River flow remote sensing inversion method based on local water area

Abstract

The invention discloses a river flow remote sensing inversion method based on a local water area, and belongs to the technical field of environment remote sensing and hydrological observation, and the method comprises the following steps: S1, obtaining a flow monitoring section position, an actually measured flow and a remote sensing image of a region of a target river; s2, determining a plurality of research areas; s3, converting the vector file and performing cloud removal processing; s4, generating a binary water surface distribution diagram; s5, obtaining historical maximum riverway boundary vector data; s6, constructing a water surface area data set; s7, forming a paired sample set; s8, constructing a river section flow inversion model, and performing training and verification; s9, outputting a river flow inversion result based on the river section flow inversion model; only the local water area of the river channel needs to be monitored, the monitoring cost is reduced, meanwhile, the dependence of a traditional remote sensing technology on the width of a large river can be broken through, the small and medium-sized river channel flow inversion problem is effectively solved, and the applicability of the remote sensing means in narrow river sections and complex terrain areas is remarkably expanded.

Description

Technical Field

[0001] The present invention belongs to the technical fields of environmental remote sensing and hydrological observation, and particularly relates to a method for remotely sensing and inverting river channel flow based on the local water area. Background Technique

[0002] River channel flow refers to the amount of water passing through a certain cross-section of a river channel per unit time, and its calculation is an important basis for water resources planning, water environment simulation, and basin comprehensive management.

[0003] Traditional methods for calculating river channel flow generally use manual on-site monitoring or observation with the aid of automated hydrological monitoring equipment. Such methods have problems such as high monitoring costs and fewer observation points.

[0004] To solve the above problems, the use of remote sensing methods to interpret river channel flow has emerged. Due to its advantages such as low monitoring costs, wide coverage, and strong adaptability, it has developed rapidly in recent years.

[0005] Currently, the remote sensing methods for interpreting river channel flow mainly obtain two river channel hydraulic parameters, water level and river width, through remote sensing monitoring, and then establish the response relationship between them and the observed flow, so as to achieve the purpose of calculating river channel flow using remote sensing parameters. Due to the limitation of the spatial resolution of remote sensing technology, it is difficult to interpret the flow in areas with narrow river channels, which greatly limits the application of remote sensing technology in such areas.

[0006] In view of this, a method for remotely sensing and inverting river channel flow based on the local water area is designed to solve the above problems. Summary of the Invention

[0007] To solve the problems raised in the above background technique, the present invention provides a method for remotely sensing and inverting river channel flow based on the local water area, which has the characteristics of only needing to monitor the local water area of the river channel, reducing the monitoring cost while being able to break through the dependence of traditional remote sensing technology on the width of large rivers, effectively solving the problem of inverting the flow of small and medium-sized river channels, and significantly expanding the applicability of remote sensing means in narrow river sections and complex terrain areas.

[0008] To achieve the above object, the present invention provides the following technical solution: A method for remotely sensing and inverting river channel flow based on the local water area, comprising the following steps:

[0009] S1: Obtain the position of the flow monitoring cross-section, the measured flow, and the remote sensing image of the area where the target river channel is located;

[0010] S2: Centered on the location of the river channel hydrological monitoring station under study, expand multiple times with the width multiple of the river channel flow monitoring cross-section where the river channel hydrological monitoring station is located or an approximate integer distance of the river channel reach width as the statistical radius to determine multiple river channel water surface area statistical ranges, and use them as multiple research areas;

[0011] S3: Obtain the remote sensing images in multiple research areas respectively, convert them into vector files, and perform cloud removal processing;

[0012] S4: Based on the vector file data of multiple research areas respectively, extract water surface pixels through the normalized difference water index, and generate a binary water surface distribution map;

[0013] S5: Perform maximum value composition on all the normalized difference water indices of multiple research areas respectively. Combine with the actual topographic map to crop the single-band image of the maximum normalized difference water index to obtain the historical maximum river channel boundary vector data in multiple research areas;

[0014] S6: Statistically calculate the effective water area in the maximum historical river channel boundary vector data of multiple research areas, and construct a water surface area dataset;

[0015] S7: Match the constructed water surface area dataset with the measured flow in time and calculate the day of the year to form a paired sample set;

[0016] S8: Use the river channel water area and the day of the year as independent variables, and the paired measured flow as the dependent variable to construct an inverse model for river channel cross-section flow inversion, and train and validate the model based on the paired sample set;

[0017] S9: Output the river channel flow inversion result based on the river channel cross-section flow inversion model.

[0018] Further, in the step S1, the positions of the target river channel flow monitoring sections include the longitude and latitude coordinate data of the target river channel flow monitoring sections. The measured flow includes the historical flow observation data of the target river channel flow monitoring sections. The remote sensing images include the green band and the near-infrared band with a spatial resolution not higher than 1 / 10 of the width of the target river channel section and a spatial range not less than 10 times the river section radius of the area where the target river channel flow monitoring section is located.

[0019] Further, in the step S2, the multiple of the width of the river channel flow monitoring section or the approximate integer distance of the river channel section width is obtained from the longitude and latitude coordinate data of the target river channel flow monitoring section.

[0020] Further, in the step S4, the specific steps of extracting water surface pixels through the normalized difference water index include:

[0021] Calculate the normalized difference water index, and the expression is:

[0022]

[0023] In the formula: NDVWI is the normalized difference water index; B3 is the green band; B8 is the near-infrared band;

[0024] Pixels with a normalized difference water index greater than 0 are assigned a value of 1, determined as water surface pixels and extracted.

[0025] Further, in the step S6, the specific steps for counting the effective water area in the maximum historical river channel boundary vector data in the research area include:

[0026] Calculate the proportion of all available pixels within the historical maximum river channel boundary of the research area, remove images with a remaining pixel proportion less than 90% after cloud removal processing, and obtain high-quality image data with a remaining pixel proportion equal to or greater than 90% after cloud removal processing;

[0027] Obtain all pixels with a normalized difference water index of 1 in the high-quality image data, and calculate the actual water area, that is, the effective water area, based on the spatial resolution of the image. The expression is:

[0028]

[0029] In the formula: A represents the water surface area of the river channel section, with the unit of km 2 or m 2 , CF represents the area conversion factor per pixel, with the unit of km 2 or m 2 , Vi represents the value of the pixel, 1 represents that the pixel is water area, 0 represents others, and i represents the pixel number in the research area.

[0030] Further, in the step S8, the ratio of the training set to the validation set in the paired sample set is 7:3.

[0031] Compared with the prior art, the beneficial effects of the present invention are:

[0032] 1. The present invention takes the location of the river channel hydrological monitoring station under study as the center, and uses the multiple of the width of the river flow monitoring section where the river channel hydrological monitoring station is located or an approximate integer distance of the river channel section width as the statistical radius for multiple expansions to determine multiple statistical ranges of the river channel water surface area, which are used as multiple research areas. A model is constructed, and based on the water areas within the multiple research areas, a response relationship between the water area and the river channel flow is established to achieve remote sensing interpretation of the river channel flow. Compared with the prior art, this method only needs to monitor the local water area of the river channel, reduces the monitoring cost while being able to break through the dependence of traditional remote sensing technology on the width of large rivers, effectively solves the problem of inverse calculation of the flow of medium and small river channels, and significantly expands the applicability of remote sensing means in narrow river sections and complex terrain areas.

[0033] 2. The present invention takes the location of the river channel hydrological monitoring station under study as the center, and uses the multiple of the width of the river flow monitoring section where the river channel hydrological monitoring station is located or an approximate integer distance of the width of the river channel section as the statistical radius for multiple expansions to determine multiple statistical ranges of the river channel water surface area, which are used as multiple research areas. Its statistical radius can be dynamically adjusted, that is, the research areas can be dynamically adjusted, having flexibility.

[0034] 3. The present invention takes the location of the river channel hydrological monitoring station under study as the center, and uses the multiple of the width of the river flow monitoring section where the river channel hydrological monitoring station is located or an approximate integer distance of the width of the river channel section as the statistical radius for multiple expansions to determine multiple statistical ranges of the river channel water surface area, which are used as multiple research areas. The water area data of its multiple research areas are all processed by combining the NDWI index with cloud removal, which can effectively improve the random error problem that may be caused by a single index, and at the same time can reduce interference such as cloud and terrain occlusion, improve the reliability of water area extraction and data quality, that is, improve the accuracy of model construction.

[0035] 4. The present invention realizes the remote sensing interpretation of river channel flow based on model inversion, can support most current remote sensing data sources, does not need to rely on direct observation of high-precision water level or river width, has scalability while reducing the threshold of data acquisition, and is also applicable to remote areas lacking hydrological stations. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is the method flow chart of the present invention;

[0037] Figure 2 is the diagram of the selection of the research area of the present invention;

[0038] Figure 3 is the importance ranking diagram of the influencing factors on the interpretation accuracy of river channel flow of the present invention;

[0039] Figure 4 is the interpretation accuracy diagram of the model of the present invention in the training set and the validation set;

[0040] Figure 5 is the diagram of the river channel flow interpretation result of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0041] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

[0042] The present invention provides the following technical solution: A remote sensing inversion method for river channel flow based on local water area, comprising the following steps:

[0043] S1: Obtain the location of the flow monitoring section of the target river channel, the measured flow rate, and the remote sensing image of the area where it is located;

[0044] S2: Centering on the location of the river channel hydrological monitoring station under study, expand multiple times with a statistical radius that is a multiple of the width of the river channel flow monitoring section where the river channel hydrological monitoring station is located or an approximate integer distance of the width of the river channel reach, determine multiple statistical ranges of the river channel water surface area, and use them as multiple study areas;

[0045] S3: Obtain the remote sensing images within multiple study areas respectively, convert them into vector files, and perform cloud removal processing;

[0046] S4: Respectively based on the vector file data of multiple study areas, extract water surface pixels through the normalized difference water index to generate a binary water surface distribution map;

[0047] S5: Respectively perform maximum value synthesis on all the normalized difference water indices of multiple study areas, and in combination with the actual topographic map, crop the single-band image of the maximum normalized difference water index to obtain the historical maximum river channel boundary vector data within multiple study areas;

[0048] S6: Statistically calculate the effective water area in the maximum historical river channel boundary vector data of multiple study areas, and construct a water surface area data set;

[0049] S7: Perform time matching on the constructed water surface area data set and the measured flow rate and calculate the day of the year to form a paired sample set;

[0050] S8: Using the river channel water area and the day of the year as independent variables and the paired measured flow rate as the dependent variable, construct a river channel section flow inversion model, and train and verify the model based on the paired sample set;

[0051] S9: Output the river channel flow inversion result based on the river channel section flow inversion model.

[0052] Specifically, in step S1, the location of the target river channel flow monitoring section includes the longitude and latitude coordinate data of the target river channel flow monitoring section, the measured flow rate includes the historical flow rate observation data of the target river channel flow monitoring section, and the remote sensing image includes the green band and the near-infrared band with a spatial resolution not higher than 1 / 10 of the width of the target river channel reach and a spatial range not less than 10 times the reach radius of the area where the target river channel flow monitoring section is located.

[0053] Specifically, in step S2, the multiple of the width of the river channel flow monitoring section or the approximate integer distance of the width of the river channel reach is obtained from the longitude and latitude coordinate data of the target river channel flow monitoring section.

[0054] Specifically, in step S4, the specific steps for extracting water surface pixels by the Normalized Difference Water Index include:

[0055] Calculate the Normalized Difference Water Index, and the expression is:

[0056]

[0057] In the formula: NDVWI is the Normalized Difference Water Index; B3 is the green band; B8 is the near-infrared band;

[0058] Assign pixels with a Normalized Difference Water Index greater than 0 as 1, determine them as water surface pixels and extract them.

[0059] Specifically, in step S6, the specific steps for statistically analyzing the effective water area in the maximum historical river channel boundary vector data within the study area include:

[0060] Calculate the proportion of all available pixels within the historical maximum river channel boundary of the study area, remove images with a remaining pixel proportion less than 90% after cloud removal processing, and obtain high-quality image data with a remaining pixel proportion equal to or greater than 90% after cloud removal processing;

[0061] Obtain all pixels with a Normalized Difference Water Index of 1 in the high-quality image data, and calculate the actual water area, that is, the effective water area, based on the spatial resolution of the image. The expression is:

[0062]

[0063] In the formula: A represents the water surface area of the river channel section, with the unit of km 2 or m 2 , CF represents the area conversion factor per pixel, with the unit of km 2 or m 2 , Vi represents the value of the pixel, 1 represents that the pixel is a water area, 0 represents other, and i represents the pixel number within the study area.

[0064] Specifically, in step S8, the ratio of the training set to the validation set in the paired sample set is 7:3.

[0065] Specific applications

[0066] Select a suitable and atmospherically corrected Harmonized Sentinel-2 MSI: MultiSpectral Instrument, Level-2A image dataset according to the geographical coordinates of the Lingxia Hydrological Monitoring Station;

[0067] Sentinel-2 consists of two satellites, A and B. Its multispectral imager (MSI) can capture multispectral images with 13 bands, covering visible light, near infrared and short-wave infrared. The image width reaches 290km. Its revisit period for the same area is about 5 days, and the spatial resolution of some bands can reach 10m. Its multispectral, high spatial resolution and short revisit period characteristics make it well suited for river water area monitoring.

[0068] This application selects available images with cloud cover less than 60% between 2018 and 2022;

[0069] Taking the location of the Lingxia hydrological monitoring station as the center, the buffer zone is obtained through the ee.Geometry.Point.buffer() function of the Google Earth Engine platform. Here, the radius is extended with an integer distance that is approximately the width of the river section as an example, that is, with a spacing of 1 km, a total of four buffer zones with diameters ranging from 1, 3, 5 and 7 km are obtained, as shown in the attached figure. Figure 2 As shown, these buffer zones are the study areas;

[0070] In order to remove the interference of cloud cover in the study area, the “QA60” mask band in the Sentinel-2 image is used to remove cloud pixels. Pixels with values of 10 and 11 in the “QA60” band are considered to be covered by clouds and cirrus clouds. The green band (B3) and near-infrared band (B8) of the Sentinel-2 image are declouded.

[0071] Considering the strong reflectivity of water in the green band and the strong absorption in the near-infrared band, the normalized difference water index (NDWI) is calculated using the B3 and B8 bands of the cloud-removed image. The range of the NDWI index is between [-1, 1]. If the NDWI value is greater than 0, it can be considered that the current pixel is covered by water.

[0072] Calculate the NDWI of all available images, assign pixels with NDWI values less than 0 to 0, and pixels with NDWI values greater than or equal to 0 to 1. Finally, a binary NDWI map is calculated for each image, i.e., a binary water surface distribution map.

[0073] Affected by the change of water volume, the water level of the river will change accordingly, which is reflected in the remote sensing image as the change of river width and area;

[0074] Taking the buffer zone with a diameter of 7km as the center of the Lingxia hydrological monitoring station as an example, the NDWI calculated from all available images from 2018 to 2021 is synthesized at the maximum value, and the current maximum NDWI single-band image is cropped in combination with the actual topographic map to obtain the historical maximum river boundary vector data within a range of 7km in diameter with the Lingxia hydrological monitoring station as the center;

[0075] The same applies to other size buffers;

[0076] A total of four maximum river boundary vector data are obtained;

[0077] Calculate the proportion of all available pixels within the maximum river boundary, and remove the images with the remaining pixel proportion less than 90% after cloud removal processing to further eliminate the influence of clouds;

[0078] Statistical analysis is performed on the remaining high-quality images. Taking the 7-km maximum river boundary vector data as an example, calculate all the pixels with an NDWI value of 1 within the boundary, calculate the actual water area according to the spatial resolution of Sentinel-2, and extract this value and the corresponding date. The final data obtained is shown in Table 1:

[0079] Table 1: Water area data samples extracted based on Sentinel-2 remote sensing satellites (unit: m 2 )

[0080]

[0081]

[0082] Match the dates of the water area data at four scales with the measured dates of the flow data at the hydrological monitoring station under the ridge. At the same time, calculate the day of the year of the matched dates to obtain the final matched data as shown in Table 2:

[0083] Table 2: Matched data based on remote sensing water area data and measured flow data

[0084]

[0085]

[0086] Use the randomForest package in R language to construct a river cross-section flow inversion model. Select model parameters based on the out-of-bag error. When the tree parameter of the model is 500, the out-of-bag error reaches the minimum, and the error does not increase with the increase of the tree. Select tree = 500 as the best parameter for the random forest model;

[0087] The independent variable data are the water area data and the day-of-year data at 1 km, 3 km, 5 km, and 7 km. The sorting results of the importance of each independent variable for flow simulation are as attached Figure 3 shown;

[0088] The dependent variable is the measured flow data at the hydrological monitoring station under the ridge;

[0089] The matching data is divided into a training set and a validation set in a ratio of 7:3. The training set is used to fully train the model, and the validation set is used to test the generalization performance of the model;

[0090] The results of model training are as shown in Appendix Figure 4 (a). R 2 = 0.93, RMSE = 84.37m 3 s -1 , MAE = 36.36m 3 s -1 , achieving a good training effect;

[0091] The results of model validation are as shown in Figure 4 (b). R 2 = 0.79, RMSE = 78.42m 3 s -1 , MAE = 53.56m 3 s -1 , compared with the training set, the accuracy of the two indicators of R2 and MAE of the model has decreased, and the RMSE has increased slightly. Generally speaking, the model has very good generalization performance;

[0092] The results show that the inversion model established based on the local water surface area data obtained from remote sensing images has high fitting accuracy and good generalization performance, providing technical support for the study of high-precision river channel flow inversion;

[0093] Based on the constructed river channel cross-section flow inversion model, the water area and the day of the year of different research areas extracted from the remote sensing image of the target river channel cross-section are input into the model, and the flow inversion results of the current river channel cross-section are output. The output results are shown in Appendix Figure 5 , and the results show that the flow of the river channel cross-section can be accurately output.

[0094] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A remote sensing inversion method for river channel flow based on local water area, characterized in that It includes the following steps: S1: Obtain the flow monitoring section location, measured flow, and remote sensing images of the target river channel in the area; S2: Centered on the location of the river channel hydrological monitoring station under study, expand multiple times with a statistical radius of several times the width of the flow monitoring section of the river channel where the river channel hydrological monitoring station is located or an approximate integer distance of the width of the river channel section, determine multiple statistical ranges of the river channel water surface area, and use them as multiple study areas; S3: Obtain the remote sensing images within multiple study areas respectively, convert them into vector files, and perform cloud removal processing; S4: Respectively based on the vector file data of multiple study areas, extract water surface pixels through the normalized difference water index to generate a binary water surface distribution map; S5: Respectively perform maximum value synthesis on all the normalized difference water indices of multiple study areas, and combine with the actual topographic map to crop the single-band image of the maximum normalized difference water index to obtain the historical maximum river channel boundary vector data within multiple study areas; S6: Statistically calculate the effective water area in the maximum historical river channel boundary vector data within multiple study areas to construct a water surface area data set; S7: Perform time matching on the constructed water surface area data set and the measured flow and calculate the day of the year to form a paired sample set; S8: With the river channel water area and the day of the year as independent variables and the paired measured flow as the dependent variable, construct a river channel section flow inversion model, and train and validate the model based on the paired sample set; S9: Output the river channel flow inversion result based on the river channel section flow inversion model.

2. The remote sensing inversion method of river channel flow based on local water area according to claim 1, characterized in that: In step S1, the target river channel flow monitoring section location includes the longitude and latitude coordinate data of the target river channel flow monitoring section, the measured flow includes the historical flow observation data of the target river channel flow monitoring section, and the remote sensing images include the green band and the near-infrared band with a spatial resolution not higher than 1 / 10 of the width of the target river channel section and a spatial range not less than 10 times the river section radius of the area where the target river channel flow monitoring section is located.

3. A remote sensing inversion method for river channel flow based on local water area according to claim 2, characterized in that: In step S2, the multiple of the width of the river channel flow monitoring section or the approximate integer distance of the width of the river channel section is obtained from the longitude and latitude coordinate data of the target river channel flow monitoring section.

4. A remote sensing inversion method for river channel flow based on local water area according to claim 3, characterized in that: In step S4, the specific steps for extracting water surface pixels through the normalized difference water index include: Calculate the normalized difference water index, and the expression is: In the formula: NDVWI is the normalized difference water index; B3 is the green band; B8 is the near-infrared band; Assign a value of 1 to the pixels with a normalized difference water index greater than 0, determine them as water surface pixels and extract them.

5. A remote sensing inversion method for river channel flow based on local water area according to claim 4, characterized in that: In step S6, the specific steps for statistically calculating the effective water area in the maximum historical river channel boundary vector data of the study area include: Calculate the proportion of all available pixels within the historical maximum river channel boundary of the study area, remove the images with the remaining pixel proportion less than 90% after cloud removal processing, and obtain high-quality image data with the remaining pixel proportion equal to or greater than 90% after cloud removal processing; Obtain the pixels with a normalized difference water index of 1 in the high-quality image data, and calculate the actual water area, that is, the effective water area, based on the spatial resolution of the image. The expression is: Where: A represents the water surface area of the river channel section, with the unit of km 2 or m 2 , CF represents the area conversion factor per pixel, with the unit of km 2 or m 2 , Vi represents the value of the pixel, 1 indicates that the pixel is water area, 0 indicates other, and i represents the pixel number in the study area.

6. The remote sensing inversion method of river channel flow based on local water area according to claim 5, characterized in that: In the step S8, the ratio of the training set and the validation set into which the paired sample set is divided is 7:3.

Images