Comprehensive drought assessment method based on multi-source data

Through multi-source data and deep neural network model, combined with precipitation, temperature, vegetation, topography and evaporation data, a comprehensive drought assessment model was constructed, which solved the problems of incomplete consideration of disaster factors and simple model structure in the existing technology, and achieved drought Accurate evaluation of grades and effects reflecting drought characteristics.

CN118917546APending Publication Date: 2024-11-08XIAN UNIV OF TECH
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202411037136.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The existing drought assessment methods consider fewer disaster-causing factors and simple structure of the assessment model, which leads to inaccurate drought assessment results.

Method used

Using a comprehensive drought assessment method based on multi-source data, the historical remote sensing data and meteorological site data are obtained, and the deep neural network (DNN) model is constructed, combining precipitation, temperature, vegetation index, elevation and evaporation data. Build a comprehensive drought assessment model and output comprehensive drought indicators that consider multiple factors.

Benefits of technology

An accurate assessment of the drought level was achieved, and a comprehensive response to the drought characteristic information of the target area was to be fully understood, which was of great significance to drought resistance and disaster reduction work.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118917546A_ABST
    Figure CN118917546A_ABST
Patent Text Reader

Abstract

The invention discloses a comprehensive drought assessment method based on multi-source data. The method is specifically implemented according to the following steps: acquiring historical remote sensing data and meteorological station data of a target area; preprocessing the remote sensing data, unifying the time and space scales of the remote sensing data, and extracting the corresponding remote sensing data according to the meteorological station coordinate data; calculating a standardized rainfall evapotranspiration index; dividing the processed remote sensing data and the standardized rainfall evapotranspiration index into a training set and a test set, training a DNN model by taking a preset drought level as a target, and constructing a comprehensive drought evaluation model; evaluating the drought condition of the target area by using the comprehensive drought evaluation model, and outputting a comprehensive drought index considering multiple factors; according to the method, the problem of inaccurate drought assessment effect caused by incomplete consideration of drought disaster-inducing factors and simple assessment model structure in existing drought assessment is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the field of drought assessment, and in particular relates to a comprehensive drought assessment method based on multi-source data. Background Art

[0002] Drought is one of the most common, widespread and complex natural disasters in the world. It occurs frequently and lasts for a long time, which has a significant impact on the development of the national economy and social stability. It has attracted widespread attention from experts in the fields of meteorology, hydrology, agriculture and ecological environment. Drought events are complex and have unclear boundaries. It is also difficult to identify their occurrence. When people realize that a drought has occurred, the drought may have developed to a serious stage. Therefore, accurate drought assessment is crucial to mitigate the adverse effects of drought on regional food production and water resource allocation.

[0003] However, most current studies consider only a few drought-causing factors, such as precipitation or temperature, which are not comprehensive and the drought assessment model is relatively simple. Drought is a complex natural disaster, and the few drought assessment indicators and simple drought assessment models have poor drought assessment effects. Therefore, it is necessary to propose a drought assessment method with better assessment effects to solve the above problems. Summary of the invention

[0004] The purpose of the present invention is to provide a comprehensive drought assessment method based on multi-source data, which solves the problem of inaccurate drought assessment results caused by incomplete consideration of drought disaster factors and simple assessment model structure in existing drought assessment.

[0005] The technical solution adopted by the present invention is a comprehensive drought assessment method based on multi-source data, which is specifically implemented according to the following steps:

[0006] Step 1, obtaining historical remote sensing data and meteorological station data of the target area;

[0007] Step 2, preprocessing the remote sensing data in step 1, unifying the time and space scales of the remote sensing data, and extracting the corresponding remote sensing data according to the meteorological station coordinate data;

[0008] Step 3, calculating the standardized precipitation evapotranspiration index based on the meteorological station data in step 1;

[0009] Step 4, constructing a DNN model, dividing the remote sensing data processed in step 2 and the standardized precipitation evapotranspiration index in step 3 into a training set and a test set, and training the DNN model with a pre-set drought level as the target to construct a comprehensive drought assessment model;

[0010] Step 5, using the comprehensive drought assessment model in step 4 to assess the drought situation in the target area, and output a comprehensive drought index that considers multiple factors;

[0011] Step 6: Analyze the evaluation results in step 5.

[0012] The present invention is also characterized in that:

[0013] The remote sensing data and meteorological station data in step 1 include historical data of several consecutive months before the date to be measured in the target area.

[0014] Remote sensing data include precipitation, temperature, vegetation index, elevation and evapotranspiration data; meteorological station data include the longitude and latitude coordinates of the meteorological station, altitude, monthly average temperature and monthly precipitation data.

[0015] In step 2, please follow the steps below:

[0016] Step 2.1, converting the remote sensing data file obtained in step 1 into a file format;

[0017] Step 2.2, batch stitching and reprojection of remote sensing data according to the vector boundary of the target area;

[0018] Step 2.3, the spatial scale of remote sensing data was resampled to 1 km using the nearest neighbor assignment method;

[0019] Step 2.4: For remote sensing images with some missing data, the inverse distance weighted method is used to fill in the missing data;

[0020] Step 2.5, the temperature and evapotranspiration data with a time resolution of 8 days are synthesized into monthly data using the maximum value synthesis method;

[0021] In step 2.6, the precipitation, temperature, vegetation index, elevation and evapotranspiration data of the corresponding meteorological station are extracted using the bilinear interpolation method, taking the longitude and latitude coordinates of each meteorological station as the center.

[0022] In step 3, the standardized precipitation evapotranspiration index of each meteorological station on a one-month scale is calculated using the Thornthwaite method based on the latitude and longitude coordinates, altitude, monthly average temperature and monthly precipitation data of each meteorological station in the target area.

[0023] In step 4, please follow the steps below:

[0024] Step 4.1, predefine the hyperparameters, optimization algorithm and loss function of the DNN model;

[0025] Step 4.2, generating a DNN model based on the hyperparameters in step 4.1;

[0026] In step 4.3, the optimization algorithm and loss function in step 4.1 are used to train the DNN model in step 4.2 to build a comprehensive drought assessment model.

[0027] In step 4.1, the hyperparameters include the number of network layers, the number of neurons, the number of hidden layers, the number of iterations, the training batch size, the learning rate, and the dropout rate;

[0028] The optimization algorithm is the Adam model optimization algorithm;

[0029] The loss function is the mean square error loss function MSE;

[0030] In step 4.2, based on the historical data of several consecutive months before the test date in the target area, the precipitation, temperature, vegetation index, elevation, evapotranspiration and standardized precipitation evapotranspiration index data are divided into training set and test set in a ratio of 4:1, and the pre-set drought level is used as the training target;

[0031] Step 5: Input the remote sensing data of precipitation, temperature, vegetation index, elevation and evapotranspiration of the target area in step 1 into the comprehensive drought assessment model in step 4, and output the comprehensive drought index MDI considering multiple factors.

[0032] Step 6: Identify the drought level according to the SPEI drought level classification standard. When the MDI value is less than or equal to -0.5, it indicates that drought has occurred.

[0033] The beneficial effects of the present invention are: based on multi-source data of precipitation, temperature, vegetation, topography and evapotranspiration, with the help of the powerful information extraction function of deep neural networks, a comprehensive drought index is constructed, which can comprehensively reflect the drought characteristic information of the target area and accurately assess the drought level. Accurate assessment of drought is of great significance to drought prevention and disaster reduction work. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a flow chart of a comprehensive drought assessment method based on multi-source data of the present invention;

[0035] Figure 2 Schematic diagram of the DNN model structure of the comprehensive drought assessment method based on multi-source data of the present invention;

[0036] Figure 3 This is a drought level distribution map of the target area in Example 3 of the present invention in each month of 2019. DETAILED DESCRIPTION

[0037] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0038] Example 1

[0039] The present invention is based on a comprehensive drought assessment method of multi-source data, the process is as follows Figure 1 As shown, the specific implementation steps are as follows:

[0040] Step 1, obtaining historical remote sensing data and meteorological station data of the target area;

[0041] Step 2, preprocessing the remote sensing data in step 1, unifying the time and space scales of the remote sensing data, and extracting the corresponding remote sensing data according to the meteorological station coordinate data;

[0042] Step 3, calculating the standardized precipitation evapotranspiration index based on the meteorological station data in step 1;

[0043] Step 4, constructing a DNN model, dividing the remote sensing data processed in step 2 and the standardized precipitation evapotranspiration index in step 3 into a training set and a test set, training the DNN model, and constructing a comprehensive drought assessment model;

[0044] Step 5, using the comprehensive drought assessment model in step 4 to assess the drought situation in the target area, and output a comprehensive drought index that considers multiple factors;

[0045] Step 6: Analyze the evaluation results in step 5.

[0046] Example 2

[0047] The comprehensive drought assessment method based on multi-source data of the present invention is implemented in the following steps:

[0048] Step 1, obtaining historical remote sensing data and meteorological station data of the target area;

[0049] Remote sensing data include precipitation, temperature, vegetation index, elevation and evapotranspiration data; remote sensing data are obtained from the MODIS database, NASA official website and China Geospatial Data Cloud;

[0050] Meteorological station data include the latitude and longitude coordinates, altitude, monthly average temperature, and monthly precipitation data of the meteorological station; the meteorological station data are obtained from the National Meteorological Science Data Center;

[0051] The remote sensing data and meteorological station data include historical data of several consecutive months before the date to be measured in the target area.

[0052] Step 2, preprocessing the remote sensing data in step 1, unifying the time and space scales of the remote sensing data, and extracting the corresponding remote sensing data according to the meteorological station coordinate data;

[0053] Step 2.1, converting the remote sensing data file obtained in step 1 into a file format;

[0054] Step 2.2, batch stitching and reprojection of remote sensing data according to the vector boundary of the target area;

[0055] Step 2.3, the spatial scale of remote sensing data was resampled to 1 km using the nearest neighbor assignment method;

[0056] Step 2.4: For remote sensing images with some missing data, the inverse distance weighted method is used to fill in the missing data;

[0057] Step 2.5, the temperature and evapotranspiration data with a time resolution of 8 days are synthesized into monthly data using the maximum value synthesis method;

[0058] In step 2.6, the precipitation, temperature, vegetation index, elevation and evapotranspiration data of the corresponding meteorological station are extracted using the bilinear interpolation method, taking the longitude and latitude coordinates of each meteorological station as the center.

[0059] Step 3, calculating the standardized precipitation evapotranspiration index based on the meteorological station data in step 1;

[0060] According to the latitude and longitude coordinates, altitude, monthly average temperature and monthly precipitation data of each meteorological station in the target area, the standardized precipitation evapotranspiration index of each meteorological station on a one-month scale was calculated using the Thornthwaite method.

[0061] Step 4, constructing a DNN model, dividing the remote sensing data processed in step 2 and the standardized precipitation evapotranspiration index in step 3 into a training set and a test set, and training the DNN model with a pre-set drought level as the target to construct a comprehensive drought assessment model;

[0062] Step 4.1, predefine the hyperparameters, optimization algorithm and loss function of the DNN model;

[0063] Hyperparameters include the number of network layers, number of neurons, number of hidden layers, number of iterations, training batch size, learning rate, and dropout rate;

[0064] The optimization algorithm is the Adam model optimization algorithm;

[0065] The loss function is the mean square error loss function MSE;

[0066] Step 4.2, generating a DNN model based on the hyperparameters in step 4.1;

[0067] Based on the historical data of several consecutive months before the test date in the target area, the precipitation, temperature, vegetation index, elevation, evapotranspiration and standardized precipitation evapotranspiration index data were divided into training set and test set in a ratio of 4:1, and the pre-set drought level was used as the training target;

[0068] In step 4.3, the Adam model optimization algorithm and mean square error loss function MSE in step 4.1 are used to train the DNN model in step 4.2 to build a comprehensive drought assessment model.

[0069] Step 5, using the comprehensive drought assessment model in step 4 to assess the drought situation in the target area, and output a comprehensive drought index that considers multiple factors;

[0070] The precipitation, temperature, vegetation index, elevation and evapotranspiration remote sensing data of the target area in step 1 are input into the comprehensive drought assessment model in step 4, and the comprehensive drought index MDI considering multiple factors is output.

[0071] Step 6: Analyze the evaluation results in step 5.

[0072] According to the SPEI drought grade classification standard, the drought grade is identified. When the MDI value is less than or equal to -0.5, it indicates that drought has occurred.

[0073] Example 3

[0074] The comprehensive drought assessment method based on multi-source data of the present invention is implemented in the following steps:

[0075] Step 1: Obtain historical remote sensing data and meteorological station data for several consecutive months before the target area is tested; obtain precipitation, temperature, vegetation index, elevation and evapotranspiration remote sensing data of the target area from the MODIS database, NASA official website and China Geospatial Data Cloud dataset; obtain the longitude and latitude coordinates, altitude, monthly average temperature and monthly precipitation data of the meteorological station from the National Meteorological Science Data Center.

[0076] Among them, the data sources of remote sensing data include MOD11A2, MOD13A3, MOD16A2, TRMM_3B43_7, and SRTM_DEM. These data sets are all public remote sensing data.

[0077] The drought disaster information obtained in the target area within the preset historical time period includes: the area of ​​crops affected and harvestless, the population affected and with difficulty in drinking water, and the direct economic losses caused by drought.

[0078] Step 2: preprocess the remote sensing data, unify the time and space scales of the remote sensing data, and extract the corresponding remote sensing data according to the meteorological station coordinate data;

[0079] The acquired remote sensing data files were converted into file formats, batch stitched and reprojected according to the vector boundaries of the target area, and the spatial scale of the remote sensing data was resampled to 1 km using the nearest neighbor assignment method. For remote sensing images with some missing data, the inverse distance weighted method was used to fill the missing data. The temperature and evapotranspiration data with a time resolution of 8 days were synthesized into monthly data using the maximum value synthesis method. With the longitude and latitude coordinates of each meteorological station as the center, the bilinear interpolation method was used to extract the precipitation, temperature, vegetation index, elevation and evapotranspiration data of the station.

[0080] In this embodiment, the time scale of remote sensing data is unified as 1 month, and the spatial scale is unified as 1 km. Drought assessment effects are different at different time and spatial scales.

[0081] Step 3, calculating the standardized precipitation evapotranspiration index based on the meteorological station data, means calculating the standardized precipitation evapotranspiration index of each meteorological station on a one-month scale using the Thornthwaite method based on the latitude and longitude coordinates, altitude, monthly average temperature and monthly precipitation data of each meteorological station in the target area. The standardized precipitation evapotranspiration index is calculated based on precipitation and evapotranspiration data.

[0082] The time scale of the calculated standardized precipitation evapotranspiration index is consistent with the time scale of the remote sensing data obtained in step 2.

[0083] Step 4, constructing a DNN model, dividing the precipitation, temperature, vegetation index, elevation and evapotranspiration data in step 2 and the standardized precipitation evapotranspiration index in step 3 into a training set and a test set, and training the DNN model with a pre-set drought level as the target to construct a comprehensive drought assessment model;

[0084] Predefine the hyperparameters, optimization algorithm and loss function of the DNN model, wherein the hyperparameters include the number of network layers, the number of neurons, the number of hidden layers, the number of iterations, the training batch size, the learning rate, and the discard rate; the optimization algorithm is the Adam model optimization algorithm; the loss function is the mean square error loss function MSE; based on the number of network layers, the number of neurons, the number of hidden layers, the number of iterations, the training batch size, the learning rate, and the discard rate, generate a DNN model, such as Figure 2 As shown;

[0085] The precipitation, temperature, vegetation index, elevation, evapotranspiration and standardized precipitation evapotranspiration index data were divided into training set and test set in a ratio of 4:1. The pre-set drought level was used as the training target. The Adam model optimization algorithm and mean square error loss function MSE were used to train the DNN model and construct a comprehensive drought assessment model.

[0086] In this embodiment, the DNN model has 5 network layers, including 3 hidden layers, and the number of neurons in each network layer is 5, 16, 32, 16, and 1 respectively. The number of iterations is 1000, the training batch size is 128, the learning rate is 0.001, and the discarding rate is 0.2.

[0087] Step 5: Use the comprehensive drought assessment model in step 4 to assess the drought situation in the target area and output a comprehensive drought index that takes multiple factors into account; input the precipitation, temperature, vegetation index, elevation and evapotranspiration data of the target area into the comprehensive drought assessment model to obtain the comprehensive drought index MDI.

[0088] Step 6: Analyze the drought assessment results and identify the drought level based on the MDI data output by the comprehensive drought assessment model according to the SPEI drought level classification standard. As shown in Table 1, when the MDI value is less than or equal to -0.5, it indicates that drought has occurred.

[0089] Table 1 Drought level classification table

[0090] MDI Range Drought level MDI>-0.5 No drought -1.0<MDI≤-0.5 Mild drought -1.5<MDI≤-1.0 Moderate Drought -2.0<MDI≤-1.5 Severe drought MDI≤-2.0 Extreme drought

[0091] like Figure 3 As shown, the drought level distribution map clearly shows the drought distribution in the current region in each month of 2019. Figure 3 (a)-(l) in the figure correspond to the 12 months in 2019. Overall, the drought in the region in 2019 mainly occurred in March in spring, with slight drought in some areas in summer and winter, and no obvious drought in autumn. Figure 3 (c) in the figure corresponds to March, when a severe drought occurred in the region, covering almost the entire region, and moderate drought or above occurred in a large area.

[0092] According to the "China Meteorological Disaster Yearbook" and related drought data, the average spring temperature in the region in 2019 was the highest since 1961, and March continued to be hot and rainy, and the drought continued to develop, with moderate to severe droughts in most areas; there were many heavy rains in summer and the intensity was high; the continuous rains in autumn lasted for a long time and the rainfall was heavy; and the precipitation in winter was less than in previous years. Overall, the evaluation results of the comprehensive drought assessment model are consistent with the actual occurrence and severity of drought, indicating that the comprehensive drought assessment model constructed in this embodiment has a good drought assessment effect.

[0093] The present invention is based on a comprehensive drought assessment method for multi-source data, which integrates remote sensing data and meteorological station data, comprehensively considers various disaster-causing factors of drought, and constructs a comprehensive drought index based on multi-source data of precipitation, temperature, vegetation, topography and evapotranspiration, with the help of the powerful information extraction function of deep neural networks, which can comprehensively reflect the drought characteristic information of the target area and accurately assess the drought level. This method is used to evaluate the drought occurrence in the target area in Example 3 in 2019, and the evaluation results of the comprehensive drought assessment model are compared with the historical drought disaster data record information. It is found that the evaluation results of the comprehensive drought assessment model are consistent with the actual occurrence and severity of drought. The comprehensive drought assessment method based on multi-source data of the present invention is highly feasible and has good evaluation effect.

Claims

1. Comprehensive drought assessment method based on multi-source data, It is characterized in that Follow the steps below to implement it: Step 1, obtaining historical remote sensing data and meteorological station data of the target area; Step 2, preprocessing the remote sensing data in step 1, unifying the time and space scales of the remote sensing data, and extracting the corresponding remote sensing data according to the meteorological station coordinate data; Step 3, calculating the standardized precipitation evapotranspiration index based on the meteorological station data in step 1; Step 4, constructing a DNN model, dividing the remote sensing data processed in step 2 and the standardized precipitation evapotranspiration index in step 3 into a training set and a test set, and training the DNN model with a pre-set drought level as the target to construct a comprehensive drought assessment model; Step 5, using the comprehensive drought assessment model in step 4 to assess the drought situation in the target area, and output a comprehensive drought index that considers multiple factors; Step 6: Analyze the evaluation results in step 5.

2. The comprehensive drought assessment method based on multi-source data according to claim 1, It is characterized in that The remote sensing data and meteorological station data in step 1 include historical data of several consecutive months that occurred before the date to be measured in the target area.

3. The comprehensive drought assessment method based on multi-source data according to claim 2, It is characterized in that The remote sensing data include precipitation, temperature, vegetation index, elevation and evapotranspiration data; the meteorological station data include the longitude and latitude coordinates of the meteorological station, altitude, monthly average temperature, and monthly precipitation data.

4. The comprehensive drought assessment method based on multi-source data according to claim 3, It is characterized in that The step 2 is specifically implemented according to the following steps: Step 2.1, converting the remote sensing data file obtained in step 1 into a file format; Step 2.2, batch stitching and reprojection of remote sensing data according to the vector boundary of the target area; Step 2.3, the spatial scale of remote sensing data was resampled to 1 km using the nearest neighbor assignment method; Step 2.4: For remote sensing images with some missing data, the inverse distance weighted method is used to fill in the missing data; Step 2.5, the temperature and evapotranspiration data with a time resolution of 8 days are synthesized into monthly data using the maximum value synthesis method; In step 2.6, the precipitation, temperature, vegetation index, elevation and evapotranspiration data of the corresponding meteorological station are extracted using the bilinear interpolation method, taking the longitude and latitude coordinates of each meteorological station as the center.

5. The comprehensive drought assessment method based on multi-source data according to claim 4, It is characterized in that In step 3, the standardized precipitation evapotranspiration index of each meteorological station on a one-month scale is calculated using the Thornthwaite method based on the latitude and longitude coordinates, altitude, monthly average temperature and monthly precipitation data of each meteorological station in the target area.

6. The comprehensive drought assessment method based on multi-source data according to claim 5, It is characterized in that The step 4 is specifically implemented according to the following steps: Step 4.1, predefine the hyperparameters, optimization algorithm and loss function of the DNN model; Step 4.2, generating a DNN model based on the hyperparameters in step 4.1; In step 4.3, the optimization algorithm and loss function in step 4.1 are used to train the DNN model in step 4.2 to build a comprehensive drought assessment model.

7. The comprehensive drought assessment method based on multi-source data according to claim 6, It is characterized in that The hyperparameters in step 4.1 include the number of network layers, the number of neurons, the number of hidden layers, the number of iterations, the training batch size, the learning rate, and the discard rate; The optimization algorithm is the Adam model optimization algorithm; The loss function is the mean square error loss function MSE.

8. The comprehensive drought assessment method based on multi-source data according to claim 7, It is characterized in that In step 4.2, based on the historical data of several consecutive months before the test date in the target area, the precipitation, temperature, vegetation index, elevation, evapotranspiration and standardized precipitation evapotranspiration index data are divided into a training set and a test set in a ratio of 4:1, and the pre-set drought level is used as the training target.

9. The comprehensive drought assessment method based on multi-source data according to claim 8, It is characterized in that In step 5, the precipitation, temperature, vegetation index, elevation and evapotranspiration remote sensing data of the target area in step 1 are input into the comprehensive drought assessment model in step 4, and a comprehensive drought index MDI considering multiple factors is output.

10. The comprehensive drought assessment method based on multi-source data according to claim 9, It is characterized in that In step 6, the drought level is identified according to the SPEI drought level classification standard. When the MDI value is less than or equal to -0.5, it indicates that drought has occurred.

Citation Information

Patent Citations

  • Agricultural drought monitoring method based on multi-source data and hybrid convolutional neural network-random forest model

    CN116258407A