Drought multi-dimensional risk partitioning method and system based on clustering algorithm
By combining self-organizing mapping networks and the K-means algorithm, a multidimensional feature of drought is constructed, which solves the problems of single-dimensional analysis and subjective classification in traditional drought research. It realizes the objective identification and spatial zoning of drought risk levels, and improves the scientificity and accuracy of drought assessment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-19
- Publication Date
- 2026-04-03
AI Technical Summary
Traditional drought research often relies on single-dimensional analysis, resulting in incomplete drought descriptions and difficulty in fully revealing the overall pattern of regional drought conditions. Furthermore, existing drought classification methods are subjective and lack a systematic identification of the intrinsic relationships between multidimensional characteristics.
A drought multidimensional risk zoning method based on clustering algorithm is adopted. Initial clustering is performed through self-organizing map network (SOM), followed by secondary clustering with K-means algorithm to construct drought multidimensional features. The optimal number of clusters is determined by using drought duration, intensity and frequency indicators, combined with clustering performance evaluation indicators, and then visualized.
It has achieved objective classification and spatial distribution identification of drought risk levels, improved classification accuracy and identification efficiency, provided scientific support for drought prevention strategies, and overcome the subjectivity and spatial identification deficiencies of traditional methods.
Smart Images
Figure CN121786525A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of climate change risk assessment, and specifically relates to a method and system for multi-dimensional drought risk zoning based on clustering algorithms. Background Technology
[0002] In recent years, the trend of global climate change has become increasingly pronounced, and the frequent occurrence of extreme weather events has become one of the important factors affecting the stability of ecosystems and the sustainable development of human society. Drought, as one of the most prevalent and destructive natural disasters under the background of climate change, is showing a trend of increasing frequency, intensity, and scope. With the intensification of drought events, their impact on natural ecosystems and socio-economic systems is becoming increasingly prominent. Drought can cause reduced crop yields and increased food security risks, and has a lasting impact on vegetation growth, ecosystem function, water resource regulation, and regional economic development. Prolonged drought can even lead to land degradation, ecosystem structural imbalance, and carbon cycle disruption, thereby exacerbating regional ecological vulnerability and environmental risks. Therefore, conducting scientific and systematic drought risk assessments is of significant theoretical and practical value for a deeper understanding of the response mechanisms and adaptability of ecosystems under the background of climate change, for rationally planning water resource allocation, and for optimizing agricultural and ecological management patterns. Quantitatively identifying the drought risk levels in different regions can provide a scientific basis for drought disaster prevention and control, ecological restoration, and regional sustainable development, thereby promoting the coordination of human-land systems and the long-term stability of the ecological environment.
[0003] Traditional drought research typically relies on a single drought index, using one characteristic—frequency, duration, or intensity—of drought events as the evaluation criterion to reflect temporal or spatial variations. However, drought processes are multidimensional, and analysis based on a single dimension can lead to incomplete descriptions and fail to comprehensively reveal the overall pattern of regional drought conditions. Current drought hazard assessment research focuses on two main aspects: the construction and identification of drought characteristic indicators, and the classification and spatial zoning of drought levels. In drought identification, researchers usually extract three types of indicators—drought frequency, drought duration, and drought intensity—from time-series data to characterize the probability, duration, and intensity of drought events. However, traditional studies often employ unidimensional analysis or empirically weighted synthesis, lacking a systematic identification of the inherent relationships between multidimensional characteristics, resulting in deficiencies in spatial continuity and objectivity in the assessment results. Regarding level classification, common methods include the analytic hierarchy process (AHP) and fuzzy comprehensive evaluation. These methods still possess strong subjectivity in weight allocation and evaluation criteria, making it difficult to fully reflect the comprehensive differences in drought characteristics across different regions. Summary of the Invention
[0004] The purpose of this invention is to provide a drought multi-dimensional risk zoning method and system based on clustering algorithms to solve the above-mentioned problems.
[0005] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a drought multi-dimensional risk zoning method based on clustering algorithms, including: Acquire drought index data with continuous time series within the study area, and obtain a spatiotemporal dataset of drought index through data quality control; Based on the aforementioned drought index spatiotemporal dataset, three drought characteristic indicators—drought duration, drought intensity, and drought frequency—are calculated to construct a multidimensional drought feature. The multidimensional features of drought are input into the self-organizing map network for preliminary clustering to obtain the neuron weight vector. The neuron weight vector is then used as input for secondary clustering, and the optimal number of clusters is determined by combining the clustering performance evaluation index. Based on the clustering results, the drought risk levels are ranked, classified, and visualized.
[0006] Furthermore, the acquisition of drought index data with continuous time series within the study area includes: Drought index data with continuous time series were selected, including the Standardized Precipitation Index (SPI), Standardized Precipitation Evapotranspiration Index (SPEI), Palmer Drought Index (PDSI), and Standardized Evapotranspiration Deficit Index (SEDI).
[0007] Furthermore, the method of obtaining the drought index spatiotemporal dataset through data quality control includes: The acquired drought index data was rasterized for the study area, and data quality control steps such as missing value imputation, outlier detection, and spatial resampling were completed to form a continuous temporal series drought index spatiotemporal dataset within the study area.
[0008] Furthermore, based on the aforementioned drought index spatiotemporal dataset, three drought characteristic indicators—drought duration, drought intensity, and drought frequency—are calculated to construct a multidimensional drought feature, including: Based on the processed drought index time series, an event identification framework using graded thresholds and runs theory was adopted to calculate three drought characteristic indicators: drought duration, drought intensity, and drought frequency. Drought duration reflects the duration of the event, drought intensity describes the intensity and cumulative impact of drought, and drought frequency is used to characterize the probability of occurrence of drought events of different levels. For different drought indices, different drought thresholds are set based on drought level classification standards, including mild drought, moderate drought, severe drought, and extreme drought, with corresponding thresholds of [missing information]. , , , When the drought index is below A drought is considered to occur when the condition is met continuously for a period of time. A drought event is defined as a period of time when this condition is met consecutively. The duration and intensity of each drought event are calculated using runs theory.
[0009] In the formula, For the first The duration of a drought event; For the first The number of months in which each drought event began; No. The number of months since the end of each drought event;
[0010] In the formula, For the first The intensity of a drought event; for The drought index value at any given time; the frequency of occurrence of different levels of drought is calculated based on different drought thresholds:
[0011] In the formula, For level The frequency of drought events; For level The number of months in which drought events occurred; This represents the total number of months in the study period.
[0012] Furthermore, the step of inputting the multidimensional drought features into a self-organizing map network for preliminary clustering to obtain neuron weight vectors includes: The three drought features calculated—drought duration, drought intensity, and drought frequency—are imported into a self-organizing map network (SOM) as input variables. Through an unsupervised learning process, the SOM network can automatically identify the potential distribution structure of the input data and map high-dimensional feature information to a low-dimensional topological space, achieving preliminary clustering and pattern recognition of regional drought features. Two indicators, quantization error and topology error, are calculated. Quantization error is used to evaluate the ability of the established network to distinguish the features of the input data; the smaller the error, the more accurately the network can reflect the spatial distribution characteristics of the input samples. Topology error is used to evaluate the network's ability to preserve the neighborhood during mapping; the lower the value, the more the network preserves the topological structure of the input space during the mapping process.
[0013] Furthermore, the step of using neuron weight vectors as input to perform secondary clustering, and determining the optimal number of clusters by combining clustering performance evaluation metrics, includes: The neuron weight vectors output by SOM are used as input data, and K-means clustering algorithm is used for secondary clustering. The K-means algorithm aims to minimize intra-class sample differences and maximize inter-class differences. Through iterative optimization, the cluster center positions are continuously adjusted to achieve refined classification of drought feature data. The Calinski-Harabasz index is introduced as a clustering performance evaluation index. The larger the value, the better the clustering effect. By calculating the CH index values under different numbers of clusters, the number of clusters corresponding to the peak value is selected as the optimal partitioning scheme.
[0014] Furthermore, based on the clustering results, the drought risk levels are ranked, classified, and visualized, including: Based on the clustering results, the comprehensive drought characteristics of each category of samples are statistically analyzed. The clustering results are ranked according to the comprehensive level of drought duration, intensity and frequency to determine the relative level of drought risk. Combined with the geographic information system (GIS) platform, the clustering results are spatially visualized to form a drought risk zoning map.
[0015] Secondly, this invention provides a drought multi-dimensional risk zoning system based on clustering algorithms, comprising: The data acquisition module is used to acquire drought index data with continuous time series within the study area, and obtain a spatiotemporal dataset of drought index through data quality control. The multidimensional feature acquisition module is used to calculate three drought feature indicators—drought duration, drought intensity, and drought frequency—based on the drought index spatiotemporal dataset, and to construct multidimensional drought features. The clustering module is used to input drought multidimensional features into the self-organizing map network for preliminary clustering, obtain neuron weight vectors, use neuron weight vectors as input for secondary clustering, and combine clustering performance evaluation indicators to determine the optimal number of clusters. The output display module is used to sort and visualize the drought risk level based on the clustering results.
[0016] Thirdly, the present invention provides a computer device, including 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 steps of the drought multidimensional risk partitioning method based on clustering algorithm.
[0017] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the drought multi-dimensional risk partitioning method based on clustering algorithm.
[0018] Compared with the prior art, the present invention has the following technical effects: This invention first utilizes a Self-Organizing Map (SOM) network to perform self-learning and pattern extraction on the input drought feature data. Through an unsupervised training process, it achieves adaptive mapping and dimensionality reduction of features, thereby preserving the topological relationships and spatial distribution characteristics of the data. Subsequently, a K-means clustering algorithm is introduced to further refine the group boundaries of the feature nodes output by the SOM, improving the stability and classification accuracy of the clustering results. This two-stage clustering method, while maintaining the pattern recognition advantages of the SOM algorithm, combines the high computational efficiency of the K-means algorithm to effectively improve the processing efficiency and classification rationality of multidimensional drought data. The clustering framework established in this invention does not require preset evaluation criteria or weights. It can objectively identify coupling patterns and spatial differences between different drought features based on data-driven approaches, while automatically classifying drought hazard levels and generating intuitive spatial distribution maps. This provides technical support for identifying high-risk areas and formulating differentiated drought prevention and resource management strategies. Attached Figure Description
[0019] Figure 1 This is a flowchart of the present invention; Figure 2 This is a spatial distribution map of SEDI from 2000 to 2018 calculated in this invention example; Figure 3 This is a spatial distribution map of drought duration and intensity from 2000 to 2018 calculated in an example of the present invention. Figure 4 This is a spatial distribution map of the frequency of different levels of drought from 2000 to 2018 calculated in an example of the present invention. Figure 5 This is a drought hazard zoning map calculated in an example of the present invention. Detailed Implementation
[0020] The present invention will be further described below with reference to the accompanying drawings: Example 1, please refer to Figure 1 This invention provides a drought multi-dimensional risk zoning method based on clustering algorithms, including: Acquire drought index data with continuous time series within the study area, and obtain a spatiotemporal dataset of drought index through data quality control; Based on the aforementioned drought index spatiotemporal dataset, three drought characteristic indicators—drought duration, drought intensity, and drought frequency—are calculated to construct a multidimensional drought feature. The multidimensional features of drought are input into the self-organizing map network for preliminary clustering to obtain the neuron weight vector. The neuron weight vector is then used as input for secondary clustering, and the optimal number of clusters is determined by combining the clustering performance evaluation index. Based on the clustering results, the drought risk levels are ranked, classified, and visualized.
[0021] This invention aims to overcome the problems of single indicators, subjective weighting, and insufficient spatial identification accuracy in traditional drought assessment methods. This method constructs a multi-dimensional feature system of drought indices and introduces a second-order clustering algorithm to achieve comprehensive identification and zoning of regional drought risk, exhibiting high scientific rigor, objectivity, and scalability.
[0022] Example 2: This invention provides a drought multi-dimensional risk zoning method based on clustering algorithms, including: (1) Acquisition and processing of drought index data Continuous time-series drought index data were selected, including Standardized Precipitation Index (SPI), Standardized Precipitation Evapotranspiration Index (SPEI), Palmer Drought Index (PDSI), and Standardized Evapotranspiration Deficit Index (SEDI). These drought indices can be obtained through two methods: first, by directly extracting data of the corresponding time scale and spatial resolution from publicly available datasets (such as global or regional drought monitoring products); second, by generating drought index sequences based on regional observational or reanalysis data of precipitation, temperature, and evapotranspiration using a unified calculation method. The obtained drought index data needs to be raster-cropped for the study area, and data quality control steps such as missing value imputation, outlier detection, and spatial resampling must be performed to ultimately form a continuous time-series spatiotemporal dataset of drought indices within the study area.
[0023] (2) Extraction and construction of multidimensional features of drought Based on the processed drought index time series, an event identification framework using a tiered threshold and runs theory was employed to calculate three drought characteristic indicators: drought duration, drought intensity, and drought frequency. Drought duration reflects the duration of the event, drought intensity describes the intensity and cumulative impact of drought, and drought frequency characterizes the probability of occurrence of drought events at different levels.
[0024] For different drought indices, different drought thresholds are set with reference to the corresponding drought level classification standards, including mild drought, moderate drought, severe drought, and extreme drought, with corresponding thresholds of [missing information]. , , , When the drought index is below A drought is considered to occur when the condition is met continuously for a period of time. A drought event is defined as a period of time when this condition is met consecutively. The duration and intensity of each drought event are calculated using runs theory. (1) In the formula, For the first The duration of a drought event; For the first The number of months in which each drought event began; No. The number of months that a drought event ended.
[0025] (2) In the formula, For the first The intensity of a drought event; for The drought index value at any given time. The frequency of occurrence of different levels of drought is calculated based on different drought thresholds: (3) In the formula, For level The frequency of drought events; For level The number of months in which drought events occurred; This represents the total number of months in the study period.
[0026] (3) SOM training and preliminary clustering The calculated three drought features (drought duration, drought intensity, and drought frequency) are input variables into a self-organizing map (SOM) network. Through an unsupervised learning process, the SOM network can automatically identify the potential distribution structure of the input data and map high-dimensional feature information to a low-dimensional topological space, achieving preliminary clustering and pattern recognition of regional drought features. To determine the rationality of the network structure and the number of output units, this invention calculates two indicators: quantization error and topology error. Quantization error is used to evaluate the ability of the established network to distinguish the features of the input data; the smaller the error, the more accurately the network can reflect the spatial distribution characteristics of the input samples. Topology error is used to evaluate the network's ability to preserve neighborhoods during mapping; the lower the value, the more the network preserves the topological structure of the input space during the mapping process.
[0027] (4) K-means re-clustering and category determination The neuron weight vectors output by the SOM (Search Engine Optimization) are used as input data, and a secondary clustering algorithm, K-means clustering, is employed. The K-means algorithm aims to minimize intra-cluster sample differences and maximize inter-cluster differences, continuously adjusting the cluster center positions through an iterative optimization process to achieve refined classification of drought characteristic data. To determine the optimal number of clusters, this invention introduces the Calinski-Harabasz index (CH-index) as a clustering performance evaluation metric; a higher CH-index value indicates better clustering results. By calculating the CH-index values under different cluster numbers, the cluster number corresponding to the peak value is selected as the optimal partitioning scheme.
[0028] (5) Drought risk classification and spatial zoning identification Based on the clustering results, statistical analysis was performed on the comprehensive drought characteristics of each category of samples. The clustering results were ranked according to the comprehensive levels of drought duration, intensity, and frequency to determine the relative level of drought risk. Combined with a Geographic Information System (GIS) platform, the clustering results were spatially visualized to create a drought risk zoning map, visually displaying the differences and spatial patterns of drought risk in different regions.
[0029] This invention constructs a multidimensional feature space for drought based on three core characteristics of the drought index: drought frequency, drought duration, and drought intensity, enabling comprehensive identification and classification of drought events. Unlike traditional methods that rely on single indicators or empirical weighting, this invention employs a data-driven clustering analysis approach, achieving an objective fusion of drought characteristics and avoiding subjective biases caused by human weighting, thereby improving the scientific rigor and reliability of drought risk assessment.
[0030] This invention combines Self-Organizing Maps (SOMs) with the K-means clustering algorithm to form a second-order clustering framework. The SOM network performs pattern recognition and dimensionality reduction mapping while preserving the data topology, and the K-means algorithm further performs feature re-clustering, thereby improving classification accuracy and computational efficiency. This method fully leverages the feature extraction capabilities of SOMs and the high computational efficiency of K-means, effectively handling large-scale drought data and improving the stability and interpretability of the results.
[0031] This invention enables automatic zoning and spatial distribution identification of drought risk based on clustering results, revealing the comprehensive differences and spatial patterns of drought characteristics in different regions. Compared with traditional methods based on weights or experience, this invention has significant advantages in spatial identification accuracy, classification objectivity, and practical application effectiveness.
[0032] The assessment framework proposed in this invention possesses strong versatility and portability. The required input data types and parameter settings are concise and clear, allowing for rapid replication and widespread application in other regions or at different time scales. This characteristic enables this invention not only to be applicable to drought hazard identification in specific regions but also to provide unified technical support for cross-regional drought comparative studies and multi-scale assessments.
[0033] Example 3: This invention provides a drought multi-dimensional risk zoning method based on clustering algorithms, including: (1) Acquisition and processing of drought index data This example calculates the standardized evapotranspiration deficit index (SEDI) for China from 2000 to 2018 for multidimensional drought risk assessment and zoning. The data has a spatial resolution of 0.25° and a temporal resolution of monthly. The spatial distribution of SEDI is shown below. Figure 2 As shown.
[0034] (2) Extraction and construction of multidimensional features of drought Table 1 shows the drought classification thresholds for SEDI. Using -1 as the boundary, all drought events from 2000 to 2018, along with their duration and intensity, were extracted pixel-by-pixel based on runs theory. These were then summed to calculate the total drought duration and total drought intensity. Figure 3 As shown in the figure, the frequency of drought events at different drought levels was calculated pixel-by-pixel based on the drought classification threshold. Since the frequency of extreme drought events was relatively low between 2000 and 2018, this example uniformly calculated the frequencies of severe and extreme drought events. The spatial distribution of the frequencies of mild, moderate, severe, and extreme drought events is shown in the figure. Figure 4 As shown.
[0035] Table 1 SEDI Drought Levels
[0036] (3) SOM training and preliminary clustering The drought feature results obtained in step 2 are used as input data for SOM training and preliminary clustering. Based on the quantization error and topology error results, the optimal SOM network is selected for preliminary training. For the drought feature results obtained in step 2, the optimal SOM network selected in this example is 22×20, with quantization error and topology error of 2.037 and 0.0004, respectively.
[0037] (4) K-means re-clustering and category determination Using the neuron weight vectors output by SOM in step 3 as input data, a second clustering process is performed using the K-means clustering algorithm. The final number of clusters is determined based on the CH index. In this example, the final number of clusters is 4, and the CH index is 205.89.
[0038] (5) Drought risk classification and spatial zoning identification The raster file of the final clustering results obtained in step 4 is imported into the geographic information system platform for spatial visualization. The resulting drought hazard zoning results are as follows: Figure 5 As shown.
[0039] In another embodiment of the present invention, a drought multi-dimensional risk zoning system based on a clustering algorithm is provided, which can be used to implement the above-mentioned drought multi-dimensional risk zoning method based on a clustering algorithm. Specifically, the system includes: The data acquisition module is used to acquire drought index data with continuous time series within the study area, and obtain a spatiotemporal dataset of drought index through data quality control. The multidimensional feature acquisition module is used to calculate three drought feature indicators—drought duration, drought intensity, and drought frequency—based on the drought index spatiotemporal dataset, and to construct multidimensional drought features. The clustering module is used to input drought multidimensional features into the self-organizing map network for preliminary clustering, obtain neuron weight vectors, use neuron weight vectors as input for secondary clustering, and combine clustering performance evaluation indicators to determine the optimal number of clusters. The output display module is used to sort and visualize the drought risk level based on the clustering results.
[0040] The module division in this embodiment of the invention is illustrative and represents only one logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional modules in the various embodiments of the invention can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.
[0041] In another embodiment of the present invention, a computer device is provided, comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions from the computer storage medium to achieve a corresponding method flow or corresponding function. The processor described in this embodiment of the present invention can be used for the operation of a drought multi-dimensional risk partitioning method based on clustering algorithms.
[0042] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both the built-in storage medium in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the drought multi-dimensional risk partitioning method based on clustering algorithms in the above embodiments.
[0043] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0044] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0045] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0046] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0047] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.
Claims
1. A drought multi-dimensional risk zoning method based on clustering algorithm, characterized in that, include: Acquire drought index data with continuous time series within the study area, and obtain a spatiotemporal dataset of drought index through data quality control; Based on the aforementioned drought index spatiotemporal dataset, three drought characteristic indicators—drought duration, drought intensity, and drought frequency—are calculated to construct a multidimensional drought feature. The multidimensional features of drought are input into the self-organizing map network for preliminary clustering to obtain the neuron weight vector. The neuron weight vector is then used as input for secondary clustering, and the optimal number of clusters is determined by combining the clustering performance evaluation index. Based on the clustering results, the drought risk levels are ranked, classified, and visualized.
2. The drought multi-dimensional risk zoning method based on clustering algorithm according to claim 1, characterized in that, The acquisition of drought index data with continuous time series within the study area includes: Drought index data with continuous time series were selected, including the Standardized Precipitation Index (SPI), Standardized Precipitation Evapotranspiration Index (SPEI), Palmer Drought Index (PDSI), and Standardized Evapotranspiration Deficit Index (SEDI).
3. The drought multi-dimensional risk zoning method based on clustering algorithm according to claim 1, characterized in that, The drought index spatiotemporal dataset obtained through data quality control includes: The acquired drought index data was rasterized for the study area, and data quality control steps such as missing value imputation, outlier detection, and spatial resampling were completed to form a continuous temporal series drought index spatiotemporal dataset within the study area.
4. The drought multi-dimensional risk zoning method based on clustering algorithm according to claim 1, characterized in that, Based on the aforementioned drought index spatiotemporal dataset, three drought characteristic indicators—drought duration, drought intensity, and drought frequency—are calculated to construct a multidimensional drought feature, including: Based on the processed drought index time series, an event identification framework using graded thresholds and runs theory was adopted to calculate three drought characteristic indicators: drought duration, drought intensity, and drought frequency. Drought duration reflects the duration of the event, drought intensity describes the intensity and cumulative impact of drought, and drought frequency is used to characterize the probability of occurrence of drought events of different levels. For different drought indices, different drought thresholds are set based on drought level classification standards, including mild drought, moderate drought, severe drought, and extreme drought, with corresponding thresholds of [missing information]. , , , When the drought index is below A drought is considered to occur when the condition is met continuously for a period of time. A drought event is defined as a period of time when this condition is met consecutively. The duration and intensity of each drought event are calculated using runs theory. In the formula, For the first The duration of a drought event; For the first The number of months in which each drought event began; No. The number of months since the end of each drought event; In the formula, For the first The intensity of a drought event; for The drought index value at any given time; the frequency of occurrence of different levels of drought is calculated based on different drought thresholds: In the formula, For level The frequency of drought events; For level The number of months in which drought events occurred; This represents the total number of months in the study period.
5. The drought multi-dimensional risk zoning method based on clustering algorithm according to claim 1, characterized in that, The step of inputting drought multidimensional features into a self-organizing map network for preliminary clustering to obtain neuron weight vectors includes: The three drought features calculated—drought duration, drought intensity, and drought frequency—are imported into a self-organizing map network (SOM) as input variables. Through an unsupervised learning process, the SOM network can automatically identify the potential distribution structure of the input data and map high-dimensional feature information to a low-dimensional topological space, achieving preliminary clustering and pattern recognition of regional drought features. Two indicators, quantization error and topology error, are calculated. Quantization error is used to evaluate the ability of the established network to distinguish the features of the input data; the smaller the error, the more accurately the network can reflect the spatial distribution characteristics of the input samples. Topology error is used to evaluate the network's ability to preserve the neighborhood during mapping; the lower the value, the more the network preserves the topological structure of the input space during the mapping process.
6. The drought multi-dimensional risk zoning method based on clustering algorithm according to claim 1, characterized in that, The process of using neuron weight vectors as input to perform secondary clustering, and determining the optimal number of clusters based on clustering performance evaluation metrics, includes: The neuron weight vectors output by SOM are used as input data, and K-means clustering algorithm is used for secondary clustering. The K-means algorithm aims to minimize intra-class sample differences and maximize inter-class differences. Through iterative optimization, the cluster center positions are continuously adjusted to achieve refined classification of drought feature data. The Calinski-Harabasz index is introduced as a clustering performance evaluation index. The larger the value, the better the clustering effect. By calculating the CH index values under different numbers of clusters, the number of clusters corresponding to the peak value is selected as the optimal partitioning scheme.
7. The drought multi-dimensional risk zoning method based on clustering algorithm according to claim 1, characterized in that, The process of ranking and visualizing drought risk levels based on clustering results includes: Based on the clustering results, the comprehensive drought characteristics of each category of samples are statistically analyzed. The clustering results are ranked according to the comprehensive level of drought duration, intensity and frequency to determine the relative level of drought risk. Combined with the geographic information system (GIS) platform, the clustering results are spatially visualized to form a drought risk zoning map.
8. A drought multi-dimensional risk zoning system based on clustering algorithm, characterized in that, include: The data acquisition module is used to acquire drought index data with continuous time series within the study area, and obtain a spatiotemporal dataset of drought index through data quality control. The multidimensional feature acquisition module is used to calculate three drought feature indicators—drought duration, drought intensity, and drought frequency—based on the drought index spatiotemporal dataset, and to construct multidimensional drought features. The clustering module is used to input drought multidimensional features into the self-organizing map network for preliminary clustering, obtain neuron weight vectors, use neuron weight vectors as input for secondary clustering, and combine clustering performance evaluation indicators to determine the optimal number of clusters. The output display module is used to sort and visualize the drought risk level based on the clustering results.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the drought multidimensional risk partitioning method based on clustering algorithm as described in any one of claims 1 to 7.
10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the drought multidimensional risk partitioning method based on clustering algorithm as described in any one of claims 1 to 7.