A drought monitoring and analysis method based on the coupling relationship between vegetation index and drought

Abstract

The present invention provides a drought monitoring and analysis method based on the coupling relationship between vegetation index and drought, which involves the field of drought remote sensing detection, starting from data collection, to model design and training, and finally model evaluation and application. By using a large amount of labeled data for training, the coupling relationship between vegetation index and drought is learned, and the convolutional neural network model gradually adjusts the weights and parameters to more accurately determine the coupling relationship between vegetation index and drought. In order to further improve the performance of the model, strategies such as hyperparameter tuning, data enhancement and model integration are used to meet the challenges of various data scenarios. The beneficial effects of the present invention are: not only effectively overcoming the limitations of traditional methods, but also providing more reliable and accurate drought monitoring results, and improving the accuracy and timeliness of drought monitoring. It provides a solid foundation and practical value for various application fields.

Description

Technical Field

[0001] The present invention relates to the field of drought remote sensing detection, and in particular to a drought monitoring and analysis method based on vegetation index and drought coupling relationship. Background Art

[0002] The intensification of climate change has profound impacts on Earth's ecosystems, one of which is the increasing frequency and severity of drought events. The impact of drought on multiple fields, including meteorology, earth sciences, ecology, agriculture, and water resources management, is becoming increasingly significant. Therefore, accurately monitoring and assessing drought conditions is of vital importance to scientists, decision makers, and society. Currently, global climate change has led to frequent droughts. Drought is a complex natural phenomenon that is affected not only by precipitation and temperature, but also by complex factors such as vegetation cover, water resources management, natural disasters, soil texture, and soil type. However, traditional drought monitoring methods often fail to fully account for these factors, resulting in inaccurate drought monitoring results and difficulty in providing timely and accurate information.

[0003] Currently, several methods have been developed for drought monitoring, primarily using vegetation indices combined with meteorological data and multi-source data combined with machine learning algorithms. The drawbacks of using vegetation indices combined with meteorological data for drought monitoring include reliance on meteorological data, such as precipitation and temperature, which can be limited in time and space, making drought monitoring susceptible to insufficient or inaccurate data. Furthermore, this approach lacks advanced analytical methods for vegetation index data, preventing a comprehensive assessment of vegetation conditions. Combining only vegetation indices with meteorological data also results in limited spatial resolution, preventing fine-grained drought monitoring. The drawbacks of using multi-source data combined with machine learning algorithms for drought monitoring include the inability to fully consider the complex nonlinear relationships between vegetation indices and meteorological data, which only combine meteorological data, underlying surface influencing factors, and machine learning methods. Consequently, it fails to capture the response of vegetation conditions to drought conditions.

[0004] Traditional drought monitoring methods are often limited by analytical methods or single-type data, making them incapable of comprehensive and accurate drought monitoring. These limitations of existing technologies necessitate the development of a more comprehensive and accurate approach to drought monitoring to better understand and respond to drought events. Summary of the Invention

[0005] In order to solve the problem that drought monitoring methods have poor evaluation effects in complex environments, the present invention provides a drought monitoring and analysis method based on the vegetation index and drought coupling relationship, which mainly includes:

[0006] S1: Using data that influences drought monitoring and analysis, these data specifically include high-resolution satellite remote sensing data, soil data, and meteorological data. The meteorological data includes precipitation data, surface temperature data, humidity data, and evapotranspiration data. The soil data includes soil moisture data and soil temperature data.

[0007] S2: Clean, denoise, correct, and normalize the data. Use the processed high-resolution remote sensing data to calculate the vegetation index. Use the processed precipitation data, temperature data, humidity data, soil moisture data, soil temperature data, and evapotranspiration data to calculate the drought index. The obtained vegetation index and drought index form a data set. Use data enhancement technology to transform the data in the data set to obtain a new data set with more data. The new data set is divided into a training set, a validation set, and a test set.

[0008] S3: Building a deep learning model based on convolutional neural networks;

[0009] S4: Use the training set to train the deep learning model, and use the validation set and test set to verify the trained model to obtain a trained deep learning model. The input of the model is the vegetation index, and the output is the drought analysis result.

[0010] S5: The vegetation index obtained from the actual high-resolution satellite remote sensing data is input into the trained deep learning model to obtain the drought analysis results.

[0011] Furthermore, in step S2, data cleaning includes processing of outliers and missing values. Outlier processing is implemented according to the 3σ rule. Missing values ​​Y i for:

[0012]

[0013] where Y i Indicates missing values; X i Indicates missing value Y i Corresponding time point; X i-1 Indicates Y i-1 Corresponding time point; Y i-1 and Y i+1 Represents X i-1 and X i+1 The observed value at the time;

[0014] Use the moving average method to perform data denoising, the moving average of the i-th data include:

[0015]

[0016] Where k is the size of the time window; Yj is the original value of the jth data point, and n represents the number of data in the time window;

[0017] The linear normalization algorithm is used to process the data, and the normalized result Y' l include:

[0018]

[0019] Among them, Y l Represents a sample value in the original data set; Y' l Represents the sample value after normalization; Y min Indicates the minimum value in the entire data set; Y max Represents the maximum value in the entire data set.

[0020] Furthermore, in step S2, the processed high-resolution remote sensing data is used to calculate the vegetation index:

[0021]

[0022] Among them, NIR represents near-infrared band remote sensing data; Red represents red light band remote sensing data; NDVI represents the calculated vegetation index;

[0023] The drought index ADI is calculated using the processed precipitation data, temperature data, humidity data, soil moisture data, soil temperature data, and evapotranspiration data:

[0024] ADI=ω1×SPI+ω2×THI+ω3×SSMI

[0025] Among them, ω1, ω2, and ω3 represent weights; SPI represents the standardized precipitation index; THI represents the temperature humidity index; SSMI represents the standardized soil moisture index;

[0026] The Standardized Precipitation Index (SPI) is:

[0027]

[0028] Where Y represents the precipitation in a certain period of time; μ represents the long-term average precipitation; σ p represents the standard deviation of precipitation;

[0029] The temperature humidity index THI is:

[0030] THI=T-0.55×(1-f)×(T-58)

[0031] Where T represents the Fahrenheit temperature data; f represents the monthly average relative humidity data, expressed in percentage;

[0032] The Standardized Soil Moisture Index SSMI is:

[0033]

[0034] Among them, SM represents the soil moisture value on a certain time scale; represents the multi-year mean soil moisture on this time scale; σ s It represents the multi-year soil moisture standard deviation on this time scale.

[0035] Furthermore, in step S2, the vegetation index is used to reflect the health status of surface vegetation, and the value range is 0 to 1, where 0 represents no vegetation and 1 represents the healthiest vegetation.

[0036] Furthermore, in step S3, the deep learning model includes a convolution layer, a pooling layer, and a fully connected layer; through multi-layer convolution and pooling, key features of the vegetation index and drought index are automatically learned and extracted, including original index data, temporal features, and spatial features. These features represent different levels of abstraction of the data, and the weight parameters are updated through the back propagation process of the training data, so that the convolutional neural network models the coupling relationship between the vegetation index and drought.

[0037] Furthermore, the deep learning model selects a loss function to measure the difference between the model output and the actual vegetation index and drought conditions.

[0038] Furthermore, during the model training process, forward propagation is used to calculate the output, backpropagation is used to calculate the gradient, and the gradient is used to update the model parameters until the model converges.

[0039] Furthermore, during the model training process, it is necessary to adjust the hyperparameters of the convolutional neural network model until the model performance is optimized. The hyperparameters include adjusting the learning rate, batch size, and weight decay parameters.

[0040] A storage device stores instructions and data for implementing a drought monitoring and analysis method based on a vegetation index and drought coupling relationship.

[0041] A drought monitoring and analysis system based on the vegetation index and drought coupling relationship comprises: a processor and a storage device; the processor loads and executes instructions and data in the storage device to implement a drought monitoring and analysis method based on the vegetation index and drought coupling relationship.

[0042] The beneficial effects of the technical solution provided by the present invention are as follows: the present invention introduces a convolutional neural network, which uses the convolutional neural network to capture the characteristics of the coupling relationship between the vegetation index and drought, and trains the coupling relationship between the vegetation index and the drought situation. Compared with the traditional model, the deep learning model can better understand the complex nonlinear relationship and improve the accuracy and generalization of monitoring. The present invention integrates high-resolution satellite data, meteorological data and soil data to construct a comprehensive, multi-dimensional data set, and calculates the corresponding vegetation index and drought index. Compared with traditional monitoring methods, the comprehensive data integration makes the monitoring more accurate and comprehensive. Finally, the present invention combines high-resolution remote sensing data, meteorological data, soil data and convolutional neural networks to improve the accuracy and timeliness of drought monitoring. The present invention not only effectively overcomes the limitations of traditional methods, but also can provide more reliable and accurate drought monitoring results, providing a solid foundation and practical value for various application fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0044] Figure 1 The present invention is a flowchart of a drought monitoring and analysis method based on the vegetation index and drought coupling relationship in an embodiment of the present invention.

[0045] Figure 2 2 is a schematic diagram of data preprocessing in an embodiment of the present invention.

[0046] Figure 3 This is a result diagram of the deep learning model in an embodiment of the present invention.

[0047] Figure 4 It is a schematic diagram of the operation of the hardware device in the embodiment of the present invention. DETAILED DESCRIPTION

[0048] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.

[0049] Existing drought monitoring methods often face problems such as insufficient or inaccurate data, insufficient model complexity, incomplete vegetation assessment, limited spatial resolution, single data sources, and a failure to consider the complex relationship between vegetation indices and meteorological data. These methods fail to fully capture the characteristics of the data, limiting the accuracy, applicability, and timeliness of drought monitoring results. To address this, the present invention employs a drought monitoring and analysis method based on the coupled relationship between vegetation indices and drought. By combining high-resolution satellite data, meteorological data, soil data, and convolutional neural networks, this method analyzes the coupled relationship between vegetation indices and drought conditions, enabling more accurate monitoring and assessment of drought conditions.

[0050] Please refer to Figure 1 , Figure 1 This is a flowchart of a drought monitoring and analysis method based on the coupling relationship between vegetation index and drought in an embodiment of the present invention, starting from data collection, to model design and training, and finally model evaluation and application. By using a large amount of labeled data for training and learning the coupling relationship between vegetation index and drought, the convolutional neural network model gradually adjusts the weights and parameters to more accurately determine the coupling relationship between vegetation index and drought. In order to further improve the performance of the model, strategies such as hyperparameter tuning, data enhancement and model integration are used to meet the challenges of various data scenarios. It covers the entire process from data collection to model training to application to provide a more accurate and comprehensive solution. Specifically including:

[0051] S1: Collect high-resolution satellite remote sensing data, including visible light and infrared spectral band information, as well as precipitation data, surface temperature data, soil moisture data, soil temperature data, evapotranspiration data, and other data for model training. These data can come from different satellites, such as Landsat, MODIS, Sentinel, etc., to ensure global coverage and high spatiotemporal resolution, forming a comprehensive, multi-dimensional data set, and then processing these data accordingly.

[0052] S2: Data preprocessing: For details on the data processing process, refer to Figure 2 . Process missing, abnormal and duplicate values ​​in the data to ensure data quality and consistency, and then clean, denoise, correct and normalize the processed data.

[0053] Data cleaning includes processing of outliers and missing values. Outlier processing is implemented according to the 3σ rule. Missing values ​​Y i for:

[0054]

[0055] where Y i Indicates missing values; X i Indicates missing value Y i Corresponding time point; X i-1 Indicates Y i-1 Corresponding time point; Y i-1 and Y i+1 Represents X i-1 and X i+1 The observation value at the moment; the missing value is processed by calculating and filling in the missing value based on the observation values ​​before and after the missing value.

[0056] Use the moving average method to perform data denoising, the moving average of the i-th data include:

[0057]

[0058] Where k is the size of the time window; Y j is the original value of the jth data point, and n represents the number of data in the time window;

[0059] The data is processed using a linear normalization algorithm, and the normalized result Y' l include:

[0060]

[0061] Among them, Y l Represents a sample value in the original data set; Y' l Represents the sample value after normalization; Y min Indicates the minimum value in the entire data set; Y max Represents the maximum value in the entire data set.

[0062] Calculate vegetation index using processed high-resolution remote sensing data:

[0063]

[0064] Among them, NIR represents near-infrared band remote sensing data; Red represents red light band remote sensing data; NDVI represents the calculated vegetation index.

[0065] These indices can reflect the health of surface vegetation, and the values ​​are usually between 0 and 1, where 0 represents no vegetation and 1 represents the healthiest vegetation.

[0066] The drought index ADI is calculated using the processed precipitation data, soil data and other meteorological data:

[0067] ADI=ω1×SPI+ω2×THI+ω3×SSMI

[0068] Among them, ω1, ω2, and ω3 represent weights; SPI represents the standardized precipitation index; THI represents the temperature humidity index; SSMI represents the standardized soil moisture index;

[0069] The Standardized Precipitation Index (SPI) is:

[0070]

[0071] Where Y represents the precipitation in a certain period of time; μ represents the long-term average precipitation; σ p represents the standard deviation of precipitation;

[0072] The temperature humidity index THI is:

[0073] THI=T-0.55×(1-f)×(T-58)

[0074] Where T represents the Fahrenheit temperature data; f represents the monthly average relative humidity data, expressed in percentage;

[0075] The Standardized Soil Moisture Index SSMI is:

[0076]

[0077] Among them, SM represents the soil moisture value on a certain time scale; represents the multi-year mean soil moisture on this time scale; σ s It represents the multi-year soil moisture standard deviation on this time scale.

[0078] These indicators are used to quantify the severity of drought. Vegetation index and drought index are integrated into the dataset. The dataset is divided into training set, validation set and test set to evaluate the performance of the model.

[0079] Data augmentation: Data augmentation technology is used to perform a series of reasonable transformations on the data to generate diverse and robust training samples. Data augmentation helps to reduce overfitting and improve the generalization ability of the model.

[0080] S3: Reference Figure 3 A deep learning model based on a convolutional neural network was constructed. This model takes into account the coupled relationship between vegetation indices and drought conditions, ensuring sufficient depth and complexity to capture the complex, nonlinear relationship between vegetation indices and drought conditions. The model includes convolutional layers, pooling layers, and fully connected layers. During training, the convolutional neural network effectively represents the correlation between vegetation indices and drought conditions through filters in the convolutional layers. Finally, the final layer of the convolutional neural network establishes a model based on the coupled relationship between vegetation indices and drought conditions, enabling accurate analysis of drought conditions.

[0081] Model Interpretation Analysis: Apply interpretable tools and techniques to gain a deeper understanding of the decision-making process of convolutional neural network models. This includes using methods such as attention mechanisms, feature importance maps, and gradient heatmaps to visualize the important areas and key features that the model focuses on in the input data.

[0082] Real-time considerations: Consider the inference speed and latency of the convolutional neural network model during model design to ensure that the model can respond quickly in real-time applications. By comprehensively considering factors such as model complexity, computing resources, and data flow, we can meet real-time requirements and ensure that the model can process inputs promptly and efficiently and produce fast outputs in actual applications.

[0083] Model ensembles combine multiple convolutional neural network models to improve overall performance and robustness. Model ensembles can employ diverse approaches, such as varying model architectures, parameter initializations, and training data, to reduce the risk of overfitting and improve model generalization. Through ensemble strategies such as voting, averaging, or concatenation, the predictions of individual models are combined to produce a more powerful and robust combined model. Model ensembles help address the limitations of individual models and improve the performance of the entire model in complex scenarios.

[0084] S4: Use the training set to train the deep learning model, and use the validation set and test set to verify the trained model to obtain a trained deep learning model. The input of the model is the vegetation index, and the output is the drought analysis result.

[0085] During the training process, the convolutional neural network automatically learns and extracts key features from the input data, which can play a key role in analyzing the coupling relationship between vegetation index and drought, and can provide a deep understanding of the coupling relationship between vegetation and drought.

[0086] Feature Extraction and Learning: Through multiple layers of convolution and pooling, the convolutional neural network automatically learns and extracts key features related to vegetation and drought indices, including raw index data, extracted temporal features, and extracted spatial features. These features represent different levels of abstraction in the data, better expressing the structure and patterns within it. The weight parameters in this process are updated through backpropagation of the training data, enabling the convolutional neural network to gradually model the coupled relationship between vegetation indices and drought. This process is a core step in deep learning, enabling the model to automatically learn effective representations of key task features from the input data.

[0087] Training and Optimization: A convolutional neural network model is trained on a pre-split training set. An appropriate loss function is selected to measure the discrepancy between the model output and the actual vegetation index and drought conditions. Subsequently, a suitable optimization algorithm is selected to minimize the loss function. During model training, the output is calculated using forward propagation, the gradient is calculated using backpropagation, and the model parameters are updated using the gradient. This process is iterated until the model converges. Throughout this process, the loss function and performance metrics are continuously monitored to ensure that the model has learned the coupled relationship between vegetation index and drought from the training data and can generalize to unseen data.

[0088] Hyperparameter Tuning: Optimize the performance of the convolutional neural network by systematically adjusting the model's hyperparameters. This includes adjusting hyperparameters such as the learning rate, batch size, and weight decay. First, multiple experiments are conducted within a predefined hyperparameter range using a grid search. The model performance under each hyperparameter set is evaluated using a validation set. Based on the experimental results, the best-performing hyperparameter combination is selected, and the model is retrained on the entire training set. This iterative process continues until the hyperparameter combination that optimizes model performance is found. The goal of hyperparameter tuning is to improve the model's generalization performance, ensure good performance on unseen data, and further enhance the modeling of the relationship between vegetation indices and drought.

[0089] Model Validation and Evaluation: A validation set is used to evaluate the performance of the trained convolutional neural network model. Data from the validation set is fed into the model for forward propagation, and the coupling relationship between the model output and the actual vegetation index and drought is calculated. The model's performance is evaluated by monitoring metrics such as loss function, accuracy, recall, and precision on the validation set to determine whether overfitting or underfitting is occurring.

[0090] Results Presentation and Application: A comprehensive evaluation of the convolutional neural network model's performance on a test set was conducted, demonstrating its accuracy and reliability in drought monitoring. Visualization methods, such as charts and heat maps, were used to intuitively present the relationship between the model's output and actual observations, facilitating user understanding and interpretation. The trained model was then applied to real-world scenarios to monitor and analyze drought conditions.

[0091] During the evaluation and application phase, a rigorous test set was used to comprehensively evaluate the model's performance, continuously monitoring changes in vegetation and drought indices. Simultaneously, emphasis was placed on interpretative analysis of the model to enhance user confidence in the prediction results. Ultimately, a user-friendly software system was developed to apply the model to real-world data, clearly demonstrating changing trends in vegetation indices and drought conditions. This supports practical applications in a wide range of fields, including agricultural management and monitoring, water resource management, and forestry management.

[0092] S5: The vegetation index obtained from the actual high-resolution satellite remote sensing data is input into the trained deep learning model to obtain the drought analysis results.

[0093] See Figure 4 , Figure 4 4 is a schematic diagram of the working of the hardware device of an embodiment of the present invention, wherein the hardware device specifically comprises: a drought monitoring and analysis system 401 based on the vegetation index and drought coupling relationship, a processor 402 and a storage device 403.

[0094] A drought monitoring and analysis system 401 based on the vegetation index and drought coupling relationship: The drought monitoring and analysis system 401 based on the vegetation index and drought coupling relationship implements the drought monitoring and analysis method based on the vegetation index and drought coupling relationship.

[0095] Processor 402: The processor 402 loads and executes the instructions and data in the storage device 403 to implement the drought monitoring and analysis method based on the vegetation index and drought coupling relationship.

[0096] Storage device 403: The storage device 403 stores instructions and data; the storage device 403 is used to implement the drought monitoring and analysis method based on the vegetation index and drought coupling relationship.

[0097] Software System Development: Design and develop a complete system that integrates the convolutional neural network model. This includes establishing the system architecture and selecting an appropriate technology stack to support model deployment and integration. By implementing the model's call interface, data input and output processes, and user interaction, we ensure that the system can effectively interact with the convolutional neural network model.

[0098] User interface design: A clear layout, intuitive graphical elements, and an easy-to-understand navigation structure enable users to easily upload data, view model results, and understand the system's operational processes.

[0099] Continuous Optimization and Updates: The system is constantly monitored and evaluated to identify potential areas for improvement. By regularly collecting user feedback, performance data, and model performance, targeted adjustments and optimizations are made to enhance system performance and stability. At the same time, we monitor technological developments and industry trends, and promptly update models, algorithms, or system components to maintain the system's competitiveness and adaptability.

[0100] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A drought monitoring and analysis method based on the coupled relationship between vegetation index and drought, characterized by: include: S1: Using data that influences drought monitoring and analysis, these data specifically include high-resolution satellite remote sensing data, soil data, and meteorological data. The meteorological data includes precipitation data, surface temperature data, humidity data, and evapotranspiration data. The soil data includes soil moisture data and soil temperature data. S2: Clean, denoise, correct, and normalize the data. The obtained vegetation index and drought index are combined into a dataset. The data in the dataset is transformed using data augmentation technology to obtain a new dataset composed of more data. The new dataset is divided into a training set, a validation set, and a test set. The drought index ADI is calculated using the processed precipitation data, temperature data, humidity data, soil moisture data, soil temperature data, and evapotranspiration data; S3: Building a deep learning model based on convolutional neural networks; S4: Use the training set to train the deep learning model, and use the validation set and test set to verify the trained model to obtain a trained deep learning model. The input of the model is the vegetation index and the drought index, and the output is the drought analysis result; S5: The vegetation index and soil data obtained from the actual high-resolution satellite remote sensing data, and the drought index obtained from the meteorological data are input into the trained deep learning model to obtain the drought analysis results.

2. The drought monitoring and analysis method based on the vegetation index and drought coupling relationship according to claim 1, wherein: In step S2, data cleaning includes processing of outliers and missing values. The rule is implemented, and the missing value Yi is: in Indicates missing values; Indicates missing values The corresponding time point; express The corresponding time point; and Respectively and The observed value at the time; Use the moving average method to perform data denoising, the moving average of the i-th data include: Where k is the size of the time window; is the original value of the jth data point, and n represents the number of data in the time window; The linear normalization algorithm is used to process the data and the normalized results are include: in, Represents a sample value in the original data set; Represents the sample value after normalization; Represents the minimum value in the entire data set; Represents the maximum value in the entire data set.

3. A drought monitoring and analysis method based on the vegetation index and drought coupling relationship according to claim 2, Its characteristics are: In step S2, the vegetation index is calculated using the processed high-resolution remote sensing data: Among them, NIR represents near-infrared band remote sensing data; Red represents red light band remote sensing data; NDVI represents the calculated vegetation index; The drought index ADI is calculated using the processed precipitation data, temperature data, humidity data, soil moisture data, soil temperature data, and evapotranspiration data: in, 、 、 represents weight; SPI represents Standardized Precipitation Index; THI represents Temperature Humidity Index; SSMI represents Standardized Soil Moisture Index; The Standardized Precipitation Index (SPI) is: Among them, Y represents the precipitation in a certain period of time; represents the long-term average precipitation; represents the standard deviation of precipitation; The temperature humidity index THI is: Where T represents the Fahrenheit temperature data; f represents the monthly average relative humidity data, expressed in percentage; The Standardized Soil Moisture Index SSMI is: Among them, SM represents the soil moisture value on a certain time scale; represents the multi-year mean soil moisture on this time scale; It represents the multi-year soil moisture standard deviation on this time scale.

4. The drought monitoring and analysis method based on the vegetation index and drought coupling relationship according to claim 1, wherein: In step S2, the vegetation index is used to reflect the health status of surface vegetation, and the value range is 0-1, where 0 represents no vegetation and 1 represents the healthiest vegetation.

5. The drought monitoring and analysis method based on the vegetation index and drought coupling relationship according to claim 1, wherein: In step S3, the deep learning model includes a convolution layer, a pooling layer, and a fully connected layer; through multi-layer convolution and pooling, key features of the vegetation index and drought index are automatically learned and extracted, including original index data, temporal features, and spatial features. These features represent different levels of abstraction of the data, and the weight parameters are updated through the back propagation process of the training data, so that the convolutional neural network models the coupling relationship between the vegetation index and drought.

6. The drought monitoring and analysis method based on the vegetation index and drought coupling relationship according to claim 3, characterized in that: The deep learning model selects a loss function to measure the difference between the model output and the actual vegetation index and drought conditions.

7. The drought monitoring and analysis method based on the vegetation index and drought coupling relationship according to claim 3, characterized in that: During model training, forward propagation is used to calculate the output, backpropagation is used to calculate the gradient, and the gradient is used to update the model parameters until the model converges.

8. The drought monitoring and analysis method based on the vegetation index and drought coupling relationship according to claim 3, characterized in that: During the model training process, the hyperparameters of the convolutional neural network model need to be adjusted until the model performance is optimized. The hyperparameters include adjusting the learning rate, batch size, and weight decay parameters.

9. A storage device, characterized in that: The storage device stores instructions and data for implementing the drought monitoring and analysis method based on the vegetation index and drought coupling relationship as described in any one of claims 1 to 8.

10. A drought monitoring and analysis system based on the coupled relationship between vegetation index and drought, characterized by: include: A processor and a storage device; the processor loads and executes instructions and data in the storage device to implement the drought monitoring and analysis method based on the vegetation index and drought coupling relationship as described in any one of claims 1 to 8.

Images