Bird diversity prediction method and apparatus

Abstract

The application provides a bird diversity prediction method and device, belongs to the field of bird diversity prediction, and is applied to an electronic device. The method comprises the following steps: acquiring multi-period land distribution data and regional development trend data of a region to be predicted, wherein the multi-period land distribution data comprises current land distribution data and historical land distribution data; determining future land distribution data of the region to be predicted at a future preset time according to the multi-period land distribution data and the regional development trend data; determining future landscape pattern data of the region to be predicted at the future preset time based on the future land distribution data; and processing the future landscape pattern data by using a bird diversity prediction model to obtain a bird diversity prediction result of the region to be predicted at the future preset time. The accuracy of the bird diversity prediction result can be improved.

Description

Technical Field

[0001] This application belongs to the field of bird diversity prediction, and specifically relates to a method and apparatus for bird diversity prediction. Background Technology

[0002] In the process of rapid urbanization, ecosystem functions are damaged and biodiversity faces serious threats. Birds are highly sensitive to environmental changes and play an important ecological indicative role; their distribution is closely related to the urban environment. Therefore, the distribution characteristics of birds can provide a scientific basis for urban ecological protection and sustainable development, and are an important aspect of urban ecological sustainable development.

[0003] The widespread collection of public birdwatching data has greatly expanded the scale and efficiency of data collection, providing rich data resources for bird diversity research. Traditional bird diversity statistics methods involve recording the distribution of environmental features such as buildings and bird distribution within the area to be surveyed, and then correlating the two data. This method can roughly infer the distribution of birds in a certain area during a specific season each year. However, the environmental distribution within an area is not static; once the environment changes, the bird distribution may also change. Currently, traditional statistical methods are relatively lacking in research on future trends, lacking foresight, and cannot accurately predict the bird diversity of the area to be surveyed in the future. Summary of the Invention

[0004] This application provides a method and apparatus for predicting bird diversity. The electronic device uses multi-period land distribution data and regional development trend data of the area to be predicted to predict the future land distribution data of the area, thereby obtaining the future landscape pattern data of the area. Then, based on the landscape pattern data of the area at a preset future time, the corresponding bird diversity prediction result is determined. The method takes into account the changes in land distribution in the area over time, making the determined bird diversity prediction result more accurate.

[0005] This application provides a method for predicting bird diversity, applied to electronic devices. The method includes: acquiring multi-period land distribution data and regional development trend data of the area to be predicted, wherein the multi-period land distribution data includes current land distribution data and historical land distribution data; determining the future land distribution data of the area to be predicted at a preset future time based on the multi-period land distribution data and regional development trend data; determining the future landscape pattern data of the area to be predicted at the preset future time based on the future land distribution data; and processing the future landscape pattern data using a bird diversity prediction model to obtain the bird diversity prediction result of the area to be predicted at the preset future time.

[0006] In one possible embodiment, the future land distribution data of the area to be predicted at a predetermined future time is determined based on multi-period land distribution data and regional development trend data. This includes: determining a land transfer probability matrix based on the differences in land distribution in the area to be predicted at different times in the multi-period land distribution data, wherein the land transfer probability matrix includes the transfer probability between any two land use types in the area to be predicted; determining the target total amount of each land use type in the area to be predicted at a predetermined future time based on the land transfer probability matrix and the total amount of each land use type in the current land distribution data; querying a predetermined data set to determine the suitability probability weight, conversion cost matrix, and neighborhood weight corresponding to the regional development trend data; determining the conversion potential group of each piece of land in the area to be predicted based on the suitability probability weight, conversion cost matrix, and neighborhood weight, wherein the conversion potential group includes the conversion potential of the land's current land use type to each other land use type; and obtaining the future land distribution data of the area to be predicted based on the target total amount of each land use type and the conversion potential group of each piece of land.

[0007] In one possible embodiment, the conversion potential group of each land parcel in the area to be predicted is determined based on suitability probability weights, a conversion cost matrix, and neighborhood weights. This includes: determining the suitability probability of each land parcel in the area to be predicted for each land use type based on suitability probability weights and spatial driving factors. The spatial driving factors are determined based on historical economic and population data of the area to be predicted, including population density distribution data and regional GDP density distribution data of the area to be predicted at a future preset time; determining the conversion allowance matrix for each land parcel based on the conversion cost matrix, which includes the conversion allowance of land use type to each target land use type; determining the neighborhood influence weight of each land parcel for each land use type based on neighborhood weights and the land use types of adjacent land parcels in the current land distribution data; and for each land parcel, weighted fusion of the suitability probability, conversion allowance matrix, and neighborhood influence weights for each land use type to obtain the land conversion potential group.

[0008] In one possible embodiment, the future landscape pattern data includes multiple feature parameters. The future landscape pattern data is processed using a bird diversity prediction model to obtain the bird diversity prediction result for the area to be predicted at a preset future time. This includes: processing the future landscape pattern data using the bird diversity prediction model to obtain an initial bird diversity prediction result for the area to be predicted; processing the bird diversity prediction result and the future landscape pattern data using a network interpretation model to obtain the contribution of each feature parameter to the bird diversity prediction result; and determining the initial bird diversity prediction result as the bird diversity prediction result for the area to be predicted in response to the contribution of each feature parameter to the bird diversity prediction result meeting a preset contribution condition.

[0009] In one possible embodiment, the method further includes: obtaining a reference contribution range for each feature parameter to the bird diversity prediction result, wherein the reference contribution range for each feature parameter is obtained by processing the bird diversity prediction result based on multi-period landscape pattern data of the benchmark area by the bird diversity prediction model; in response to the existence of at least one feature parameter whose contribution to the bird diversity prediction result is not in the corresponding reference contribution range, determining that the contribution of each feature parameter to the bird diversity prediction result does not meet the preset contribution condition.

[0010] In one possible embodiment, after determining that the contribution of each feature parameter to the bird diversity prediction result does not meet the preset contribution condition, the method further includes: taking the feature parameter whose contribution to the bird diversity prediction result is not in the corresponding reference contribution interval as the target feature parameter; obtaining the difference between the contribution of the target feature parameter to the bird diversity prediction result and the corresponding reference contribution interval; adjusting the bird diversity prediction result using the difference to obtain the final bird diversity prediction result for the area to be predicted; or, adjusting the parameters in the bird diversity prediction model using the difference to obtain an updated bird diversity prediction model; and using the updated bird diversity prediction model to process the future landscape pattern data to obtain the final bird diversity prediction result for the area to be predicted.

[0011] In one possible embodiment, based on future land distribution data, determining the future landscape pattern data of the area to be predicted at a preset future time includes: extracting features from the future land distribution data to obtain a pattern parameter set and a set of parameters for the area to be predicted. The pattern parameter set includes at least one of the following feature parameters: area, shape, density, and aggregation of each patch in the area to be predicted. The set of parameters includes at least one of the following feature parameters: the synergistic diversity index, fragmentation index, and shape index of the patches in the area to be predicted. The pattern parameter set and the set of parameters for the area to be predicted are then combined to obtain the future landscape pattern data.

[0012] In one possible embodiment, the bird diversity prediction results for the area to be predicted include bird diversity levels of several local areas within the area to be predicted. The method further includes: determining the ecological status data of the area to be predicted according to the bird diversity levels of each local area in the bird diversity prediction results; and determining ecological restoration measures for the area to be predicted based on the ecological status data and the contribution of each characteristic parameter to the bird diversity prediction results.

[0013] In one possible embodiment, the method further includes a training step for the bird diversity prediction model: acquiring bird diversity sample data, which includes bird diversity labels and landscape pattern sample data; dividing the bird diversity sample data into grids according to a preset size to obtain bird diversity labels and landscape pattern sample data under each grid; using a five-fold cross-validation method to determine the training dataset and validation dataset for each iteration of training from the bird diversity labels and landscape pattern sample data under each grid; for each iteration, training the bird diversity prediction model using the training dataset and validating the bird diversity prediction model using the validation dataset.

[0014] This application also provides a bird diversity prediction device applied to an electronic device. The bird diversity prediction device includes: a data acquisition unit, a land distribution prediction unit, a landscape pattern data determination unit, and a bird diversity prediction unit. The data acquisition unit is used to acquire multi-period land distribution data and regional development trend data of the area to be predicted. The multi-period land distribution data includes current land distribution data and historical land distribution data. The land distribution prediction unit is used to determine the future land distribution data of the area to be predicted at a preset future time based on the multi-period land distribution data and the regional development trend data. The landscape pattern data determination unit is used to determine the future landscape pattern data of the area to be predicted at a preset future time based on the future land distribution data. The bird diversity prediction unit is used to process the future landscape pattern data using a bird diversity prediction model to obtain the bird diversity prediction result of the area to be predicted at a preset future time.

[0015] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement any of the bird diversity prediction methods described above.

[0016] This application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the bird diversity prediction methods described above.

[0017] This application also provides a computer program product, including a computer program that, when executed by a processor, implements any of the bird diversity prediction methods described above.

[0018] The technical solution of the bird diversity prediction method and apparatus provided in this application takes into account that bird distribution is closely related to the landscape in the environment, and the landscape in the area to be predicted often changes with the development trend of the area. For example, if the development trend of the area to be predicted is based on benchmarks, economic priority, ecological protection or sustainable development, new buildings such as factories or farmland conversion to forest may be implemented in the area to be predicted, which may cause the distribution of birds in the area to change accordingly. Therefore, this solution first obtains multi-period land distribution data and regional development trend data in the area to be predicted, determines the land distribution data of the area to be predicted under the specific regional development trend at a certain time in the future, and then determines the future landscape pattern data based on the land distribution data. Then, based on the landscape pattern data, the bird diversity prediction result under the specific landscape pattern is obtained, which improves the accuracy of the bird diversity prediction result of the area to be predicted at a preset time in the future. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is one of the flowcharts of a bird diversity prediction method provided in this application;

[0021] Figure 2 This is the second flowchart of a bird diversity prediction method provided in this application;

[0022] Figure 3 This is the third flowchart of a bird diversity prediction method provided in this application;

[0023] Figure 4 This is a schematic diagram of the average SHAP values ​​of various feature parameters under multiple aspects provided in this application;

[0024] Figure 5 This is a schematic diagram of bird diversity in the area to be predicted at a future preset time, output by the bird prediction model provided in this application.

[0025] Figure 6 This is a schematic diagram of the bird diversity levels of the area to be predicted in 2020, provided in this application.

[0026] Figure 7 This is a schematic diagram of the bird diversity levels of the area to be predicted at future prediction times, based on the regional development trend provided in this application.

[0027] Figure 8 This is a schematic diagram of the bird diversity levels of the area to be predicted at a future prediction time under the condition that the regional development trend is ecological protection, provided in this application;

[0028] Figure 9 This is a schematic diagram of the bird diversity levels of the area to be predicted at the future prediction time under the condition that the regional development trend is economic priority.

[0029] Figure 10 This is a schematic diagram of the bird diversity levels of the area to be predicted at future prediction times under the condition that the regional development trend is sustainable.

[0030] Figure 11 This is one of the functional unit block diagrams of a bird diversity prediction device provided in this application;

[0031] Figure 12 This is the second functional unit block diagram of a bird diversity prediction device provided in this application;

[0032] Figure 13 This is a schematic diagram of the structure of an electronic device provided in this application. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0034] The terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0035] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0036] Given that current bird forecasting methods rely on predicting future patterns based on the distribution patterns of birds in specific seasons each year, but the regional environment is not static, bird distribution may change as well. The bird distribution patterns in specific seasons of the region next year or in the future may not be consistent with the bird distribution patterns of the region this year or in previous years. This means that the current method cannot accurately predict the future bird distribution in the region.

[0037] To address the aforementioned problems, this application provides a method and apparatus for predicting bird diversity. Considering that bird distribution is closely related to the landscape, and that the landscape within the predicted area often changes with the region's development trends (e.g., based on baseline, economic priority, ecological protection, or sustainable development), the predicted area may undergo new construction such as factories or reforestation projects, potentially altering bird distribution. Therefore, this solution first acquires multi-period land distribution data and regional development trend data within the predicted area. It then determines the land distribution data under a specific regional development trend at a future point in time, thereby determining future landscape pattern data based on this land distribution data. Finally, it determines the bird diversity prediction result under the specific landscape pattern based on this landscape pattern data. The embodiments of this application are described in detail below with reference to the accompanying drawings.

[0038] Please see Figure 1 , Figure 1 This is one of the flowcharts illustrating a bird diversity prediction method provided in this application. The bird diversity prediction method may include the following steps:

[0039] S101: Obtain multi-period land distribution data and regional development trend data for the area to be predicted.

[0040] The area to be predicted can be any area requiring bird diversity prediction. Multi-period land distribution data includes current land distribution data and historical land distribution data. Historical land distribution data can include land distribution data within the area to be predicted at at least one historical point in time. This land distribution data may include, but is not limited to, at least one of the following: land use category, geographical location, size, and landscape pattern of each plot of land within the area to be predicted. For example, the landscape pattern of each plot can be the distribution of various landscape types on that plot, such as a plot having a building with a floor area of ​​M every N meters. Regional development trend data may include, but is not limited to, data used to indicate benchmarks, economic priorities, ecological protection, or sustainable development trends; that is, the regional development trend data can be used to indicate the future development trend of the area to be predicted.

[0041] S102, Based on multiple periods of land distribution data and regional development trend data, determine the future land distribution data of the area to be predicted at a future preset time.

[0042] Under different regional development trends, the changes in land distribution within the predicted area will vary. For example, if economic priority is the development trend of the predicted area, more industry may be developed, resulting in more land being converted to construction land. Conversely, if ecological protection is the development trend of the predicted area, measures such as returning farmland to forest may be implemented, leading to more land being converted to forest land.

[0043] Future land distribution data can include the landscape patterns of each land plot within the area to be predicted, and the landscape patterns can include the distribution data of each landscape. Different land types will result in different landscapes. For example, if the land use type is construction land, then the landscape of this land will mostly consist of factory buildings. If the land use type is forest land, then the landscape of this land will mostly consist of coniferous forests or broad-leaved forests, etc.

[0044] The implementation of S102 can be as follows: First, a land distribution prediction model is established. Then, multi-period land distribution data and regional development trend data are processed based on this model to obtain future land distribution data for the area to be predicted at a predetermined time. Alternatively, the transfer probability between different land use types is first determined based on multi-period land distribution data, and then the future land distribution data for each land in the area to be predicted at a predetermined time is determined by combining this data with regional development trend data.

[0045] S103, based on future land distribution data, determines the future landscape pattern data of the area to be predicted at a preset future time.

[0046] The future landscape pattern data of the area to be predicted at a predetermined future time may include the future landscape pattern data of each land parcel in the area at the predetermined future time. The area to be predicted can be divided into grids to obtain multiple land parcels. The area of ​​each land parcel can be 1km*1km. The above-mentioned S103 can be implemented by extracting features from the distribution data of each landscape in the future land distribution data to obtain future landscape pattern data. For example, the future landscape pattern data may include feature parameters such as the category of each landscape in each land parcel in the area to be predicted, the area and shape of the area where the landscape is located, and the degree of aggregation between the areas where the various landscapes are located.

[0047] S104. The bird diversity prediction model is used to process the future landscape pattern data to obtain the bird diversity prediction results of the area to be predicted at a preset time in the future.

[0048] Bird diversity prediction models can be of various types, including but not limited to Bayesian regression models, support vector regression, random forest models, extreme augmented regression trees, or multilayer perceptron models. One approach is to use future landscape pattern data for all land parcels within the prediction area as input to the bird diversity prediction model, predicting the bird diversity of each parcel at a predetermined future time. Alternatively, the future landscape pattern data for each parcel within the prediction area can be used as input to a separate bird diversity prediction model, yielding individual predictions for each parcel at a predetermined future time. The bird diversity prediction model can be pre-trained to learn the relationship between landscape pattern data and bird diversity, allowing it to output the predicted bird diversity for the area at the predetermined future time when future landscape pattern data is used as input.

[0049] In the above scheme, considering that bird distribution is closely related to the landscape pattern in the environment, and that the landscape in the area to be predicted often changes with the development trend of the area, for example, if the area to be predicted is based on benchmark, economic priority, ecological protection or sustainable development, new buildings such as factories or reforestation may be implemented in the area, which may cause the distribution of birds in the area to change accordingly. Therefore, this scheme first obtains multi-period land distribution data and regional development trend data in the area to be predicted, determines the land distribution data under the specific regional development trend at a certain time in the future, and then determines the future landscape pattern data based on the land distribution data. Then, based on the landscape pattern data, the bird diversity prediction result under the specific landscape pattern is obtained, which improves the accuracy of the bird diversity prediction result at the preset time in the future of the area to be predicted.

[0050] In one possible embodiment, see Figure 2 The above S102 may include the following steps:

[0051] S201. Based on the differences in land distribution in the expected prediction area at different times in the multi-period land distribution data, determine the land transfer probability matrix.

[0052] The land transfer probability matrix includes the transfer probability between any two land use types in the area to be predicted. For example, land use types may include, but are not limited to: forest land, construction land, arable land, grassland and / or water bodies.

[0053] The differences in land distribution in the area to be predicted at different times in S201 above include the differences in land use type for each piece of land at different times. S201 can be implemented as follows: based on the differences in land use type for each piece of land at different times, count the number of transfers between each land use type and all other land use types. Based on the number of transfers between each land use type and all other land use types, determine the land transfer probability matrix. For example, if there are N pieces of land in the area to be predicted that are cultivated land, and 10% × N of these cultivated lands have been converted to construction land, then the probability of cultivated land being converted to construction land is 0.1. These transfer probabilities between land use types constitute a land transfer probability matrix.

[0054] S202, based on the land transfer probability matrix and the total amount of each land use type in the current land distribution data, determine the target total amount of each land use type in the area to be predicted at a future preset time.

[0055] The total amount of each land use type in the current land distribution data can be the quantity and / or area of ​​land under each land use type. In one possible embodiment, the land transfer probability matrix can be multiplied by the total amount of each land use type in the current land distribution data to obtain the target total amount of each land use type in the area to be predicted at a future preset time. Alternatively, the spatial driving factor, the land transfer probability matrix, and the total amount of each land use type in the current land distribution data can be combined to determine the target total amount of each land use type in the area to be predicted at a future preset time.

[0056] Spatial driving factors can be determined based on historical economic and population data of the region to be predicted. These factors can include population density distribution data and GDP density distribution data for the region at a predetermined future time. Population density distribution data can represent the population size per plot of land within the region, while GDP density distribution data can represent the GDP per plot of land. For example, spatial driving factors can be predicted using a time-series forecasting model (such as the XGBoost time-series forecasting model). The inputs to this model can include historical economic data (such as GDP and fixed asset investment data), historical population data, and regional development trend data for the region to be predicted. The output is the spatial driving factor for the region at a predetermined future time. Through pre-training, the time-series forecasting model can capture the impact of time and regional development trends on spatial driving factors. Therefore, this model can output relatively accurate spatial driving factors based on historical economic data (such as GDP and fixed asset investment data), historical population data, and regional development trend data for the region to be predicted.

[0057] The method described above, which combines spatial driving factors, the land transfer probability matrix, and the total amount of each land use type in the current land distribution data to determine the target total amount of each land use type in the predicted area at a predetermined future time, can be as follows: A Markov model (also known as a Markov model) is used to process the spatial driving factors, the land transfer probability matrix, and the total amount of each land use type in the current land distribution data to obtain the target total amount of each land use type in the predicted area at a predetermined future time. The Markov model uses the total amount of each land use type in the current land distribution data as the starting point for prediction, while the land transfer probability matrix reflects the historical transformation patterns of land types, and the spatial driving factors characterize future economic and population conditions. By combining these three sets of data, the target total amount of each land use type in the predicted area at a predetermined future time can be obtained.

[0058] S203, query the preset data group to determine the suitability probability weight, conversion cost matrix and neighborhood weight corresponding to the regional development trend data.

[0059] The preset data set stores the suitability probability weights, conversion cost matrices, and neighborhood weights corresponding to each regional development trend. The suitability probability weights include the influence weights of different influencing factors on the applicability of each land use type. The suitability probability weights can be the same or different under different development trends.

[0060] The influencing factors can include various parameters from the spatial driving factors (such as GDP density distribution data), or they can include various parameters from the static driving factors of land. Static driving factors can include, but are not limited to, geographical static factors and man-made facility static factors. Geographical static factors can include, but are not limited to, land slope, altitude, soil type, and distance to water systems. Man-made facility static factors can include, but are not limited to, distance from main roads / highways and distance from historical and cultural protection zones / ecological red lines. For example, the influence weight of each parameter on the land's suitability for construction land can be: GDP density has an influence weight of 0.4, distance from main roads has an influence weight of 0.3, slope has an influence weight of 0.2, and distance from ecological red lines has an influence weight of 0.1. The higher the influence weight for each land use type, the greater the influence of that influencing factor on the suitability of the land for construction land.

[0061] The conversion cost matrix includes the ease or difficulty of converting between any two land types. The difficulty level ranges from 0 to 1, where 1 indicates a complete prohibition on conversion (e.g., the difficulty of converting forest land within ecological red lines to construction land is 1), and 0 indicates a complete permission for conversion (e.g., the difficulty of converting abandoned wasteland to construction land is 0). The middle value indicates moderate difficulty. The conversion cost matrix can be the same or different under different regional development trends. For example, the difficulty of converting forest land to construction land under the regional development trend of ecological protection is greater than the difficulty of converting forest land to construction land under the regional development trend of economic priority.

[0062] Neighborhood weight is a parameter characterizing the degree of clustering influence of surrounding similar land on the current land, typically ranging from 0 to 1. A higher neighborhood weight indicates a greater influence on the current land, suggesting that the land is more prone to "following the trend" (stronger clustering effect). Conversely, a lower neighborhood weight indicates a smaller influence from surrounding similar land, suggesting that the current land is more likely to be dispersed. For example, the neighborhood weight is 0.9 for construction land, 0.5 for forest land, and 0.3 for cultivated land. This indicates that among the surrounding similar land, construction land has a greater influence on the current land, meaning that construction land tends to cluster, while cultivated land is relatively dispersed.

[0063] S204 determines the conversion potential group of each land parcel in the area to be predicted based on suitability probability weights, conversion cost matrix, and neighborhood weights.

[0064] The conversion potential group includes the conversion potential of the land’s current land use type to each of the other land use types. For example, if there are three land use types in total, the conversion potential group for each piece of land includes the conversion potential of the land’s current land use type to the other two land use types.

[0065] The suitability probability weight is used to determine whether land is suitable for land use type conversion, the conversion cost matrix is ​​used to determine whether land can undergo land use type conversion, and the neighborhood weight is used to determine whether the surrounding area will drive land use type conversion. Specifically, the impact of the suitability probability weight, conversion cost matrix, and neighborhood weight on the land's conversion potential for each land use type can be calculated separately. For each land use type, the impact of the three conversion potentials is summarized to obtain the land's conversion potential for each land use type.

[0066] In one possible embodiment, S204 above may include the following steps:

[0067] First, based on the suitability probability weights and spatial driving factors, the suitability probability of each land parcel in the area to be predicted with respect to each land use type is determined. The spatial driving factors are determined based on historical economic and population data of the area to be predicted, including population density distribution data and regional GDP density distribution data for the area at a predetermined future time. As mentioned above, the suitability probability weights specify the influence weight of each parameter in the spatial driving factors on each land use type. By combining the population size and regional GDP of each land parcel in the spatial driving factors, the suitability probability of each land parcel with respect to each land use type can be determined. Alternatively, the suitability probability of each land parcel with respect to each land use type can also be determined by combining at least one parameter in the static driving factors of each land parcel, the suitability probability weights, and the spatial driving factors.

[0068] Then, based on the conversion cost matrix, a conversion allowance matrix is ​​determined for each piece of land. The conversion allowance matrix includes the conversion allowance for converting land use type to each target land use type. For example, under the development trend of prioritizing economic development, the conversion cost of converting arable land A to construction land B is 0.6 (where 1 represents complete prohibition and 0 represents complete permission). The difference between 1 and each conversion cost in the cost matrix is ​​taken as the conversion allowance. Therefore, the conversion allowance for converting arable land A to construction land B = 1 - 0.6 = 0.4 (the higher the value, the more permissible the conversion).

[0069] Then, based on the neighborhood weights and the land use types of adjacent land parcels in the current land distribution data, the neighborhood influence weight of each land parcel with respect to each land use type is determined. Specifically, the number of land parcels of each land use type within their neighborhood is counted, and the product of the proportion of land parcels of each land use type and the neighborhood weight of the land use type is used as the neighborhood influence weight of that land parcel with respect to each land use type. Continuing the example above, the neighborhood weight of construction land is 0.9. If the set neighborhood range is a 3×3 land grid centered on the current land parcel, it means there are a total of 9 land parcels within the neighborhood, including the current land parcel. Counting the number of grid cells within the 3×3 grid that are already construction land, assuming that 6 out of the 9 grid cells are construction land, the proportion of construction land is 6 / 9 ≈ 0.67. The neighborhood influence weight of the current land parcel with respect to construction land is 0.67 × 0.9, approximately equal to 0.603.

[0070] For each piece of land, the suitability probability of the land with respect to each land use type, the conversion allowance matrix, and the neighborhood influence weight are weighted and fused to obtain the land conversion potential group.

[0071] The conversion potential group includes the conversion potential of land from its current land use type to each of the other land use types. The suitability probability is weighted at 0.5, the conversion allowance matrix at 0.2, and the neighborhood influence weight at 0.3. The conversion potential group of land is obtained by multiplying the suitability probability of the current land for each land use type by 0.5, adding the conversion allowance matrix by 0.2, and adding the neighborhood influence weight by 0.3.

[0072] In the above scheme, it is more accurate to determine the suitability probability of each land parcel in the area to be predicted with respect to each land use type by combining spatial driving factors, and then determine the conversion potential group of each land parcel based on the suitability probability.

[0073] S205, based on the target total amount for each land use type and the conversion potential group for each piece of land, obtains the future land distribution data for the area to be predicted.

[0074] For each land use type, based on the conversion potential of each plot of land to that land use type, and according to the target total amount of that land use type, some land is converted in descending order of conversion potential, ensuring that the total land area of ​​that land use type after conversion is less than or equal to the target total amount. The conversion order of each land use type can be ranked according to regional development trend data. For example, under the trend of ecological protection and development, priority is given to forest land conversion; that is, based on the conversion potential of each plot of land to forest land and the target total amount of forest land, some land is converted in descending order of conversion potential, ensuring that the total amount of forest land reaches the target total amount. After the forest land conversion is completed, other land use types are converted. Optionally, land use types that have already been converted within the predicted area at a future preset time will not be converted to other land use types. For example, land already converted to forest land will not be converted to other land use types.

[0075] After obtaining the land use type of each plot of land at a future preset time, the landscape pattern of each plot of land at the future preset time can be determined based on its land use type and the current landscape pattern of each plot of land in the area to be predicted. For example, if the land use type of a plot of land remains unchanged, the current landscape pattern of the plot can be directly used as its landscape pattern at the future preset time. If the land use type of plot A changes to land use type C, the landscape pattern of one of the plots of land S belonging to land use type C in the area to be predicted can be used as the landscape pattern of plot A. For example, if the landscape pattern of plot S is that there is a building with a floor area of ​​M every N meters, then the landscape pattern of plot A can also be set to have a building with a floor area of ​​M every N meters. Alternatively, the landscape pattern of plot S can be adjusted and used as the landscape pattern of plot A. Alternatively, a landscape pattern prediction model can be used to process the land use type of each plot of land at the future preset time and the current landscape pattern of each plot of land to obtain the landscape pattern of each plot of land at the future preset time. Optionally, the future land distribution data can be presented in raster data format.

[0076] Optionally, S204 and S205 above can be performed by a Future Land Use Simulation model (FLUS model). The bird diversity prediction method may further include a training process for the future land use simulation model. This training process may include the following steps: based on the current land use status and regional development trends of the area to be predicted, an objective function and constraints are formulated, four development scenarios are set: baseline, economic priority, ecological protection, and sustainable development, and the land structure under different development scenarios is calculated as the training labels for FLUS; then, using the FLUS model, by calculating the land use distribution suitability probability, setting the conversion cost matrix and neighborhood weights, the accuracy of the FLUS model is constructed and verified. Then, the land use of the area to be predicted at a preset future time is predicted under the four scenarios of baseline, economic priority, ecological protection, and sustainable development. By performing steps such as calculating the land use distribution suitability probability, setting the conversion cost matrix, setting neighborhood weights, and verifying accuracy for each scenario, changes in land use (e.g., changes in land use types) and changes in pattern (e.g., changes in landscape pattern) under different future scenarios are formed. The above scheme integrates FLUS multi-scenario land use simulation (such as economic priority / ecological protection development), time series forecasting, and the limitations of existing Markov model technology that relies on static variables, thereby improving the accuracy of bird diversity prediction results at future forecast times and providing a proactive planning basis for ecological restoration measures.

[0077] In another possible application scenario, the future land use simulation model, based on multi-period land distribution data and development trend data for each region within the area to be predicted, determines the future land distribution data for each development trend within the area. Then, based on the difference between the future land distribution data under each development trend and the current land distribution data, a transition probability matrix is ​​determined. This transition probability matrix is ​​input into a Markov model to obtain the impact of different regional development trends on land use quantity.

[0078] In the above scheme, by determining the target total amount of each land use type at a future preset time based on multiple periods of land distribution data, and then predicting the conversion between land types at the future preset time based on the constraints of the target total amount and regional development trend data, the accuracy of the determined future land distribution data can be improved.

[0079] In one possible embodiment, the future landscape pattern data includes multiple feature parameters. S103 described above may include the following steps:

[0080] First, feature extraction is performed on the future land distribution data to obtain a pattern parameter set and an ensemble parameter set for the area to be predicted. The pattern parameter set includes at least one of the following feature parameters: area, shape, density, and clustering of each patch in the area to be predicted. The ensemble parameter set includes at least one of the following feature parameters: Shannon diversity index, fragmentation, and shape index of patches in the area to be predicted. Landscape analysis software (such as Fragstats) can be used to perform the feature extraction steps on the future land distribution data. Each patch in the area to be predicted can correspond to a landscape. The area and shape of each patch characterize its size and morphological features. Patch density represents the degree of landscape segmentation, and clustering reveals whether the landscape in the area to be predicted tends to cluster or disperse. The Shannon diversity index characterizes the number of landscape types in the area to be predicted, and fragmentation includes the overall segmentation degree of each type of landscape. The shape index quantifies the irregularity of individual patches. Landscapes can include roads, buildings, vegetation, water bodies, etc. The aforementioned clustering can include road density, building density, vegetation density, water system density, etc.

[0081] Then, the pattern parameter group and set parameter group of the area to be predicted are combined to obtain the future landscape pattern data.

[0082] The future landscape pattern data is obtained by combining the pattern parameter group and the ensemble parameter group within the area to be predicted. Optionally, the future landscape pattern data can also be obtained by combining the pattern parameter group and the ensemble parameter group with the following data within the area to be predicted: population density, average annual temperature, nighttime light density, nighttime light, etc.

[0083] In the above scheme, by obtaining multiple feature parameters within the region to be predicted, the predicted bird diversity results can be made more accurate.

[0084] As described above, future landscape pattern data includes multiple characteristic parameters. In one possible embodiment, S104 may include... Figure 3 The steps shown are as follows:

[0085] S301. Bird diversity prediction model is used to process future landscape pattern data to obtain initial bird diversity prediction results for the area to be predicted.

[0086] The initial bird diversity prediction results can include the initial bird diversity prediction results for each land parcel within the area to be predicted. For example, the initial bird diversity prediction results for each land parcel can include the species of birds contained in that land parcel. Alternatively, the initial bird diversity prediction results for each land parcel can also include the quantity of each species in that land parcel. The resulting initial bird diversity prediction results are not necessarily accurate, so the accuracy of the initial diversity prediction results can be assessed.

[0087] S302 uses a network interpretation model to process bird diversity prediction results and future landscape pattern data to obtain the contribution of each feature parameter to the bird diversity prediction results.

[0088] The network interpretation model can be a SHAP model (SHapley Additive exPlanations). A SHAP model outputs the SHAP value for each feature parameter, which represents the contribution of each feature parameter to the bird diversity prediction results. Based on the SHAP values ​​of each feature parameter output by the SHAP model, we can obtain... Figure 4 The SHAP value graph shown. Figure 4 The table shows characteristic parameters related to anthropogenic factors, habitat factors, topographic factors, and climatic factors, along with their average SHAP values. The average SHAP value represents the average impact of a characteristic on bird diversity; a higher value indicates a more significant impact. Figure 4 The horizontal bar chart of the environmental factor values ​​and the SHAP value indicates the degree of influence of the feature on bird diversity at the current value. A SHAP value greater than 0 indicates that the feature has a positive impact on the result, and vice versa. The larger the absolute value, the greater the impact. Figure 4The data shows the anthropogenic factors and their SAP values: road density (RD): 2.98, population density (PD): 2.11, nighttime light density (NIL): 1.87, and urban density (UD): 1.84. Among these, road density, with an average SAP value of 2.98, has the greatest impact on bird diversity. Habitat factors and their SAP values ​​include: building density (CD): 2.24, water system density (WD): 1.11, grassland proportion (GM): 0.13, normalized difference vegetation index (NDVI): 0.22, tree cover proportion (TBE): 0.21, terrestrial natural ecosystem proportion (TNE): 0.12, greenness index (GN): 0.39, solar radiation (SBE): 0.03, wind speed (WIN): 0.03, and land cover type (LC): 0.04. The characteristic parameters and their SAP values ​​for topographic factors include the Digital Elevation Model (DEM): 1.01, Slope (S): 0.06, and Aspect (A): 0.06. The characteristic parameters and their SAP values ​​for climatic factors include Annual Precipitation (AP): 0.54, and Annual Average Temperature (AAT): 0.02. Among these, anthropogenic factors account for 54% of the total characteristic parameters, habitat factors account for 35%, topographic factors account for 7%, and climatic factors account for 4%. Clearly, anthropogenic factors have the greatest impact on bird diversity.

[0089] S303, in response to the fact that the contribution of each feature parameter to the bird diversity prediction result meets the preset contribution condition, the initial bird diversity prediction result is determined as the bird diversity prediction result of the area to be predicted.

[0090] In one possible embodiment, determining whether the contribution of each feature parameter to the bird diversity prediction result meets a preset contribution condition may include the following steps: obtaining a reference contribution range for each feature parameter to the bird diversity prediction result. The reference contribution range corresponding to each feature parameter is obtained by processing multi-period landscape pattern data of a baseline area using a bird diversity prediction model. In response to the existence of at least one feature parameter whose contribution to the bird diversity prediction result is not within the corresponding reference contribution range, it is determined that the contribution of each feature parameter to the bird diversity prediction result does not meet the preset contribution condition.

[0091] The baseline region and the region to be predicted can be the same region or different regions. For example, the baseline region can be the area where the validation dataset used after the bird diversity prediction model has been trained. For instance, the bird diversity prediction model can be validated using multi-period landscape pattern data and bird diversity labels from the baseline region. Then, if the accuracy of the bird diversity prediction model meets the accuracy requirements, it can be deployed in the baseline region and other regions to predict bird diversity in both regions.

[0092] The reference contribution range of each feature parameter can be determined by using landscape pattern data in the validation dataset as input during the validation process of the bird diversity prediction model, outputting the corresponding bird diversity prediction results, and then using the network interpretation model to obtain the range of SHAP values ​​of each feature parameter, which serves as the reference contribution range of each feature parameter to the bird diversity prediction results.

[0093] In the above scheme, the contribution of each feature parameter in the landscape pattern data to the bird diversity prediction result should be consistent. If the bird diversity prediction model processes landscape pattern data from different regions and the SHAP values ​​of each feature parameter are significantly different, it indicates that the bird diversity prediction result may be inaccurate. Only when the SHAP values ​​of each feature parameter are within their respective reference contribution ranges can the initial bird diversity prediction result be used as the bird diversity prediction result for the region to be predicted, thus ensuring the accuracy of the output bird diversity prediction result for the region to be predicted.

[0094] In one possible embodiment, after determining that the contribution of each feature parameter to the bird diversity prediction results does not meet the preset contribution condition, the method further includes:

[0095] Feature parameters whose contribution to bird diversity prediction does not fall within the corresponding reference contribution range are selected as target feature parameters. In other words, feature parameters with abnormal SHAP values ​​are selected as target feature parameters.

[0096] Obtain the difference between the contribution of the target feature parameter to the bird diversity prediction results and the corresponding reference contribution range. The difference can be the distance between the contribution and the reference contribution.

[0097] By leveraging these differences, the bird diversity prediction results are adjusted to obtain the final bird diversity prediction results for the area to be predicted. In other words, the bird diversity prediction results can be directly adjusted so that the contribution of the target feature parameters to the adjusted bird diversity prediction results falls within the corresponding reference contribution range. Specific adjustment methods can include increasing or decreasing the number of one or more bird species in one or more plots of land within the area to be predicted; for example, increasing the number of bird species in one plot and decreasing the number of bird species in another plot.

[0098] Alternatively, the differences can be used to adjust the parameters in the bird diversity prediction model to obtain an updated model. This updated model can then be used to process future landscape pattern data to obtain the final bird diversity prediction for the area to be predicted. For example, the differences can be used to determine a loss, which can then be used to adjust the parameters in the bird diversity prediction model. The adjusted model can then be reused to process future landscape pattern data to obtain the final bird diversity prediction for the area to be predicted. Dynamically adjusting parameters through transfer learning can improve the accuracy of the bird diversity prediction model in predicting bird diversity within the area to be predicted.

[0099] The final bird diversity prediction results for the area to be predicted can be as follows: Figure 5 The predicted raster map shown shows that the higher the diversity, the closer the color is to red. The raster can be obtained by dividing the area to be predicted into 1km*1km grids.

[0100] In one possible embodiment, the bird diversity prediction result for the area to be predicted includes bird diversity levels for several local areas within the area to be predicted. Based on this, the method may further include the following steps:

[0101] Based on the bird diversity levels of each local area in the bird diversity prediction results, the ecological status data of the area to be predicted are determined.

[0102] Please see Figures 6 to 10 , Figure 6 The image shows the bird diversity distribution in the area to be predicted in 2020. Using the natural discontinuity method, the bird diversity level in 2020 was divided into five levels, which served as the criteria for classifying habitat quality. These levels are defined as follows: habitat hotspots with a bird diversity index of 30 or higher; secondary habitat hotspots with an index between 23 and 29; general habitats with an index between 17 and 22; secondary habitat coldspots with an index between 11 and 16; and habitat coldspots with a bird diversity index below 11. Figures 7 to 10 The study uses current land distribution data from 2020 to predict the bird diversity distribution in the target region in 2035 under four different regional development trends. Specifically, Figure 7 This shows the distribution of ecological types in 2035 based on regional development trends (BD), which is the bird diversity in 2035 under natural evolution trends. Figure 8 The distribution of ecological types in 2035 is shown under the condition that the regional development trend is ecological protection (PD). Figure 9 The distribution of ecological types in 2035 is shown under the condition that the regional development trend is economic priority (ED). Figure 10 This shows the distribution of ecosystem types in 2035 under the regional development trend of Sustainable Development (SD). Clearly, as... Figures 7 to 10 As shown, the area and spatial distribution of habitat hotspots differ under different regional development trends.

[0103] Ecological status data can include the ecological level of the area to be predicted. Specifically, the ecological level can be determined based on the distribution of bird diversity levels within the area to be predicted. A low ecological level indicates an urgent need for ecological protection. For example, if the area of ​​secondary habitat cold spots and habitat cold spots within the area to be predicted exceeds a preset proportion, it indicates a low ecological level.

[0104] Then, based on the ecological status data and the contribution of each characteristic parameter to the bird diversity prediction results, ecological restoration measures for the area to be predicted are determined.

[0105] Specifically, if the ecological level of the area to be predicted is lower than a preset level, ecological restoration measures can be determined based on the contribution of ecological status data and various characteristic parameters to the bird diversity prediction results. If the ecological level of the area to be predicted is high, it indicates that the bird diversity in the area is high and the ecological environment is acceptable, and ecological restoration is not required at this time.

[0106] Based on ecological status data and the contribution of various characteristic parameters to bird diversity prediction results, the method for determining ecological restoration measures for the area to be predicted can be as follows: Based on the ecological status data and the contribution of various characteristic parameters to bird diversity prediction results, determine the characteristic parameters within the area that can be restored. For example, if the characteristic parameter that can be restored in the area is a small forest area, then reforestation can be implemented; or if the industrial density in the area is high, then measures such as slowing down industrialization or relocating some industries to other areas can be taken. Restoring the ecology at a predetermined future time can be considered a preventative ecological protection decision.

[0107] In the above scheme, the ecological status of the area to be predicted can be determined by the bird diversity prediction results. If the ecological status indicates that ecological restoration is needed, ecological restoration measures for the area to be predicted can be determined based on the ecological status data and the contribution of each characteristic parameter to the bird diversity prediction results, so as to achieve precise positioning of ecological restoration.

[0108] In one possible embodiment, the method further includes a training step for a bird diversity prediction model:

[0109] Obtain bird diversity sample data. This bird diversity sample data includes bird diversity labels and landscape pattern sample data. For example, bird diversity labels and landscape pattern sample data within a baseline area can be obtained as bird diversity sample data.

[0110] Bird diversity sample data is divided into grids of a preset size to obtain bird diversity labels and landscape pattern sample data for each grid. The baseline area can include densely distributed land. For example, the bird diversity sample data is divided into grids of 1km × 1km, with each grid corresponding to a piece of land, resulting in bird diversity labels and landscape pattern sample data for each grid. 70% of the bird diversity labels and landscape pattern sample data is used as the training set, and 30% is used as the test set.

[0111] The training and validation datasets for each iteration of training are determined from bird diversity labels and landscape pattern sample data under each grid using a five-fold cross-validation method.

[0112] For each iteration, the bird diversity prediction model is trained using the training dataset and validated using the validation dataset.

[0113] In other words, during each iteration, the bird diversity prediction model is trained using the training dataset from each iteration, and then validated using the validation dataset. Alternatively, in each iteration, the bird diversity prediction model can be trained using only the training dataset, and then validated using the validation dataset after confirming that the difference between the output of the bird diversity prediction model and the label meets the stopping training condition.

[0114] In the above scheme, by dividing the landscape pattern sample data and bird diversity sample data within the area to be predicted into grids, the spatiotemporal difference and resolution of bird observation point and land distribution data are normalized, reducing data bias caused by differences in sampling frequency. The use of five-fold cross-validation ensures that every data point in the dataset has the opportunity to participate in both training and testing, making full use of the relatively small dataset while ensuring a balanced data distribution, thus effectively improving the accuracy of evaluating the model's generalization ability.

[0115] In one possible implementation, five types of machine learning models suitable for few-shot learning (including Bayesian regression, support vector regression, random forest, extreme augmentation regression tree, and multilayer perceptron) can be used to learn and predict the relationship between bird diversity and environmental features, determining the model most suitable for bird diversity prediction. The model learns the relationships between different features and their complex mappings to the target variable by observing samples in the training set. The test set is then input into the trained model for prediction, and the output predicted values ​​and the original true values ​​are used to calculate MSE and R0. 2 To evaluate the performance of different models and identify the model with the best prediction accuracy and robustness (using MLP as an example), the bird diversity samples used for statistical analysis were divided into 1km*1km grids, with 70% serving as the training set and 30% as the test set. These samples were then input into the five machine learning models for training. After multiple experiments, the optimal parameters were selected through parameter optimization methods such as grid optimization, ultimately determining the parameter settings for the five models.

[0116] For bird diversity prediction models of the aforementioned type (random forest), a geographically weighted random forest model can be used. This model constructs a complex relationship between bird diversity and various landscape pattern indices, setting a bandwidth of 10 km. Within this range, samples are weighted using a biquadratic summation function to calculate the weight matrix, thereby constructing a local random forest model. R0 is then used. 2 The indicators were evaluated using a model. Specifically, the geographically weighted random forest model was interpreted, revealing the spatial heterogeneity of the impact of forest density and shape on bird diversity. By fusing the geographically weighted random forest model with the SHAP interpretation model, the shortcomings of traditional machine learning in neglecting the spatial non-stationarity of environmental factors were addressed. The spatially variable coefficient function of the geographically weighted random forest model accurately quantifies the location-specific influence weights of the landscape pattern index, while the SHAP value resolves the threshold of single-factor nonlinear effects, significantly improving prediction accuracy and reducing prediction bias in urban gradient areas.

[0117] The above scheme utilizes the SHAP model to interpret various types of bird diversity prediction models. Specifically, by measuring the contribution of various environmental characteristics to the prediction results, it reveals the influence mechanism of urban environmental factors on bird diversity and clarifies the important roles of anthropogenic disturbance factors and habitat factors. Finally, the trained bird diversity prediction model is applied to the environmental characteristic data of all land within the prediction area to obtain a spatial distribution prediction map of bird diversity, providing a scientific basis and decision-making reference for urban ecological planning and bird conservation.

[0118] The following describes a bird diversity prediction device provided in this application. The bird diversity prediction device described below corresponds to the method of the bird diversity prediction device described above.

[0119] Please see Figure 11 , Figure 11 This is one of the functional unit block diagrams of a bird diversity prediction device provided in this application. The bird diversity prediction device 500 is applied to an electronic device. The bird diversity prediction device 500 includes: a data acquisition unit 501, a land distribution prediction unit 502, a landscape pattern data determination unit 503, and a bird diversity prediction unit 504. The data acquisition unit 501 is used to acquire multi-period land distribution data and regional development trend data of the area to be predicted. The multi-period land distribution data includes current land distribution data and historical land distribution data. The land distribution prediction unit 502 is used to determine the future land distribution data of the area to be predicted at a preset future time based on the multi-period land distribution data and the regional development trend data. The landscape pattern data determination unit 503 is used to determine the future landscape pattern data of the area to be predicted at a preset future time based on the future land distribution data. The bird diversity prediction unit 504 is used to process the future landscape pattern data using a bird diversity prediction model to obtain the bird diversity prediction result of the area to be predicted at a preset future time.

[0120] In one possible embodiment, the land distribution prediction unit 502 determines the future land distribution data of the area to be predicted at a preset future time based on multi-period land distribution data and regional development trend data. This includes: determining a land transfer probability matrix based on the differences in land distribution in the area to be predicted at different times in the multi-period land distribution data, the land transfer probability matrix including the transfer probability between any two land use types in the area to be predicted; determining the target total amount of each land use type in the area to be predicted at a preset future time based on the land transfer probability matrix and the total amount of each land use type in the current land distribution data; querying a preset data group to determine the suitability probability weight, conversion cost matrix, and neighborhood weight corresponding to the regional development trend data; determining the conversion potential group of each piece of land in the area to be predicted based on the suitability probability weight, conversion cost matrix, and neighborhood weight, the conversion potential group including the conversion potential of the land's current land use type to each other land use type; and obtaining the future land distribution data of the area to be predicted based on the target total amount of each land use type and the conversion potential group of each piece of land.

[0121] In one possible embodiment, the land distribution prediction unit 502 determines the conversion potential group of each land parcel in the area to be predicted based on suitability probability weights, a conversion cost matrix, and neighborhood weights. This includes: determining the suitability probability of each land parcel in the area to be predicted for each land use type based on suitability probability weights and spatial driving factors. The spatial driving factors are determined based on historical economic and population data of the area to be predicted, including population density distribution data and regional GDP density distribution data of the area to be predicted at a future preset time; determining the conversion allowance matrix for each land parcel based on the conversion cost matrix, which includes the conversion allowance of land use type to each target land use type; determining the neighborhood influence weight of each land parcel for each land use type based on neighborhood weights and the land use types of adjacent land parcels in the current land distribution data; and for each land parcel, weighted fusion of the suitability probability, conversion allowance matrix, and neighborhood influence weights for each land use type to obtain the land conversion potential group.

[0122] In one possible embodiment, the future landscape pattern data includes multiple feature parameters. The bird diversity prediction unit 504 processes the future landscape pattern data using a bird diversity prediction model to obtain the bird diversity prediction result for the area to be predicted at a preset future time. This includes: processing the future landscape pattern data using the bird diversity prediction model to obtain the initial bird diversity prediction result for the area to be predicted; processing the bird diversity prediction result and the future landscape pattern data using a network interpretation model to obtain the contribution of each feature parameter to the bird diversity prediction result; and determining the initial bird diversity prediction result as the bird diversity prediction result for the area to be predicted in response to the contribution of each feature parameter to the bird diversity prediction result meeting a preset contribution condition.

[0123] In one possible embodiment, the bird diversity prediction unit 504 is further configured to: obtain the reference contribution range of each feature parameter to the bird diversity prediction result, wherein the reference contribution range corresponding to each feature parameter is obtained by the bird diversity prediction model based on the multi-period landscape pattern data of the benchmark area; and in response to the existence of at least one feature parameter whose contribution to the bird diversity prediction result is not in the corresponding reference contribution range, determine that the contribution of each feature parameter to the bird diversity prediction result does not meet the preset contribution condition.

[0124] In one possible embodiment, after determining that the contribution of each feature parameter to the bird diversity prediction result does not meet the preset contribution condition, the bird diversity prediction unit 504 is further configured to: take the feature parameter whose contribution to the bird diversity prediction result is not in the corresponding reference contribution interval as the target feature parameter; obtain the difference between the contribution of the target feature parameter to the bird diversity prediction result and the corresponding reference contribution interval; adjust the bird diversity prediction result using the difference to obtain the final bird diversity prediction result for the area to be predicted; or, adjust the parameters in the bird diversity prediction model using the difference to obtain an updated bird diversity prediction model; and process the future landscape pattern data using the updated bird diversity prediction model to obtain the final bird diversity prediction result for the area to be predicted.

[0125] In one possible embodiment, the landscape pattern data determination unit 503 determines the future landscape pattern data of the area to be predicted at a preset future time based on future land distribution data, including: extracting features from the future land distribution data to obtain a pattern parameter set and a set of parameters for the area to be predicted. The pattern parameter set includes at least one of the following feature parameters: the area of ​​each patch in the area to be predicted, the shape of each patch, the patch density, and the aggregation degree. The set of parameters includes at least one of the following feature parameters: the synergistic diversity index, fragmentation degree, and shape index of the patches in the area to be predicted. The pattern parameter set and the set of parameters for the area to be predicted are combined to obtain the future landscape pattern data.

[0126] In one possible embodiment, the bird diversity prediction result of the area to be predicted includes the bird diversity level of several local areas in the area to be predicted. The bird diversity prediction unit 504 is further used to: determine the ecological status data of the area to be predicted according to the bird diversity level of each local area in the bird diversity prediction result; and determine the ecological restoration measures for the area to be predicted based on the ecological status data and the contribution of each characteristic parameter to the bird diversity prediction result.

[0127] In one possible embodiment, the bird diversity prediction device 500 further includes a model training unit (not shown), which performs the training steps of the bird diversity prediction model: acquiring bird diversity sample data, which includes bird diversity labels and landscape pattern sample data; dividing the bird diversity sample data into grids according to a preset size to obtain bird diversity labels and landscape pattern sample data under each grid; determining the training dataset and validation dataset for each iteration of training from the bird diversity labels and landscape pattern sample data under each grid using a five-fold cross-validation method; and for each iteration, training the bird diversity prediction model using the training dataset and validating the bird diversity prediction model using the validation dataset.

[0128] It is understood that since the method embodiments and the device embodiments are different presentations of the same technical concept, the content of the method embodiment section in this application should be adapted to the device embodiment section in a synchronous manner, and will not be repeated here.

[0129] In the case of using integrated units, please refer to Figure 12 , Figure 12 This is the second functional unit block diagram of a bird diversity prediction device provided in this application. The bird diversity prediction device is applied to electronic equipment. Figure 12The bird diversity prediction device 500 includes a processing module 512 and a communication module 511. The processing module 512 controls and manages the operation of the bird diversity prediction device 500, for example, executing the steps of the data acquisition unit 501, the land distribution prediction unit 502, the landscape pattern data determination unit 503, and the bird diversity prediction unit 504, and / or performing other processes of the technology described herein. The communication module 511 is used for interaction between the bird diversity prediction device 500 and other devices. Figure 12 As shown, the bird diversity prediction device 500 may also include a storage module 513, which is used to store the program code and data of the bird diversity prediction device 500.

[0130] The processing module 512 can be a processor or controller, such as a central processing unit (CPU), a general-purpose processor, a digital signal processor (DSP), an ASIC, an FPGA, or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. The processor can also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc. The communication module 511 can be a transceiver, RF circuitry, or a communication interface, etc. The storage module 513 can be a memory.

[0131] All relevant content in each scenario involved in the above method embodiments can be referenced from the functional descriptions of the corresponding functional modules, and will not be repeated here. The above-mentioned bird diversity prediction device 500 can perform the above-mentioned... Figure 1 The method for predicting bird diversity is shown.

[0132] Please see Figure 13 , Figure 13 This is a schematic diagram of the structure of an electronic device provided in this application. For example... Figure 13 As shown, the electronic device may include a processor 610, a communications interface 620, a memory 630, and a communication bus 640, wherein the processor 610, the communications interface 620, and the memory 630 communicate with each other via the communication bus 640. The processor 610 can call logical instructions in the memory 630 to execute the aforementioned bird diversity prediction method.

[0133] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0134] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the bird diversity prediction methods provided in the above embodiments.

[0135] This application also provides a computer program product, including a computer program that, when executed by a processor, implements any of the bird diversity prediction methods described above.

[0136] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0137] This application also provides a computer storage medium storing a computer program for electronic data interchange, which causes a computer to perform some or all of the steps of any of the methods described in the above method embodiments, wherein the computer includes an electronic device.

[0138] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0139] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0140] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical or other forms.

[0141] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0142] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0143] If the aforementioned integrated units are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0144] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage device, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0145] The embodiments of this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for predicting bird diversity, characterized in that, Applied to electronic devices, the method includes: Acquire multi-period land distribution data and regional development trend data for the area to be predicted. The multi-period land distribution data includes current land distribution data and historical land distribution data. Based on the multi-period land distribution data and the regional development trend data, the future land distribution data of the area to be predicted at a future preset time is determined; Based on the future land distribution data, determine the future landscape pattern data of the area to be predicted at the preset future time. The future landscape pattern data is processed using a bird diversity prediction model to obtain the bird diversity prediction results for the area to be predicted at the preset future time. The step of determining the future landscape pattern data of the area to be predicted at the preset future time based on the future land distribution data includes: extracting features from the future land distribution data to obtain a pattern parameter set and a set of parameters for the area to be predicted. The pattern parameter set includes at least one of the following feature parameters: the area of ​​each patch in the area to be predicted, the shape of each patch, the patch density, and the aggregation degree. The set of parameters includes at least one of the following feature parameters: the Shannon diversity index, fragmentation degree, and shape index of the patches in the area to be predicted. The pattern parameter set and the set of parameters for the area to be predicted are then combined to obtain the future landscape pattern data. The future landscape pattern data includes multiple feature parameters. The step of processing the future landscape pattern data using a bird diversity prediction model to obtain the bird diversity prediction result for the area to be predicted at the preset future time includes: processing the future landscape pattern data using the bird diversity prediction model to obtain an initial bird diversity prediction result for the area to be predicted; processing the bird diversity prediction result and the future landscape pattern data using a network interpretation model to obtain the contribution of each feature parameter to the bird diversity prediction result; and determining the initial bird diversity prediction result as the bird diversity prediction result for the area to be predicted in response to the contribution of each feature parameter to the bird diversity prediction result meeting a preset contribution condition.

2. The method according to claim 1, characterized in that, The step of determining the future land distribution data of the area to be predicted at a preset future time based on the multi-period land distribution data and the regional development trend data includes: Based on the differences in land distribution in the predicted area at different periods in the multi-period land distribution data, a land transfer probability matrix is ​​determined, which includes the transfer probability between any two land use types in the predicted area. Based on the land transfer probability matrix and the total amount of each land use type in the current land distribution data, determine the target total amount of each land use type in the area to be predicted at the preset future time. Query the preset data group to determine the suitability probability weight, conversion cost matrix and neighborhood weight corresponding to the regional development trend data; Based on the suitability probability weight, conversion cost matrix, and neighborhood weight, a conversion potential group is determined for each piece of land in the area to be predicted. The conversion potential group includes the conversion potential of the land's current land use type to each of the other land use types. The future land distribution data for the area to be predicted is obtained based on the target total for each land use type and the conversion potential group for each piece of land.

3. The method according to claim 2, characterized in that, The step of determining the conversion potential group for each piece of land in the area to be predicted based on the suitability probability weight, conversion cost matrix, and neighborhood weight includes: Based on the suitability probability weights and spatial driving factors, the suitability probability of each piece of land in the area to be predicted with respect to each land use type is determined. The spatial driving factors are determined based on the historical economic data and historical population data of the area to be predicted, and include the population density distribution data and the density distribution data of the regional GDP of the area to be predicted at a future preset time. Based on the conversion cost matrix, a conversion allowance matrix is ​​determined for each piece of land, the conversion allowance matrix including the conversion allowance of the land use type to each target land use type; Based on the neighborhood weights and the land use types of the adjacent land parcels in the current land distribution data, determine the neighborhood influence weight of each land parcel with respect to each land use type; For each piece of land, the suitability probability, conversion allowance matrix, and neighborhood influence weight of the land for each land use type are weighted and fused to obtain the conversion potential group of the land.

4. The method according to any one of claims 1 to 3, characterized in that, The method further includes: The reference contribution range of each of the aforementioned feature parameters to the bird diversity prediction results is obtained. The reference contribution range corresponding to each of the aforementioned feature parameters is obtained by the bird diversity prediction model based on multi-period landscape pattern data of the benchmark area. In response to the existence of at least one of the feature parameters whose contribution to the bird diversity prediction result is not within the corresponding reference contribution range, it is determined that the contribution of each of the feature parameters to the bird diversity prediction result does not meet the preset contribution condition.

5. The method according to claim 4, characterized in that, After determining that the contribution of each of the aforementioned feature parameters to the bird diversity prediction result does not meet the preset contribution condition, the method further includes: The feature parameters whose contribution to the bird diversity prediction results is not in the corresponding reference contribution range are taken as target feature parameters. The difference between the contribution of the target feature parameter to the bird diversity prediction result and the corresponding reference contribution range is obtained; Using the aforementioned differences, the bird diversity prediction results are adjusted to obtain the final bird diversity prediction results for the area to be predicted; or, Using the aforementioned differences, the parameters in the bird diversity prediction model are adjusted to obtain an updated bird diversity prediction model; The updated bird diversity prediction model is used to process the future landscape pattern data to obtain the final bird diversity prediction results for the area to be predicted.

6. The method according to any one of claims 1 to 3, characterized in that, The bird diversity prediction result for the area to be predicted includes bird diversity levels for several local areas within the area to be predicted. The method further includes: Based on the bird diversity levels of each local area in the bird diversity prediction results, determine the ecological status data of the area to be predicted; Based on the ecological status data and the contribution of each of the aforementioned characteristic parameters to the bird diversity prediction results, ecological restoration measures for the area to be predicted are determined.

7. The method according to any one of claims 1 to 3, characterized in that, The method also includes a training step for the bird diversity prediction model: Obtain bird diversity sample data, which includes bird diversity labels and landscape pattern sample data; The bird diversity sample data is divided into grids according to a preset size to obtain bird diversity labels and landscape pattern sample data under each grid. The training dataset and validation dataset for each iteration of training are determined from bird diversity labels and landscape pattern sample data under each grid using a five-fold cross-validation method. For each iteration, the bird diversity prediction model is trained using the training dataset and validated using the validation dataset.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the bird diversity prediction method as described in any one of claims 1-7.

Images