decision trees for predictive analysis. It has strong noise immunity and efficient feature selection capabilities.

[0072] During model training, a cross-validation strategy was employed for optimization to ensure the model's robustness and generalization capabilities. Cross-validation effectively avoids overfitting and improves the model's predictive accuracy by dividing the dataset into multiple subsets and rotating the data between them for training and validation. The trained random forest model outputs an assessment of the frequency of matching changes based on the input characteristics of vegetation cover degradation and ecological restoration potential, reflecting the underlying patterns and trends of vegetation cover changes during ecological restoration. Its specific advantages are as follows:

[0073] First, the combined input of vegetation cover degradation and ecological restoration potential characteristics can comprehensively describe the impact of land use transition on the ecosystem, avoiding the limitations of single factor analysis; second, the random forest model can better capture the nonlinear relationship between features by integrating multiple decision trees for prediction, avoiding the linear assumption limitations in traditional regression analysis, and enhancing the accuracy and robustness of the model; third, the cross-validation strategy can effectively improve the generalization ability of the model, reduce errors caused by insufficient or biased training data, and ensure the reliability of the evaluation results; finally, by evaluating the frequency of changes in matching degree, this method provides a quantitative basis for the ecological restoration potential of land use transition, which can provide scientific decision-making support for land managers and ecological restoration planners, and has important practical application value.

[0074] S5, correcting the vegetation coverage prediction value of the preset first prediction model according to the data output by the evaluation model.

[0075] In this example, the vegetation coverage prediction value of the preset first prediction model is corrected according to the data output by the evaluation model, specifically:

[0076] Based on the evaluation model, the matching degree change evaluation value is obtained;

[0077] In the three-dimensional coordinate system, a three-dimensional prediction surface is constructed with the matching degree change frequency, matching degree change evaluation value, and prediction coverage as coordinate axes;

[0078] Calculate the feature points on the three-dimensional prediction surface where the gradient change rate exceeds the threshold, and obtain a feature point set;

[0079] Cluster the feature point set to obtain several mutation areas;

[0080] The correction weight is calculated according to the area of ​​the mutation region, and the predicted value of vegetation coverage is corrected according to the correction weight.

[0081] It should be noted that based on the evaluation model, a matching change assessment value is obtained. This value is obtained through a dynamic analysis of the land use transition rate and the hysteresis rate of vegetation recovery, and can comprehensively reflect the impact of land use change on vegetation coverage. Next, a three-dimensional prediction surface is constructed in a three-dimensional coordinate system, with the matching change frequency, matching change assessment value, and predicted coverage as the coordinate axes. The construction of this three-dimensional surface effectively integrates multiple influencing factors, enabling the prediction model to more accurately describe vegetation coverage in multiple dimensions of space and time.

[0082] Furthermore, by calculating the gradient change rate on the three-dimensional prediction surface and identifying characteristic points exceeding a set threshold, we can accurately capture areas of sudden changes in vegetation cover. These sudden changes may represent important ecological changes or key land use transitions, and are areas that require special attention in prediction. By performing cluster analysis on these characteristic point sets, we can divide them into several sudden change areas, thereby identifying the key drivers or trends of vegetation cover changes.

[0083] Finally, correction weights are calculated based on the areas of these mutation zones and used to refine the predicted vegetation cover. This correction process further adjusts and optimizes the initial predictions by incorporating the geographic characteristics and ecological changes in the mutation zones, ensuring that the predicted values ​​more closely reflect actual changes. This method not only considers dynamic changes in time and space, but also incorporates important factors such as land use transitions and delayed vegetation recovery, significantly improving the accuracy of the prediction model. Its specific advantages are as follows:

[0084] By constructing a three-dimensional prediction surface and analyzing the gradient rate of change, the team was able to accurately identify areas of sudden changes in vegetation coverage, overcoming the limitations of traditional methods that cannot handle sudden changes. Cluster analysis and the introduction of correction weights further optimized the predicted vegetation coverage, enhancing the model's accuracy and adaptability. Furthermore, the calculation of correction weights allows for differentiated corrections to be made to the predicted values ​​for different regions based on the actual conditions of the ecological region, making the predictions more precise in terms of spatial distribution and providing greater practical application value and operability.

[0085] Example 2, Figure 2 The present invention provides a vegetation coverage prediction system based on multi-source data fusion, which includes a data acquisition module, a frequency analysis module, a disturbance generation module, a model construction module, and a prediction correction module:

[0086] Data acquisition module, used to obtain historical vegetation cover data and land use type data, and obtain matching time series of land use transition and vegetation restoration lag;

[0087] Frequency analysis module, used to extract frequency bands from matching time series based on Fourier transform to obtain the initial change frequency of matching degree;

[0088] A disturbance generation module, configured to randomly perturb the initial change frequency of the matching degree based on a preset first constraint condition to obtain a plurality of different matching degree change frequencies;

[0089] The model building module is used to obtain vegetation cover degradation impact data and ecological restoration potential data corresponding to different matching degree change frequencies, and to build an evaluation model;

[0090] The prediction correction module is used to correct the vegetation coverage prediction value of the preset first prediction model according to the data output by the evaluation model.

[0091] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0092] The above embodiments may be implemented in whole or in part through software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0093] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0094] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0095] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0096] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A vegetation coverage prediction method based on multi-source data fusion, characterized in that: The following steps are involved: Obtain historical vegetation cover data and land use type data, and obtain matching time series of land use transition and vegetation restoration lags; Construct land use type transfer matrix; Based on the land use type transition matrix, the trace of the area change of each land use type in adjacent periods is calculated to obtain the land use transition rate time series; Construct the time-lagged cross-correlation function between the land use area change rate and vegetation cover change rate of each land use type; Obtain the time delay when the time-lag cross-correlation function first reaches a peak, and use the time delay as the lag period; The lag period length is obtained and the vegetation recovery lag rate series is calculated. The land use transition rate time series and the vegetation recovery lag rate series are then aligned using the cubic spline interpolation method to obtain a matching time series. Randomly perturb the initial change frequency of the matching degree based on the preset first constraint condition to obtain a number of different matching degree change frequencies; Based on Fourier transform, the frequency band of the matching time series is extracted to obtain the initial change frequency of the matching degree, which is specifically: A sliding window mechanism is used to Fourier transform the matching degree time series, and the main frequency band whose amplitude ratio exceeds the preset threshold in each window is extracted; Calculate the frequency mean of the main frequency band and use it as the initial change frequency; Obtain vegetation cover degradation impact data and ecological restoration potential data corresponding to different matching degree change frequencies, and construct an assessment model; The vegetation coverage prediction value of the preset first prediction model is modified according to the data output by the evaluation model, specifically: Based on the evaluation model, the matching degree change evaluation value is obtained; In the three-dimensional coordinate system, a three-dimensional prediction surface is constructed with the matching degree change frequency, matching degree change evaluation value, and prediction coverage as coordinate axes respectively; Calculate the feature points on the three-dimensional prediction surface where the gradient change rate exceeds the threshold, and obtain a feature point set; Cluster the feature point set to obtain several mutation areas; The correction weight is calculated according to the area of ​​the mutation region, and the predicted value of vegetation coverage is corrected according to the correction weight.

2. The vegetation coverage prediction method based on multi-source data fusion according to claim 1 is characterized in that: The initial change frequency of the matching degree is randomly perturbed based on the preset first constraint condition to obtain several different matching degree change frequencies, specifically: Calculate the maximum vegetation recovery hysteresis rate, land use transformation rate and average hysteresis period, and establish the first constraint condition; According to the first constraint, obtain the disturbance amplitude of the Poisson distribution; Use Poisson distribution for random perturbation to obtain several random perturbation terms within the perturbation amplitude; The random perturbation term is added to the initial change frequency of the matching degree to obtain several different matching degree change frequencies.

3. The vegetation coverage prediction method based on multi-source data fusion according to claim 2 is characterized in that: The vegetation cover degradation impact data includes vegetation cover degradation impact characteristics, and the specific method for obtaining the vegetation cover degradation impact characteristics is as follows: Obtain multi-temporal remote sensing NDVI data and calculate the vegetation cover change rate for each phase; Analyze the vegetation cover change rate data, identify vegetation degradation hotspots, and calculate the standard deviation and mean of the NDVI change rate in vegetation degradation hotspots; Based on the standard deviation and mean of the NDVI change rate, the degradation fluctuation value is obtained; Obtain climate data, calculate the ratio of precipitation during the drought period to the annual average precipitation, and obtain the drought frequency index; The impact characteristics of vegetation cover degradation were calculated based on the degradation fluctuation value and drought frequency index.

4. The vegetation coverage prediction method based on multi-source data fusion according to claim 3 is characterized in that: The ecological restoration potential data includes ecological restoration potential characteristics, and the specific method for obtaining the ecological restoration potential characteristics is as follows: Obtain the species distribution database and calculate the overlap between historical vegetation coverage and species suitable habitat to obtain the first potential value; Obtain remote sensing precipitation data and calculate the water collection rate corresponding to different slopes; The mean and variance of the water collection rate were calculated, and the ratio of the mean and variance of the water collection rate was used as the terrain wetness index; The ecological restoration potential characteristics were calculated based on the first potential value and terrain wetness index.

5. The vegetation coverage prediction method based on multi-source data fusion according to claim 4 is characterized in that: The specific formula of the first constraint condition is as follows: ， in, is the frequency disturbance, is the initial change frequency of matching degree, is the maximum vegetation recovery hysteresis rate, is the land use transition rate, is the average lag period.

6. A multi-source data fusion vegetation coverage prediction system, applied to a multi-source data fusion vegetation coverage prediction method according to any one of claims 1 to 5, characterized in that: It includes data acquisition module, frequency analysis module, disturbance generation module, model building module and prediction correction module: Data acquisition module, used to obtain historical vegetation cover data and land use type data, and obtain matching time series of land use transition and vegetation restoration lag; Frequency analysis module, used to extract frequency bands from matching time series based on Fourier transform to obtain the initial change frequency of matching degree; A disturbance generation module, configured to randomly perturb the initial change frequency of the matching degree based on a preset first constraint condition to obtain a plurality of different matching degree change frequencies; The model building module is used to obtain vegetation cover degradation impact data and ecological restoration potential data corresponding to different matching degree change frequencies, and to build an evaluation model; The prediction correction module is used to correct the vegetation coverage prediction value of the preset first prediction model according to the data output by the evaluation model.

Images