Ecosystem monitoring method and system based on large model

Abstract

The present application provides an ecosystem monitoring method and system based on a large model, the method comprising: analyzing ecological data based on a macro-ecological model, multiple single-species distribution models, and a joint species distribution model, and outputting prediction content; fusing the output content of the three types of models with the linkage of environmental data and geographic location, and processing the fusion results using sequential Bayesian filtering to obtain an estimate of the target state of species in the ecological protection area; aligning the estimated value of the target state of species and other multi-source monitoring data according to timestamps, and converting them into input data sets, inputting the target time series and key data features included in the input data set into the target large model according to a preset prompt template, and using the target large model to predict the ecological status in the ecological protection area to monitor the ecological protection area. The present application can make full use of multi-dimensional data and multiple categories of models to predict the ecological status of species and ensure the reliability and accuracy of the prediction.

Description

Technical Field

[0001] The present application relates to the field of ecological technology, and in particular to a large-scale model-based ecosystem monitoring method and system. Background Art

[0002] Ecology, as a scientific discipline that studies the interaction between organisms and the environment, focuses on biodiversity, ecosystem structure and function, and human impact on the environment. In order to protect the ecological environment and maintain biodiversity, it is necessary to monitor and manage the ecological environment of ecological reserves in real time. By monitoring the ecological environment in the reserve in real time, we can understand the composition, quantity, distribution and ecological function of organisms in the reserve; by managing the reserve, we can maintain the stability of the ecological environment.

[0003] At present, when conducting ecological and environmental monitoring in protected areas, analysis is often based on a single data source, which can easily lead to information loss and affect monitoring results. In addition, there are few methods in existing technologies that can accurately predict species distribution and habitat suitability, which to some extent limits ecological and environmental protection work. Summary of the invention

[0004] In view of the above problems, the embodiments of the present application provide a large-model-based ecosystem monitoring method and system that overcome the above problems or at least partially solve the above problems.

[0005] In a first aspect, the present application provides an ecosystem monitoring method based on a large model, comprising:

[0006] Analyze ecological data in the ecological protection area based on the macro-ecological model to predict species richness, species evenness and species distribution changes at different observation points in the ecological protection area, wherein the ecological data includes environmental data and species distribution data;

[0007] Analyze the ecological data based on aggregated multiple single-species distribution models to predict comprehensive assessment information of species diversity corresponding to the ecological protection area;

[0008] Analyze the ecological data based on the joint species distribution model to predict the comprehensive information of the associated species community structure and its environmental response corresponding to the ecological protection area;

[0009] The prediction contents output by the three types of models are integrated with the linkage of environmental data and geographic location, and the integrated results are processed by sequential Bayesian filtering to obtain the estimated value of the target state of species in the ecological protection area;

[0010] Aligning the estimated value of the target state of species in the ecological protection zone and other multi-source monitoring data according to timestamps and converting them into an input data set, wherein the input data set includes: a target time series extracted from the estimated value of the target state of species in the ecological protection zone, and key data features extracted from other multi-source monitoring data;

[0011] The target time series and the key data features are input into the target macro model according to a preset prompt template, and the target macro model is used to obtain the predicted target time series corresponding to the species in the ecological protection zone, so as to monitor the ecological protection zone according to the predicted target time series.

[0012] In the second aspect, an embodiment of the present application provides an ecosystem monitoring system based on a large model, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein when the processor executes the computer program, the ecosystem monitoring method based on the large model described in the first aspect is implemented.

[0013] The technical solution of the embodiment of the present application analyzes ecological data based on a macro-ecological model, multiple single-species distribution models, and a joint species distribution model, combines the prediction content output by the three types of models with the linkage between environmental data and geographical location, and uses sequential Bayesian filtering for processing to obtain an estimate of the target state of species in the ecological protection zone, aligns the estimate of the target state of species with other multi-source monitoring data according to timestamps, and converts them into an input data set including target time series and key data features, inputs the target time series and key data features into the target large model according to a preset prompt template, and uses the target large model to analyze and predict the ecological status of species in the ecological protection zone in the next stage. By making full use of multi-dimensional data and multiple categories of prediction models, the ecological status of species can be predicted, the reliability and accuracy of the prediction can be guaranteed, and comprehensive, efficient and accurate monitoring of ecological protection zones can be achieved, providing support for ecological environmental protection work. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A schematic diagram showing an ecosystem monitoring method based on a large model provided in an embodiment of the present application;

[0015] Figure 2 A flowchart showing the overall implementation of the large model-based ecosystem monitoring method provided in an embodiment of the present application;

[0016] Figure 3 A schematic diagram showing an ecosystem monitoring device based on a large model provided in an embodiment of the present application;

[0017] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0019] It should be understood that the "one embodiment" or "an embodiment" mentioned throughout the specification means that the specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. The multiple in the embodiments of the present application can include two and more than two.

[0020] In the various embodiments of the present application, it should be understood that the size of the serial numbers of the following processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0021] The present application embodiment provides an ecosystem monitoring method based on a large model, such as Figure 1 As shown, including:

[0022] Step 101: Analyze ecological data in the ecological protection area based on the macro-ecological model to predict species richness, species evenness and species distribution changes at different observation points in the ecological protection area. The ecological data includes environmental data and species distribution data.

[0023] Data collection is conducted on ecological protection areas within a set period of time to obtain ecological data within the ecological protection areas during that period. The collected ecological data include environmental data and species distribution data, where environmental data include at least temperature, humidity and soil pH; species distribution data include species occurrence frequency and geographical location. The species occurrence frequency can be understood as the number of occurrences or observations of a species at a specific location. Since species distribution data can provide information on the distribution of species under different environmental conditions, it is crucial for the construction of species distribution models.

[0024] When collecting environmental data, temperature sensors and humidity sensors are used to collect temperature data and humidity data from different observation points in the ecological protection area; soil pH sensors are used to measure the acidity and alkalinity of the soil at different observation points in the ecological protection area to assess the health of the soil; environmental data collection can be periodic or real-time, such as using various sensors to monitor the temperature, humidity, and soil conditions at different observation points in the ecological protection area in real time, or controlling sensors to collect data periodically.

[0025] When collecting species distribution data, it is necessary to use imaging equipment, such as using infrared cameras to capture the activities of animals in ecological reserves, and using high-resolution cameras mounted on drones to obtain the geographic information and vegetation coverage of ecological reserves in order to analyze the frequency of species occurrence and their geographical locations.

[0026] After obtaining environmental data and species distribution data based on data collection, the data is preprocessed, including data cleaning, data standardization, and data conversion. In data cleaning, outliers and missing values ​​are removed, for example, data points with abnormally high or low temperature sensor readings are excluded; in data standardization, the data provided by the sensor is converted to a unified metric, for example, the temperature is converted from Celsius to Fahrenheit; in data conversion, non-numeric data (such as soil type) is converted to numerical data for further analysis.

[0027] After completing data preprocessing, applicable ecological data are obtained, and the ecological data are input into the pre-constructed macro-ecological model. The macro-ecological model is used to analyze the ecological data and predict the species richness, species evenness and species distribution changes at different observation points in the ecological protection area. Among them, the macro-ecological model can analyze α-diversity and β-diversity. α-diversity is an indicator to measure the species richness and evenness in a specific location, and β-diversity is an indicator to measure the difference in species composition between different locations. The species turnover between different sites is estimated by analyzing α-diversity and β-diversity. That is, the macro-ecological model analyzes the input ecological data, predicts the species richness and species evenness of different observation points in the ecological protection area, and predicts the species turnover between different sites based on the species richness and species evenness of different observation points, that is, analyzes the distribution changes of species between different locations.

[0028] Step 102: Analyze ecological data based on the aggregated multiple single-species distribution models to predict comprehensive assessment information of species diversity corresponding to the ecological protection area.

[0029] After preprocessing the collected data and obtaining applicable ecological data, the ecological data were analyzed based on multiple aggregated single species distribution models. Among them, the single species distribution model (s-SDM) is a model used in ecology to describe the relationship between the geographical distribution of a species and its environmental factors, and is used to predict the distribution of species under different environmental conditions.

[0030] After analyzing the ecological data based on multiple single-species distribution models, the comprehensive assessment of species diversity in the ecological protection area can be obtained by aggregating the output contents of multiple single-species distribution models, that is, the comprehensive assessment information of species diversity corresponding to the ecological protection area can be predicted. The predicted comprehensive assessment information on diversity includes the number of species and their uniformity of distribution in the area, so as to obtain the overall diversity index of the region.

[0031] Among them, the macroecological model focuses on comparing the differences in species diversity between different locations, which helps to identify biodiversity hotspots and conservation priority areas; the single species distribution model focuses on predicting and understanding the distribution of specific species under different environmental conditions; through the combined use of these two models, we can more comprehensively understand and evaluate the biodiversity status in protected areas, and provide a basis for the management and protection of protected areas.

[0032] Step 103: Analyze the ecological data based on the joint species distribution model to predict the comprehensive information of the associated species community structure and its environmental response corresponding to the ecological protection area.

[0033] After preprocessing the collected data and obtaining applicable ecological data, the ecological data were analyzed based on Joint Species Distribution Models (JSDMs), which are used to simulate the impact of single species and communities, including common responses between species and other information, such as phylogenetic relationships, characteristics, and spatiotemporal dynamics.

[0034] After analyzing ecological data based on the joint species distribution model, the comprehensive information of the associated species community structure and its environmental response in the ecological protection area can be predicted, thereby analyzing the interactions between different species in the protection area and how they jointly respond to environmental changes.

[0035] Specifically, when predicting the comprehensive information of related species community structure and its environmental response, the ecological data are input into the joint species distribution model, and the output includes comprehensive information including the interaction relationship between species, environmental driving factors of species distribution, species co-occurrence patterns, species richness and evenness, and species distribution, so as to realize the prediction of comprehensive information.

[0036] Among them, species interactions can reveal the interactions between species, such as competition and symbiosis; environmental drivers of species distribution can identify key environmental variables that affect species distribution; species co-occurrence patterns can analyze the spatial co-occurrence patterns of species and how these patterns are affected by environmental factors; species richness and evenness can provide estimates of species richness and evenness, which reflect the diversity and stability of the community; species distribution can reflect the potential distribution of species under different environmental conditions.

[0037] Through the above outputs, the joint species distribution model can provide important scientific basis for the monitoring and management of protected areas, help protected area managers understand the structure and dynamics of species communities, assess the impact of environmental changes on species distribution, and formulate corresponding conservation strategies.

[0038] Step 104: Fuse the prediction contents output by the three types of models with the linkage between environmental data and geographic location, and process the fusion results using sequential Bayesian filtering to obtain an estimated value of the target status of species in the ecological protection area.

[0039] The fusion of the output content of the three types of models can include information such as predicted species distribution, species richness, species diversity index, etc. This information can be used to construct a comprehensive species distribution information, such as a comprehensive species distribution map, which is used to show the distribution of different species under different environmental conditions.

[0040] The linkage between environmental data and geographic location (also known as the function that characterizes the difference between environmental and geographic distance) is used to describe how environmental variables change with geographic location. It can involve a variety of different functions or models, depending on the relationship between the environmental variables and geographic location being studied. Among them, the linkage between environmental data and geographic location can be expressed as:

[0041] 1. Relationship between environmental variables and geographical location: describes how environmental variables (such as temperature, humidity, soil pH, etc.) change with changes in geographical location (such as longitude and latitude). Geographic distance can be used as an independent variable and environmental variables as dependent variables.

[0042] 2. Spatial autocorrelation: In ecology and geography, spatial autocorrelation means that places close in space tend to have similar environmental conditions. Functions that characterize differences in environment and geographic distance can be used to quantify spatial autocorrelation, that is, the degree of spatial correlation of environmental variables.

[0043] 3. Spatial interpolation methods: When processing geographic data, spatial interpolation methods are often needed to estimate the values ​​of environmental variables at unknown locations. Common spatial interpolation methods include Kriging, Inverse Distance Weighted (IDW), etc. These methods can be used to construct functions that characterize environmental and geographic distance differences.

[0044] 4. Geographically weighted regression (GWR): GWR is a spatial statistical method used to analyze the relationship between environmental variables and geographic location. It can be used to build local regression models to consider spatial heterogeneity, that is, the relationship between environmental variables and geographic location may be different in different locations.

[0045] When combining the outputs of the three types of models with the linkage between environmental data and geographic location, this can be achieved by calculating the correlation between species distribution and environmental variables, for example, using geographically weighted regression to analyze the spatial relationship between species distribution and environmental variables; through the combination of the above content, the relationship between environmental variables and geographic location can be quantified, which helps to more accurately predict species distribution and assess the health of ecosystems.

[0046] By combining the predictions output by the three types of models with the linkage between environmental data and geographic location, and using sequential Bayesian filtering, we can obtain estimates of the target status of species in ecological protection areas, providing more accurate and comprehensive information for the monitoring and management of protected areas. These estimates can also help protected area managers develop more effective conservation strategies, such as habitat restoration and the establishment of species migration corridors, to protect and restore biodiversity.

[0047] Step 105: align the estimated values ​​of the target status of species in the ecological protection zone and other multi-source monitoring data according to timestamps and convert them into an input data set, wherein the input data set includes: a target time series extracted based on the estimated values ​​of the target status of species in the ecological protection zone, and key data features extracted based on other multi-source monitoring data.

[0048] After obtaining the estimated value of the target state of species in the ecological protection area, it is aligned with other multi-source monitoring data corresponding to the ecological protection area according to the timestamp. Other multi-source monitoring data include image video data, remote sensing drone image data, geographic information data and environmental data collected for the ecological protection area. After data alignment, the target time series associated with the species is constructed based on the estimated value of the target state of species in the ecological protection area, and the key data features are determined based on other multi-source monitoring data. The obtained target time series and key data features are associated in the time dimension, so as to obtain an input data set including the target time series and key data features.

[0049] Step 106: Input the target time series and key data features into the target macro model according to the preset prompt template, and use the target macro model to obtain the predicted target time series corresponding to the species in the ecological protection area, so as to monitor the ecological protection area according to the predicted target time series.

[0050] After obtaining the input data set for model input, the target time series and key data features in the input data set are input into the target large model according to the preset prompt template. The target large model understands and analyzes the input content to predict the corresponding ecological status of species in the ecological protection area in the next stage.

[0051] Among them, when the target time series and key data features are input into the target big model and the target big model is used for prediction, specifically: the key data features are filled into the prompt template, and then the prompt template and target time series filled with data are input into the target big model, and the prompt template filled with data is used to assist the target big model to understand the target time series; based on the target big model, the target time series corresponding to the species in the ecological protection area in the next time period is predicted.

[0052] Based on the prompt of the prompt template, key data features are filled in the prompt template, and then the prompt template filled with data is input into the target big model. The target big model understands the need to understand the target time series based on the content filled in the prompt template. After understanding the target time series, the target time series related to the species in the ecological protection area in the next stage (such as the next 3 months) is predicted. The predicted target time series can feedback the ecological status of species in the ecological protection area, so as to realize the prediction of the ecological status of species in the ecological protection area using the target big model, and provide support for ecological environmental protection work.

[0053] Among them, after predicting the target time series related to the species in the ecological protection zone in the next stage based on the target big model, the predicted target time series can also be processed by the text-image big model to display the ecological status of the predicted species in the ecological protection zone in the form of images through text-image conversion.

[0054] The above implementation plan of the present application analyzes ecological data based on macro-ecological models, multiple single-species distribution models, and joint species distribution models, combines the prediction content output by the three types of models with the linkage between environmental data and geographical location, and uses sequential Bayesian filtering for processing to obtain the estimated value of the target state of species in the ecological protection zone, aligns the estimated value of the target state of species with other multi-source monitoring data according to timestamps, and converts them into an input data set including target time series and key data features, inputs the target time series and key data features into the target large model according to a preset prompt template, and uses the target large model to analyze and predict the ecological status of species in the ecological protection zone in the next stage. By making full use of multi-dimensional data and multiple categories of prediction models, the ecological status of species can be predicted to ensure the reliability and accuracy of the prediction, thereby realizing comprehensive, efficient and accurate monitoring of ecological protection zones and providing support for ecological environmental protection work.

[0055] The process of model building is introduced below. In an optional embodiment, the method further includes:

[0056] Collect ecological sample data from ecological protection areas, and pre-process the collected data to determine the training sample data set;

[0057] A macro-ecological model is constructed based on the training sample data set, wherein the macro-ecological model includes a first sub-model for identifying species richness and species evenness at each observation point and a second sub-model for identifying species distribution changes at different observation points;

[0058] Separating training sample subsets corresponding to multiple species from the training sample data set, and constructing matching single-species distribution models based on the training sample subsets corresponding to each species to determine multiple single-species distribution models;

[0059] A joint species distribution model for comprehensive analysis of community ecological data was constructed based on the training sample dataset.

[0060] Before using macroecological models, single species distribution models, and joint species distribution models to analyze ecological data, it is necessary to collect ecological sample data for the current ecological protection area, such as collecting data within one year, and preprocessing the collected data to determine the training sample data set for model construction.

[0061] A macroecological model is constructed based on the determined training sample data set, wherein the macroecological model includes a first sub-model (α-diversity sub-model) for identifying species richness and species evenness at each observation point and a second sub-model (β-diversity sub-model) for identifying species distribution changes at different observation points. The α-diversity sub-model analyzes the distribution of species in different locations and provides estimates of species richness and evenness at each location; the β-diversity sub-model estimates species turnover between different sites (species distribution changes between different locations). Based on the coordination of the α-diversity sub-model and the β-diversity sub-model, the macroecological model analyzes species diversity and changes in species distribution between different locations. Among them, macroecological models are usually constructed using a top-down approach to analyze species diversity and species turnover in the entire ecological reserve. Top-down approaches usually start from a macroscopic perspective, considering the dynamics of the entire ecological region or community, and how environmental factors affect the structure and function of the entire community.

[0062] After determining the training sample data set, separate the training sample subsets corresponding to multiple species from the training sample data set, and build matching single-species distribution models based on the training sample subsets corresponding to each species, so as to build multiple single-species distribution models. Single-species distribution models are usually constructed using a bottom-up approach to analyze the distribution of a single species under different environmental conditions. The bottom-up approach starts from a single species or the interaction between species and gradually builds to the community or ecosystem level.

[0063] After determining the training sample data set, a joint species distribution model for comprehensive analysis of community ecological data is constructed based on the training sample data set. The construction of the joint species distribution model combines top-down and bottom-up methods, which are different methods and perspectives for describing model construction. When constructing a joint species distribution model, combining these two methods can more comprehensively simulate the impact of single species and communities.

[0064] By collecting data from ecological protection areas, training sample data sets are determined, and different types of prediction models are constructed based on the training sample data sets. In the model application stage, different types of models can be used to analyze the ecological data to be processed, so as to determine the prediction information associated with the ecological protection area through data prediction from different angles.

[0065] In an optional embodiment, in the model application stage, for the constructed multiple single species distribution models, when analyzing ecological data based on the aggregated multiple single species distribution models to predict the species diversity comprehensive assessment information corresponding to the ecological protection area, it includes:

[0066] Ecological data are divided based on species categories to determine the data to be analyzed corresponding to multiple species. For each species, the data to be analyzed corresponding to the species are input into the matching single-species distribution model to obtain the distribution of the species under different environmental conditions. The distribution of multiple species under different environmental conditions is aggregated to obtain the comprehensive assessment information on species diversity corresponding to the ecological protection area.

[0067] In the model application stage, when using the constructed multiple single-species distribution models to process the ecological data to be analyzed, the ecological data is first divided based on species categories to obtain the data to be analyzed corresponding to each species. For example, the ecological data is divided based on the 1,000 species corresponding to the ecological protection area. For each species, the data to be analyzed corresponding to the species is input into the adapted single-species distribution model, and the adapted single-species distribution model predicts the distribution of the species under different environmental conditions. After analyzing the ecological data based on multiple single-species distribution models to obtain the distribution of multiple species under different environmental conditions, the distribution of multiple species under different environmental conditions is aggregated to obtain the comprehensive assessment information of species diversity corresponding to the ecological protection area.

[0068] By aggregating the predicted distribution of each species, we can obtain the number of species in the ecological protection area and the uniformity of their distribution in the protection area to obtain the overall diversity index of the region.

[0069] The following is an introduction to the process of obtaining an estimated value of the target state of species in an ecological protection zone. Optionally, when the prediction content output by the three types of models is fused with the linkage of environmental data and geographic location, the fusion result is processed using sequential Bayesian filtering to obtain an estimated value of the target state of species in an ecological protection zone, including:

[0070] The prediction contents of the three types of model outputs are integrated to predict the comprehensive information of species distribution corresponding to the ecological protection area; the sequential Bayesian filter is used to integrate the linkage between the comprehensive information of species distribution, environmental data and geographical location to obtain the estimated value of the target status of species in the ecological protection area.

[0071] By integrating the prediction contents output by the three types of models, more comprehensive ecological information can be provided. Based on the more comprehensive ecological information provided (such as the predicted distribution of species, species richness, species diversity index, etc.), the comprehensive species distribution information corresponding to the ecological protection area can be predicted, such as a comprehensive species distribution map, which can be used to show the distribution of different species under different environmental conditions.

[0072] Specifically, the output of the joint species distribution model, the output of the macroecological model, and the output of multiple single-species distribution models can provide the following information:

[0073] 1. Interactions between species

[0074] Joint species distribution models can reveal interactions between species, such as competition, symbiosis, etc.; this information can be combined with detailed information about the distribution of individual species in single species distribution models to better understand how interactions between species affect the structure of the entire community.

[0075] 2. Environmental drivers of species distribution

[0076] Joint species distribution models can identify key environmental variables that influence species distributions, and these environmental drivers can be combined with species diversity and species turnover across ecological reserves analyzed in macroecological models to provide deeper ecological insights.

[0077] 3. Species co-occurrence patterns

[0078] Joint species distribution models can analyze the co-occurrence patterns of species in space and how these patterns are affected by environmental factors. These co-occurrence patterns can be combined with detailed information about the distribution of individual species in single species distribution models to better understand the spatial distribution patterns of species.

[0079] 4. Species richness and evenness

[0080] Joint species distribution models can provide estimates of species richness and evenness, which can be combined with species diversity and species turnover across ecological reserves analyzed in macroecological models to provide a more comprehensive assessment of biodiversity.

[0081] Among them, when using sequential Bayesian filtering to integrate and analyze the linkage of species distribution information, environmental data and geographical location to obtain the estimated value of the target status of species in the ecological protection area, it includes:

[0082] Based on the comprehensive information of species distribution, the prior probability of species in the ecological protection area corresponding to different observation points is predicted. The prior probability indicates the probability of species existence.

[0083] For each observation point, the observation probability of the species being observed at the current observation point is calculated based on the geographical location and environmental data corresponding to the current observation point and the geographical location corresponding to the previous observation point;

[0084] Determine the posterior probability of the species based on the prior probability and observation probability corresponding to the species at the observation point;

[0085] According to the posterior probabilities corresponding to species at different observation points, the estimated values ​​of the target states of species in the ecological protection area are obtained.

[0086] Based on the fusion of the output content of the three types of models, the comprehensive information of species distribution corresponding to the ecological protection area can be predicted, and the prior probability of species in the ecological protection area at different observation points can be predicted based on the comprehensive information of species distribution, that is, the probability of species existing at different observation points can be predicted. Then, for each observation point, the likelihood function is used to calculate the observation probability of the species being observed at the current observation point. For the likelihood function in this embodiment, it is used to calculate the possibility of the species existing under given environmental conditions (temperature, humidity, soil state) and geographical location (latitude and longitude of the current and previous observation points) after the geographical location of the species changes (such as animal migration). The likelihood function in this embodiment is a function of environmental and geographical data, which is expressed as follows:

[0087] likelihood(T, H, P, lat, lon, prev_lat, prev_lon)

[0088] Among them, T represents temperature, H represents humidity, P represents soil pH value, lat represents the latitude of the current observation point, lon represents the longitude of the current observation point, prev_lat represents the latitude of the previous observation point, and prev_lon represents the longitude of the previous observation point.

[0089] The likelihood function likelihood in this embodiment is used to calculate the probability of observing a certain species under specific environmental and geographical conditions. In theory, a likelihood function can be defined for each species, or a more complex function can be designed to handle the observation probabilities of multiple species at the same time. This embodiment takes the likelihood function for a single species as an example for explanation.

[0090] Among them, when calculating based on the likelihood function, it is necessary to obtain the distance between the two observation points. When the longitude and latitude of the two points are known, the Haversine formula can be used to calculate the great circle route distance between the two points on the earth. When the latitude and longitude of the first point (lat1, lon1) and the latitude and longitude of the second point (lat2, lon2) are known, convert the latitude and longitude from degrees to radians, and then calculate the latitude difference Δlat = lat2 - lat1 and the longitude difference Δlon = lon2 - lon1, where the latitude difference and longitude difference are both in radians. For the parameters in the Haversine formula ,parameter , which is calculated as follows:

[0091] parameter

[0092] parameter

[0093] After obtaining the parameter c, the distance distance = R * c is calculated based on the earth radius R and the parameter c to obtain the great circle distance between two points on the earth.

[0094] When the great circle distance between two observation points is known, the observation probability is calculated based on the following formula:

[0095]

[0096] The above formula uses the form of Gaussian function (or normal distribution) to calculate the observation probability, which takes into account the deviation of environmental variables (temperature T, humidity H, soil pH P) from ideal conditions (20°C, 80% humidity, 6.5 pH), as well as changes in geographical location (calculated by distance). Among them, temperature T is in degrees Celsius (°C) and humidity H is in percentage (%). exp is the natural exponential function, which is used to convert the output of the Gaussian function into a probability value. This function ensures that the probability value is between 0 and 1. The calculation based on the above formula can obtain the calculation result of the likelihood function, which represents the probability of observing a species under specific environmental and geographical conditions.

[0097] After the prior probability corresponding to the known species at the observation point and the observation probability determined based on the likelihood function, the posterior probability of the species is determined based on the product of the prior probability and the observation probability. The prior probability is the probability predicted based on the comprehensive information of the species distribution. The likelihood function provides a new observation probability. By updating the prior probability by multiplication, the posterior probability can be obtained. That is, the posterior probability refers to the updated estimate of the probability of an event occurring after considering the observation probability, which can also be understood as a probability that combines prior knowledge and new observation data.

[0098] As an example, prior_A, prior_B, and prior_C represent the prior probabilities of species A, species B, and species C appearing at a certain location, respectively. For species A, the corresponding likelihood function is used to calculate the corresponding observation probability likelihood_values_A under specific environmental and geographical conditions, for species B, the corresponding likelihood function is used to calculate the corresponding observation probability likelihood_values_B under specific environmental and geographical conditions, and for species C, the corresponding likelihood function is used to calculate the corresponding observation probability likelihood_values_C under specific environmental and geographical conditions.

[0099] The posterior probability of species A appearing at a certain location is calculated based on prior_A*likelihood_values_A; the posterior probability of species B appearing at a certain location is calculated based on prior_B*likelihood_values_B; the posterior probability of species C appearing at a certain location is calculated based on prior_C*likelihood_values_C. posterior_A, posterior_B, and posterior_C represent the posterior probabilities of species A, B, and C appearing at a certain location, respectively. The posterior probabilities reflect the likelihood of a species appearing at a certain location after considering the observed probability.

[0100] When the posterior probability of the species is determined, the estimated value of the target state of the species in the ecological protection zone is obtained according to the posterior probability of the species at different observation points. In particular, when the estimated value of the target state of the species in the ecological protection zone is obtained according to the posterior probability of the species at different observation points, it includes:

[0101] Based on the posterior probabilities of species at different observation points and the environmental data of the observation points, analyze at least one of the following: species distribution, species diversity, ecosystem health, interactions between species, the relationship between species distribution and diversity and environmental changes, and changing trends in species richness and evenness to obtain estimates of the target status of species in ecological protection areas.

[0102] The posterior probability, as a probability that combines prior knowledge and new observation data, can relatively accurately estimate the existence of species. For different species, based on the posterior probability of species at different observation points and the environmental data of the observation points, a comprehensive analysis can be conducted to obtain an estimate of the target status of species in the ecological protection area.

[0103] The estimated value of the target state of species in ecological reserves refers to the prediction and evaluation of species distribution, diversity and ecosystem health status within ecological reserves obtained through model analysis and data processing in ecological reserve monitoring. These estimates can help reserve managers understand the current status of the reserve and predict future trends. Specifically, the estimated value of the target state of species can include the following parameters:

[0104] 1. Species distribution prediction: Predict the distribution of species within the protected area, including where and how many species they are likely to appear.

[0105] 2. Species diversity index: includes estimates of species richness (number of species) and species evenness (how evenly species are distributed).

[0106] 3. Ecosystem health: Assess whether the ecosystem within the ecological protection area is healthy, including species diversity, community structure and the stability of ecological processes.

[0107] 4. Interactions between species: Analyze the interactions between species, such as competition, symbiosis, etc., and how these interactions affect the structure and function of the entire community.

[0108] 5. Impact of environmental change on species distribution: Assess how environmental change (e.g. climate change, land use change) affects the distribution and diversity of species.

[0109] 6. Trends in species richness and evenness: Predict the changing trends in species richness and evenness over time within ecological reserves.

[0110] By obtaining estimates of the target status of species within ecological reserves, a basis can be provided for subsequent large-scale model predictions of the final target, and these estimates can assist reserve managers in formulating more effective conservation strategies, such as habitat restoration and the establishment of species migration corridors, to protect and restore biodiversity.

[0111] It should be noted that the estimated value of the species target state is the species state estimation information combined with time information. Accordingly, when making predictions based on the three-category model, the output prediction content is also related to the time information, and the time information here is the collection time corresponding to the ecological data input into the three-category model.

[0112] In the above implementation plan for obtaining the estimated value of the target status of species, the output content of the three types of models is combined with the linkage between environmental data and geographical location, and the estimated value of the target status of species is obtained through sequential Bayesian filtering. This can more accurately predict and track the distribution changes of species, help us better understand and predict the distribution and diversity of species in protected areas, and provide a scientific basis for the management and protection of protected areas.

[0113] The following is a further introduction to the process of obtaining the input data set for the input target large model. When the estimated values ​​of the target status of species in the ecological protection area and other multi-source monitoring data are aligned according to timestamps and converted into the input data set, it includes:

[0114] Based on timestamps, the estimated values ​​of the target status of species in the ecological protection area are aligned with other multi-source monitoring data; feature extraction, feature preprocessing and transformation are performed on the aligned data to generate target time series associated with the species status and key data features associated with other multi-source monitoring data.

[0115] After obtaining the estimated value of the target state of species in the ecological protection area, data alignment is first performed to align the estimated value of the target state of species with the image video data, remote sensing drone image data, geographic information data, and environmental data collected for the ecological protection area according to the timestamp to ensure that they have the same time step and order. Among them, the image video data, remote sensing drone image data, geographic information data, and environmental data collected for the ecological protection area have the same collection period as the ecological data input into the three types of models, so data alignment can be performed based on timestamps.

[0116] After completing the data alignment, feature selection and feature engineering are performed. In the feature selection stage, based on the understanding and needs of the data, select features that are meaningful to the target time series. These features can be statistical features, lag features, correlation features, etc. of the original data. In the feature engineering stage, the selected features are preprocessed and transformed, such as standardization and normalization, so that the model can better identify them. After the above processing, the target time series associated with the species status and the key data features associated with other multi-source monitoring data are generated. Then, the target time series and key data features are integrated into the input data set.

[0117] Among them, key data features are extracted from other multi-source monitoring data. Key data features can provide additional contextual information to help the target large model better understand and predict the target time series. It should be noted that for image and video data in other multi-source monitoring data, visual modalities can be used as a bridge mechanism to convert these data into a format that can be processed by the target large model, such as extracting key information through image recognition technology and inputting this information as features into the target large model.

[0118] The process of integrating input data sets can include the following main steps:

[0119] 1. Determine the target time series. The target time series is the main analysis object. For example, the target time series represents the distribution data of species at different time points.

[0120] 2. Extract environmental data features, including environmental variables such as temperature, humidity, and soil pH, which can affect the distribution and behavior of species.

[0121] 3. Extract remote sensing data features. These features come from remote sensing drone image data, such as vegetation index, land cover type, etc., which can provide detailed information about environmental conditions.

[0122] 4. Extract geographic information data features, including geographic location (longitude, latitude), altitude, etc. These data help understand the spatial patterns of species distribution.

[0123] 5. Image and video data features, image and video data can provide direct observations on species behavior and numbers.

[0124] The purpose of integrating the above data features is to create a comprehensive input dataset so that the target large model can utilize these multi-source data to improve the accuracy and reliability of predictions.

[0125] As an example, suppose you are studying the seasonal distribution changes of a species. The corresponding input dataset may include the following:

[0126] Target time series: the number of observations of the species each month.

[0127] Environmental data characteristics: average temperature, humidity, and soil pH value for each month.

[0128] Remote sensing data characteristics: vegetation index and land cover type for each month.

[0129] Geographic information data features: geographical location (longitude, latitude) of species observations.

[0130] Image and video data features: Image or video data of species behavior and abundance for each month.

[0131] By integrating the above features, a comprehensive input dataset can be constructed to provide rich information for the target large model, thereby better understanding and predicting the target time series.

[0132] Among them, after the integrated input data set is processed by the target large model, the final output result depends on the training target and application scenario of the model. In the context of ecological protection area monitoring in this embodiment, the output content of the target large model may include the following types of information:

[0133] 1. Species distribution prediction: The model can predict the future distribution of specific species, including where and how many they are likely to appear.

[0134] 2. Species diversity index: The model can output predicted values ​​of species richness (number of species) and species evenness (how evenly species are distributed).

[0135] 3. Ecosystem health assessment: The model can assess whether the ecosystem within the ecological protection area is healthy, including species diversity, community structure and the stability of ecological processes.

[0136] 4. Impact of environmental change on species distribution: The model can predict how environmental changes (such as climate change and land use change) affect the distribution and diversity of species.

[0137] 5. Trends in species richness and evenness: The model can predict the temporal trends in species richness and evenness within ecological reserves.

[0138] 6. Prediction of interspecies interactions: The model can predict interspecies interactions, such as competition, symbiosis, etc., and how these interactions affect the structure and function of the entire community.

[0139] 7. Risk assessment and early warning: The model can identify potential ecological risk areas, such as species invasion, habitat destruction, etc., and provide early warning.

[0140] 8. Management recommendations: Based on the output of the model, it can provide a scientific basis for the management and protection of ecological reserves, such as habitat restoration and the establishment of species migration corridors.

[0141] In practical applications, the output of the target large model can be specific numerical predictions, classification results, probability distributions or other forms of analysis results. These outputs can help protected area managers better understand the current status of the protected area and predict future trends, thereby formulating more effective protection strategies.

[0142] In the above implementation process, the estimated values ​​of the species target status obtained by sequential Bayesian filtering and other multi-source monitoring data are effectively input into the target large model for intelligent analysis, which can realize comprehensive monitoring and prediction of ecological protection areas.

[0143] The following is an introduction to the ecosystem monitoring method based on a large model provided by this application through an overall implementation process. Figure 2 As shown:

[0144] Step 201: Collect ecological sample data from an ecological protection area, and pre-process the collected data to determine a training sample data set.

[0145] Step 202: construct a macro-ecological model, multiple single-species distribution models, and a joint species distribution model based on the training sample data set.

[0146] Step 203: Analyze the collected ecological data in the ecological protection area based on the macro-ecological model to predict the species richness, species evenness and species distribution changes at different observation points in the ecological protection area.

[0147] Step 204: Analyze the ecological data based on the aggregated multiple single-species distribution models to predict the comprehensive assessment information of species diversity corresponding to the ecological protection area.

[0148] Step 205: Analyze the ecological data based on the joint species distribution model to predict the comprehensive information of the associated species community structure and its environmental response corresponding to the ecological protection area.

[0149] Step 206: Integrate the prediction contents output by the three types of models to predict the comprehensive information of species distribution corresponding to the ecological protection area.

[0150] Step 207: Use sequential Bayesian filtering to perform a fusion analysis on the linkage between species distribution comprehensive information, environmental data and geographical location to obtain an estimated value of the target status of species in the ecological protection area.

[0151] Step 208: align the estimated values ​​of the target status of species in the ecological protection zone and other multi-source monitoring data according to timestamps and convert them into an input data set, wherein the input data set includes: a target time series extracted based on the estimated values ​​of the target status of species in the ecological protection zone, and key data features extracted based on other multi-source monitoring data.

[0152] Step 209, fill the key data features into the prompt template, input the data-filled prompt template and the target time series into the target large model, and predict the target time series corresponding to the species in the ecological protection area in the next time period based on the target large model to monitor the ecological protection area.

[0153] The above implementation process makes full use of multi-dimensional data and multiple categories of models to predict the ecological status of species, ensuring the reliability and accuracy of the prediction, and can achieve comprehensive, efficient and accurate monitoring of ecological protection areas.

[0154] The present application embodiment provides an ecosystem monitoring device based on a large model, such as Figure 3 As shown, including:

[0155] The first prediction module 301 is used to analyze ecological data in the ecological protection area based on the macro-ecological model to predict species richness, species evenness and species distribution changes at different observation points in the ecological protection area, wherein the ecological data includes environmental data and species distribution data;

[0156] The second prediction module 302 is used to analyze the ecological data based on the aggregated multiple single species distribution models to predict the comprehensive assessment information of species diversity corresponding to the ecological protection area;

[0157] The third prediction module 303 analyzes the ecological data based on the joint species distribution model to predict the comprehensive information of the associated species community structure and its environmental response corresponding to the ecological protection area;

[0158] The processing acquisition module 304 is used to fuse the prediction contents output by the three types of models with the linkage conditions of the environmental data and the geographical location, and process the fusion results using sequential Bayesian filtering to obtain the estimated value of the target state of the species in the ecological protection area;

[0159] The processing module 305 is used to align the estimated value of the target state of the species in the ecological protection zone and other multi-source monitoring data according to timestamps and convert them into an input data set, wherein the input data set includes: a target time series extracted according to the estimated value of the target state of the species in the ecological protection zone, and key data features extracted according to other multi-source monitoring data;

[0160] The acquisition module 306 is used to input the target time series and the key data features into the target macro model according to a preset prompt template, and use the target macro model to obtain the predicted target time series corresponding to the species in the ecological protection area, so as to monitor the ecological protection area according to the predicted target time series.

[0161] Optionally, the device further comprises:

[0162] A collection and determination module is used to collect ecological sample data from the ecological protection area and pre-process the collected data to determine a training sample data set;

[0163] A first construction module is used to construct a macro-ecological model based on the training sample data set, wherein the macro-ecological model includes a first sub-model for identifying species richness and species evenness at each observation point and a second sub-model for identifying changes in species distribution at different observation points;

[0164] A second construction module is used to separate training sample subsets corresponding to multiple species from the training sample data set, and to construct matching single-species distribution models based on the training sample subsets corresponding to each species, so as to determine multiple single-species distribution models;

[0165] The third construction module is used to construct a joint species distribution model for comprehensive analysis of community ecological data based on the training sample data set.

[0166] Optionally, the second prediction module includes:

[0167] A division determination submodule is used to divide the ecological data based on species categories and determine the data to be analyzed corresponding to multiple species respectively;

[0168] An input acquisition submodule is used to input the to-be-analyzed data corresponding to each species into a matching single-species distribution model to obtain the distribution of the species under different environmental conditions;

[0169] The aggregation acquisition submodule is used to aggregate the distribution conditions of multiple species under different environmental conditions to obtain comprehensive assessment information on species diversity corresponding to the ecological protection area.

[0170] Optionally, the third prediction module is further used for:

[0171] The ecological data are input into the joint species distribution model, and the output includes comprehensive information on species interactions, environmental drivers of species distribution, species co-occurrence patterns, species richness and evenness, and species distribution.

[0172] Optionally, the processing acquisition module includes:

[0173] An integrated prediction submodule is used to integrate the prediction contents output by the three types of models to predict the comprehensive information of species distribution corresponding to the ecological protection area;

[0174] The analysis and acquisition submodule is used to use sequential Bayesian filtering to perform a fusion analysis on the linkage between the comprehensive information on species distribution, the environmental data and the geographical location, so as to obtain an estimated value of the target status of species in the ecological protection area.

[0175] Optionally, the analysis and acquisition submodule includes:

[0176] A prediction unit, configured to predict the prior probability of species in the ecological protection area corresponding to different observation points based on the comprehensive information on species distribution, wherein the prior probability indicates the probability of existence of the species;

[0177] A calculation unit, for calculating, for each observation point, the observation probability of the species being observed at the current observation point based on the geographical location corresponding to the current observation point, environmental data, and the geographical location corresponding to the previous observation point;

[0178] A determination unit, used to determine the posterior probability of the species based on the prior probability and the observation probability corresponding to the species at the observation point;

[0179] The acquisition unit is used to obtain the estimated value of the target state of the species in the ecological protection area according to the posterior probability corresponding to the species at different observation points.

[0180] Optionally, the acquisition unit is further used for:

[0181] Based on the posterior probability of species at different observation points and the environmental data of the observation points, analyze at least one of the species distribution, species diversity, ecosystem health, species interaction, the relationship between species distribution and diversity and environmental change, and the changing trend of species richness and evenness to obtain an estimate of the target status of species in the ecological protection area.

[0182] Optionally, the processing module is further used for:

[0183] Aligning the estimated values ​​of the target status of species in the ecological reserve with other multi-source monitoring data based on timestamps;

[0184] Feature extraction, feature preprocessing and transformation are performed on the aligned data to generate target time series associated with species status and key data features associated with other multi-source monitoring data.

[0185] Optionally, the acquisition module is further used for:

[0186] Filling the key data features into the prompt template;

[0187] Inputting the data-filled prompt template and the target time series into the target macro model, wherein the data-filled prompt template is used to assist the target macro model in understanding the target time series;

[0188] The target time series corresponding to the species in the ecological protection area in the next time period is predicted based on the target large model.

[0189] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0190] An embodiment of the present application also provides an electronic device, including: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the various processes of the above-mentioned large-model-based ecosystem monitoring method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be described here.

[0191] For example, Figure 4 FIG. 1 shows a schematic diagram of the physical structure of an electronic device. Figure 4 As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communications interface 420, and the memory 430 communicate with each other through the communication bus 440. The processor 410 may call the logic instructions in the memory 430, and the processor 410 is used to execute each process of the ecosystem monitoring method based on the large model in the embodiment of the present application, which will not be described one by one here.

[0192] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application.

[0193] The embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, each process of the above-mentioned large-model-based ecosystem monitoring method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it is not repeated here. The computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0194] An embodiment of the present application also provides an ecosystem monitoring system based on a large model, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the various processes of the above-mentioned ecosystem monitoring method embodiment based on the large model are implemented.

[0195] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0196] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0197] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

Claims

1. A method for monitoring an ecosystem based on a large model, characterized in that: include: Analyze ecological data in the ecological protection area based on the macro-ecological model to predict species richness, species evenness and species distribution changes at different observation points in the ecological protection area, wherein the ecological data includes environmental data and species distribution data; Analyze the ecological data based on aggregated multiple single-species distribution models to predict comprehensive assessment information of species diversity corresponding to the ecological protection area; Analyze the ecological data based on the joint species distribution model to predict the comprehensive information of the associated species community structure and its environmental response corresponding to the ecological protection area; The prediction contents output by the three types of models are integrated with the linkage of environmental data and geographic location, and the integrated results are processed by sequential Bayesian filtering to obtain the estimated value of the target state of species in the ecological protection area; Aligning the estimated value of the target state of species in the ecological protection zone and other multi-source monitoring data according to timestamps and converting them into an input data set, wherein the input data set includes: a target time series extracted from the estimated value of the target state of species in the ecological protection zone, and key data features extracted from other multi-source monitoring data; The target time series and the key data features are input into the target macro model according to a preset prompt template, and the target macro model is used to obtain the predicted target time series corresponding to the species in the ecological protection zone, so as to monitor the ecological protection zone according to the predicted target time series.

2. The ecosystem monitoring method based on a large model according to claim 1, characterized in that: The method further comprises: Collecting ecological sample data from the ecological protection area, and preprocessing the collected data to determine a training sample data set; Constructing a macro-ecological model based on the training sample data set, the macro-ecological model comprising a first sub-model for identifying species richness and species evenness at each observation point and a second sub-model for identifying species distribution changes at different observation points; Separating training sample subsets corresponding to a plurality of species from the training sample data set, and constructing matching single-species distribution models based on the training sample subsets corresponding to each species to determine a plurality of single-species distribution models; A joint species distribution model for comprehensive analysis of community ecological data is constructed based on the training sample data set.

3. The ecosystem monitoring method based on a large model according to claim 1, characterized in that: The multiple single-species distribution models based on aggregation analyze the ecological data to predict the comprehensive assessment information of species diversity corresponding to the ecological protection area, including: Dividing the ecological data based on species categories to determine the data to be analyzed corresponding to the multiple species; For each species, the data to be analyzed corresponding to the species is input into a matching single-species distribution model to obtain the distribution of the species under different environmental conditions; The distribution conditions of multiple species under different environmental conditions are aggregated to obtain comprehensive assessment information on species diversity corresponding to the ecological protection area.

4. The ecosystem monitoring method based on a large model according to claim 1, characterized in that: The ecological data is analyzed based on the joint species distribution model to predict the comprehensive information of the associated species community structure and environmental response corresponding to the ecological protection area, including: The ecological data are input into the joint species distribution model, and the output includes comprehensive information on species interactions, environmental drivers of species distribution, species co-occurrence patterns, species richness and evenness, and species distribution.

5. The ecosystem monitoring method based on a large model according to claim 1, characterized in that: The prediction contents output by the three types of models are integrated with the linkage of environmental data and geographic location, and the integrated results are processed by sequential Bayesian filtering to obtain the estimated value of the target state of species in the ecological protection area, including: Integrate the prediction contents output by the three types of models to predict the comprehensive information of species distribution corresponding to the ecological protection area; Sequential Bayesian filtering is used to perform a fusion analysis on the linkage between the comprehensive information on species distribution, the environmental data and the geographical location, so as to obtain an estimated value of the target status of species in the ecological protection area.

6. The method for monitoring an ecosystem based on a large model according to claim 5, characterized in that: The sequential Bayesian filtering is used to perform a fusion analysis on the linkage between the species distribution comprehensive information, the environmental data and the geographical location to obtain an estimated value of the target status of the species in the ecological protection area, including: Predicting the prior probability of species in the ecological protection area at different observation points based on the comprehensive information on species distribution, wherein the prior probability indicates the probability of existence of the species; For each observation point, the observation probability of the species being observed at the current observation point is calculated based on the geographical location and environmental data corresponding to the current observation point and the geographical location corresponding to the previous observation point; Determine the posterior probability of the species based on the prior probability and observation probability corresponding to the species at the observation point; According to the posterior probabilities corresponding to the species at different observation points, an estimated value of the target state of the species in the ecological protection area is obtained.

7. The ecosystem monitoring method based on a large model according to claim 6 is characterized in that: The step of obtaining an estimated value of the target state of species in the ecological protection zone according to the posterior probabilities corresponding to the species at different observation points includes: Based on the posterior probability of species at different observation points and the environmental data of the observation points, analyze at least one of the species distribution, species diversity, ecosystem health, species interaction, the relationship between species distribution and diversity and environmental change, and the changing trend of species richness and evenness to obtain an estimate of the target status of species in the ecological protection area.

8. The method for monitoring an ecosystem based on a large model according to claim 1, characterized in that: The estimated values ​​of the target status of species in the ecological protection area and other multi-source monitoring data are aligned according to timestamps and converted into input data sets, including: Aligning the estimated values ​​of the target status of species in the ecological reserve with other multi-source monitoring data based on timestamps; Feature extraction, feature preprocessing and transformation are performed on the aligned data to generate target time series associated with species status and key data features associated with other multi-source monitoring data.

9. The method for monitoring an ecosystem based on a large model according to claim 8, characterized in that: Inputting the target time series and the key data features into the target macro model according to a preset prompt template, and using the target macro model to obtain the predicted target time series corresponding to the species in the ecological protection area, including: Filling the key data features into the prompt template; Inputting the data-filled prompt template and the target time series into the target macro model, wherein the data-filled prompt template is used to assist the target macro model in understanding the target time series; The target time series corresponding to the species in the ecological protection area in the next time period is predicted based on the target large model.

10. An ecosystem monitoring system based on a large model, characterized in that: The method comprises a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the steps of the method according to any one of claims 1 to 9 when executing the computer program.

Images