A drought monitoring and early warning method, system and device based on characteristic water area changes

Abstract

The present invention discloses a drought monitoring and early warning method, system, and equipment based on changes in characteristic water areas. First, a water body precision extraction technology based on random deep forest classification and a comprehensive threshold segmentation model is used to automatically and accurately extract surface water areas in the target area. Then, characteristic water areas are identified and separated based on prior knowledge and area thresholds, and the changes in the area of ​​characteristic water areas over multiple years are monitored in real time. Finally, drought early warnings and comprehensive assessments are performed based on the dynamic changes in the characteristic water areas. When the area of ​​characteristic water areas decreases significantly during the monitoring period, a conclusion is drawn that drought may occur based on the actual situation and a drought early warning is issued. The present invention directly uses surface water as an indicator and factor to characterize the drought state, which is simpler than other models and methods and can more intuitively reflect the development and evolution of droughts. The present invention performs real-time drought monitoring based on changes in the area of ​​characteristic water areas, and can more accurately and timely perform drought early warnings and comprehensive assessments.

Description

Technical Field

[0001] The present invention belongs to the technical field of water and drought disaster monitoring and prevention, and relates to a drought monitoring and early warning method, system and equipment, and in particular to a method, system and equipment for drought monitoring and drought early warning based on changes in characteristic water area. Background Art

[0002] Drought is one of the most costly, widespread, and complex natural disasters globally. It is the leading cause of food production losses, severely impacting water resources, agricultural production, ecosystems, and human life. Research indicates that the frequency and duration of droughts have increased by 29% globally since the beginning of this century, and droughts are expected to become more severe, frequent, and prolonged in the future. It is estimated that by 2050, over 75% of the world's population may be affected by drought. Global climate change will lead to significant shifts in the spatial and temporal distribution of droughts, creating uncertainty at various scales.

[0003] Drought monitoring and early warning are of vital importance to food security, ecological security, economic and social stability, and ensuring basic living conditions for vulnerable groups. However, due to the wide range, long duration, and multiple influencing factors of drought, monitoring and early warning remain a global challenge. Current drought monitoring relies primarily on ground-based statistical and remote sensing data and models. These methods utilize single or multiple indicators, such as precipitation, evapotranspiration, runoff, soil moisture, temperature, and vegetation conditions, to calculate relevant index models, thereby directly or indirectly representing drought conditions. However, these methods still have shortcomings in certain situations. For one thing, in areas where ground-based stations are scarce and unevenly distributed, accurate monitoring of drought conditions over large areas is impossible. Furthermore, some methods, due to difficulties in obtaining real-time data and complex models, may not be able to directly and promptly reflect the development of drought conditions, thus failing to meet the needs of water conservancy and drought relief efforts. Therefore, exploring more comprehensive, accurate, and timely drought monitoring and early warning methods is a crucial task.

[0004] Compared to other indicators, surface water and drought have a more direct connection, with changes in surface water more directly reflecting the development of drought conditions. However, relevant research has shown that small water bodies have poor self-regulation capabilities when the environment and climate undergo dramatic changes, making them more sensitive to drought. Therefore, obtaining dynamic information about small water bodies for drought monitoring may be a more effective approach. Therefore, the primary challenge addressed by this invention is how to accurately monitor and provide early warning of drought conditions through changes in small water bodies. Summary of the Invention

[0005] In response to the complexity of drought generation and impact mechanisms, and the inability of many existing drought monitoring methods to directly, accurately, and promptly reflect the development of large-scale droughts, this invention, based on previous related research, proposes through a series of experimental investigations a mechanism by which smaller water bodies have poor self-regulation capabilities and are more sensitive to drought evolution. Small water bodies with an area smaller than a certain threshold are defined as characteristic water bodies, and a pioneering method, system, and equipment for drought monitoring and drought warning based on changes in the area of ​​characteristic water bodies are proposed.

[0006] The technical solution adopted by the method of the present invention is: a drought monitoring and early warning method based on changes in characteristic water areas, comprising the following steps:

[0007] Step 1: Combine random forest classification and comprehensive threshold segmentation to automatically and accurately extract the surface water area of ​​the target area, including water target extraction in cloud-free areas and water target extraction in cloud-covered areas, as well as post-processing;

[0008] The water body target extraction in the cloud-free area is carried out by querying all available Sentinel-2 image data within the monitoring time range of the target area according to the drought monitoring requirements, and using the pre-prepared water body classification samples to perform random forest classification to obtain the water body area W1 and the area classified as cloud-covered area C; the water body area W1 is the water body target range initially extracted in the cloud-free area;

[0009] The water body target extraction in the cloud-covered area is as follows: for the area C classified as cloud-covered, the Sentinel-1 image data within the corresponding spatiotemporal range is queried for supplementation, and threshold classification is performed using a comprehensive threshold method to extract a preliminary water body area; a series of masking operations are performed on the preliminary water body area using several comprehensive threshold discrimination conditions to remove interference from some water-like features, and the water body area W2 is obtained, which is the water body target range of the cloud-covered area;

[0010] The post-processing is to mask the initial water area W3 synthesized from the water areas W1 and W2 using prior data on human settlements, slope masking using DEM data, and maximum water area data formed through long-term monitoring, thereby removing building shadows, mountain shadows, and other water-like features that are mistakenly classified as water areas;

[0011] Step 2: Identify and separate characteristic water area targets from the water body extraction results by combining prior knowledge that conforms to the actual situation of the target area and the area threshold;

[0012] The characteristic water area is a small water body target with an area smaller than a certain threshold; a water body with a smaller area is more sensitive to drought evolution. The present invention adopts the discrimination condition shown in formula (1) and sets the area smaller than the threshold T a Small water bodies are defined as characteristic water areas.

[0013] Area <T a (1)

[0014] Where Area is the water area in km 2 , T a The threshold for determining characteristic water areas needs to be determined based on the actual scale characteristics of the surface water in the target area.

[0015] Step 3: Based on the actual needs of drought monitoring services, conduct real-time monitoring of changes in the area of ​​characteristic water areas over multiple years, and then issue drought warnings based on the dynamic changes in the characteristic water areas.

[0016] The technical solution adopted by the system of the present invention is: a drought monitoring and early warning system based on characteristic water area changes, including the following modules:

[0017] Module 1 is used to automatically and accurately extract surface water bodies in the target area by combining random forest classification and comprehensive threshold segmentation, including water body target extraction in cloud-free areas and water body target extraction in cloud-covered areas, as well as post-processing;

[0018] The water body target extraction in the cloud-free area is carried out by querying all available Sentinel-2 image data within the monitoring time range of the target area according to the drought monitoring requirements, and using the pre-prepared water body classification samples to perform random forest classification to obtain the water body area W1 and the area classified as cloud-covered area C; the water body area W1 is the water body target range initially extracted in the cloud-free area;

[0019] The water body target extraction in the cloud-covered area is as follows: for the area C classified as cloud-covered, the Sentinel-1 image data within the corresponding spatiotemporal range is queried for supplementation, and threshold classification is performed using a comprehensive threshold method to extract a preliminary water body area; a series of masking operations are performed on the preliminary water body area using several comprehensive threshold discrimination conditions to remove interference from some water-like features, and the water body area W2 is obtained, which is the water body target range of the cloud-covered area;

[0020] The post-processing is to mask the initial water area W3 synthesized from the water areas W1 and W2 using prior data on human settlements, slope masking using DEM data, and maximum water area data formed through long-term monitoring, thereby removing building shadows, mountain shadows, and other water-like features that are mistakenly classified as water areas;

[0021] Module 2 is used to identify and separate characteristic water body targets from the water body extraction results by combining prior knowledge that conforms to the actual situation of the target area and an area threshold; the characteristic water body targets are small water body targets with an area smaller than a certain threshold;

[0022] Module 3 is used to monitor the changes in the area of ​​characteristic water areas over multiple years in real time according to the actual needs of drought monitoring services, and then issue drought warnings based on the dynamic changes in characteristic water areas.

[0023] The technical solution adopted by the device of the present invention is: a drought monitoring and early warning device based on characteristic water area changes, comprising:

[0024] one or more processors;

[0025] A storage device is used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the drought monitoring and early warning method based on changes in characteristic water areas.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] (1) The present invention directly uses surface water as an indicator and factor to characterize the drought state, which is simpler than other models and methods and can more intuitively reflect the development and evolution of drought.

[0028] (2) The present invention proposes a mechanism in which changes in characteristic water areas are more sensitive to the evolution of drought conditions. Real-time drought monitoring based on changes in the area of ​​characteristic water areas can provide more accurate and timely drought warnings and comprehensive assessments.

[0029] (3) The implementation process, calculation method and principle of the method of the present invention are relatively simple, and data acquisition is relatively easy, so it is suitable for precise drought monitoring in large areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 is a flow chart of a method according to an embodiment of the present invention;

[0031] Figure 2 This is a flowchart showing the accurate extraction of water targets based on random deep forest classification and integrated threshold segmentation model in an embodiment of the present invention;

[0032] Figure 3 This is a graph showing changes in characteristic water area in the Dongjiang and Hanjiang River basins from 2017 to July 2021 according to an embodiment of the present invention;

[0033] Figure 4 This is a graph showing changes in characteristic water area in the Dongjiang and Hanjiang River basins from 2017 to August 2021 according to an embodiment of the present invention;

[0034] Figure 5 This is a graph showing changes in characteristic water area in the Dongjiang and Hanjiang River basins from 2017 to September 2021 according to an embodiment of the present invention;

[0035] Figure 6This is a graph showing changes in characteristic water area in the Dongjiang and Hanjiang River basins from 2017 to October 2021 according to an embodiment of the present invention;

[0036] Figure 7 This is a graph showing changes in characteristic water area in the Dongjiang and Hanjiang River basins from 2017 to November 2021 according to an embodiment of the present invention;

[0037] Figure 8 This is a graph showing changes in characteristic water area in the Dongjiang and Hanjiang River basins from December 2017 to 2021 according to an embodiment of the present invention. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and implementation examples. It should be understood that the specific implementations described herein are intended to help better understand the content of the present invention, but these specific implementations do not limit the scope of protection of the present invention in any way.

[0039] The target area of ​​the present invention example is the Dongjiang River and Hanjiang River basins in the Pearl River basin, the image data is Sentinel-2 and Sentinel-1, and the programming tool is the Google Earth Engine (GEE) cloud computing platform.

[0040] Please see attached Figure 1 The present invention provides a drought monitoring and early warning method based on characteristic water area changes, comprising the following steps:

[0041] Step 1: Using a water target extraction technique that combines random forest classification and comprehensive threshold segmentation, the surface water areas of the target area are automatically and accurately extracted, including water target extraction in cloud-free areas and water target extraction in cloud-covered areas, as well as post-processing.

[0042] In this embodiment, the surface waters of the Dongjiang and Hanjiang River basins are automatically and accurately extracted. The water extraction technology flow chart is shown in the attached figure. Figure 2 shown.

[0043] Among them, the target extraction of water bodies in cloudless areas in this embodiment is based on the needs of drought monitoring and queries all available Sentinel-2 image data within the Dongjiang and Hanjiang River basins according to a specific time scale. The method of the present invention uses the surface reflectance (Surface Reflectance) product of the image, which can be directly called in GEE through the ee.ImageCollection('COPERNICUS / S2_SR') code. Before extracting water bodies, it is necessary to prepare a sufficient number of ground object classification samples suitable for the target area according to actual conditions, including ground object categories such as water bodies, land, and clouds, and consider the influence of different seasons; use the prepared samples to perform random forest classification through the ee.Classifier.smileRandomForest() function in the GEE platform to obtain the water body area W1 and the area C classified as cloud-covered. The water body area W1 is the preliminary extraction result of the water body in the cloudless area.

[0044] The water body target extraction in the cloud-covered area of ​​this embodiment is to query and obtain the Sentinel-1 image data within the corresponding time and space range for the area C classified as cloud-covered to ensure complete coverage; the method of the present invention uses the IW mode of the GRD product of the image. First, the Sentinel-1 image is threshold-segmented using the comprehensive threshold method shown in formula (2) to obtain a preliminary water area extraction result; on this basis, mask processing is performed using the comprehensive threshold discrimination condition shown in formula (3) to remove interference from water-like features such as deserts and ice and snow, thereby obtaining a more accurate water body extraction result W2 in the cloud-covered area. By using Sentinel-1 images to supplement the water body extraction in the cloud-covered area, the water body extraction is more complete and accurate than using only optical images for water body extraction.

[0045] VV≤T1 and VH≤T2 (2)

[0046] NDWI>T3 and NIR≤T4 (3)

[0047] In formulas (2) and (3), VV and VH are the dual-frequency cross-polarization band combinations of Sentinel-1 images, NDWI is the normalized difference water index, NIR is the near-infrared band, and T1, T2, T3, and T4 are empirical thresholds that need to be determined experimentally according to the actual conditions of the target area.

[0048] This example combines the water areas extracted using Sentinel-2 and Sentinel-1 imagery to create the initial water area W3. To achieve more precise and accurate water extraction results for the Dongjiang and Hanjiang River basins, this example designed the following three different processing mechanisms to perform reasonable post-processing on the initial water extraction results:

[0049] ① The initial water area W3 is masked using globally available human settlement grid data to remove interference from building shadows and other water-like features. The present invention uses the GHSL (Global Human Settlement Layers) dataset released by the European Commission. Its human settlement layer can be directly called on the GEE platform through the ee.Image('JRC / GHSL / P2016 / BUILT_LDSMT_GLOBE_V1').select('built') code.

[0050] ② Based on the method in step 1, slope data is calculated using publicly available DEM data. Slope masking is performed using a specific altitude and slope threshold to remove the interference of mountain shadows when extracting water bodies in mountainous areas. This method uses the SRTM dataset provided by NASA, which has a resolution of 30 meters and can be directly accessed on the GEE platform using the ee.Image('USGS / SRTMGL1_003') function. It should be noted that this step is not necessary for some low-altitude target areas located in plains.

[0051] ③Finally, the water body extraction results are masked using the publicly available maximum water area data to remove interference from ice and snow pixels and other noise when extracting water bodies in cloudless areas. The present invention selects the maximum water area layer data in the JRC global surface water dataset, which can be directly called on the GEE platform through the ee.Image('JRC / GSW1_3 / GlobalSurfaceWater').select('max_extent') function code.

[0052] The mask operations described above can all be implemented through the updateMask() function in the GEE platform.

[0053] Step 2: Identify and separate characteristic water area targets from the water body extraction results by combining prior knowledge that conforms to the actual situation of the target area and the area threshold; the characteristic water area is the area smaller than the characteristic water area judgment threshold T a Small water bodies;

[0054] In this embodiment, the characteristic water area determination threshold T is first determined based on the actual scale characteristics of the surface water in the Dongjiang and Hanjiang River basins. aThe size of the water area is then accurately identified and separated through the following two steps.

[0055] Step 2.1: Eliminate large rivers and lakes based on prior knowledge.

[0056] In this embodiment, in addition to the waters of the Dongjiang and Hanjiang rivers themselves, large water bodies such as the Xinfengjiang Reservoir, Baipenzhu Reservoir, and Mianhuatan Reservoir are also included. This embodiment obtains the water areas of these large rivers, reservoirs, and lakes through relevant methods, generates appropriate buffer zones as prior knowledge, and performs masking on the water body extraction results. This not only removes large water bodies, but also further eliminates noise interference around the water bodies.

[0057] Step 2.2: For the water area obtained in step 2.1, continue to filter out all water areas with an area greater than or equal to the threshold T. a The non-characteristic water areas are separated and then the complete characteristic water area targets are separated. This step can be screened and implemented in GEE through relevant codes.

[0058] Step 3: Based on the actual needs of drought monitoring services, conduct real-time monitoring of changes in the area of ​​characteristic water areas over multiple years, and then issue drought warnings based on the dynamic changes in the characteristic water areas.

[0059] This embodiment monitors changes in the area of ​​characteristic waters in a target area in real time. The area of ​​characteristic waters within a specific time period is compared with the area of ​​the same period in previous years. If the area of ​​characteristic waters within the monitoring period decreases significantly compared to the same period in previous years and remains at a low level, it can be concluded that drought is imminent in the area, and a drought warning can be issued.

[0060] This embodiment extracts and separates the precise characteristic water area results of the Dongjiang and Hanjiang River basins from 2017 to 2021 through the above series of steps. Based on the complexity of the water body extraction process in the Hanjiang and Dongjiang River basins, real-time monitoring of the changes in the characteristic water area over the same period over many years is selected.

[0061] This embodiment conducts drought warning and comprehensive assessment based on the changes in the characteristic water area. It is monitored that the characteristic water area in the embodiment has decreased significantly from July to December in 2018 and 2019 compared with the same period in 2017, and has remained at a low level in the same period in 2020 and 2021. Figures 3 to 8 As shown; therefore, based on the actual situation, it is concluded that the Dongjiang and Hanjiang River basins will experience severe droughts from July to December from 2019 to 2021, and a drought warning is issued.

[0062] This paper uses Sentinel-2 and Sentinel-1 images to automatically and accurately extract surface water areas in target areas, separate characteristic water targets based on certain prior knowledge and area thresholds, and then conducts real-time monitoring of the changes in the area of ​​characteristic water areas over multiple years, thereby achieving drought warning and comprehensive assessment based on the dynamic changes in characteristic water areas.

[0063] By extracting water bodies in cloudless and cloud-covered areas separately, the present invention's initial water body extraction results may contain areas where building shadows, mountain shadows, and other water-like features are mistakenly identified as water bodies. Here, publicly available global human settlement grid data is used to mask the initial water bodies to remove interference from building shadows in built-up areas. For mountain shadows in mountainous areas, publicly available DEM data is used to perform slope masking on the water body extraction results to remove interference. Finally, publicly available data on the maximum water area is used for further masking to remove interference from other ground noise. This post-processing step effectively improves the accuracy of water body extraction, thereby making drought warning and monitoring more timely and accurate.

[0064] This method is applicable to drought monitoring and early warning scenarios in most regions under normal circumstances. Implementation requires setting empirical thresholds for water extraction and drought monitoring and early warning based on the actual conditions of the target area. Using the GEE cloud computing platform, a program was designed to accurately extract characteristic water areas and implement a comprehensive drought monitoring and early warning process.

[0065] The above description is merely a partial embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that the present invention is susceptible to various modifications. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of the present invention.

Claims

1. A drought monitoring and early warning method based on changes in characteristic water areas, characterized in that: The following steps are involved: Step 1: Combine random forest classification and comprehensive threshold segmentation to automatically and accurately extract the surface water area of ​​the target area, including water target extraction in cloud-free areas and water target extraction in cloud-covered areas, as well as post-processing; The water body target extraction in the cloud-free area is carried out by querying all available Sentinel-2 image data within the monitoring time range of the target area according to the drought monitoring requirements, and using the pre-prepared water body classification samples to perform random forest classification to obtain the water body area W1 and the area classified as cloud-covered area C; the water body area W1 is the water body target range initially extracted in the cloud-free area; For the water body target extraction in the cloud-covered area, for the area C classified as cloud-covered, the Sentinel-1 image data within the corresponding time and space range is queried for supplementation, and threshold classification is performed using the comprehensive threshold method to extract the preliminary water body area; A series of masking operations are performed on the preliminary water area through several comprehensive threshold discrimination conditions to remove the interference of some water-like objects, and the water area W2 is obtained, which is the target water range of the cloud coverage area. The post-processing is to mask the initial water area W3 synthesized from the water areas W1 and W2 using prior data on human settlements, slope masking using DEM data, and maximum water area data formed through long-term monitoring, thereby removing building shadows, mountain shadows, and other water-like features that are mistakenly classified as water areas; Step 2: Identify and separate characteristic water area targets from the water body extraction results by combining prior knowledge that conforms to the actual situation of the target area and the area threshold; The characteristic water area is a small water body target with an area smaller than a certain threshold; Step 3: Based on the actual needs of drought monitoring services, conduct real-time monitoring of changes in the area of ​​characteristic water areas over multiple years, and then issue drought warnings based on the dynamic changes in the characteristic water areas.

2. The drought monitoring and early warning method based on characteristic water area changes according to claim 1 is characterized by: In step 1, for the extraction of water body targets in the cloud-covered area, the comprehensive threshold segmentation model of Sentinel-1 image data is shown in formula (1), and the comprehensive threshold mask shown in formula (2) is performed on the initially extracted water body area; VV≤T1 and VH≤T2 (1) NDWI>T3 and NIR≤T4 (2) VV and VH are the dual-frequency cross-polarization band combinations of Sentinel-1 images, NDWI is the normalized difference water index, NIR is the near-infrared band, and T1, T2, T3, and T4 are empirical thresholds.

3. The drought monitoring and early warning method based on characteristic water area changes according to claim 1 is characterized by: In step 1, for the post-processing of the initial water area extraction results, the initial water area W3 is first masked using the global public human settlement grid data to remove the interference of building shadows and other water-like features; On this basis, slope data was calculated using publicly available DEM data, and slope masking was performed based on the preset altitude and slope thresholds to remove the interference of mountain shadows when extracting water bodies in mountainous areas. Finally, the water body extraction results are masked using the publicly available maximum water area data to remove interference from ice and snow pixels and other noise when extracting water bodies in cloudless areas.

4. The drought monitoring and early warning method based on characteristic water area changes according to any one of claims 1 to 3, characterized in that: The specific implementation of step 2 includes the following sub-steps: Step 2.1: First, based on the actual situation of the target area and the needs of drought monitoring, obtain the water areas of large rivers, reservoirs, and lakes in the area, and create appropriate buffer zones as prior knowledge. Then, use masking to remove large rivers, reservoirs, and lakes from the water body extraction results; Step 2.2: Determine the characteristic water area judgment threshold T based on the actual situation of the target area and drought monitoring needs a , and then remove all the areas greater than or equal to T a non-characteristic waters, and then separate the characteristic water targets.

5. A drought monitoring and early warning system based on characteristic water area changes, characterized in that: Includes the following modules: Module 1 is used to automatically and accurately extract surface water bodies in the target area by combining random forest classification and comprehensive threshold segmentation, including water body target extraction in cloud-free areas and water body target extraction in cloud-covered areas, as well as post-processing; The water body target extraction in the cloud-free area is carried out by querying all available Sentinel-2 image data within the monitoring time range of the target area according to the drought monitoring requirements, and using the pre-prepared water body classification samples to perform random forest classification to obtain the water body area W1 and the area classified as cloud-covered area C; the water body area W1 is the water body target range initially extracted in the cloud-free area; For the water body target extraction in the cloud-covered area, for the area C classified as cloud-covered, the Sentinel-1 image data within the corresponding time and space range is queried for supplementation, and threshold classification is performed using the comprehensive threshold method to extract the preliminary water body area; A series of masking operations are performed on the preliminary water area through several comprehensive threshold discrimination conditions to remove the interference of some water-like objects, and the water area W2 is obtained, which is the target water range of the cloud coverage area. The post-processing is to mask the initial water area W3 synthesized from the water areas W1 and W2 using prior data on human settlements, slope masking using DEM data, and maximum water area data formed through long-term monitoring, thereby removing building shadows, mountain shadows, and other water-like features that are mistakenly classified as water areas; Module 2 is used to identify and separate characteristic water area targets from the water body extraction results by combining prior knowledge that conforms to the actual situation of the target area and the area threshold; The characteristic water area is a small water body target with an area smaller than a certain threshold; Module 3 is used to monitor the changes in the area of ​​characteristic water areas over multiple years in real time according to the actual needs of drought monitoring services, and then issue drought warnings based on the dynamic changes in characteristic water areas.

6. A drought monitoring and early warning device based on characteristic water area changes, characterized in that: include: one or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, enables the one or more processors to implement the drought monitoring and early warning method based on characteristic water area changes as described in any one of claims 1 to 4.

Images