Grassland ecological assessment early warning monitoring system based on artificial intelligence

Abstract

The invention relates to the technical field of ecological monitoring, and discloses a grassland ecological assessment early warning monitoring system based on artificial intelligence. Collecting vegetation indexes, soil parameters, livestock tracks and topographic data in real time through satellite remote sensing, an unmanned aerial vehicle and Internet of Things equipment; an intelligent algorithm is adopted to process missing values and standardized data, and vegetation phenological characteristics are extracted; the weight is dynamically calculated by fusing the index volatility, the correlation and the ecological coordination rule, and the net influence of the grazing activity on the ecology is accurately quantified based on the contrast area with strict geographical constraints, so that a degradation risk score is generated; high, medium and low risk areas are divided according to scores, and a livestock capacity optimization scheme and targeted measures of a matching policy library are dynamically output. Natural interference and artificial grazing influence are accurately separated, the problem that the risk level is misjudged in a traditional method is solved, and the evaluation adaptability is improved through a dynamic weight model.

Description

Technical Field

[0001] This invention relates to the field of ecological monitoring technology, and in particular to an artificial intelligence-based grassland ecological assessment, early warning and monitoring system. Background Technology

[0002] The current grassland ecological assessment and early warning system faces three major technical bottlenecks: First, the data collection dimension is singular, mainly relying on the isolated operation of satellite remote sensing and ground sensors, lacking a multi-platform collaborative collection mechanism, resulting in incomplete identification of vegetation phenological characteristics; second, the assessment model is rigid, using preset weights to calculate ecological indicators, and cannot dynamically respond to the spatiotemporal variation characteristics of soil parameters and microclimate data; third, the quantitative impact of grazing is distorted, because the lack of a strictly matched ecological control area screening standard leads to the incorrect attribution of natural factor interference to grazing activities, resulting in misjudgment of risk levels.

[0003] In existing technologies, static threshold early warning mechanisms have significant drawbacks: when satellites detect abnormal fluctuations in NDVI, the system only triggers an early warning by comparing historical averages. It neither extracts phenological time-series features through LSTM networks nor calculates the dynamic weights and synergistic relationships between indicators, directly leading to two misjudgments: first, it misidentifies temporary vegetation wilting caused by short-term climate fluctuations as grassland degradation caused by overgrazing; second, it ignores the nonlinear relationship between soil organic matter content and carrying capacity, and underestimates the lag effect of ecological restoration in continuously grazing areas. Summary of the Invention

[0004] The technical problem to be solved by this invention is the shortcomings of the existing static threshold early warning mechanism. To address this, we propose an artificial intelligence-based grassland ecological assessment, early warning and monitoring system.

[0005] To achieve the above objectives, this application adopts the following technical solution: an artificial intelligence-based grassland ecological assessment, early warning and monitoring system, including a data acquisition module, a data processing module, a dynamic assessment module and an early warning decision module; Data acquisition module: Collects vegetation index, soil parameters, meteorological data, and livestock trajectories through satellite remote sensing interface, UAV sensor network and IoT devices, and integrates terrain slope and historical carrying capacity to build a benchmark database; Data processing module: KNN is used to impute missing values, the data is standardized by the range method, and LSTM network is used to extract vegetation phenological characteristics; Dynamic evaluation module: This module integrates indicator dispersion, correlation, and ecological synergy rules to calculate dynamic weights, and quantifies the net impact of grazing based on the control difference method. Generate a degradation risk score ; Early warning decision module: based on Divide the areas into high, medium, and low risk zones, and output optimized carrying capacity plans and matching policies.

[0006] Preferably, data collection includes: Satellite data acquisition: Obtain the normalized vegetation index of Modis, and trigger high-frequency acquisition when the daily precipitation is greater than the historical 95th percentile value; UAV data collection: The UAV scans twice a week at a height of 120±12 meters and an image overlap rate of more than 80%, and outputs a vegetation cover raster map. Ground data acquisition: Soil sensors monitor pH values ​​at a frequency of 10 minutes per cycle, and livestock collars record the coordinates of grazing sites.

[0007] Preferably, the data processing module uses the KNN algorithm to fill in missing values ​​in the satellite data using historical means of similar terrain units. For positive indicators, such as the vegetation index, For negative indicators, such as soil erosion rate , Where x is the original value, For the standard result, x comes from the raw monitoring value of the data acquisition module. and These are the maximum and minimum values ​​in the historical statistics.

[0008] Preferably, in the dynamic evaluation module, the formula for calculating the index dispersion is: , This is used to measure the degree of fluctuation of the j-th indicator in the sample. The higher the dispersion, the higher the sensitivity of the indicator to ecological changes, and the greater the weight should be. This represents the j-th index value of the i-th sample, such as the vegetation index value of 5 adjacent grassland units. This represents the sample mean of indicator j, where m is the sample size, such as the number of grassland units. This data originates from the raw monitoring values ​​in the data acquisition module; the correlation coefficient between indicators is calculated. , The linear correlation between quantifiable indicators j and k is used to avoid redundant calculation of highly correlated indicators during weight allocation. This is represented by the Pearson correlation coefficient between indicators j and k. The closer the absolute value is to 1, the stronger the correlation; generating the final weights. ,in The synergy strength coefficient amplifies the weight of indicators with ecological synergy relationships. This indicates the synergistic relationship between indicators, with 1 representing positive synergy, -1 representing negative synergy, and 0 representing no relationship. Defined by the expert knowledge base.

[0009] Preferably, the control area selected for quantifying the net impact of grazing meets the following criteria: For example, the grazing area is less than 5km away, the slope difference is less than 3°, the soil type is the same, and the grazing ban period is more than 3 years; Calculate the influence coefficient: ,in and The values ​​are for the grazing area and the control area, respectively.

[0010] Preferred degradation risk score ,in As weight, This represents the grazing impact coefficient.

[0011] Preferably, the risk zoning plan is as follows: for It has been designated as a high-risk area, with grazing prohibited and ecological restoration initiated. for These areas are designated as medium-risk areas, and livestock numbers are reduced by 20%-50%. for The area was designated as a low-risk zone, and livestock numbers were reduced while rotational grazing was implemented, with the rotational grazing period lasting 1.5 times the vegetation restoration period.

[0012] Preferred strategy for dynamic optimization of carrying capacity: For high-risk areas, For low-risk areas ,in The baseline carrying capacity is determined by historical data or policy.

[0013] The preferred policy matching method is: S1. Constructing the feature vector of ecological shortcomings: Identify the ecological weaknesses that require the most intervention, such as those with a high weighting for soil erosion rate; S2. Calculate policy text similarity: Match the policy entries in the policy library that are most relevant to the shortcomings, where P is the feature vector of the policy text, and extract the keyword weights through NLP; S3. Output satisfies Policy entries.

[0014] Preferably, the biomass is remeasured by drone after 8 weeks of adjustment of the carrying capacity. If the response rate Update the synergy coefficients according to the following rules: .

[0015] The technical effects and advantages of this invention are as follows: In this invention, a spatiotemporal database is constructed through collaborative data collection from satellites, drones, and ground sensors. Combined with dynamic weight calculation and a strict control area screening mechanism, this invention solves the problems of single data dimension, fixed model, and misjudgment of grazing impact in traditional methods. It has the advantages of improving the accuracy of grassland degradation risk identification, dynamically responding to spatiotemporal changes in ecological indicators, and reducing the misjudgment rate of grazing impact. Attached Figure Description

[0016] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts: Figure 1 This is a schematic diagram of the modules of the present invention. Detailed Implementation

[0017] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0018] Reference Figure 1 As shown, the present invention provides a technical solution: a grassland ecological assessment, early warning and monitoring system based on artificial intelligence, including a data acquisition module, a data processing module, a dynamic assessment module and an early warning decision module; Data acquisition module: Collects vegetation index, soil parameters, meteorological data, and livestock trajectories through satellite remote sensing data interface, UAV remote sensing and IoT devices, and integrates terrain slope and historical carrying capacity to build a benchmark database; Data processing module: KNN is used to impute missing values, the data is standardized by the range method, and LSTM network is used to extract vegetation phenological characteristics; Dynamic evaluation module: This module integrates indicator dispersion, correlation, and ecological synergy rules to calculate dynamic weights, and quantifies the net impact of grazing based on the control difference method. Generate a degradation risk score ; Early warning decision module: based on Divide the areas into high, medium, and low risk zones, and output optimized carrying capacity plans and matching policies.

[0019] Data collection includes: Satellite data acquisition: Real-time acquisition of Modis' normalized vegetation index. High-frequency acquisition is triggered when the daily precipitation is greater than the historical 95th percentile. The historical 95th percentile refers to the probability distribution threshold calculated based on long-term precipitation observation data. Specifically, the sliding window method can be used to calculate the quantile of precipitation data over the past ten years, which is used to dynamically judge extreme precipitation events. UAV data collection: The UAV scans twice a week at a height of 120±12 meters and an image overlap rate of more than 80%. It is equipped with a multispectral sensor for mesoscale monitoring and outputs a vegetation cover raster map. Ground data acquisition: Soil sensors monitor pH values ​​at a frequency of 10 minutes per cycle, and livestock collars record the coordinates of grazing sites.

[0020] The data processing module uses the KNN algorithm to fill in missing values ​​in satellite data using historical means of similar terrain units. For positive indicators, such as vegetation index, For negative indicators, such as soil erosion rate , Where x is the original value, For the standard result, x comes from the raw monitoring value of the data acquisition module. and To determine the maximum and minimum values ​​of historical statistics, during the data preprocessing stage, standardization operations are performed on ecological indicators with different attributes. For positive indicators, each data point is scaled to the 0-1 range by calculating the difference between the minimum and maximum values ​​of the original data. For negative indicators, a reverse scaling method is used, with the difference between the maximum value of the original data and the current value as the numerator, keeping the scaled value still within the 0-1 range. Through this processing method, ecological indicators with different dimensions and different directions of change are uniformly converted into positively comparable values, providing a consistent data foundation for subsequent dynamic weight calculations.

[0021] In the dynamic evaluation module, the formula for calculating the index dispersion is: , This is used to measure the degree of fluctuation of the j-th indicator in the sample. The higher the dispersion, the higher the sensitivity of the indicator to ecological changes, and the greater the weight should be. This represents the j-th index value of the i-th sample, such as the vegetation index value of 5 adjacent grassland units. This represents the sample mean of indicator j, where m is the sample size, such as the number of grassland units. This data originates from the raw monitoring values ​​in the data acquisition module; the correlation coefficient between indicators is calculated. , The linear correlation between quantifiable indicators j and k is used to avoid redundant calculation of highly correlated indicators during weight allocation. This is represented by the Pearson correlation coefficient between indicators j and k. The closer the absolute value is to 1, the stronger the correlation; generating the final weights. ,in The synergy strength coefficient amplifies the weight of indicators with ecological synergy relationships. This indicates the synergistic relationship between indicators, with 1 representing positive synergy, -1 representing negative synergy, and 0 representing no relationship. Defined by an expert knowledge base; specifically, when calculating the weights of each indicator, the importance of the indicator itself is first determined by calculating the standard deviation, then the weight ratio of multicollinear indicators is reduced by the product term of the correlation coefficient, and finally the weight of indicators with ecological synergistic or antagonistic relationships is corrected by combining the synergistic rules; compared with the existing technology, the existing method adopts a fixed weight allocation method, which cannot dynamically adjust the weights according to the data distribution characteristics and the relationship between indicators, resulting in the evaluation results deviating from the actual ecological situation. This method dynamically integrates dispersion, correlation and synergistic rules, so that the weight allocation can respond in real time to the changing trend and interaction relationship of ecological indicators.

[0022] To quantify the net impact of grazing, control areas meeting the following criteria were selected: For example, the grazing area is less than 5km away, the slope difference is less than 3°, the soil type is the same, and the grazing ban period is more than 3 years. The influence coefficient is generated by calculating the difference in indicators between the grazing area and the control area. When the indicator value of the control area is non-zero, the calculation method is to subtract the indicator value of the control area from the indicator value of the grazing area and then divide by the absolute value of the indicator value of the control area. When the indicator value of the control area is zero, the influence coefficient is zero. Calculate the influence coefficient: ,in and The values ​​are for the grazing area and the control area, respectively.

[0023] Degradation risk score ,in As weight, The grazing impact coefficient is a parameter reflecting the importance of different ecological indicators in degradation assessment. It can be calculated using a combination of standard deviation and correlation coefficient. Dispersion measures the volatility of indicator data, correlation coefficient eliminates redundant indicators, and ecological synergy rules strengthen the role of key indicators. This ensures that the weight allocation conforms to the dynamic characteristics of the actual ecosystem. The grazing impact coefficient refers to the net impact of grazing activities on ecological indicators after excluding natural factors. It can be calculated using a geographically matched control area difference. By selecting a grazing-prohibited area with similar topography and soil conditions as a benchmark, the interference of environmental variables such as precipitation and slope on the assessment results is eliminated, thus accurately separating... The study examines the independent impacts of grazing activities on vegetation cover and soil quality. Specifically, when calculating the degradation risk score, the dispersion of each ecological indicator is first calculated based on its data distribution characteristics. Correlation analysis between indicators is then used to screen out assessment factors with strong independence. The weight allocation ratio is adjusted through collaborative rules set by ecological experts. Simultaneously, by comparing the differences in ecological indicators between grazing areas and grazing-prohibited control areas, the interference of environmental factors on the assessment results is eliminated. Finally, the dynamic weights and net impact coefficients are weighted and summed to form a comprehensive scoring index that reflects the cumulative impact of human grazing activities on the ecosystem. This scoring mechanism dynamically adjusts the weight parameters, enabling the assessment model to adapt to the ecological change characteristics of different seasons and regions.

[0024] The risk zoning plan is as follows: for It has been designated as a high-risk area, with grazing prohibited and ecological restoration initiated. for These areas are designated as medium-risk areas, and livestock numbers are reduced by 20%-50%. for The area was designated as a low-risk zone, and livestock numbers were reduced while rotational grazing was implemented, with the rotational grazing period lasting 1.5 times the vegetation restoration period.

[0025] Strategies for dynamic optimization of carrying capacity: For high-risk areas, For low-risk areas ,in The baseline carrying capacity is determined by historical data or policy. After the system calculates the degradation risk score for each region, it automatically classifies the risk level according to predefined interval thresholds. For high-risk areas, a grazing ban order is immediately triggered and the ecological restoration solution library is invoked. For medium-risk areas, control parameters limiting grazing intensity are generated using the carrying capacity calculation formula. For low-risk areas, a grassland rotation plan is output based on the rotational grazing cycle extension rules.

[0026] The policy matching method is as follows: S1. Constructing the feature vector of ecological shortcomings: Identify the ecological weaknesses that require the most intervention, such as those with a high weighting for soil erosion rate; S2. Calculate policy text similarity: Match the policy entries in the policy library that are most relevant to the shortcomings, where P is the feature vector of the policy text, and extract the keyword weights through NLP; S3. Output satisfies The policy entries are calculated based on dynamic weights and grazing impact coefficients to generate feature vectors. The components of each dimension are normalized to eliminate dimensional differences. Then, keywords are extracted and vectorized from each policy text in the policy library to form a set of policy feature vectors. By calculating the cosine similarity between the ecological weakness vector and each policy vector, target policies with a matching degree exceeding a preset threshold are selected.

[0027] Eight weeks after the livestock carrying capacity adjustment period, the biomass was remeasured by drone. If the response rate Update the synergy coefficients according to the following rules: Specifically, eight weeks after the livestock carrying capacity adjustment, the biomass of the target area was remeasured using a drone equipped with a multispectral sensor to obtain measured biomass data. The measured values ​​were compared with the initial biomass before adjustment and the model prediction values ​​to calculate the recovery rate index. When the recovery rate was lower than the preset threshold of 0.8, it indicated that the current model parameters failed to accurately reflect the ecological restoration law. At this time, the synergy strength coefficient was dynamically attenuated and adjusted according to the formula. This adjustment process was completed by multiplying the original coefficient by a correction factor that included the recovery rate parameter, ensuring that the model reduced its dependence on the failure synergy rules in subsequent calculations.

[0028] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A grassland ecological assessment, early warning, and monitoring system based on artificial intelligence, characterized in that, It includes a data acquisition module, a data processing module, a dynamic evaluation module, and an early warning decision-making module; Data acquisition module: Collects vegetation index, soil parameters, and livestock trajectories through satellite remote sensing data interface, UAV remote sensing and IoT devices, and integrates meteorological data, terrain slope and historical carrying capacity to build a benchmark database; Data processing module: KNN is used to impute missing values, the data is standardized by the range method, and LSTM network is used to extract vegetation phenological characteristics; Dynamic evaluation module: This module integrates indicator dispersion, correlation, and ecological synergy rules to calculate dynamic weights, and quantifies the net impact of grazing based on the control difference method. Generate a degradation risk score ; Early warning decision module: based on Divide the areas into high, medium, and low risk zones, and output optimized carrying capacity plans and matching policies.

2. The grassland ecological assessment, early warning, and monitoring system based on artificial intelligence according to claim 1, characterized in that: The data collection includes: Satellite data acquisition: Obtain the normalized vegetation index of Modis, and trigger high-frequency acquisition when the daily precipitation is greater than the historical 95th percentile value; UAV data collection: The UAV scans twice a week at a height of 120±12 meters and an image overlap rate of more than 80%, and outputs a vegetation cover raster map. Ground data collection: Soil sensors monitor moisture content at a frequency of 10 minutes per cycle, and livestock collars record the coordinates of grazing sites.

3. The grassland ecological assessment, early warning, and monitoring system based on artificial intelligence according to claim 1, characterized in that: The data processing module uses the KNN algorithm to fill in missing values ​​in satellite data using historical averages of similar terrain units. For positive indicators, For negative indicators, such as soil erosion rate , Where x is the original value, For the standard result, x comes from the raw monitoring value of the data acquisition module. and These are the maximum and minimum values ​​in the historical statistics.

4. The grassland ecological assessment, early warning, and monitoring system based on artificial intelligence according to claim 1, characterized in that: In the dynamic evaluation module, the formula for calculating the index dispersion is as follows: , This is used to measure the degree of fluctuation of the j-th indicator in the sample. The higher the dispersion, the higher the sensitivity of the indicator to ecological changes, and the greater the weight should be. This represents the j-th index value of the i-th sample, such as the vegetation index value of 5 adjacent grassland units. This represents the sample mean of indicator j, where m is the sample size, such as the number of grassland units. This data originates from the raw monitoring values ​​in the data acquisition module; the correlation coefficient between indicators is calculated. , The linear correlation between quantifiable indicators j and k is used to avoid redundant calculation of highly correlated indicators during weight allocation. This is represented by the Pearson correlation coefficient between indicators j and k. The closer the absolute value is to 1, the stronger the correlation. Generate final weights ,in The synergy strength coefficient amplifies the weight of indicators with ecological synergy relationships. This indicates the synergistic relationship between indicators, with 1 representing positive synergy, -1 representing negative synergy, and 0 representing no relationship. Defined by the expert knowledge base.

5. The grassland ecological assessment, early warning, and monitoring system based on artificial intelligence according to claim 4, characterized in that: The net impact quantification of grazing selected control areas that met the following conditions: For example, the grazing area is less than 5km away, the slope difference is less than 3°, the soil type is the same, and the grazing ban period is more than 3 years; Calculate the influence coefficient: ,in and The values ​​are for the grazing area and the control area, respectively.

6. The grassland ecological assessment, early warning, and monitoring system based on artificial intelligence according to claim 5, characterized in that: The degradation risk score ,in As weight, This represents the grazing impact coefficient.

7. The grassland ecological assessment, early warning, and monitoring system based on artificial intelligence according to claim 1, characterized in that: The risk zoning plan is specifically as follows: for It has been designated as a high-risk area, with grazing prohibited and ecological restoration initiated. for These areas are designated as medium-risk areas, and livestock numbers are reduced by 20%-50%. for The area was designated as a low-risk zone, and livestock numbers were reduced while rotational grazing was implemented, with the rotational grazing period lasting 1.5 times the vegetation restoration period.

8. The grassland ecological assessment, early warning, and monitoring system based on artificial intelligence according to claim 7, characterized in that: The strategy for dynamically optimizing carrying capacity is as follows: For high-risk areas, For low-risk areas ,in The baseline carrying capacity is determined by historical data or policy.

9. The grassland ecological assessment, early warning, and monitoring system based on artificial intelligence according to claim 1, characterized in that: The policy matching method is as follows: S1. Constructing the feature vector of ecological shortcomings: Identify the ecological weaknesses that most urgently need intervention; S2. Calculate policy text similarity: Match the policy entries in the policy library that are most relevant to the shortcomings, where P is the feature vector of the policy text, and extract the keyword weights through NLP; S3. Output satisfies Policy entries.

10. The grassland ecological assessment, early warning, and monitoring system based on artificial intelligence according to claim 1, characterized in that: Eight weeks after the livestock carrying capacity adjustment period, the biomass was remeasured by drone. If the response rate Update the synergy coefficients according to the following rules: .

Images