Determination method and application method of risk prediction model and related device

Abstract

The invention discloses a risk prediction model determination method, an application method and a related device, and relates to the technical field of biodiversity, and the risk prediction model determination method comprises the steps: obtaining a data set; the data set comprises landscape diversity risk assessment indexes and risk early warning values of a plurality of continuous years; training a CNN-LSTM model by taking the landscape diversity risk assessment indexes of the previous two years as input and the risk early warning value of the later year as output to obtain a risk prediction model; the CNN-LSTM model comprises two convolution layers, one LSTM layer and three full connection layers. According to the method, the CNN-LSTM model is adopted, the strong extraction capability of the convolutional neural network on the spatial features and the deep analysis capability of the long and short-term memory network on the time sequence data are combined, and an innovative method is provided for predicting the future landscape diversity risk.

Description

Technical Field

[0001] The present invention relates to the field of biodiversity technology, and in particular to a method for determining a risk prediction model, an application method and related devices. Background Art

[0002] Biodiversity is a broad concept that describes the diversity of species in the entire natural world, including complex life diversity at all scales, including genes, species, ecosystems, and landscapes. It summarizes the structure, function, distribution, characteristics, and composition of all organisms. In recent years, there has been a phenomenon of varying degrees of biodiversity loss around the world, which has affected the normal operation of multiple functions of ecosystems. Many studies have emphasized that biodiversity is crucial in maintaining stable and productive ecosystems. The loss of biodiversity will lead to a decrease in the efficiency of ecological communities in obtaining basic biological resources, producing biomass, and decomposing and recycling basic biological nutrients. Therefore, it is very necessary to obtain biodiversity information in a timely manner and conduct effective assessments and early warnings to maintain the stability of the ecosystem, so as to implement targeted biodiversity protection and restoration measures.

[0003] In ecology, biodiversity assessment is often carried out from three scales: species, ecosystems, and landscapes. The acquisition of species diversity information is mainly based on ground survey data, recording the types and quantities of various organisms in the sample plot. Some scholars have also proposed DNA barcoding technology to achieve rapid, accurate and standardized identification of species, and have established a relatively complete species database. However, compared with species diversity assessment, the assessment of landscape and ecosystem diversity is much more complicated. It is necessary to consider and quantify multiple factors such as land cover type, climate, and human activities, so as to establish a complete and suitable assessment technology system for the area to be evaluated. One method to quantify landscape diversity is to calculate α-diversity (referring to the number of land cover types in a small area, as well as their area and distribution), β-diversity (α-diversity calculated based on a single land cover type but at multiple scales, characterizing the spatial configuration of the study area), and γ-diversity (the number of land cover types in a large area, as well as their area and distribution) at the landscape scale. However, this method only replaces "species" with "land cover type" and applies the diversity index of community ecology to landscape ecology, which cannot highlight the landscape characteristics well. Starting from the smallest homogeneous unit of the landscape, landscape diversity can be comprehensively evaluated from multiple perspectives, including landscape heterogeneity, fragmentation, aggregation and dispersion, and stability. In recent years, scholars have also proposed to establish a comprehensive evaluation index system for regional biodiversity based on the DPSIR (Driving, Pressure, State, Impact, Response) model. The DPSIR model provides a structured method to evaluate biodiversity and provide information for policy making by analyzing the five elements of driving force, pressure, state, impact, and response. With the help of remote sensing technology, it is a commonly used method to obtain a variety of indicators that can characterize the characteristics of landscape structure through time series analysis, spatial statistics, etc. to conduct systematic evaluation. The evaluation method of ecosystem diversity is similar to that of landscape diversity. Corresponding evaluation indicators are obtained from aspects such as type richness, species diversity, and ecological environment diversity for systematic evaluation.

[0004] Humanity is facing a huge challenge of biodiversity loss. A large number of studies have shown that due to climate change, urban expansion, and the conversion of natural habitats to agricultural and forestry land, biodiversity loss is increasingly considered one of the greatest risks facing humanity. This irreversible loss of biodiversity occurs at all levels of biodiversity assessment, including ecosystems, species, genes, and other levels. As humanity's awareness of the importance of biodiversity deepens, the need to protect and manage biodiversity is becoming increasingly urgent, and the biodiversity early warning system has become a hot topic of research. It is very necessary to conduct biodiversity risk assessment and early warning based on the assessment of the current status of biodiversity, which can make early predictions of biodiversity safety and forecast dangerous situations, and is of great significance to the study of the coordinated development of the regional economy, society, and ecological environment.

[0005] Therefore, how to predict future regional landscape risks has become a technical problem that needs to be urgently solved in this field. Summary of the invention

[0006] The purpose of the present invention is to provide a method for determining a risk prediction model, an application method and related devices, which can predict future regional landscape risks.

[0007] To achieve the above object, the present invention provides the following solutions:

[0008] In a first aspect, the present invention provides a method for determining a risk prediction model, the method for determining a risk prediction model comprising:

[0009] A data set is obtained; the data set includes: landscape diversity risk assessment index and risk warning value for several consecutive years.

[0010] The landscape diversity risk assessment index of the previous two years is used as input, and the risk warning value of the next year is used as output. The CNN-LSTM model is trained to obtain a risk prediction model; the CNN-LSTM model includes: two convolutional layers, one LSTM layer and three fully connected layers.

[0011] In a second aspect, the present invention provides an application method of a risk prediction model, the application method of the risk prediction model comprising:

[0012] Obtain the landscape diversity risk assessment index for the current two consecutive years.

[0013] The landscape diversity risk assessment index for the current two consecutive years is input into a risk prediction model to obtain a risk warning value for the next year; the risk prediction model is a model trained according to the risk prediction model determination method described above.

[0014] In a third aspect, the present invention provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for determining the risk prediction model or the method for applying the risk prediction model described above.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for determining the risk prediction model or the method for applying the risk prediction model described above.

[0016] In a fifth aspect, the present invention provides a computer program product, comprising a computer program, which, when executed by a processor, implements the method for determining the risk prediction model or the method for applying the risk prediction model described above.

[0017] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0018] The present invention provides a method for determining a risk prediction model, an application method and related devices. The risk prediction model is obtained by training a CNN-LSTM model using the landscape diversity risk assessment index of the previous two years as input and the risk warning value of the next year as output. The CNN-LSTM model is used in the present invention, which combines the powerful extraction capability of convolutional neural networks for spatial features and the in-depth analysis capability of long and short-term memory networks for time series data, and provides an innovative method for predicting future landscape diversity risks. It has important practical significance for biodiversity conservation, sustainable development, landscape planning and management, and ecological environmental protection, and provides useful reference and inspiration for practical work in related fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 A flowchart of a method for determining a risk prediction model provided in one embodiment of the present invention.

[0021] Figure 2 A location map of a study area provided in accordance with an embodiment of the present invention.

[0022] Figure 3 This is a PCR framework diagram of a landscape diversity risk assessment system provided in one embodiment of the present invention.

[0023] Figure 4 A schematic diagram of a landscape diversity risk warning scoring process at each level provided in an embodiment of the present invention.

[0024] Figure 5 A flowchart of a method for applying a risk prediction model provided in one embodiment of the present invention.

[0025] Figure 6 A schematic diagram of landscape diversity risk warning value prediction based on CNN-LSTM provided in one embodiment of the present invention.

[0026] Figure 7 A box plot of the spatial distribution and interannual changes of rainfall erosivity, dryness, surface temperature, night light intensity and population density in Inner Mongolia from 2010 to 2020 provided in one embodiment of the present invention.

[0027] Figure 8 This is a graph showing the spatial distribution and change curves of the aggregation index, polymerization index, spread index and split index of Inner Mongolia from 2010 to 2020, provided in one embodiment of the present invention.

[0028] Fig. 9 This is a graph showing the spatial distribution and change curves of the Shannon Diversity Index, Simpson Diversity Index, Shannon Evenness Index and Simpson Evenness Index in Inner Mongolia from 2010 to 2020, provided in one embodiment of the present invention.

[0029] Fig.10 A spatial distribution map of the road network density in Inner Mongolia from 2010 to 2020 provided as an embodiment of the present invention.

[0030] Fig.11 A spatial distribution map of risk warning values ​​in Inner Mongolia in 2010, 2015 and 2020 and a schematic diagram of risk warning forecast values ​​in 2025 are provided as an embodiment of the present invention.

[0031] Fig.12 A schematic diagram of the predicted values ​​of landscape diversity risk warning for typical areas in Inner Mongolia in 2025 provided by one embodiment of the present invention.

[0032] Fig.13 A schematic diagram of the structure of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0035] The development of machine learning has provided a new research method for biodiversity early warning. Computer systems can analyze and infer patterns in data by using algorithms and statistical models, and make predictions about the future based on sufficient historical data. Machine learning methods can effectively analyze available data and predict the risks to biodiversity in future time series. These data include historical observations and records, satellite images and other forms of remote sensing data, and manually collected species observation records. Further improving the accuracy and timeliness of early warnings and reducing the operating costs of early warning systems are the main challenges currently faced.

[0036] In an exemplary embodiment, Figure 1 As shown, a method for determining a risk prediction model is provided, the method comprising the following steps:

[0037] A1: Obtain a data set; the data set includes: landscape diversity risk assessment index and risk warning value for several consecutive years.

[0038] A2: The landscape diversity risk assessment index of the previous two years is used as input, and the risk warning value of the next year is used as output. The CNN-LSTM model is trained to obtain a risk prediction model; the CNN-LSTM model includes: two convolutional layers, one LSTM layer and three fully connected layers.

[0039] By implementing the above steps A1 to A2, the present invention adopts the CNN-LSTM model, which combines the powerful extraction capability of convolutional neural network for spatial features and the in-depth analysis capability of long short-term memory network for time series data, and provides an innovative method for predicting future landscape diversity risks. It has important practical significance for biodiversity conservation, sustainable development, landscape planning and management, and ecological environmental protection, and provides useful reference and inspiration for practical work in related fields.

[0040] Inner Mongolia is one of the most complete natural ecosystem areas in China and a key area for biodiversity conservation, covering a variety of terrestrial ecosystems including grasslands, farmlands, forests, cities, deserts, and wetlands. Although some scholars have evaluated the diversity of various ecosystems in Inner Mongolia, they lack systematicity and integrity, and there are few risk warning studies based on the current status of biodiversity assessment. This paper constructs a landscape biodiversity risk assessment system for the Inner Mongolia Autonomous Region based on the PCR (Potential, Connectedness, Resilience) framework, and deeply analyzes the current status and changing trends of landscape biodiversity risks. In particular, this paper introduces advanced machine learning technology and predicts landscape diversity risks based on the CNN-LSTM model. The research is of great significance for timely grasping the current status, changing trends and degree of threat of biodiversity in Inner Mongolia, so as to carry out targeted biodiversity protection and restoration.

[0041] Among the mainstream machine learning methods, LSTM (Long Short-Term Memory, LSTM) is good at capturing long-term dependencies in sequence data, while CNN (Convolutional Neural Network, CNN) is good at extracting local features of image data. By combining the advantages of both, the model can take into account both the temporal and spatial information of the data, reduce parameters and reduce the risk of overfitting, thereby providing more accurate predictions, better performance and higher training efficiency.

[0042] like Figure 2 As shown, the present invention takes the Inner Mongolia Autonomous Region in China as the research area. The unique geographical location and topography of Inner Mongolia have created a complex and diverse ecosystem, mainly including zonal ecosystems such as forests, shrubs, grasslands and deserts, and non-zonal ecosystems such as wetlands and sandy lands. It is one of the regions in China with the most complete natural ecosystem types and is distributed with a variety of endangered species and protected species.

[0043] In another exemplary embodiment of the present invention, in step A1, determining the landscape diversity risk assessment index of each year specifically includes:

[0044] a1: Obtain risk assessment indicators for each year; the risk assessment indicators include: a number of first-level indicators and a number of second-level indicators; the first-level indicators include: potential risk, connectivity risk and resilience risk; each of the second-level indicators includes: a number of exposure indicators and a number of interference indicators.

[0045] a2: Preprocessing the risk assessment indicators of each year to obtain preprocessed data; the preprocessing includes: normalization processing and standardization processing.

[0046] a3: Using entropy method to determine the weight of each risk assessment indicator in the preprocessed data.

[0047] a4: Based on the weights of each risk assessment indicator, the landscape diversity risk assessment index for each year is calculated.

[0048] Based on the landscape diversity status assessment indicator system, this paper establishes a risk assessment indicator system from three dimensions: potential, connectivity, and resilience risk according to the PCR risk assessment framework ( Figure 3 ). Potential indicators represent multiple static attributes of the landscape and reflect the intrinsic characteristics of landscape units. Connectivity indicators refer to the interactions between adjacent patches in the landscape structure. Resilience indicators indicate the resistance and maintenance capacity of an ecosystem when disturbed, as well as its ability to tolerate and mitigate disturbances and restore stability. These can be used to predict possible landscape risks in the future. Among them, potential and connectivity indicators characterize the risk effects of the landscape in the spatial dimension, while resilience indicators reflect its dynamic change trends in the temporal dimension. Landscape diversity risk assessment involves three first-level indicators: potential risk, connectivity risk, and resilience risk; and 28 second-level indicators such as soil erosion sensitivity and aggregation index. The risk assessment indicators are shown in Table 1.

[0049] Table 1 Risk assessment indicators

[0050]

[0051]

[0052] In order to better assess the risk changes of landscape diversity in the study area, the present invention increases the time span of the assessment years, and conducts risk assessment and early warning on the landscape diversity in 2010, 2015, and 2020. The hierarchical risk assessment analysis of landscape diversity is divided into the following steps: First, the data is normalized. The present invention uses the range normalization method to standardize the assessment indicators of biodiversity; secondly, the entropy method is used to determine the indicator weights; finally, the evaluation values ​​of each level are obtained through a comprehensive evaluation method, and then an effective quantitative assessment of the biodiversity risk situation in the Inner Mongolia Autonomous Region is conducted to observe the changing development trends within the sample period. The indicator confirmation methods are mainly divided into subjective weighting methods and objective weighting methods. The entropy method selected in the present invention is one of the objective weighting methods. Compared with the subjective weighting method, this method can better eliminate the influence of subjective factors and obtain more objective judgment criteria. Moreover, the entropy method can also obtain estimation results more accurately and effectively among similar objective weighting methods. The calculation formula of the landscape diversity risk assessment index is:

[0053]

[0054] Among them, P ij is the information entropy of the jth indicator in the i-th year, x' ij is the normalized value after processing, g j is the utility value of the jth indicator, m is the number of years of data, and w j is the weight of the jth indicator, n is the number of indicators, and RI is the landscape diversity risk assessment index.

[0055] Referring to the national “Regional Biodiversity Assessment Standards” (2011), this study divided the risk assessment index of landscape diversity and ecosystem diversity in Inner Mongolia Autonomous Region into different segments to represent different risk levels, as shown in Table 2: For landscape diversity, when RI < 0.45, the level is low; when 0.45 ≤ RI < 0.70, the level is moderate; when RI ≥ 0.70, the level is high; for ecosystem diversity, when RI < 0.35, the level is low; when 0.35 ≤ RI < 0.60, the level is moderate; when RI ≥ 0.60, the level is high.

[0056] Table 2 Segmentation standards for landscape diversity risk assessment index in Inner Mongolia Autonomous Region

[0057] Landscape diversity risk level Risk Assessment Index high RI ≥ 0.70 Moderate 0.45≤RI<0.70 Low RI<0.45

[0058] In another exemplary embodiment of the present invention, the determination of the risk warning value for each year specifically includes:

[0059] b1: Obtain risk warning indicators for each year; the risk warning indicators are indicators determined based on potential risks and connectivity risks.

[0060] b2: Determine the risk assessment level of each risk warning indicator; the risk assessment levels include: low, moderate and severe.

[0061] b3: Determine the score of each risk warning indicator according to the risk assessment level of each risk warning indicator.

[0062] b4: Use the entropy method to determine the weight of each risk warning indicator.

[0063] b5: After multiplying the scores of the risk warning indicators and the corresponding weights, normalize them to obtain the risk warning values ​​for each year.

[0064] Based on the existing landscape biodiversity risk warning assessment results, the present invention uses the CNN-LSTM model to predict the biodiversity risk warning value of the study area in 2025. Considering that the predicted risk warning value is an independent year, in order to avoid the model being too complicated, a landscape diversity risk warning index system is further screened and established on the basis of the biodiversity risk assessment index system, involving 14 risk warning indicators such as average soil erosion sensitivity and average desertification sensitivity, mainly involving potential risks and connectivity risks. The risk warning indicators are shown in Table 3.

[0065] Table 3 Risk warning indicators

[0066]

[0067] The risk warning scoring of landscape diversity is divided into the following steps: First, score each risk warning indicator according to the diversity risk assessment level, with the risk levels of low, medium and severe being 1, 2 and 3 points respectively. Then, the weight of each risk warning indicator is determined according to the entropy method, and the weight is normalized after multiplying it with the score of each risk warning indicator to obtain the landscape diversity risk warning value. For detailed calculation process, please refer to Figure 4 .

[0068] In another exemplary embodiment of the present invention, Figure 5 As shown, the present invention also provides an application method of the risk prediction model, and the application method of the risk prediction model includes:

[0069] B1: Obtain the landscape diversity risk assessment index for the current two consecutive years;

[0070] B2: Inputting the landscape diversity risk assessment index for the current two consecutive years into a risk prediction model to obtain a risk warning value for the next year; the risk prediction model is a model trained according to the risk prediction model determination method described above.

[0071] In an exemplary embodiment, obtaining the landscape diversity risk assessment index for the current two consecutive years specifically includes:

[0072] c1: Obtain risk assessment indicators for the current two consecutive years; the risk assessment indicators include: a number of first-level indicators and a number of second-level indicators; the first-level indicators include: potential risk, connectivity risk and resilience risk; each of the second-level indicators includes: a number of exposure indicators and a number of interference indicators.

[0073] c2: preprocessing the risk assessment indicators for the current two consecutive years to obtain preprocessed data; the preprocessing includes: normalization processing and standardization processing.

[0074] c3: Using entropy method to determine the weight of each risk assessment indicator in the preprocessed data.

[0075] c4: Based on the weights of each risk assessment indicator, the landscape diversity risk assessment index for the current two consecutive years is calculated.

[0076] The calculation formula for the landscape diversity risk assessment index is shown in formula (1) to formula (4), which will not be repeated here.

[0077] The specific process of predicting the landscape diversity risk warning value in 2025 based on the CNN-LSTM model is as follows: Figure 6 As shown in the figure, it mainly includes data preprocessing, model definition, model training and evaluation, model storage and application. The CNN-LSTM model consists of two convolutional layers (Conv1, Conv2), one LSTM layer and three fully connected layers (Fc1, Fc2, Fc3). The first convolutional layer Conv1 has 32 3*3 convolution kernels. The second layer Conv2 increases the number of output channels to 64. The size of the convolution kernel remains unchanged. The convolutional layers all use the ReLU activation function. The convolutional layer is connected to a fully connected layer Fc1 to convert the features into a format suitable for LSTM processing. Then it passes through an LSTM layer that processes sequence data and applies the Dropout ratio to reduce overfitting. Finally, it is transformed through two other fully connected layers Fc2 and Fc3 to output the prediction results. The model combines the local feature extraction capability of the convolutional neural network and the sequence processing capability of the long short-term memory network, and is suitable for tasks that need to consider both spatial features and temporal dependencies.

[0078] The data used in the present invention include slope topography, land use / land cover, vegetation coverage, precipitation, temperature, surface temperature and night light data. Table 4 lists the data sources and their resolutions. All raster data are interpolated into 1km×1km units through spatial resampling, and the indicator window involved in the neighborhood analysis operation in data processing is set to 5km.

[0079] Table 4 Basic data details

[0080]

[0081] The prediction results are analyzed below.

[0082] 1. Regional landscape diversity risk assessment based on PCR framework.

[0083] (1) Potential risks.

[0084] Among the potential indicators, the exposure indicators of soil erosion sensitivity, desertification sensitivity, and vegetation coverage showed consistent spatial performance in the study area in 2010, 2015, and 2020, with the peak and low values ​​remaining unchanged. In terms of interference indicators, it was observed that from 2010 to 2020, the low-value areas of rainfall erosivity were mainly concentrated in the southwest ( Figure 7 a, b and c in the figure). It is worth noting that, except for the western part of the study area and a few areas in the northeast and southeast, the rainfall erosivity in other areas has increased over the past 10 years, and the distribution of rainfall erosivity values ​​in 2020 has also changed from median aggregation to higher value aggregation, and the mean value shows an overall gradual upward trend, which indicates that the potential for soil erosion caused by rainfall is increasing. The spatial distribution of dryness values ​​is higher in the central region and lower in the northeast. Except for the northeast and the west, the dryness values ​​in other areas have increased over the past 10 years ( Figure 7 d, e and f in Figure 2). Numerically, the dryness increased from 0.7077 in 2010 to 0.7317 in 2020, indicating that the dryness is getting worse. Most areas have seen an increase in surface temperature over the past 10 years, and the numerical distribution shows two peaks, with the peak in the high value range being larger ( Figure 7 g, h and i in ). From the average value, the surface temperature has a significant upward trend from 27.17℃ to 30.03℃ between 2010 and 2015. The high values ​​of night lights are mainly concentrated in the southern region, and the peak value has continued to rise between 2010 and 2020, indicating that the intensity of human activities is increasing ( Figure 7 The high population density is mainly concentrated in the southern and eastern regions, and shows a clustering phenomenon ( Figure 7 m, n and o in the figure). Inner Mongolia Autonomous Region has a vast area, and the average population density has not changed significantly, increasing from 14 people / km2 to 15 people / km2 in 10 years. In addition, the distribution of night lights and population density values ​​is similar, with most areas showing zero, which is clearly shown in the box plot. Overall, the potential risks in climate and human activities should not be underestimated. In order to prevent further deterioration of ecological problems, more systematic and comprehensive ecological protection and governance strategies need to be adopted.

[0085] (2) Connectivity risk.

[0086] Among the exposure indicators of connectivity risk, the aggregation index AI ( Figure 8 a) and the degree of polymerization index CONTIG( Figure 8b) in the figure are similar in spatial distribution, with lower values ​​in the southeast and higher values ​​in the northwest, which is logical. The maximum values ​​of the two indicators have not changed, the minimum values ​​have increased first and then decreased, and the average values ​​have shown a downward trend, indicating that the proximity of the same type of patches in the landscape of the study area and the complexity of the patch edges have weakened from 2010 to 2020, and various landscape types or patches have become more dispersed or distributed more evenly, indicating that the diversity of the landscape has increased. On the contrary, the spread index CONTAG( Figure 8 c) and the division index DIVISION ( Figure 8 The values ​​of d) in the figure are higher in the southeast and lower in the northwest, and the spatial variation between different phases is also small. Their average values ​​increase first and then decrease, showing an overall decreasing trend, indicating that the structure of the landscape is more continuous and concentrated, which can be regarded as a positive trend.

[0087] There are four other indicators of connectivity risk exposure that represent the richness and uniformity of the landscape. Among them, the Shannon Diversity Index (SHDI) Fig. 9 a) and Simpson diversity index SIDI ( Fig. 9 b) in the figure shows consistent spatial performance, with higher values ​​in the southeast and lower values ​​in the northwest, and small spatial changes between different phases. In terms of values, the average values ​​of SHDI and SIDI first increased slightly and then decreased, and overall decreased from 0.3057 and 0.1824 in 2010 to 0.2885 and 0.1709 in 2020, respectively, indicating that the overall landscape diversity has decreased during this period. On the other hand, the Shannon Evenness Index (SHEI) Fig. 9 c) and Simpson uniformity index SIEI ( Fig. 9 The average value of d) in the figure also shows a downward trend, and the decline is more obvious between 2015 and 2020. The landscape uniformity decreases, and the dominance of dominant patches increases. Both human activities and climate change have a negative impact on biodiversity, which requires people to pay attention and take measures to protect and restore biodiversity.

[0088] Road network density is an interference indicator of connectivity risk. The road network density of Inner Mongolia Autonomous Region in 2010, 2015 and 2020 showed consistent spatial performance, with high values ​​mainly concentrated in the eastern and southern regions, and most regions had values ​​of ( Fig.10 ). In terms of numerical values, the peak and average values ​​of road network density continued to rise, from 11.909km / km in 2010 to 2 、0.0606km / km 2 ( Fig.10 a) in the figure increased to 19.9244km / km in 2015 2 、0.0721km / km 2 ( Fig.10 b) in the figure, and further increased to 25.8839km / km in 2020 2 、0.1359km / km 2 ( Fig.10 In c), the average value increased by 93% from 2015 to 2020. This indicates that human disturbance in the region has increased, the landscape has become fragmented, and the negative impact on the ecological environment has increased.

[0089] (3) Elasticity risk.

[0090] The elasticity index mainly reflects the trend of potential and connectivity indicators. From the analysis of the previous subsection, it can be seen that the vegetation coverage has a more obvious trend in the exposure potential indicators. The average vegetation coverage increased by 3.96% and 10.69% during 2010-2015 and 2015-2020, respectively. This is due to the fact that the Inner Mongolia Autonomous Region has attached great importance to the ecological protection of grassland areas in just a few decades, and has proposed relevant management systems such as the "Inner Mongolia Autonomous Region Basic Grassland Protection Regulations". Among the interference potential indicators, the change rate of the average rainfall erosivity from 2015 to 2020 is as high as 40.27%, which may lead to problems such as intensified soil erosion, insufficient water supply, and deterioration of the ecological environment. In terms of connectivity indicators, the average SHDI, average SIDI, average SHEI, and average SIEI all show a downward trend, indicating that the connectivity of the landscape is gradually increasing and the richness of land use in the landscape system is weakening. The average road network density is generally on the rise. From 2010 to 2015 and from 2015 to 2020, the average road network density increased by 12.95% and 70.82% respectively, and the negative anthropogenic interference with landscape diversity increased.

[0091] 2. Landscape biodiversity risk warning based on CNN-LSTM model.

[0092] First, based on the risk assessment indicator system of each year, the distribution of risk warning values ​​in the study area in 2010, 2015, and 2020 was obtained through risk warning scoring, weight assessment, and normalization. Then, the CNN-LSTM model was trained and the accuracy was verified with the help of the values ​​of the risk assessment indicators in 2010 and 2015 and the risk warning values ​​in 2020. After multiple experiments, the model performed best when the number of features in the hidden layer was 128, the number of features in the fully connected layer was 32, the learning rate was 0.002, and the number of batch iteration samples was 512, and the verification accuracy reached 0.85. With the help of the best performing model and the values ​​of the risk assessment indicators in 2015 and 2020, the risk warning score of the study area in 2025 was predicted.

[0093] From the distribution of landscape biodiversity risk warning values ​​in 2010, 2015 and 2020 ( Fig.11ac), the warning value of the western part of the study area - Alxa League is in the middle position, but it tends to increase over time, and by 2020 it is in the middle and high position. This phenomenon may be related to human activities such as overgrazing in the region, resulting in serious grassland desertification and the risk of deterioration of landscape diversity. The western part of Hulunbuir City and Xilin Gol League are blue in color, belonging to areas with low risk warning values, but they also increased slightly in 2020, and relevant ecological protection work still needs to be continuously and efficiently implemented. In the eastern part of the study area, that is, Hulunbuir City, Xing'an League, Tongliao City, and Chifeng City, except for the west, the risk warning values ​​are staggered, mainly related to factors such as road density and vegetation coverage, and the landscape in the area is relatively fragmented. There is a small red area in the southeast of the study area, which is Wuhai City, and the overall risk warning value is very high. As an emerging industrial city dominated by the coal industry, Wuhai City cannot ignore the issue of ecological protection while the coal chemical industry is developing rapidly. Overall, the average value of the risk warning for landscape diversity in Inner Mongolia decreased from 0.3752 to 0.3621 from 2010 to 2015, mainly due to a series of ecological protection measures taken by the state and local governments. However, the average value of the risk warning increased slightly from 2015 to 2020, reaching 0.3625, indicating that relevant departments should not relax and should always pay attention to changes in the regional ecological situation. Fig.11 The overall numerical distribution pattern remains the same as before, with risk warning values ​​in high-value areas showing a downward trend, while those in low-value areas have increased, especially in Xing'an League, Tongliao City, and Chifeng City in the east. The average risk warning forecast for 2025 is 0.3526, down 2.73% from 2020. This positive change indicates that the overall status of regional landscape biodiversity is relatively optimistic, but continued efforts and appropriate ecological management measures are still needed to maintain and improve this trend.

[0094] The current status assessment, risk evaluation and risk warning research of landscape biodiversity is based on the thinking of landscape ecology, which is holistic and dynamic. As a representative evaluation method, the landscape ecological risk index method directly or indirectly reveals the degree of landscape ecological risk through quantitative analysis of landscape pattern. The results under the PCR risk assessment framework show that the potential risk of landscape diversity has increased between 2010 and 2020, which is reflected in the increase of indicators such as rainfall erosivity, dryness, average surface temperature, and night light ( Fig.12). Although connectivity risks have weakened and landscape structures have become more continuous and concentrated, increased road network density and increased human interference may lead to landscape fragmentation. Therefore, when responding to the challenges of climate change, it is necessary to limit human interference, avoid overdevelopment, and protect the ecological environment and landscape pattern to reduce the negative impact on the ecosystem. The overall landscape diversity risk warning value in Inner Mongolia is on a downward trend, but the overall warning value in 2020 has increased slightly, indicating that the risk situation is not stable and still requires continued attention. Based on the risk warning value results predicted by the CNN-LSTM model in 2025, there are still some areas where the landscape diversity risks continue to increase. The more typical ones are the Moon Lake Tourist Area in Alxa Left Banner and the Hulun Lake area in Hulunbuir City ( Fig.12 ). For the Moon Lake tourist area, considering the pressure that tourism activities may cause on the ecological environment, ecological restoration projects such as vegetation restoration and water body protection should be implemented, and sustainable tourism management strategies should be formulated. For the Hulun Lake area, the focus should be on wetland protection and water resource management to prevent the shrinkage of the lake and the deterioration of water quality. Using landscape ecology methods to conduct in-depth risk assessment and risk warning can help decision makers identify key ecologically sensitive areas and formulate corresponding management plans and risk mitigation measures.

[0095] As a popular research direction in the field of machine learning in recent years, deep learning can learn and understand huge and complex data through multi-layer neural networks, and has been widely used in various fields. Because it can automatically and efficiently mine the patterns and laws hidden in the data, better capture the mutual influence between various factors, and thus simulate and predict future scenarios, it is of great significance to introduce deep learning methods into the field of landscape diversity risk warning prediction for the field of biodiversity conservation. The present invention predicts the landscape diversity risk warning value in 2025 with the help of the CNN-LSTM model, analyzes the development trend of future landscape diversity, and warns of possible ecological problems. The CNN-LSTM model can effectively extract the spatial features of the input data (spatial distribution data of landscape diversity risk warning indicators) using the convolutional layer of the CNN network structure, while the LSTM network can learn the temporal information in the data (multi-period landscape diversity risk warning indicator data), thereby comprehensively utilizing the advantages of the two networks and improving the ability of feature extraction.

[0096] The present invention uses the PCR framework to evaluate the risk status of Inner Mongolia's landscape diversity in 2010, 2015, and 2020 and calculates the risk warning value. The most critical thing is that the present invention adopts the CNN-LSTM model, combines the powerful extraction ability of convolutional neural networks for spatial features and the in-depth analysis ability of long short-term memory networks for time series data, and predicts the risk warning value of landscape diversity in the study area in 2025, providing an innovative method for predicting future landscape diversity risks. The research results show that the risk of landscape diversity potential increased from 2010 to 2020, the connectivity risk showed a weakening trend, the degree of fragmentation decreased, the landscape structure became more continuous and concentrated, and regional human interference increased. The predicted value of landscape diversity risk warning in 2025 is similar to that in 2020 in terms of spatial distribution characteristics, the local differences are reduced, and the mean value decreases by 2.73%. The present invention evaluates the regional landscape diversity risk based on the PCR framework, and predicts the future landscape diversity risk warning value with the help of the CNN-LSTM model, which has important practical significance for biodiversity conservation, sustainable development, landscape planning and management, and ecological environmental protection, and provides useful reference and inspiration for practical work in related fields.

[0097] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Fig.13 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data sets. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for determining a risk prediction model or an application method of a risk prediction model is implemented.

[0098] Those skilled in the art will understand that Fig.13 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0099] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the above-mentioned method embodiments are implemented when the processor executes the computer program.

[0100] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the above-mentioned method embodiments are implemented.

[0101] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the above-mentioned method embodiments are implemented.

[0102] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in the present invention are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0103] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided by the present invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0104] The database involved in each embodiment provided by the present invention may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided by the present invention may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0105] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0106] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for determining a risk prediction model, characterized in that: The method for determining the risk prediction model includes: Acquire a data set; the data set includes: landscape diversity risk assessment index and risk warning value for several consecutive years; The landscape diversity risk assessment index of the previous two years is used as input, and the risk warning value of the next year is used as output. The CNN-LSTM model is trained to obtain a risk prediction model; the CNN-LSTM model includes: two convolutional layers, one LSTM layer and three fully connected layers.

2. The method for determining the risk prediction model according to claim 1, characterized in that: The determination of the landscape diversity risk assessment index for each year includes: Obtain risk assessment indicators for each year; the risk assessment indicators include: a number of primary indicators and a number of secondary indicators; the primary indicators include: potential risk, connectivity risk and resilience risk; each of the secondary indicators includes: a number of exposure indicators and a number of interference indicators; Preprocessing the risk assessment indicators of each year to obtain preprocessed data; the preprocessing includes: normalization processing and standardization processing; Determining the weight of each risk assessment indicator in the preprocessed data using an entropy method; Based on the weight of each risk assessment indicator, the landscape diversity risk assessment index of each year was calculated.

3. The method for determining the risk prediction model according to claim 1, characterized in that: The calculation formula of the landscape diversity risk assessment index is: Among them, P ij is the information entropy of the jth indicator in the i-th year, x' ij is the normalized value after processing, g j is the utility value of the jth indicator, m is the number of years of data, and w j is the weight of the jth indicator, n is the number of indicators, and RI is the landscape diversity risk assessment index.

4. The method for determining the risk prediction model according to claim 1, characterized in that: The determination of risk warning values ​​for each year includes: Obtain risk warning indicators for each year; the risk warning indicators are indicators determined based on potential risks and connectivity risks; Determine the risk assessment level of each risk warning indicator; the risk assessment levels include: low, moderate and severe; Determine the score of each risk warning indicator according to the risk assessment level of each risk warning indicator; The entropy method is used to determine the weight of each risk warning indicator; After multiplying the scores of the risk warning indicators and the corresponding weights, normalization is performed to obtain the risk warning values ​​for each year.

5. An application method of a risk prediction model, characterized in that: The application method of the risk prediction model includes: Obtain the landscape diversity risk assessment index for the current two consecutive years; The landscape diversity risk assessment index for the current two consecutive years is input into a risk prediction model to obtain a risk warning value for the next year; the risk prediction model is a model trained according to the method for determining a risk prediction model according to any one of claims 1-4.

6. The application method of the risk prediction model according to claim 5, characterized in that: Obtain the current two consecutive years of landscape diversity risk assessment index, including: Obtain risk assessment indicators for the current two consecutive years; the risk assessment indicators include: a number of primary indicators and a number of secondary indicators; the primary indicators include: potential risk, connectivity risk and resilience risk; each of the secondary indicators includes: a number of exposure indicators and a number of interference indicators; Preprocessing the risk assessment indicators for the current two consecutive years to obtain preprocessed data; the preprocessing includes: normalization processing and standardization processing; Determining the weight of each risk assessment indicator in the preprocessed data using an entropy method; Based on the weight of each risk assessment indicator, the landscape diversity risk assessment index for the current two consecutive years is calculated.

7. The application method of the risk prediction model according to claim 5, characterized in that: The calculation formula of the landscape diversity risk assessment index is: Among them, P ij is the information entropy of the jth indicator in the i-th year, x' ij is the normalized value after processing, g j is the utility value of the jth indicator, m is the number of years of data, and w j is the weight of the jth indicator, n is the number of indicators, and RI is the landscape diversity risk assessment index.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the method for determining the risk prediction model described in any one of claims 1 to 4 or the method for applying the risk prediction model described in any one of claims 5 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, it implements the method for determining the risk prediction model described in any one of claims 1 to 4 or the method for applying the risk prediction model described in any one of claims 5 to 7.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, it implements the method for determining the risk prediction model described in any one of claims 1 to 4 or the method for applying the risk prediction model described in any one of claims 5 to 7.

Images