Remote sensing virtual station-based river water level time sequence encryption method for observation station missing region

Abstract

The invention discloses a river water level time sequence encryption method for an observation station missing region based on a remote sensing virtual station, and the method comprises the steps: taking a measurement foot point of a satellite-borne satellite altimeter as a virtual station of a river channel, and extracting a river reach water area based on the satellite height measurement data of a river reach where the virtual station is located at the same time and a remote sensing image; and constructing an empirical model of the water level and the water area of the river virtual station. And encrypting a long-time-period, high-time-frequency and near-real-time river water level time sequence by using the empirical model and a water area time sequence observed by a remote sensing virtual constellation in a high-time-frequency manner. The method is mainly oriented to an observation station missing area, the defect of low time resolution of satellite height measurement is overcome through the high time frequency observation advantage of a virtual constellation constructed by multi-source optics and radar remote sensing satellites, and method support is provided for understanding the river hydrological situation time-space evolution characteristics of the observation station missing area.

Description

Technical Field

[0001] The invention belongs to the field of hydrological remote sensing, and in particular relates to a time series encryption method for river water levels in a station-missing area based on a remote sensing virtual station. Background Art

[0002] As an important part of the earth's water cycle, rivers are of great significance to social and economic development, ecological sustainable development and climate stability. In the context of global change, climate change and human activities have intensified the global hydrological cycle, causing most rivers in the world to become more fragile and unstable. Furthermore, these unstable rivers control the material transportation channels, affecting the hydrological cycle and the safety of life and property of downstream residents.

[0003] Real-time monitoring of river water level fluctuations is a key technical issue in the current research field of hydrological remote sensing. This problem is particularly evident in areas where the distribution of observation stations is sparse or missing. Understanding the hydrological situation of rivers is an important basis for revealing the hydrological cycle process and preventing flood disasters. At the same time, the dynamic and changeable characteristics of river hydrological situation highlight the importance of high-frequency and real-time monitoring. The deployment of hydrological stations is the most accurate method to obtain river hydrological parameters, but on a large spatial scale, the deployment of hydrological stations cannot take into account multiple factors such as economy, efficiency and reality. The development of remote sensing observation technology has not only improved the spatial refinement of river hydrological monitoring, but also increased the frequency of river monitoring, achieving a balance between economy and efficiency. River water level is one of the key parameters reflecting river hydrological situation. Satellite altimetry technology has been widely used in river water level monitoring. Common altimetry satellites include Topex / Poseidon, Jason-1 / 2 / 3, ERS-1 / 2, Envisat, ICESat, ICESat-2, CryoSat-2 and Saral / AltiKa. However, the low revisit period of altimetry satellites limits their ability to conduct high-frequency monitoring of river water levels, making the current satellite altimetry methods unable to meet the technical requirements for continuous and high-frequency monitoring of dynamic changes in river water levels.

[0004] Compared with altimetry satellites, optical remote sensing satellites and microwave radar remote sensing satellites have the advantages of high revisit period and high temporal resolution imaging. The construction of a virtual satellite constellation by combining multiple optical and radar remote sensing observation platforms can achieve daily-scale monitoring of most rivers in the world, but the monitoring technology of how to convert two-dimensional observation images into three-dimensional water level elevation measurement still needs to be broken through. Summary of the invention

[0005] The object of the present invention is to break through the technical bottleneck of long-term, high-time-frequency, and near-real-time monitoring of river water levels in existing remote sensing technologies, and provide a method for encrypting the time series of river water levels in areas lacking gauging stations. The high-time-frequency observation advantages of constructing a virtual constellation with multi-source optical and radar remote sensing satellites are used to make up for the low temporal resolution defect of satellite altimetry. Combining with the empirical model of water level-water area of virtual stations, the dense time series of river water levels of virtual stations is encrypted. The present invention proposes a dynamic monitoring framework for river water levels based on remote sensing virtual stations, constructs an empirical model of water level-water area based on the potential function relationship between synchronously observed river water levels and water areas, and applies this model to the time series of water areas observed with high time frequency to encrypt the time series of river water levels. The present invention mainly targets areas lacking gauging stations, encrypts the dense time series of river water levels based on remote sensing observation data, and provides methodological support for understanding the spatio-temporal evolution characteristics of river hydrological regimes in areas lacking gauging stations.

[0006] To achieve the above technical objectives, the present invention adopts the following technical solutions: A method for encrypting the time series of river water levels in areas lacking gauging stations based on remote sensing virtual stations, comprising: Generating buffer zones along the upstream and downstream of the river channel as the control range of virtual stations, and acquiring optical remote sensing images and radar remote sensing images within the control range of the virtual stations; Constructing a virtual constellation image time series from the acquired optical remote sensing images and radar remote sensing images according to the imaging time; Extracting the water areas of each scene image from the optical remote sensing images and radar remote sensing images respectively; Matching the water areas with the water area values of virtual stations with the same or adjacent virtual station water level observation dates, and constructing an empirical model of water level-water area for each virtual station; Based on the empirical model and the time series of water areas of the observed virtual constellation images, encrypting the time series of river water levels.

[0007] As a preferred implementation manner, the method further includes: preprocessing the measurement footpoints of the satellite altimeter, removing redundant points with close spatial proximity and invalid points outside the river channel range, and generating buffer zones along the upstream and downstream of the river channel as the control range of virtual stations. Further, after fusing the multi-source altimetry datasets HYDROWEB and DAHITI, removing redundant points with close spatial proximity and invalid points outside the river channel range, and generating buffer zones upstream and downstream along the river channel centerline with the remaining points as the center.

[0008] As a preferred implementation manner, calculating a remote sensing vegetation index for water body extraction for the optical remote sensing images, and extracting the water areas of each scene image from the calculated remote sensing vegetation index images based on threshold segmentation; Further, the remote sensing vegetation index for water body extraction is the normalized difference water index NDWI.

[0009] As a preferred implementation manner, the initial water area range is extracted from the radar remote sensing image by using a preset initial segmentation threshold. After extracting the edge of the initial water area range, a buffer zone is created through morphological closing operation; The frequency distribution characteristics of the pixels in the buffer zone of each scene image are respectively counted, and the water area range of each scene image is extracted from the gray-scale image of the backscattering coefficient of the radar remote sensing image based on threshold segmentation.

[0010] Furthermore, the Kernel size of the morphological closing operation is set to 15 - 25 pixels.

[0011] As a preferred implementation manner, the Otsu method is used for threshold segmentation.

[0012] As a preferred implementation manner, the water level - water area empirical model is constructed as follows: the water level time series of each virtual station is matched with the water area values of the same or adjacent observation dates. Taking the water area as the independent variable and the water level as the dependent variable, a fitting function is established, that is, the water level - water area empirical model.

[0013] As a preferred implementation manner, the method further includes selecting the optimal fitting model with the coefficient of determination as the evaluation index for the established fitting function as the water level - water area empirical model.

[0014] As a preferred implementation manner, the optical remote sensing images are Landsat 8 / 9 and Sentinel - 2 A / B remote sensing images; the radar remote sensing images are Sentinel - 1 A / B remote sensing images.

[0015] As a preferred implementation manner, the method further includes preprocessing the Sentinel - 1 A / B images. After removing the speckle noise of the Sentinel - 1 A / B IW - mode VV - polarized SAR images, the optimal water - land segmentation threshold is predicted based on the adaptive threshold method to extract water bodies. Furthermore, according to the imaging mode and polarization mode of the images (IW - mode VV - polarization), the GRD data products in the available Sentinel - 1 SAR images are selected, and the speckle noise of each scene of SAR images is removed by using the Refined Lee filter to complete the preprocessing.

[0016] The present invention has the following beneficial effects: (1) The river water level time - series encryption method proposed by the present invention greatly reduces the manual measurement and station - setting costs, and can be completed based on publicly available remote sensing data, which is especially suitable for areas with difficult measurement operations and areas lacking measurement stations; (2) The algorithm of the present invention is easy to implement and can quickly encrypt the river water level time series. It is applied to the monitoring of large-scale river dynamic changes and has good applicability to rivers with different flow states and different geomorphic environments. It can be extended to large regional and even global scale research, providing a method support for understanding the spatio-temporal evolution characteristics of river hydrological situations in areas lacking gauging stations under the background of global change. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a map of the sample area and the spatial distribution of virtual stations in the embodiment of the present invention.

[0018] Figure 2 It is a flowchart of the algorithm of the present invention.

[0019] Figure 3 It is the fitting effect of the water level-water area empirical model of each virtual station in the embodiment of the present invention.

[0020] Figure 4 It is the encryption accuracy of the virtual station water level in the embodiment of the present invention.

[0021] Figure 5 It is the real-time monitoring potential of the virtual station water level in the embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0022] The following will further describe in detail the specific implementation manners of the present invention in conjunction with the drawings and embodiments. The following embodiments are used to illustrate the present invention, but not to limit the scope of the present invention.

[0023] The embodiments of the present application are nine major rivers originating from the high mountains in Asia, including the Amu Darya, Yarlung Zangbo River, Ganges River, Ili River, Indus River, Mekong River, Salween River, Yellow River and Yangtze River. The high mountains in Asia include the Qinghai-Tibet Plateau known as the "Asian Water Tower" and its surrounding mountains, with the terrain showing a trend of high in the southwest and low in the northeast. Affected by the complex terrain and atmospheric circulation, the climate shows the characteristics of severe cold and dryness in the northwest and warm and humid in the southeast. In recent years, the imbalance of the Asian Water Tower under the influence of climate change has led to sudden changes in the hydrological situations of these rivers. The hydrological stations of these rivers are sparse and mostly concentrated in the downstream populated areas, resulting in the problem of lack of monitoring data in the upstream. Considering the terrain and climate characteristics of the high mountains in Asia, as well as the challenges such as insufficient hydrological stations and lack of observational data of the large rivers originating from this area, the embodiments of the present invention are representative to a certain extent.

[0024] This embodiment includes the following steps: Step 1. Preprocessing of river virtual stations. The altimetry datasets HYDROWEB and DAHITI obtained from multi-source satellite altimeters are data-fused, and the attribute information of the fused data is retained, including spatial location, measurement date, elevation, etc. According to the spatial proximity search algorithm, redundant points within 5 km are removed, and invalid points outside the river channel water area are manually deleted, and a total of 157 virtual stations are retained. Taking the altimetry foot points as the center, 2.5 km buffers are generated along the river centerline in the upstream and downstream directions as the control range of the virtual stations.

[0025] Step 2. Constructing the observation image set of the remote sensing virtual constellation. Landsat 8 / 9, Sentinel-1 A / B, and Sentinel-2 A / B remote sensing images from January 2017 to December 2022 are obtained from the Google Earth Engine (GEE) platform (https: / / earthengine.google.com / ), and the spatial constraints for image screening refer to the control range of the virtual stations generated in Step 1. The results of atmospheric correction of optical images can be directly obtained by screening OLI and MSI surface reflectance products, and optical images with cloud and cloud shadow occlusion within the control range of the virtual stations are removed. The retained optical images and SAR images are used to construct the virtual constellation image time series according to the imaging time.

[0026] Step 3. Extracting the water area range within the control of the virtual stations from optical images. The Normalized Difference Water Index (NDWI) is calculated using the green band and near-infrared band surface reflectance of each Landsat 8 / 9 OLI image and Sentinel-2 A / B MSI image. The frequency distribution characteristics of the water index grayscale images generated for each scene are statistically analyzed respectively, the segmentation threshold with the maximum between-class variance is searched using an equal-interval step size, and the water area range of each scene image is extracted based on this threshold to segment the water index NDWI grayscale image.

[0027] Step 4. Extracting the water area range within the control of the virtual stations from SAR images. The GRD data product is selected according to the IW imaging mode and VV polarization mode of the Sentinel-1 SAR image, and the speckle noise in the GRD data product is filtered using the Refined Lee filter. The initial segmentation threshold of the SAR image is set to -15 dB to obtain the initial water area range, the Canny edge filter is used to extract the initial water edge line, and the morphological closing operation kernel size is set to 20 pixels to create a buffer. The distribution characteristics of the pixel values within the buffer of each scene image are statistically analyzed respectively, the segmentation threshold with the maximum between-class variance is searched using an equal-interval step size, and the SAR image is re-segmented based on this updated threshold to extract the water area range of each scene image.

[0028] Step 5: Construct a water level - water area empirical model. Match the water area values corresponding to the water levels at each moment for 157 virtual stations one by one. Taking the water area as the independent variable x and the water level as the dependent variable y, fit a linear function or a quadratic polynomial function, and use the coefficient of determination as the evaluation index to select a water level - water area empirical model with better fitting performance for each of the 157 virtual stations.

[0029] Step 6: Densify the water level time series of river virtual stations. For the 157 virtual stations, based on the water level - water area empirical model selected for each and the water area time series observed by the high - time - frequency remote sensing virtual constellation, predict the water levels of the virtual stations corresponding to the water area monitoring dates, and densify the water level time series of river virtual stations.

[0030] Step 7: Verify the accuracy of the densified water level. Randomly select the predicted water level values at any moment of each virtual station, and compare them with the true values regarded as the water levels observed by the ICESat - 2 satellite altimeter and the water levels observed by hydrological stations. Calculate the root - mean - square error and relative error of the predicted water levels respectively to verify the accuracy of the densified water level.

Claims

1. A method for encrypting river water levels in areas without remote sensing stations based on remote sensing virtual stations, characterized in that: include: Generate a buffer zone along the upstream and downstream of the river as the control range of the virtual station, and obtain optical remote sensing images and radar remote sensing images within the control range of the virtual station; Constructing a virtual constellation image time series according to the imaging time using the acquired optical remote sensing images and radar remote sensing images; Extracting the water area of ​​each scene from the optical remote sensing image and the radar remote sensing image; Match the water area range with the water area value of the virtual station with the same or adjacent water level observation date, and construct a water level-water area empirical model for each virtual station; Based on the empirical model and the water area time series of the observed virtual constellation images, the river water level time series is encrypted.

2. The method according to claim 1, characterized in that The satellite altimeter measurement points are preprocessed to eliminate redundant points in close space and invalid points outside the river channel, and a buffer zone is generated along the upstream and downstream of the river as the control range of the virtual station.

3. The method according to claim 1, characterized in that A remote sensing vegetation index for water body extraction is calculated for the optical remote sensing image, and the water area of ​​each scene is extracted from the calculated remote sensing vegetation index image based on threshold segmentation.

4. The method according to claim 3, characterized in that The remote sensing vegetation index used for water body extraction is the Normalized Difference Water Index (NDWI).

5. The method according to claim 1, characterized in that Extracting an initial water area range from the radar remote sensing image using a preset initial segmentation threshold, extracting edges from the initial water area range, and creating a buffer zone through a morphological closing operation; The frequency distribution characteristics of the pixels in the buffer of each image are counted respectively, and the water area of ​​each image is extracted from the grayscale image of the backscatter coefficient of the radar remote sensing image based on threshold segmentation.

6. The method according to claim 5, characterized in that The kernel size of the morphological closing operation is set to 15-25 pixels.

7. The method according to claim 3 or 5, characterized in that: The maximum inter-class variance method is used for threshold segmentation.

8. The method according to claim 1, characterized in that The water level-water area empirical model is constructed as follows: the water level time series of each virtual station is matched with the water area value of the same or adjacent observation date, the water area is taken as the independent variable and the water level is taken as the dependent variable, and a fitting function is established, namely the water level-water area empirical model.

9. The method according to claim 8, characterized in that It also includes selecting the best fitting model from the established fitting function using the determination coefficient as an evaluation index as the water level-water area empirical model.

10. The method according to claim 1, characterized in that The optical remote sensing images are Landsat 8 / 9 and Sentinel-2 A / B remote sensing images; the radar remote sensing images are Sentinel-1 A / B and remote sensing images.

Images