Integrated construction method for drought-enduring plant community in desert area

Abstract

The invention discloses a desert area drought-enduring plant community integrated construction method, and the method comprises the steps: obtaining soil desertification degree distribution data and environment dynamic data of a target desert area, and carrying out the region division; constructing a drought-enduring plant community space configuration model of each region; carrying out multi-dimensional ecological data monitoring, and constructing a soil desertification dynamic change matrix; predicting the future desertification trend of each region; and dynamically adjusting a plant community configuration strategy. According to the method, the success rate and sustainability of vegetation restoration in the desert area are remarkably improved, the self-restoration process of an ecological system is accelerated, and the ecological restoration cost is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of ecological restoration, more particularly, the present application relates to a method for constructing a drought-tolerant plant community in a desert area. BACKGROUND

[0002] Desertification is one of the serious ecological and environmental problems faced by the world. At present, desert ecological restoration mainly adopts vegetation restoration methods, which slow down wind erosion, fix dunes, and improve soil environment by planting drought-tolerant plants. Traditional desert vegetation restoration techniques mainly include mechanical sand fixation, chemical sand fixation, and biological sand fixation. Mechanical sand fixation slows down wind speed through physical means such as setting up sand barriers and laying grass squares to create conditions for plant growth; chemical sand fixation uses chemicals such as polymers to form a protective layer to temporarily stabilize the sand surface; biological sand fixation uses drought-tolerant plants to achieve long-term stable ecological restoration effect by fixing sand with plant roots and preventing wind with the aboveground part.

[0003] However, traditional desert management methods generally use single-function plant configuration strategies, lack overall ecological system design thinking, and result in single plant community structure and one-sided ecological function, making it difficult to form a stable wind-preventing and sand-fixing system. For example, in desert management, large-scale single-species planting increases vegetation coverage in the short term, but neglects water balance and soil improvement, leading to serious vegetation decline in the later period. Plant configuration schemes rely heavily on empiricism and lack precise assessment and quantitative analysis of desert microenvironment differences, failing to develop differentiated configuration strategies for different desertification degree areas. For example, in the edge zone of the Taklimakan Desert, the same vegetation restoration scheme has significantly different effects in different desertification degree areas. At the same time, existing technologies use a "one-time design, static implementation" mode, which cannot adapt to the dynamic characteristics of the desert environment, especially under the background of climate change, extreme climate events frequently occur in the desert ecosystem, and static restoration strategies are difficult to maintain long-term restoration results. In addition, there is a serious lack of use of ecological monitoring data, and a large amount of valuable monitoring data cannot be converted into decision-making basis for guiding ecological restoration, resulting in an "information island" phenomenon between monitoring and restoration; the lack of prediction ability leads to passive response of management measures, and it is difficult to deploy prevention and control strategies in advance in response to future desertification trends, such as in the process of Mu Us Sandland management, due to the lack of effective prediction models, it is difficult to respond to the damage of seasonal sandstorms to newly grown vegetation. The lack of research on the synergistic effect of plants and soil microorganisms ignores the key role of microorganisms in the desert ecosystem, resulting in difficulty in plant colonization and slow growth, and the inability to form a self-sustaining ecosystem, ultimately requiring continuous human intervention to maintain, significantly increasing the cost of ecological restoration and reducing sustainability. SUMMARY

[0004] The main purpose of the present application is to provide a method for constructing a drought-tolerant plant community in a desert area to overcome the above-mentioned deficiencies of the prior art.

[0005] To achieve the above-mentioned application purposes, the present application provides the following technical solutions.

[0006] Some embodiments of the present application provide a method for integrated construction of a drought-tolerant plant community in a desert area, comprising:

[0007] Obtain soil desertification degree distribution data and environmental dynamic data of a target desert area, divide the target desert area into regions according to the soil desertification degree distribution data, and obtain a light desertification region, a moderate desertification region, and a severe desertification region:

[0008] Based on the region division result and the environmental dynamic data, construct a drought-tolerant plant community spatial configuration model for each region, wherein the spatial configuration model includes an outer windproof layer, a middle water-locking layer, and an inner soil improvement layer;

[0009] Monitor the drought-tolerant plant community spatial configuration model of each region with multi-dimensional ecological data, obtain a time sequence of multi-dimensional ecological data of each region, and construct a soil desertification dynamic change matrix for each region based on the time sequence of multi-dimensional ecological data;

[0010] Based on the soil desertification dynamic change matrix, predict the future desertification trend of each region, and generate a regional desertification prediction result;

[0011] According to the regional desertification prediction result, dynamically adjust the plant species distribution and spatial layout strategy of the drought-tolerant plant community spatial configuration model, and realize integrated construction of a drought-tolerant plant community in a desert area.

[0012] Further, the dividing the target desert area into regions according to the soil desertification degree distribution data comprises:

[0013] Obtain soil particle composition data, soil moisture content data, and soil organic matter content data of the target desert area;

[0014] According to the proportion of fine particles with a particle size less than 0.05 mm in the soil particle composition data, calculate the soil desertification index of each sampling point;

[0015] Weight the soil desertification index, wherein the weighting includes taking the soil moisture content data as a negative correlation weight and taking the soil organic matter content data as a negative correlation weight, to obtain a comprehensive desertification degree value of each sampling point;

[0016] Based on a preset desertification degree threshold interval, divide the comprehensive desertification degree value into a light desertification interval, a moderate desertification interval, and a severe desertification interval, which correspond to the light desertification region, the moderate desertification region, and the severe desertification region, respectively;

[0017] Smooth the divided region boundary to obtain a continuous region division result.

[0018] In one embodiment, the drought-tolerant plant community spatial configuration model of each region includes:

[0019] According to the soil desertification degree and wind speed distribution data of each region, the plant species and planting density of the outer circle wind protection layer are determined, the plant species include trees and shrubs with high wind resistance, and the planting density is adaptively adjusted based on the gradient change of the wind speed distribution data;

[0020] According to the soil water evaporation rate and surface runoff data of each region, the plant species and root system distribution characteristics of the middle circle water locking layer are determined, the plant species include herbaceous plants with deep root system and high water locking capacity, and the root system distribution characteristics are characterized by the ratio between root depth and root density;

[0021] According to the soil organic matter content and microbial activity data of each region, the plant species and symbiotic flora configuration of the inner circle soil improvement layer are determined, the plant species include leguminous plants with nitrogen fixation ability, and the symbiotic flora configuration is optimized by soil microbial diversity index;

[0022] Based on the terrain fluctuation data and hydrological connectivity data of each region, the spatial layering width and transition zone width of the outer circle wind protection layer, the middle circle water locking layer and the inner circle soil improvement layer are determined, and the transition zone width is normalized calculated by the product of the standard deviation of the terrain fluctuation data and the watershed convergence index of the hydrological connectivity data.

[0023] In one embodiment, the drought-tolerant plant community spatial configuration model of each region is monitored by multi-dimensional ecological data to obtain a time series sequence of multi-dimensional ecological data of each region, including:

[0024] Soil moisture sensors, wind speed sensors and soil organic matter sensors are respectively arranged in the outer circle wind protection layer, the middle circle water locking layer and the inner circle soil improvement layer of each region to obtain soil moisture time series data, wind speed time series data and soil organic matter time series data of each layer;

[0025] Based on unmanned aerial vehicle remote sensing technology, surface vegetation coverage rate time series data and surface temperature time series data of each region are obtained;

[0026] The soil moisture time series data, the wind speed time series data, the soil organic matter time series data, the surface vegetation coverage rate time series data and the surface temperature time series data are time axis aligned to generate a time series sequence of multi-dimensional ecological data of each region;

[0027] Performing outlier detection on the multi-dimensional ecological data time series, the outlier detection comprising calculating a local outlier factor for each time series data, marking data points with a local outlier factor greater than a preset outlier threshold as outliers and performing interpolation repair.

[0028] In one embodiment, constructing a soil desertification dynamic change matrix for each region based on the multi-dimensional ecological data time series comprises:

[0029] Performing principal component analysis on the multi-dimensional ecological data time series to extract dominant ecological factors for each region, the dominant ecological factors comprising a soil moisture change rate, a surface vegetation coverage change rate, and a wind speed change rate;

[0030] Constructing an ecological factor time series vector for each region based on the dominant ecological factors, the ecological factor time series vector being composed of time series values of each dominant ecological factor arranged in chronological order;

[0031] Calculating a change rate of the ecological factor time series vector for each region within adjacent time windows, the change rate being represented by a ratio of a Euclidean distance of adjacent time windows to a time window length;

[0032] Matrixing the change rate for each region in chronological order and spatial position to generate a soil desertification dynamic change matrix for each region, the rows of the soil desertification dynamic change matrix representing the time dimension and the columns representing the spatial dimension.

[0033] In one embodiment, predicting a future desertification trend for each region based on the soil desertification dynamic change matrix comprises:

[0034] Performing time series decomposition on the soil desertification dynamic change matrix to obtain a trend component, a periodic component, and a noise component;

[0035] Constructing a desertification trend prediction model for each region based on the trend component, the desertification trend prediction model being trained using a long short-term memory network, the training input being historical data of the trend component, and the training output being a future prediction value of the trend component;

[0036] Performing Fourier transform on the periodic component to extract a desertification periodic feature for each region, the desertification periodic feature comprising a periodic length and a periodic amplitude;

[0037] Superimposing the future prediction value of the desertification trend prediction model and the desertification periodic feature to generate a desertification trend prediction sequence for each region;

[0038] Generating a region desertification prediction result based on the desertification trend prediction sequence and a preset desertification degree threshold interval, the region desertification prediction result comprising a change in desertification degree level for each region within a future time period.

[0039] In one embodiment, the plant species distribution and spatial layout strategy of the drought-tolerant plant community spatial configuration model is dynamically adjusted according to the regional desertification prediction results, including:

[0040] According to the change of the desertification degree level of each region in the regional desertification prediction results, determine the region that needs to be adjusted and the adjustment priority, and the adjustment priority is calculated by normalizing the product of the amplitude and the speed of the change of the desertification degree level;

[0041] For the region with deteriorating desertification degree, increase the planting density of the outer wind protection layer and the root density of the middle water locking layer, and adjust the proportion of nitrogen-fixing plants in the inner soil improvement layer;

[0042] For the region with improved desertification degree, reduce the planting density of the outer wind protection layer, increase the diversity of plant species in the middle water locking layer, and optimize the configuration of symbiotic bacteria in the inner soil improvement layer;

[0043] Based on the adjusted plant species distribution and spatial layout strategy, the spatial layer width and transition zone width of the outer wind protection layer, the middle water locking layer and the inner soil improvement layer are recalculated to generate an updated drought-tolerant plant community spatial configuration model.

[0044] In one embodiment, the surface vegetation coverage time series data and surface temperature time series data of each region are obtained based on unmanned aerial vehicle remote sensing technology, including:

[0045] Use a multi-spectral camera and a thermal imaging camera carried by a drone to obtain multi-spectral images and thermal imaging images of each region;

[0046] Perform normalized vegetation index calculation on the multi-spectral images to obtain surface vegetation coverage time series data of each region;

[0047] Temperature calibration is performed on the thermal imaging images to obtain surface temperature time series data of each region;

[0048] The surface vegetation coverage time series data and the surface temperature time series data are spatially registered, and the spatial registration includes affine transformation based on terrain feature points and non-rigid registration based on vegetation edges.

[0049] In one embodiment, the principal component analysis is performed on the multi-dimensional ecological data time series to extract the dominant ecological factors of each region, including:

[0050] Standardize the multi-dimensional ecological data time series to generate a standardized time series matrix;

[0051] obtaining eigenvalues and eigenvectors of the covariance matrix;

[0052] According to the size of the eigenvalues, selecting principal components corresponding to the first several eigenvectors with a cumulative contribution rate greater than a preset contribution rate threshold;

[0053] Linearly combining the principal components and original ecological factors in the multi-dimensional ecological data time sequence to extract the dominant ecological factors, the number of the dominant ecological factors being determined by the number of the principal components.

[0054] Compared with the prior art, the desert region drought-tolerant plant community integrated construction method provided by the present application can significantly improve the success rate and sustainability of desert ecological restoration, so that the restored vegetation community has stronger stress resistance and self-repairing ability, and can effectively cope with the impact of extreme climate events. By accurately predicting the desertification trend and intervening in advance, the overall cost of desert governance is greatly reduced, and the manpower and material resources are reduced. Systematic spatial configuration design accelerates the self-succession process of the ecosystem, shortens the ecological restoration period, and makes the desert region form stable vegetation coverage faster. The differential governance strategy optimizes the resource allocation efficiency and avoids resource waste. The synergistic configuration of plants and microorganisms promotes soil health recovery, forms a virtuous cycle of the ecosystem, and reduces the later maintenance cost. The closed-loop optimization mechanism changes the ecological restoration from passive response to active management, improving the overall governance efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 A schematic diagram of a desert region drought-tolerant plant community integrated construction method in an embodiment of the present application. DETAILED DESCRIPTION

[0056] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0057] The present embodiment provides a desert region drought-tolerant plant community integrated construction method, and the execution subjects of the method include but are not limited to computer systems, mobile monitoring stations, edge computing devices, plant ecological monitoring devices, etc. that implement the method. These execution subjects can also be regarded as general computing nodes of the present application. Among them, the computer system includes but is not limited to at least one of a server, a workstation, and an embedded controller.

[0058] Generally, the method for constructing a drought-tolerant plant community in a desert region can include: obtaining soil desertification degree distribution data and environmental dynamic data of a target desert region, dividing the target desert region into regions according to the soil desertification degree distribution data; based on the region division result and the environmental dynamic data, constructing a drought-tolerant plant community spatial configuration model for each region; monitoring the drought-tolerant plant community spatial configuration model of each region with multi-dimensional ecological data, obtaining a time sequence of multi-dimensional ecological data, and constructing a soil desertification dynamic change matrix; predicting the future desertification trend of each region based on the soil desertification dynamic change matrix, and generating a regional desertification prediction result; and dynamically adjusting the plant species distribution and spatial layout strategy of the drought-tolerant plant community spatial configuration model according to the regional desertification prediction result.

[0059] The method realizes the three-dimensional construction of the desert ecosystem through hierarchical plant community spatial configuration models, improves the resistance and self-repairing ability of the ecological system by adopting a three-layer structure design of "outer windproof layer, middle water-locking layer, and inner soil improvement layer", establishes a desertification dynamic change matrix through multi-dimensional ecological data monitoring and analysis to realize accurate prediction of the change trend of the desert ecosystem, dynamically adjusts the plant community configuration strategy based on the prediction result, forms a closed-loop optimization mechanism, and significantly improves the success rate and sustainability of vegetation restoration in the desert region.

[0060] Specifically, the detailed implementation steps of the method for constructing a drought-tolerant plant community in a desert region include:

[0061] First, soil desertification degree distribution data and environmental dynamic data of a target desert region are obtained through soil sampling and environmental monitoring equipment. The soil desertification degree distribution data includes soil particle composition, water content, organic matter content, and other key indicators, covering the target region in a grid sampling manner. The environmental dynamic data includes wind speed distribution, precipitation pattern, surface temperature change, and other meteorological factors, which are obtained through automatic weather stations and remote sensing data fusion. These data are transmitted to the data processing center through standardized data collection protocols to ensure the accuracy and integrity of the data.

[0062] According to the soil desertification degree distribution data, the target desert region is divided into regions using a hierarchical clustering algorithm, obtaining a light desertification region, a moderate desertification region, and a severe desertification region. The division process is based on the soil desertification index, considering the weighted influence of soil water content and organic matter content, to generate a continuous region division result. This region division method based on multiple indicators avoids the limitations of traditional single-index division, and can more accurately reflect the complexity of the desert ecosystem.

[0063] Based on the regional division results and environmental dynamic data, a spatial configuration model of drought-tolerant plant community for each region is constructed. The spatial configuration model includes an outer windproof layer, a middle water-locking layer, and an inner soil improvement layer, forming a three-dimensional ecological protection structure. The outer windproof layer selects trees and shrubs with high wind resistance, the middle water-locking layer selects herbaceous plants with deep root systems and high water-locking capacity, and the inner soil improvement layer selects legume plants and their symbiotic flora with nitrogen fixation capacity. This multi-level spatial configuration fully considers the synergistic effect and niche complementarity principle among plants, and constructs a plant community structure with high resistance.

[0064] A multi-dimensional ecological data monitoring system is used to monitor the spatial configuration model of drought-tolerant plant community in each region, and obtain the time series sequence of multi-dimensional ecological data in each region. The monitoring system includes a soil sensor network, a weather monitoring station, and a UAV remote sensing platform, which realizes the continuous monitoring of key indicators such as soil moisture, wind speed, organic matter content, vegetation coverage, and ground temperature. The obtained data is aligned with the time axis and processed for abnormal values to form a time series sequence of multi-dimensional ecological data, providing a high-quality data basis for subsequent analysis.

[0065] Based on the time series sequence of multi-dimensional ecological data, the dominant ecological factors are extracted through principal component analysis, and a soil desertification dynamic change matrix for each region is constructed. This matrix unifies the time and space dimensions and can directly reflect the dynamic change law of the desertification process. In the matrix construction process, the interaction and cumulative effect between ecological factors are considered, which improves the representation ability of complex ecological system changes.

[0066] Based on the soil desertification dynamic change matrix, a long short-term memory network is used to establish a desertification trend prediction model, and combined with periodic analysis, the regional desertification prediction results are generated. The prediction process not only considers the historical trend, but also integrates seasonal changes and climate cycle effects, realizing multi-scale prediction of the desertification trend. The obtained prediction results include the change of desertification degree level in each region in the future time period, providing a scientific basis for dynamic adjustment of plant community.

[0067] According to the regional desertification prediction results, the plant species distribution and spatial layout strategy of the drought-tolerant plant community spatial configuration model are dynamically adjusted. For regions with deteriorating desertification, the windproof and water-locking functions are enhanced; for regions with improved desertification, the species diversity and symbiotic flora configuration are optimized. This dynamic adjustment strategy based on prediction forms a closed-loop optimization mechanism, significantly improving the self-adaptation ability and recovery efficiency of the desert ecosystem.

[0068] Further, in the method, the detailed implementation steps of dividing the target desert region according to the soil desertification degree distribution data can include:

[0069] Firstly, the soil particle composition data, soil moisture content data and soil organic matter content data of the target desert area are obtained by soil sampling equipment. The sampling equipment includes a portable soil particle analyzer, a TDR soil moisture sensor and a carbon-nitrogen analyzer. The sampling points cover the target area in a grid layout, and the grid density is dynamically adjusted according to the complexity of the terrain to ensure the representativeness of the sampling. The soil particle composition data includes the percentage of soil particles of different particle sizes, especially the percentage of fine particles with a particle size less than 0.05 mm, which is directly related to the soil wind erosion resistance and water retention capacity. The soil moisture content data is collected by layering to reflect the vertical distribution characteristics of the soil profile, and the soil organic matter content data reflects the soil fertility status and microbial activity potential.

[0070] According to the percentage of fine particles with a particle size less than 0.05 mm in the soil particle composition data, the soil desertification index of each sampling point is calculated. The calculation formula is:

[0071] Soil desertification index = 1 - (percentage of particles with a particle size less than 0.05 mm) / (reference standard value);

[0072] Wherein, the reference standard value is the average percentage of fine particles of the same soil type in the non-desertification area. The larger the index value, the more serious the soil desertification; the smaller the value, the more stable the soil structure. In the calculation process, the weight of the abnormal sampling point data is corrected to avoid the influence of local abnormal values on the overall evaluation.

[0073] The soil desertification index is weighted to improve the comprehensiveness and accuracy of the evaluation. The weighted processing includes using the soil moisture content data as a negative correlation weight and using the soil organic matter content data as a negative correlation weight to obtain the comprehensive desertification degree value of each sampling point. The weighted calculation formula is:

[0074]

[0075] Wherein, ZH is the comprehensive desertification degree value, TZ is the soil desertification index, SH is the moisture content, SF is the moisture reference value, YH is the organic matter content, CH is the organic matter reference value, and a and β are weight coefficients, which reflect the influence strength of moisture factors and organic matter factors on the desertification degree, and are determined by historical data regression analysis. The moisture reference value and the organic matter reference value are the standard values of the non-desertification soil in the region. This multi-factor weighted processing method considers the inhibitory effect of soil moisture and organic matter on the desertification process, making the evaluation result more consistent with the ecological principle.

[0076] Based on the preset threshold interval of desertification degree, the comprehensive desertification degree value is divided into a mild desertification interval, a moderate desertification interval and a severe desertification interval, which correspond to a mild desertification region, a moderate desertification region and a severe desertification region respectively. The threshold interval setting adopts a quantile method combined with expert knowledge, to ensure that the division result conforms to the statistical distribution law and has ecological significance. In special terrain regions (such as mountain-plain transition zones), the threshold interval is corrected according to the terrain, to improve the adaptability of the division.

[0077] The divided region boundary is smoothed to obtain a continuous region division result. The smoothing process adopts a geographic spatial interpolation method combined with morphological processing, to eliminate the region boundary sawtooth effect caused by limited sampling points, and to generate a continuous region interface with ecological significance. In the smoothing process, a terrain factor is introduced as an auxiliary variable, to ensure that the region boundary is coordinated with the terrain characteristics, and to enhance the practicability of the division result. The smoothed region division result is stored in the form of grid data and vector data, to facilitate subsequent analysis and application.

[0078] Further, in the method, the detailed implementation steps of constructing the spatial configuration model of the drought-tolerant plant community of each region can include:

[0079] According to the soil desertification degree and wind speed distribution data of each region, the plant species and planting density of the outer windproof layer are determined. The plant species include trees and shrubs with high wind resistance, such as tamarix, elaeagnus angustifolia, haloxylon ammodendron, etc. These plants have developed root systems and leaf structures adapted to wind-sand environments. The planting density is adaptively adjusted based on the gradient change of the wind speed distribution data, and the calculation formula is:

[0080]

[0081] wherein ZG is the planting density, JG is the basic density, FG is the wind speed value, FH is the wind speed threshold value, the basic density is the recommended planting density under standard conditions, and γ is the wind speed response coefficient. The wind speed threshold value is 80% of the historical maximum wind speed of the region. This design increases the planting density of plants in areas with high wind speed, enhances the windproof effect, and appropriately reduces the density in areas with low wind speed, optimizing resource utilization. In special areas with extremely high wind speed, a "dense-sparse-dense" strip structure is adopted to form a wind force buffer zone, further improving the windproof effect.

[0082] According to the soil moisture evaporation rate and surface runoff data of each region, the plant species and root distribution characteristics of the middle circle water locking layer are determined. The plant species includes herbaceous plants with deep root system and high water locking capacity, such as needle grass, sand sage, and ice grass, which can reduce water evaporation loss through special physiological structure. The root distribution characteristics are characterized by the ratio between root depth and root density, which reflects the relationship between plant water acquisition strategy and soil water use efficiency. In the area with high water evaporation rate, plant species with high root ratio are selected to develop vertical roots to obtain deep water; in the area with large surface runoff, plant species with low root ratio are selected to develop horizontal roots to intercept surface runoff. This plant configuration strategy based on hydrological characteristics greatly improves the water use efficiency and is the core innovation point of the design of the middle circle water locking layer.

[0083] According to the soil organic matter content and microbial activity data of each region, the plant species and symbiotic flora configuration of the inner circle soil improvement layer are determined. The plant species includes leguminous plants with nitrogen fixation ability, such as sand punch, sweet clover, and sand wintergreen, which can improve soil fertility through symbiotic mycorrhizal and nitrogen-fixing bacteria. The symbiotic flora configuration is optimized by the soil microbial diversity index, and the calculation formula is:

[0084] Flora configuration index = nitrogen-fixing bacteria proportion × (1 + δ × microbial diversity index);

[0085] Where δ is the diversity response coefficient, which is determined by experiment. In the area with extremely low organic matter content, the proportion of nitrogen-fixing bacteria is increased to accelerate the accumulation of organic matter; in the area with high microbial activity, the optimization of microbial community structure is focused on to improve the stability of the ecosystem. The design concept of the inner circle soil improvement layer is to promote soil health through the "plant-microorganism" mutualistic symbiotic system and provide a good soil environment for the entire plant community.

[0086] Based on the terrain relief data and hydrological connectivity data of each region, the spatial layer width and transition zone width of the outer circle windproof layer, the middle circle water locking layer, and the inner circle soil improvement layer are determined. The transition zone width is normalized by the product of the standard deviation of terrain relief data and the watershed convergence index of hydrological connectivity data, and the calculation formula is:

[0087] Transition zone width = basic width × (terrain relief standard deviation × watershed convergence index) / normalization factor;

[0088] wherein the base width is the standard transition zone width under flat terrain conditions, and the normalization factor ensures that the calculation result is within a reasonable range. In areas with severe terrain changes, the transition zone width is increased to buffer the environmental gradient changes; in areas with high hydrological connectivity, the transition zone width is also increased accordingly to promote the effective transfer of water between layers. This spatial layout optimization based on terrain and hydrological characteristics greatly improves the ecological adaptability and anti-interference ability of plant community structure.

[0089] Further, in the method, the detailed implementation steps for monitoring the multi-dimensional ecological data of each region to obtain the time series of multi-dimensional ecological data of each region can include:

[0090] Soil moisture sensors, wind speed sensors, and soil organic matter sensors are deployed in the outer windproof layer, the middle water-locking layer, and the inner soil improvement layer of each region, respectively, to construct a multi-level and full-coverage ecological monitoring network. The soil moisture sensors use TDR technology and have multi-depth monitoring capability to capture the vertical distribution characteristics of soil profile water; the wind speed sensors use ultrasonic principle and have low power consumption and high precision to monitor the near-surface wind field distribution; the soil organic matter sensors are based on near-infrared spectroscopy analysis technology to realize in-situ rapid determination of soil organic matter content. These sensors are connected to the edge computing unit through a wireless sensor network to form a distributed monitoring system, obtaining the time series data of soil moisture, wind speed, and soil organic matter of each layer.

[0091] The time series data of surface vegetation coverage and the time series data of surface temperature of each region are obtained based on unmanned aerial vehicle remote sensing technology. The unmanned aerial vehicle is equipped with a multi-spectral camera and a thermal imaging camera, and regularly performs aerial survey on the target region according to the preset flight route to obtain high-resolution remote sensing data. The vegetation coverage data is obtained by calculating the normalized difference vegetation index (NDVI), and the surface temperature data is obtained by thermal infrared radiation correction. In the data collection process, the sunlight angle compensation algorithm and the atmospheric influence correction model are introduced to improve the accuracy of the remote sensing data. To ensure the spatial and temporal continuity of the data, a fixed flight route and a fixed time period are established for collection procedures to minimize the data fluctuations caused by non-ecological factors.

[0092] The time series data of soil moisture, wind speed, soil organic matter, surface vegetation coverage, and surface temperature are time-axis aligned to generate the time series of multi-dimensional ecological data of each region. The dynamic time warping algorithm is used for time-axis alignment to solve the problem of inconsistent time stamps of different monitoring data, ensuring data comparability. In the alignment process, considering the sensor data delay and the difference in remote sensing data acquisition period, a time compensation mechanism is introduced to realize accurate fusion of heterogeneous data sources. The aligned multi-dimensional ecological data time series is organized in a unified time scale to provide a structured data basis for subsequent analysis.

[0093] Anomaly value detection is performed on the multi-dimensional ecological data time series to improve data quality. The anomaly value detection includes calculating a local outlier factor of each time series data, marking a data point with a local outlier factor greater than a preset outlier threshold as an anomaly value and performing interpolation repair. The local outlier factor calculation is based on the statistical distance of a data point and its time neighborhood points, which can effectively identify mutation points and anomaly values in the time series data. In the anomaly value repair, a constraint interpolation algorithm that maintains the trend of ecological change is used to avoid introducing new data distortion in the repair process. The repaired multi-dimensional ecological data time series has high integrity and continuity, providing reliable data support for constructing the soil desertification dynamic change matrix.

[0094] Further, in the method, the detailed implementation steps of constructing the soil desertification dynamic change matrix of each region based on the multi-dimensional ecological data time series can include:

[0095] Principal component analysis is performed on the multi-dimensional ecological data time series to extract the dominant ecological factors of each region. Principal component analysis can identify key variables from high-dimensional data, reduce data redundancy, and improve analysis efficiency. The dominant ecological factors include soil moisture change rate, surface vegetation coverage change rate, and wind speed change rate, which are directly related to the dynamic change of the desertification process. In the principal component analysis process, the Kaiser criterion combined with the scree plot method is used to determine the optimal number of principal components, which not only retains key information but also avoids oversimplification. To improve the interpretability of the principal components, the Varimax orthogonal rotation is performed on the principal component loading matrix, making the correspondence between the principal components and the original ecological factors more explicit.

[0096] Based on the dominant ecological factors, an ecological factor time series vector of each region is constructed, which is composed of the time series values of each dominant ecological factor in chronological order. The time series vector uses multi-resolution representation, which contains both original high-frequency sampling data and aggregated statistical values of different time scales, and can capture both short-term fluctuations and long-term trends. In the vector construction process, environmental background variables (such as seasonal cycles and climate anomaly events) are introduced as auxiliary information to enhance the ecological interpretability of the time series data. The constructed ecological factor time series vector is a mathematical representation of the dynamic characteristics of the desertification process, providing a basis for subsequent change rate calculation.

[0097] The change rate of the ecological factor time series vector of each region within adjacent time windows is calculated to quantify the dynamic characteristics of the desertification process. The change rate is characterized by the ratio of the Euclidean distance of adjacent time windows to the length of the time window, and the calculation formula is:

[0098]

[0099] where VH is the rate of change, V(t) and V(t+1) are the ecological factor time series of adjacent time windows, ||·|| is the Euclidean norm, and Δt is the length of the time window. This calculation method can comprehensively consider the coordinated changes of multiple ecological factors and provide an overall measure of the dynamics of the desertification process. In the rate of change calculation, an adaptive time window technique is used to dynamically adjust the window length according to the data change speed, use a shorter window for fast-changing processes, and use a longer window for slow-changing processes, thereby improving the sensitivity and adaptability of the rate of change calculation.

[0100] The rate of change of each region is arranged in matrix form according to the time sequence and spatial position to generate a soil desertification dynamic change matrix for each region. The rows of the soil desertification dynamic change matrix represent the time dimension, the columns represent the spatial dimension, and the matrix element values represent the desertification change rate at a specific spatiotemporal point. This matrix representation method unifies time evolution and spatial distribution into one mathematical structure, facilitating global analysis and pattern recognition. In the matrix construction process, the spatial position is normalized using a geographic weighting method to eliminate the effects of region size and shape differences, making the matrix elements of different regions comparable. The constructed soil desertification dynamic change matrix is a complete expression of the spatiotemporal characteristics of the desertification process, providing a key input for desertification trend prediction.

[0101] Further, in the method, the detailed implementation steps for predicting the future desertification trend of each region based on the soil desertification dynamic change matrix can include:

[0102] The soil desertification dynamic change matrix is decomposed in time sequence to obtain trend components, periodic components, and noise components. The time sequence decomposition uses a combination of wavelet transform and empirical mode decomposition, which can effectively process non-stationary time series and separate the change characteristics of different time scales. The trend component reflects the long-term evolution direction of the desertification process, the periodic component reflects the seasonal and interannual fluctuation characteristics, and the noise component contains random fluctuations and measurement errors. In the decomposition process, consistent decomposition parameters are used for matrices of different regions to ensure the comparability of the component extraction; at the same time, the optimal decomposition level is set according to the characteristic time scale of the desert ecosystem, balancing the decomposition precision and computational complexity.

[0103] The sandification trend prediction model of each region is constructed based on the trend component. The long short-term memory network is used to train the sandification trend prediction model. The long short-term memory network is a special recurrent neural network with the ability to capture long-term dependencies, which is suitable for processing time series with long-term memory characteristics such as sandification process. The sliding window method is used to generate training samples, and the historical data sequence is used as input and the future data point is used as output. The network parameters are optimized by the back propagation algorithm. In the training process, the early stopping strategy and regularization technique are introduced to prevent model overfitting. At the same time, the learning rate decay strategy is adopted to improve the training stability and convergence speed. The trained model can predict the future trend evolution based on the historical trend component, providing a main framework for sandification trend prediction.

[0104] The sandification cycle characteristics of each region are extracted by Fourier transform of the cycle component. The sandification cycle characteristics include cycle length and cycle amplitude, which reflect the time scale and intensity of the periodic change of the sandification process, respectively. Fourier transform can convert time domain signals into frequency domain representation, which can intuitively display the periodic structure of the signal. In the cycle analysis, the power spectral density estimation method is used to identify significant cycle components. For non-integer period signals, the Lomb-Scargle periodogram method is used to improve the accuracy of cycle detection. The extracted cycle characteristics not only include seasonal cycles, but also include long-period components related to climate cycles (such as El Nino Southern Oscillation), fully capturing the periodic variation of the sandification process.

[0105] The future prediction value of the sandification trend prediction model and the sandification cycle characteristics are superimposed to generate the sandification trend prediction sequence of each region. The superposition process considers the interaction between trend and cycle, and uses phase locking technology to ensure the time alignment of cycle component and trend component. The generation formula of the prediction sequence is:

[0106]

[0107] where XU is the prediction sequence, YT is the trend prediction value, Ai, Ti and are the amplitude, cycle length and phase of the i-th cycle component, respectively. This trend-cycle superposition method not only retains the prediction ability of long-term trend, but also integrates the influence of periodic fluctuations, improving the comprehensiveness and accuracy of prediction.

[0108] According to the sandification trend prediction sequence and the preset sandification degree threshold interval, a regional sandification prediction result is generated. The regional sandification prediction result includes the sandification degree grade change of each region in a future time period, and the time point and change rate of the grade change. In the result generation process, prediction uncertainty analysis is introduced, a prediction interval is generated by a Monte Carlo simulation method, and the reliability of the prediction result is quantified; at the same time, combined with sensitivity analysis, the key factors affecting the prediction accuracy are identified, and a scientific basis is provided for the prediction result explanation. The final prediction result is presented in a spatio-temporal visualization manner, which intuitively displays the spatio-temporal evolution mode of the sandification trend of each region, and provides decision support for the dynamic adjustment of plant community configuration.

[0109] Further, in the method, the detailed implementation steps of dynamically adjusting the plant species distribution and spatial layout strategy of the drought-tolerant plant community spatial configuration model according to the regional sandification prediction result can include:

[0110] According to the sandification degree grade change of each region in the regional sandification prediction result, the regions that need to be adjusted and the adjustment priority are determined. The adjustment priority is normalized by the product of the sandification degree grade change amplitude and the change speed, and the calculation formula is:

[0111] Adjustment priority = (grade change amplitude x change speed) / normalization factor

[0112] Wherein, the grade change amplitude is the maximum change of the sandification grade in the prediction period, the change speed is the slope of the grade change, and the normalization factor ensures that the priority value is in the interval [0, 1]. This priority calculation method based on change characteristics can effectively identify the regions that need urgent intervention and optimize the resource allocation efficiency. In the priority determination process, the ecological sensitivity and protection importance of the region are also considered, and higher priority is given to ecologically sensitive regions or regions with important protection functions to ensure the stability of key nodes of the ecological system.

[0113] For the regions with deteriorating sandification degree, an enhanced adjustment strategy is adopted to increase the planting density of the outer windproof layer and the root density of the middle water-locking layer, and to adjust the proportion of nitrogen-fixing plants in the inner soil improvement layer. Specifically, the increase amplitude of the planting density of the outer windproof layer is proportional to the predicted increase amplitude of the wind speed, the increase amplitude of the root density of the middle water-locking layer is proportional to the predicted decrease amplitude of the water, and the increase amplitude of the proportion of nitrogen-fixing plants in the inner soil improvement layer is proportional to the predicted decrease amplitude of the organic matter. This "symptomatic treatment" adjustment strategy can specifically enhance the ability of the plant community to cope with deteriorating conditions and effectively curb the sandification process. In terms of plant species selection, the proportion of species with strong stress resistance is increased, such as Tamarix with high drought resistance coefficient and Psammochloa with strong wind erosion resistance, to improve the ecological resilience of the community.

[0114] For the areas with improved desertification, an optimization adjustment strategy is adopted, which reduces the planting density of the outer windproof layer, increases the diversity of plant species in the middle water-locking layer, and optimizes the configuration of the symbiotic bacteria community in the inner soil improvement layer. The increase in plant diversity follows the principle of niche complementarity, selecting species with different niche characteristics, such as different root depths and different photosynthetic pathways, to maximize resource utilization efficiency. The optimization of the symbiotic bacteria community focuses on the functional diversity of the bacteria community, increasing the proportion of functional bacteria such as phosphorus-dissolving bacteria and growth-promoting bacteria to promote soil nutrient cycling and plant growth. This optimization strategy based on improvement trends aims to promote the transition of the ecosystem from "resistance" to "sustainability", and improve the self-maintenance ability and ecological value of the system.

[0115] Based on the adjusted plant species distribution and spatial layout strategy, the spatial layering width of the outer windproof layer, the middle water-locking layer, and the inner soil improvement layer, and the transition zone width are recalculated to generate an updated spatial configuration model of the drought-tolerant plant community. The calculation of the spatial layering width considers the wind speed gradient, water gradient, and soil property gradient to establish a mapping relationship between "environmental gradient-space structure", so that the spatial structure of the plant community matches the environmental characteristics. The optimization of the transition zone width is based on the strength of species interaction, increasing the transition zone width in areas with strong mutualistic interaction to promote positive interaction, and reducing the transition zone width in areas with strong competitive interaction to reduce negative interaction. The updated spatial configuration model is verified through computer simulation to evaluate its adaptability to the predicted environmental conditions, ensuring the scientificity and effectiveness of the adjustment scheme.

[0116] Further, in the method, the detailed implementation steps for obtaining the time series data of ground vegetation coverage and the time series data of ground temperature of each region based on unmanned aerial vehicle remote sensing technology can include:

[0117] A multi-spectral camera and a thermal imaging camera are mounted on the unmanned aerial vehicle to obtain multi-spectral images and thermal imaging images of each region. The unmanned aerial vehicle adopts a hybrid design of fixed wings and rotors, combining cruise efficiency and hovering capability, suitable for long-distance monitoring in desert environments. The multi-spectral camera has five wavebands of red, green, blue, near-infrared, and red edge, with a resolution better than 10 cm / pixel, capable of capturing subtle vegetation changes; the thermal imaging camera has a temperature resolution better than 0.1°C and a spatial resolution better than 30 cm / pixel, capable of accurately measuring the distribution of ground temperature. Data collection is performed according to the pre-set flight route, which takes into account the area, terrain changes, and monitoring accuracy requirements to ensure complete data coverage and appropriate overlap. In special weather conditions (such as strong winds), the system automatically adjusts the flight parameters or re-plans the flight route to ensure data quality.

[0118] The multi-spectral images are subjected to normalized difference vegetation index (NDVI) calculation to obtain the time series data of ground vegetation coverage of each region. The normalized difference vegetation index (NDVI) calculation is based on the reflectance difference between the near-infrared and red light wavebands, with the formula:

[0119]

[0120] where NIR and RED represent the reflectance of near-infrared and red light bands, respectively. The NDVI value ranges between [-1, 1], and the higher the value, the better the vegetation coverage. To improve the accuracy of NDVI calculation, atmospheric scattering correction and soil background adjustment are introduced to reduce the interference of non-vegetation factors. In the condition of sparse vegetation in the desert, the enhanced vegetation index (EVI) is also used as an auxiliary index. EVI has lower sensitivity to soil background and atmospheric influence, and can more accurately represent low-coverage vegetation. Based on the calculation results of NDVI and EVI, the absolute vegetation coverage is converted through a regression model to form standardized vegetation coverage time series data.

[0121] The thermal imaging image is temperature calibrated to obtain the surface temperature time series data of each region. The temperature calibration process includes three key steps: radiation correction, atmospheric influence compensation, and surface emissivity correction. Radiation correction converts the digital value output by the thermal imager into radiance; atmospheric influence compensation eliminates the influence of atmospheric absorption and scattering on thermal radiation; surface emissivity correction considers the emissivity difference of different ground surface types to improve the accuracy of temperature calculation. In the desert environment, surface emissivity correction is particularly important because different degrees of sandy surface have different thermal radiation characteristics. This method innovatively introduces an emissivity estimation model based on spectral characteristics to automatically estimate the surface emissivity based on multispectral data, greatly improving the accuracy of temperature calibration. The calibrated temperature data is subjected to spatial filtering and time smoothing processing to form high-quality surface temperature time series data.

[0122] The surface vegetation coverage time series data and the surface temperature time series data are spatially registered to ensure spatial consistency of the data. Spatial registration includes two stages: affine transformation based on terrain feature points and non-rigid registration based on vegetation edges. Affine transformation solves the overall geometric deformation, and non-rigid registration handles local deformation. The combination of the two achieves high-precision registration. In feature point selection, time-stable terrain features (such as ridge lines and valley lines) are preferred as control points to improve the reliability of registration; in non-rigid registration, a mutual information maximization method based on vegetation edges is used to adapt to the challenge of sparse features in the desert environment. The registration accuracy is controlled by the root mean square error (RMSE) to ensure that the error is less than one pixel. The registered data has a pixel-level spatial correspondence, supporting subsequent pixel-level analysis and spatiotemporal pattern recognition.

[0123] Further, in the method, the detailed implementation steps of performing principal component analysis on the multi-dimensional ecological data time series to extract the dominant ecological factors of each region can include:

[0124] The multi-dimensional ecological data time series is standardized to generate a standardized time series matrix. Standardization eliminates the dimensional differences between different ecological factors, making the data comparable. Z-score method is used for standardization. This standardization makes the mean of each ecological factor 0 and the standard deviation 1, eliminating the influence of different measurement scales. In the standardization process, considering the possible non-normal distribution characteristics of ecological data, Box-Cox transformation is used for pre-processing of data with obvious skewness, so as to make it closer to normal distribution and improve the effectiveness of principal component analysis. The standardized time series matrix retains the time structure of the original data, with each row representing a time point and each column representing an ecological factor.

[0125] The covariance matrix of the standardized time series matrix is calculated to obtain the eigenvalues and eigenvectors of the covariance matrix. The covariance matrix reflects the correlation between different ecological factors and is the core data structure of principal component analysis. The calculation of eigenvalues and eigenvectors uses singular value decomposition (SVD) algorithm, which has better numerical stability than traditional eigenvalue decomposition method, especially suitable for processing high-dimensional ecological data. In the calculation process, sparse matrix optimization technology is used to improve the efficiency of large-scale data processing; at the same time, parallel computing is used to speed up the eigenvalue decomposition to meet the real-time analysis requirements. Eigenvalues represent the variance explained by each principal component, and eigenvectors define the mapping relationship of original ecological factors to principal component space.

[0126] According to the size of the eigenvalues, the principal components corresponding to the first several eigenvectors with cumulative contribution rate greater than the preset contribution rate threshold are selected. The cumulative contribution rate is the proportion of the total variance explained by the selected principal components to the total variance of the original data, which is a key indicator for determining the number of principal components. The preset contribution rate threshold is usually set to 85% to 95%, which reasonably reduces the dimension while retaining the key information. In the selection of eigenvectors, in addition to considering the cumulative contribution rate, the Kaiser criterion (eigenvalue greater than 1) and the scree plot method are also used to determine the optimal number of principal components. This multi-criteria selection method can achieve a good balance between information retention and dimension reduction, avoiding excessive simplification or redundant retention caused by simple application of a single criterion.

[0127] The principal components are linearly combined with the original ecological factors in the multi-dimensional ecological data time series to extract the dominant ecological factors. The dominant ecological factors are a dimension-reduced representation of the original multi-dimensional data, which not only retains the key information of the original data but also simplifies the data structure, facilitating subsequent analysis and interpretation. The weights of linear combination are determined based on the element values of eigenvectors, and the calculation formula is:

[0128] Z_i = ∑(aij × Y_j)

[0129] Wherein, aij is the jth element of the ith eigenvector, Z_i is the ith dominant ecological factor, Y_j is the jth original ecological factor. In order to improve the interpretability of the dominant ecological factor, the principal component loading matrix is rotated by Varimax orthogonal rotation, so that each principal component is mainly related to a few original variables, reduces the cross load, and clarifies the ecological significance of the principal component. The rotated dominant ecological factor usually corresponds to the index with clear ecological significance such as soil moisture variation rate, surface vegetation coverage rate variation rate and wind speed variation rate, which provides a scientific framework for sandification dynamic analysis.

[0130] Through the above embodiments, the application constructs a drought-tolerant plant community space configuration model with a three-layer structure of "outer circle windproof layer, middle circle water locking layer, and inner circle soil improvement layer", and realizes dynamic optimization and adjustment of the plant community through multidimensional ecological data monitoring, sandification dynamic analysis and trend prediction, significantly improving the repair efficiency and sustainability of the desert ecosystem.

[0131] It should be noted that the above formulas are dimensionless and the numerical values are calculated, the formula is obtained by software simulation of a large number of collected data to reflect the current real situation, and the preset parameters and threshold values in the formula are set by the person skilled in the art according to the actual situation.

[0132] The above is only the preferred embodiment of the application, and the protection scope of the application is not limited to the above-mentioned embodiments. Any technical solution falling within the scope of the application is within the protection scope of the application. It should be noted that for ordinary technical users in the technical field, some improvements and refinements without departing from the principles of the application are also considered as the protection scope of the application.

Claims

1. A method for the integrated construction of drought-resistant plant communities in desert areas, characterized in that, The method comprises the following steps: acquiring soil desertification degree distribution data and environmental dynamic data of a target desert area, and dividing the target desert area according to the soil desertification degree distribution data to obtain a slight desertification region, a moderate desertification region and a severe desertification region; based on the region division result and the environmental dynamic data, a spatial configuration model of a drought-resistant plant community in each region is constructed, the spatial configuration model comprising an outer windproof layer, a middle water-locking layer and an inner soil improvement layer; multi-dimensional ecological data of the drought-resistant plant community in each region is monitored to obtain a time sequence of multi-dimensional ecological data of each region, and a soil desertification dynamic change matrix of each region is constructed based on the time sequence of multi-dimensional ecological data; based on the soil desertification dynamic change matrix, a future desertification trend of each region is predicted to generate a region desertification prediction result; according to the region desertification prediction result, a plant species distribution and spatial layout strategy of the spatial configuration model of the drought-resistant plant community is dynamically adjusted to realize integrated construction of the drought-resistant plant community in the desert area.

2. The method according to claim 1, wherein the method is characterized by, The region division of the target desert area according to the soil desertification degree distribution data comprises the following steps: acquiring soil particle composition data, soil moisture content data and soil organic matter content data of the target desert area; calculating a soil desertification index of each sampling point according to a proportion of fine particles with a particle size less than 0.05 mm in the soil particle composition data; performing weighted processing on the soil desertification index, the weighted processing comprising taking the soil moisture content data as a negative correlation weight and taking the soil organic matter content data as a negative correlation weight to obtain a comprehensive desertification degree value of each sampling point; based on a preset desertification degree threshold interval, the comprehensive desertification degree value is divided into a slight desertification interval, a moderate desertification interval and a severe desertification interval, which correspond to the slight desertification region, the moderate desertification region and the severe desertification region respectively; the boundaries of the divided regions are smoothed to obtain a continuous region division result.

3. The method according to claim 1, wherein the method is characterized by, The construction of the spatial configuration model of the drought-resistant plant community in each region comprises the following steps: determining plant species and planting density of the outer windproof layer according to soil desertification degree and wind speed distribution data of each region, the plant species comprising trees and shrubs with high wind resistance, and the planting density being adaptively adjusted based on gradient changes of wind speed distribution data; determining plant species and root distribution characteristics of the middle water-locking layer according to soil water evaporation rate and surface runoff data of each region, the plant species comprising herbaceous plants with deep root systems and high water-locking capacity, and the root distribution characteristics being represented by a ratio between root depth and root density; determining plant species and symbiotic flora configuration of the inner soil improvement layer according to soil organic matter content and microbial activity data of each region, the plant species comprising leguminous plants with nitrogen fixation capacity, and the symbiotic flora configuration being optimized by a soil microbial diversity index. Based on the terrain relief data and hydrological connectivity data of each region, the spatial layering width of the outer circle wind prevention layer, the middle circle water locking layer and the inner circle soil improvement layer and the transition zone width are determined, and the transition zone width is normalized by the product of the standard deviation of the terrain relief data and the watershed convergence index of the hydrological connectivity data.

4. The method according to claim 1, wherein the method is characterized by, The multi-dimensional ecological data time series of each region is obtained by monitoring the spatial configuration model of the drought-resistant plant community in each region, including: In each region, soil moisture sensors, wind speed sensors and soil organic matter sensors are arranged in the outer circle wind prevention layer, the middle circle water locking layer and the inner circle soil improvement layer respectively, and soil moisture time series data, wind speed time series data and soil organic matter time series data of each layer are obtained; Based on the unmanned aerial vehicle remote sensing technology, the surface vegetation coverage time series data and the surface temperature time series data of each region are obtained; The time axis of the soil moisture time series data, the wind speed time series data, the soil organic matter time series data, the surface vegetation coverage time series data and the surface temperature time series data is aligned to generate the multi-dimensional ecological data time series of each region; The multi-dimensional ecological data time series is subjected to outlier detection, which includes calculating the local outlier factor of each time series data, marking the data points with local outlier factor greater than the preset outlier threshold as outliers and performing interpolation repair.

5. The method according to claim 1, wherein the method is characterized by, The soil desertification dynamic change matrix of each region is constructed based on the multi-dimensional ecological data time series, including: The multi-dimensional ecological data time series is subjected to principal component analysis, and the dominant ecological factors of each region are extracted, including soil moisture change rate, surface vegetation coverage change rate and wind speed change rate; Based on the dominant ecological factors, the ecological factor time series vector of each region is constructed, which is composed of the time series values of each dominant ecological factor in time sequence; The change rate of the ecological factor time series vector of each region in the adjacent time window is calculated, which is characterized by the ratio of the Euclidean distance of adjacent time windows to the length of time window; The change rate of each region is arranged in matrix according to time sequence and spatial position to generate the soil desertification dynamic change matrix of each region, and the row represents time dimension and the column represents space dimension.

6. The method according to claim 1, wherein the method is characterized by, The future desertification trend of each region is predicted based on the soil desertification dynamic change matrix, including: The soil desertification dynamic change matrix is subjected to time series decomposition to obtain trend component, periodic component and noise component; Based on the trend component, the desertification trend prediction model of each region is constructed, which is trained by long short-term memory network, the training input is the historical data of the trend component, and the training output is the future prediction value of the trend component; The periodic component is subjected to Fourier transform to extract the desertification periodic feature of each region, including the period length and the period amplitude; The future prediction value of the desertification trend prediction model and the desertification periodic feature are superimposed to generate the desertification trend prediction sequence of each region; According to the sandification trend prediction sequence and a preset sandification degree threshold interval, a regional sandification prediction result is generated, which includes sandification degree level changes of each region in a future time period.

7. The method according to claim 1, wherein the method is characterized by, The plant species distribution and spatial layout strategy of the drought-tolerant plant community spatial configuration model is dynamically adjusted according to the regional sandification prediction result, including: According to the sandification degree level changes of each region in the regional sandification prediction result, a region that needs to be adjusted and an adjustment priority are determined, and the adjustment priority is normalized by the product of the amplitude and the speed of the sandification degree level change; For a region with deteriorating sandification degree, the planting density of the outer windproof layer and the root density of the middle water-locking layer are increased, and the proportion of nitrogen-fixing plants in the inner soil improvement layer is adjusted; For a region with improved sandification degree, the planting density of the outer windproof layer is reduced, the diversity of plant species in the middle water-locking layer is increased, and the configuration of symbiotic bacteria in the inner soil improvement layer is optimized; Based on the adjusted plant species distribution and spatial layout strategy, the spatial layering width and the transition zone width of the outer windproof layer, the middle water-locking layer and the inner soil improvement layer are recalculated to generate an updated drought-tolerant plant community spatial configuration model.

8. The method according to claim 4, wherein the method is characterized by, When the unmanned aerial vehicle remote sensing technology is used to obtain the ground vegetation coverage rate time series data and the ground temperature time series data of each region, it includes: A multi-spectral camera and a thermal imaging camera are carried by an unmanned aerial vehicle to obtain multi-spectral images and thermal imaging images of each region; The normalized vegetation index of the multi-spectral image is calculated to obtain the ground vegetation coverage rate time series data of each region; The temperature of the thermal imaging image is calibrated to obtain the ground temperature time series data of each region; The ground vegetation coverage rate time series data and the ground temperature time series data are spatially registered, and the spatial registration includes affine transformation based on terrain feature points and non-rigid registration based on vegetation edges.

9. The method according to claim 5, wherein the method is characterized by, The principal component analysis of the multi-dimensional ecological data time series is performed to extract the dominant ecological factors of each region, including: The multi-dimensional ecological data time series is standardized to generate a standardized time series matrix; The covariance matrix of the standardized time series matrix is calculated to obtain the eigenvalues and eigenvectors of the covariance matrix; According to the size of the eigenvalues, the principal components corresponding to the first several eigenvectors with a cumulative contribution rate greater than a preset contribution rate threshold are selected; The principal components and the original ecological factors in the multi-dimensional ecological data time series are linearly combined to extract the dominant ecological factors, and the number of dominant ecological factors is determined by the number of principal components.

Images