A method for constructing a spatiotemporal dynamic monitoring model for river water quality based on multi-source data

Abstract

The embodiment of the present invention discloses a method for constructing a spatiotemporal dynamic monitoring model of river water quality based on multi-source data, including: obtaining satellite image data, climate data and land use data of a river basin, as well as water quality parameters of multiple stations in the basin; determining the optimal spectral characteristics of each station according to the spectral bands and spectral band combinations in the satellite image data; constructing a buffer zone centered on a continuous area of ​​each land use type according to the land use data; and constructing land use type characteristics of each station according to the coverage of each station by the buffer zone corresponding to each land use type; fitting a water quality parameter monitoring model applicable to the entire basin with the optimal spectral characteristics, land use type characteristics and meteorological characteristics of each station as independent variables and the water quality parameters of each station as dependent variables. This embodiment realizes large-scale and accurate monitoring of river water quality parameters based on multi-source data.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of data processing technology, and in particular to a method for constructing a spatiotemporal dynamic monitoring model for river water quality based on multi-source data. Background Art

[0002] Rivers are an important carrier of terrestrial freshwater resources and one of the most important ecosystems in nature. With the improvement of industrialization and urbanization, rivers are affected by factors such as urban and rural residential areas and industrial and mining land. Problems such as organic pollution and eutrophication in water bodies are deteriorating, and the water quality is not optimistic. Strengthening river water quality monitoring is the basis for achieving water prevention and control.

[0003] Traditional river water quality monitoring methods rely on on-site sampling and laboratory analysis. Due to the limitations of manpower, material resources, terrain, climate and hydrological characteristics, it is difficult or even impossible to collect a large number of samples. Moreover, the samples only represent local areas, and it is impossible to achieve full coverage and real-time monitoring of water bodies, and it is impossible to timely reflect the changes in water quality in large areas of water bodies. Remote sensing technology has the advantages of wide coverage, fast information acquisition, and repeatable observation, and can achieve continuous and dynamic monitoring of water bodies.

[0004] In the prior art, although there have been many studies on the use of remote sensing technology to monitor the water quality parameters of rivers and lakes, the data source is mainly limited to the field of drones, and the main technology is to construct linear or nonlinear inversion models of different water quality parameters by screening sensitive bands. Since rivers are narrow, long and tortuous, they are easily affected by various factors. It is difficult to achieve large-scale monitoring of river water quality parameters using only remote sensing images. Moreover, drone images have a small coverage area and high cost, and can only achieve spatial visualization of some river sections, which is difficult to meet the needs of river water quality monitoring over a larger range. Patent CN117422987A provides a river water quality monitoring method based on satellite remote sensing images and machine learning, and patent CN111161121B provides a method and system for determining the sudden change in the response of river water quality to the composition of land use in the basin. Both have limitations in the large-scale and accurate monitoring of river water quality parameters. Summary of the invention

[0005] The embodiment of the present invention provides a method for constructing a spatiotemporal dynamic monitoring model of river water quality based on multi-source data, which utilizes multi-source data such as satellite image data, climate data, and land use data to achieve large-scale and accurate monitoring of river water quality parameters.

[0006] In a first aspect, an embodiment of the present invention provides a method for constructing a spatiotemporal dynamic monitoring model of river water quality based on multi-source data, comprising:

[0007] Obtain satellite imagery data, climate data, and land use data for the river basin, as well as water quality parameters at multiple sites within the basin;

[0008] Determining optimal spectral characteristics of each station based on spectral bands and spectral band combinations in the satellite image data;

[0009] According to the land use data, a buffer zone centered on a continuous area of ​​each land use type is constructed; and according to the coverage of each station by the buffer zone corresponding to each land use type, a land use type feature of each station is constructed;

[0010] Taking the optimal spectral characteristics, land use type characteristics and meteorological characteristics of each station as independent variables and the water quality parameters of each station as dependent variables, a water quality parameter monitoring model suitable for the entire basin is fitted.

[0011] In a second aspect, an embodiment of the present invention provides a method for spatiotemporal dynamic monitoring of river water quality based on multi-source data, comprising:

[0012] Obtain satellite imagery data, climate data, and land use data for river basins at any time period;

[0013] Based on the acquired data, the optimal band characteristics, land use type characteristics and meteorological characteristics at each location in the basin are determined;

[0014] Input all the features at each position into the water quality parameter monitoring model constructed by the method described in any of the above embodiments to obtain the two-dimensional spatial distribution of water quality parameters in the river basin at any time period;

[0015] The two-dimensional spatial distribution of the water quality parameters is used as a visualization layer, which is superimposed with the visualization layers of other covariates in the model to achieve spatial visualization of the river basin.

[0016] In a third aspect, an embodiment of the present invention further provides an electronic device, the electronic device comprising:

[0017] one or more processors;

[0018] a memory for storing one or more programs,

[0019] When the one or more programs are executed by the one or more processors, the one or more processors implement the method for constructing a river water quality spatiotemporal dynamic monitoring model based on multi-source data, or the method for spatiotemporal dynamic monitoring of river water quality based on multi-source data as described in any embodiment.

[0020] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for constructing a river water quality spatiotemporal dynamic monitoring model based on multi-source data, or the method for spatiotemporal dynamic monitoring of river water quality based on multi-source data as described in any embodiment.

[0021] In summary, the embodiment of the present invention provides a method for constructing a spatiotemporal dynamic monitoring model for river water quality based on multi-source data. First, satellite image data, climate data, and land use data of the watershed to be detected are collected. Then, the measured water quality parameter data is used as the dependent variable, and the satellite image data, climate data, and land use data are used as independent variables. Modeling is performed based on a machine learning regression model to obtain a spatiotemporal dynamic monitoring model for river water quality. The embodiment of the present invention can achieve the following benefits:

[0022] 1. Currently, there is no related technology for inverting river water quality parameters using multi-source data. This embodiment couples river water quality parameters, satellite image data, climate data, and land use data to build an ETR machine learning model, which achieves the prediction of multiple water quality parameters under the same scenario, greatly improving the accuracy of river water quality parameter inversion;

[0023] 2. In the regression analysis, this embodiment constructs the optimal band characteristics through band transformation for the inversion of specific water quality parameters (ammonia nitrogen and total phosphorus); and uses the ETR model in the construction of the model, which has great advantages in water quality monitoring of rivers and lakes;

[0024] 3. This embodiment constructs a buffer zone with land use type as the center, forming a two-dimensional spatial distribution related to the land use type, and constructs land use type characteristics based on the two-dimensional spatial distribution, so that the model can accurately reflect the impact of land use type on water quality parameters, further improving the accuracy of water quality prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0026] Figure 1 It is a flow chart of a method for constructing a spatiotemporal dynamic monitoring model of river water quality based on multi-source data provided by an embodiment of the present invention;

[0027] Figure 2 It is a spatial inversion distribution map of ammonia nitrogen in the Weihe River Basin in April 2020 and a measured concentration distribution map provided by an embodiment of the present invention;

[0028] Figure 3 It is a spatial inversion distribution map of total phosphorus in the Weihe River Basin in April 2020 and a measured concentration distribution map provided by an embodiment of the present invention;

[0029] Figure 4is a flow chart of another method for constructing a spatiotemporal dynamic monitoring model of river water quality based on multi-source data provided by an embodiment of the present invention;

[0030] Figure 5 is a flow chart of a method for spatiotemporal dynamic monitoring of river water quality based on multi-source data provided by an embodiment of the present invention;

[0031] Figure 6 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0032] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described clearly and completely below. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0033] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0034] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0035] Figure 1 is a flow chart of a method for constructing a spatiotemporal dynamic monitoring model of river water quality based on multi-source data provided by an embodiment of the present invention. The method uses measured water quality parameter data as dependent variables, satellite image data, climate data and land use data as independent variables, and models based on a machine learning regression model to achieve global dynamic monitoring of water quality parameters in a large river basin. The method is executed by an electronic device, such as Figure 1 As shown, the specific steps include:

[0036] S110. Obtain satellite image data, climate data, and land use data of the river basin, as well as water quality parameters at multiple stations within the basin.

[0037] The river basin may be a specific river basin to be monitored for water quality, or a certain section within the river basin. The satellite image data may be the Sentinel-2 image data of the basin, and the satellite image data includes the two-dimensional spatial distribution of characteristic data of multiple bands in the river basin. Climate data is used to reflect climate factors related to river water quality, and climate data may include the two-dimensional spatial distribution of at least one of precipitation, temperature, evapotranspiration and wind speed in the river basin. Land use data may include the two-dimensional spatial distribution of land use types that affect changes in river water quality in the river basin. Specifically, the land use types that affect changes in river water quality include at least one of construction land, urban land, rural settlements, cultivated land, forest land and grassland. The site is a water quality parameter monitoring point, and the monitored water quality parameters or water quality indicators include ammonia nitrogen (NH3-N) content, total phosphorus (TP) content, etc.

[0038] Sentinel-2 image data has the advantages of short revisit period, high spatial resolution and free data acquisition. It has great advantages in remote sensing water quality monitoring and has been successfully used in lake remote sensing water quality monitoring, but is rarely used in river water quality monitoring. Considering that river water quality is also affected by a variety of external factors such as local climate conditions and surrounding land use, this step obtains satellite image data, climate data, land use data and water quality parameters of the same period for the river basin to be tested, which are used to construct the sample set of the model. Optionally, after obtaining the original data, first remove the outlier data, and then divide it into training set, test set and validation set for training, testing and validation of the regression model.

[0039] In a specific implementation, the Xi'an section of the Weihe River Basin was used as the section to be tested. This step collected and sorted out four aspects of data: (1) Non-optical activity parameters in the Xi'an section of the Weihe River Basin in spring 2020 were obtained based on field sampling, including two water quality parameters, ammonia nitrogen and total phosphorus. There were 41 actual measurement sites in the section to be tested. (2) Sentinel-2A and Sentinel-2 satellite image data were downloaded from the European Space Agency Copernicus Data Center (https: / / scihub.copernicus.eu / ), and 14 scenes of available data were included in the study. (3) Meteorological data with a resolution of 1km×1km were obtained from the China Meteorological Data Network (https: / / www.geodata.cn / ), including monthly average temperature (℃), precipitation (mm), evapotranspiration (mm) and wind speed (m / s). (4) The 2020 land use and land cover change data (LUCC) were obtained through the Resource and Environmental Science Data Platform (http: / / www.resdc.cn). The six land use types used were urban land, rural residential land, construction land, cultivated land, forest land and grassland. The data are detailed in Table 1.

[0040] Table 1

[0041]

[0042] S120: Determine the optimal band characteristics of each site according to the bands and band combinations in the satellite image data.

[0043] Due to the complexity of the water environment and the optical characteristics of the water quality parameters to be detected, this embodiment improves the accuracy of water quality parameter prediction through a specific spectral processing method and band combination. Optionally, the process may include the following steps:

[0044] Step 1: Correlation analysis is performed on each spectral band in the satellite image data and the water quality parameters, and multiple sensitive bands of the water quality parameters are determined based on the correlation and the spectral characteristics of the water body and the water quality parameters. Optionally, due to the complexity of the water environment and the optical properties of nitrogen, it is usually necessary to improve the accuracy of the total nitrogen content estimation through specific spectral processing methods and band combinations. Due to the spectral characteristics of the water body, strong absorption occurs in the short-wave infrared and thermal infrared bands, so the bands selected are the visible light and near-infrared parts, especially the range of 400~900nm. Nitrogen is an important indicator of eutrophication of water bodies. Previous studies have found that the characteristic bands of related indicators of nitrogen in water bodies include 404-477nm, 494-496nm, 589-591nm, 590-650nm, and 690-880nm. Phosphorus is also related to eutrophication of water bodies. Studies have found that there is a high correlation between the 524nm band, 574nm band, 608nm band, 638nm band, and 701nm band and the phosphorus concentration, and the near-infrared band has a better spectral response to phosphorus. Sentinel-2 satellite images have 9 bands in the visible and near-infrared parts, of which band 1 is the coastal / aerosol spectral range (central wavelength 443nm), band 2 is the blue light band (central wavelength 490nm), band 3 is the green light band (central wavelength 560nm), band 4 is the red light band (central wavelength 665nm), bands 5, 6, and 7 are vegetation red edge bands (central wavelengths are 705nm, 740nm, and 785nm respectively), band 8 is the near-infrared band (central wavelength 842nm), and band 8a is the narrow-edge near-infrared band (central wavelength 865nm). According to the empirical sensitive bands of the two water quality parameters analyzed above, combined with the correlation analysis results of the Sentienl-2 bands of water quality parameters (Table 2), bands 1, 2, 3, 4, 8 and 8a were first selected as sensitive bands for the water quality parameter ammonia nitrogen, and bands 1, 2, 3, 4, 5, 6, 7, 8 and 8a were selected as sensitive bands for total phosphorus to participate in modeling.

[0045] Table 2 Correlation analysis results of single band

[0046]

[0047] Step 2: Add, subtract and ratio operations are performed on each sensitive band to obtain multiple spectral band combinations. This step performs band transformation operations on each band. The band changes include addition operations (enhancing specific spectral characteristics), subtraction operations (eliminating background noise) and ratio operations (standardizing spectral information), which help to highlight the image water body information. The results of the band transformation generate new feature spaces, which provide richer information for subsequent model construction. Optionally, the band transformation (addition, subtraction, ratio) can be performed in combination with the correlation between the bands. According to the inter-band correlation matrix (a matrix with each band as rows and columns and the correlation coefficients of row bands and column bands as elements), two bands with high correlation are selected for addition or subtraction, and two bands with low correlation are selected for ratio calculation, which can help highlight certain water quality characteristics or reduce noise. In a specific embodiment, the correlation threshold can be set to 0.6 (>0.6 indicates that the correlation between the two bands is high, and <0.6 indicates that the correlation between the two bands is low). Correspondingly, the basis for selecting band addition is: select bands with high band correlation and sensitive to water quality parameters for band addition, there is a high positive correlation between the bands, and specific characteristics can be enhanced. Then B1+B2, B4+B8, B4+B8a, and B8+B8a are selected as the band addition combinations of ammonia nitrogen, and B1+B2, B5+B6, B5+B7, B5+B8, B5+B8a, B6+B7, B6+B8, B6+B8a, B7+B8, B7+B8a, and B8+B8a are selected as the band addition combinations of total phosphorus. The basis for selecting band subtraction is: selecting a combination in which one band is sensitive to water quality parameters and the other band is insensitive can remove background effects or enhance contrast. B1-B2, B4-B8, and B4-B8a are selected as the band subtraction combinations for ammonia nitrogen, and B1-B2, B2-B5, B3-B5, B4-B5, B4-B6, B4-B7, B4-B8, and B4-B8a are selected as the band subtraction combinations for total phosphorus. The basis for selecting the band ratio is: select bands with low correlation and different sensitivity to water quality parameters. The correlation between bands is low to ensure that the bands provide unique spectral information. B1 / B3, B1 / B4, B2 / B8, B2 / B8a, B3 / B8, and B3 / B8a are selected as the band ratio combinations for ammonia nitrogen, and B1 / B3, B1 / B4, B2 / B6, B2 / B7, B2 / B8, B2 / B8a, B3 / B6, B3 / B7, B3 / B8, and B3 / B8a are selected as the band ratio combinations for total phosphorus.

[0048] Step 3. After obtaining each spectral band combination, perform correlation analysis on each spectral band combination and water quality parameters to determine the degree of correlation between each spectral band combination and water quality parameters; and select spectral band combinations whose correlation with water quality parameters meets the set conditions and passes the significance test, and use the data of the spectral band combination at each station as the optimal spectral feature of each station. Optionally, after generating the band transformation results, evaluate the degree of correlation between these new features and water quality parameters through correlation analysis, and use the Pearson correlation coefficient to measure the correlation between the features and the target variable. Through the significance test (p=0.05), the characteristic band combinations with high correlation with the target variable and passing the significance test are selected, and the data of these combinations at the same station location are all the optimal spectral features of the station.

[0049] S130. Constructing a buffer zone centered on a continuous area of ​​each land use type based on the land use data; and constructing land use type characteristics of each site based on the coverage of each site by the buffer zone corresponding to each land use type.

[0050] The buffer zone in this embodiment is obtained by expanding outward from various land use type blocks in the river basin as the center, representing the scope of influence of various land use types on the surrounding environment. Compared with the buffer zone centered on the site, the buffer zone in this embodiment can form the spatial distribution of the land use type buffer zone in the entire river basin in subsequent operations, and realize the spatial visualization of the buffer zone; while the buffer zone centered on the site can only reflect the proportion of land use types at the site location, and cannot form the spatial distribution of various land use types, and cannot realize spatial visualization.

[0051] Based on such a buffer zone, this step can quantitatively represent the land use type characteristics of each site. In a specific embodiment, the land use data is land use vector data, and the construction of the buffer zone and land use type characteristics can include the following steps:

[0052] Step 1: Taking the land use type of cultivated land as an example, a buffer zone is established based on the vector data of cultivated land according to the river width of the main stream, first-level tributaries and second-level tributaries of the river basin and the location of the station. Optionally, the maximum range of the buffer zone (for example, 500 meters) can be roughly determined based on the distance from the river and the station to the cultivated land; within this range, a certain distance is extended around a continuous cultivated land block as the center, and the extended area is the buffer zone of the cultivated land type. All buffer zones in the river basin together constitute the two-dimensional plane distribution of the buffer zone.

[0053] Step 2: Intersect the river vector data with the vector data of each cultivated land buffer zone. River vector data is a GIS data format used to describe the shape and characteristics of rivers. It mainly includes the geometric information and attribute information of rivers, such as the length and area of ​​rivers. The longitude and latitude of each position of the river can be extracted from the river vector data. Similarly, the longitude and latitude of each position of all buffer zones can also be extracted from the vector data of the buffer zone.

[0054] Step 3: Convert the intersection of each land use type into a river raster layer of each land use type, where the intersection value in each river raster layer is 1 and the non-intersection value is 0. Optionally, keep the intersection and update it to the river vector layer (the intersection is set to 1 and the non-intersection is set to 0), and convert the updated river vector layer to a raster layer to obtain a river raster layer with only 0 and 1 values. The updated raster layer will be superimposed with other covariate layers for subsequent spatial visualization analysis.

[0055] Step 4: Extract the 0 / 1 value at each station location from the river raster layer of the cultivated land to obtain the cultivated land use situation at each station location.

[0056] After executing the above steps 1 to 4 for each land use type, the 0 / 1 values ​​at the same site location are extracted from the river raster maps corresponding to all land use types, thus obtaining the characteristics of each land use type for the site.

[0057] Furthermore, whether the spatial scale of the buffer zone is reasonable also affects the accuracy of the water quality parameter model to a certain extent. In order to determine a reasonable buffer zone, in a specific implementation, multiple buffer distances can be determined first, which are used to characterize the influence range of the land use type, that is, the spatial scale of the buffer zone. Optionally, five buffer distances of 100m, 200m, 300m, 400m and 500m are determined.

[0058] Then, taking the continuous area of ​​any land use type as the center, buffer zones of different spatial scales are delineated according to each buffer distance, namely, 100m buffer zone, 200m buffer zone, 300m buffer zone, 400m buffer zone and 500m buffer zone. The same operation is performed on each land use type to obtain the distribution of buffer zones of different spatial scales under each land use type.

[0059] Finally, the optimal scale of the buffer is determined according to the proportion of sites covered by buffers of different spatial scales under each land use type. The basic principle for determining the scale of the buffer here is: in the subsequent regression analysis, the goal of the model is to find the relationship between the independent variable and the dependent variable. When the values ​​of a certain independent variable are the same, it does not help to explain the changes in the dependent variable, so the contribution rate to the model is very small. In the modeling process, this situation is usually regarded as a redundant independent variable and may be removed to simplify the model. Since the land use data extraction result is binary data of 0 and 1, the ratio of 0 and 1 needs to be close to balance, that is, the number of samples of the two is similar, and the contribution to the model will be higher. At this time, the model can better learn the relationship between the two values. According to this principle, when determining the optimal buffer scale, this embodiment can first calculate the proportion of the number of sites covered by buffers of different spatial scales under each land use type in the total number of sites, that is, the proportion of positive samples. Taking the specific implementation method of the Xi'an section of the Weihe River Basin as an example, the statistical situation of the proportion of the intersection of the six land use types and the river based on the sampling points (i.e., stations) is shown in Table 3:

[0060] Table 3

[0061]

[0062] Then, the change range of the proportion corresponding to the buffer zones of different spatial scales under the same land use type is calculated respectively, that is, whether the above proportion has a significant change with the change of buffer distance under various land use types. If the proportion does not change much with the increase of buffer distance under a certain land use type, indicating that the buffer distance under this land use type has little effect on sample balance, the same land use type can be marked and excluded from the range of land use types that need to be considered in determining the optimal buffer distance. Exemplarily, referring to Table 3, within the 100-500 meter buffer zone, the proportion of urban land changes little (the change range is less than 0.05). This is because there are fewer sampling points of rivers around towns, so the 100-meter buffer zone covers almost all the sampling points of rivers near towns, so the land use type of urban land can be ignored in subsequent operations.

[0063] The positive sample ratios of the remaining land use types all show an increase that cannot be ignored as the buffer zone increases, so the optimal buffer distance should continue to be selected according to their sample conditions. Optionally, according to the balance of the independent variable samples, it is determined whether the ratio corresponding to the buffer zone of the unmarked land use type at each spatial scale exceeds the set range; if so, all buffer zones at the corresponding spatial scale are deleted, and the remaining buffer zones are used as the buffer zones of the optimal scale. This step checks whether the sample ratios of buffer zones of different spatial scales under the remaining land use types are balanced, or whether unacceptable imbalances have occurred. Ideally, the sample ratios of 0 and 1 should be close to 0.5 when the samples are balanced, but it can be seen from Table 3 that due to the large area of ​​cultivated land, the proportion of cultivated land in the 100-meter buffer zone is already greater than 50%, that is, in practical applications, it is too ideal to control the positive and negative sample ratios within a smaller interval centered on 0.5. Therefore, this embodiment selects 0.8 as the judgment threshold for sample balance, and it is expected that the sample ratio of 1 should not be greater than 0.8 or less than 0.2. If the proportion of positive samples greater than 0.8 is used as the standard for sample imbalance, combined with Table 3, since the proportion of cultivated land in the 400-meter buffer zone reaches 92.7%, close to 100%, the contribution to the model will be reduced. Therefore, the 400-meter and 500-meter buffer zones under all land use types are deleted, and the 100-meter, 200-meter and 300-meter buffer zones are selected for subsequent operations. Correspondingly, if the proportion of positive samples less than 0.2 is used as the standard for sample imbalance, combined with Table 3, all buffer zones from 100 meters to 500 meters will be deleted, and subsequent operations cannot be performed. Therefore, this embodiment ultimately uses the proportion of positive samples greater than 0.8 as the unacceptable range of sample balance, and determines the optimal scale buffer zones to be 100, 200 and 300 meters.

[0064] S140. Taking the optimal band characteristics, land use type characteristics and meteorological characteristics of each station as independent variables and the water quality parameters of each station as dependent variables, a water quality parameter monitoring model suitable for the entire basin is fitted.

[0065] Through the operations of S120 and S130, each site has at least one optimal band feature, at least one land use type feature and at least one meteorological feature (based on existing research, at least one item in the meteorological data that has a strong correlation with water quality parameters can be selected as at least one meteorological feature). This step uses the site data as a sample to fit a regression model with these features as independent variables and water quality parameters as dependent variables. The fitted model can be used to predict water quality parameters at any location in the entire river basin. Optionally, this embodiment uses the Extreme Trees Regressor (ETR) algorithm for model fitting, which has great potential in water quality monitoring of rivers and lakes.

[0066] Specifically, the extreme random tree regression model can be constructed using the sklearn package of python. The ETR algorithm constructs an unpruned regression tree set based on a top-down process. It has two main differences from other tree-based ensemble methods. The algorithm splits the nodes by completely randomly selecting the cut point and uses the entire training sample to generate the tree. The ETR algorithm directly uses all samples, randomly selects K features, and for these K features, randomly selects 1 split node for each feature, thereby obtaining K classification nodes. Then the scores of these K split nodes are calculated, and the node with the highest score is selected as the split node. Since the division point of the feature value is randomly selected instead of the optimal point, the size of the generated decision tree is generally larger than that generated by the random forest. Compared with the random forest algorithm, the variance of this algorithm is further reduced, the bias is further increased, and the generalization ability of the model is better.

[0067] Furthermore, after the ETR model is built, the feature importance evaluation method of the tree model can be used to further screen the feature variables. During the training process, these tree models will calculate the contribution of each feature to the model prediction results, that is, the feature importance score. By dividing the feature importance score threshold (≥0.05), the key features that have a greater impact on the model prediction results are screened out. Based on these key features, the original feature set can be streamlined to improve the computational efficiency and prediction performance of the model. After repeated operations, the optimal feature variable combination is screened out, including the screened feature band combination, meteorological factors, and land use types. Use these optimal feature variables to reconstruct the ETR regression model, and perform model training and verification. During the model verification process, methods such as 5-fold cross-validation can be used to evaluate the performance of the model, and the model parameters can be adjusted according to the verification results to achieve the best prediction effect. Finally, an inversion model based on the optimal feature variable combination and the ETR regression model is constructed for the inversion prediction of water quality parameters in practical applications.

[0068] In summary, this embodiment completes feature screening through band transformation, correlation analysis, and feature importance assessment, and combines buffer zone selection to successfully construct an inversion model based on the ETR regression model. This model can effectively use the band information and other relevant feature variables of remote sensing images to predict physical parameters, and has high prediction accuracy and generalization ability. In a specific implementation, using the ETR model, a water quality parameter regression model can be constructed through three steps and 7 scenarios:

[0069] Step 1: Use the sensitive bands and band operation characteristics screened by correlation analysis and significance analysis, couple water quality parameters (ammonia nitrogen and total phosphorus), build the ETR model (scenario 1, abbreviated as RS), and screen key features through feature importance scores (threshold ≥ 0.05).

[0070] Step 2: Based on step 1, add climate factors as independent variables, build the ETR model (scenario 2, RS+M), and screen key features through feature importance scores (threshold ≥ 0.05).

[0071] Step 3. Based on step 2, add land use types within the buffer range of 100m, 200m, and 300m as independent variables, build the ETR model (different buffers correspond to scenarios 3-5, abbreviated as RS+M+B100m, RS+M+B200m, and RS+M+B300m, respectively), and screen key features through feature importance scores.

[0072] Optionally, select the coefficient of determination (R 2 ) and root mean square error (RMSE) as evaluation indicators of model performance. Based on the two evaluation indicators of the two water quality parameter inversion models under various scenarios, the best inversion model can be selected. Still taking the specific implementation method of the Weihe River Basin in Xi'an as an example, the model fitting results are shown in Table 4, where RS represents remote sensing image data, M represents meteorological data, and B100 represents land use data under 100m buffer conditions. The extreme random tree model constructed by image data, meteorological data and land use data under 100m buffer conditions is selected as the best NH3-N model, and the extreme random tree model constructed by image data, meteorological data and land use data under 200m buffer conditions is selected as the best TP model.

[0073] Table 4 Comparison of model results

[0074]

[0075] In addition, this example also combines the verification data under the same scenario in the same quarter of 2020 to simulate the spatiotemporal prediction results of the Xi'an section of the Weihe River Basin, and draws the spatial inversion distribution map of water quality parameters and the measured concentration distribution map of the Weihe River Basin in the same quarter of 2020. Figure 2 ) and TP( Figure 3 ) as an example to show the inversion results. The figure shows the inversion results of the two water quality parameters based on river distribution. The scattered points of different sizes represent the concentration changes and spatial distribution of the measured water quality parameters in the validation set. Comparing the inversion results of the two water quality parameters with the concentration distribution map of the measured water quality parameters shows that the high and low values ​​of the inversion prediction and the measured water quality parameters are consistent, indicating that the model is robust and portable. The inversion results also show that the areas with higher concentrations of the two water quality parameters in the Xi'an section of the Weihe River Basin are the southern urban areas and some narrow river sections, especially Zaohe, Taipinghe, Xinhe and Laohe.

[0076] In summary, the embodiment of the present invention provides a method for constructing a spatiotemporal dynamic monitoring model for river water quality based on multi-source data. First, according to the characteristics of the river and the water quality conditions, multi-phase water quality data is collected, and outlier abnormal data is eliminated to divide the training set, test set and validation set; in addition, satellite image data (Sentinel-2A and Sentinel-2B) are downloaded and data preprocessing is completed to collect climate elements (temperature, rainfall, evapotranspiration and wind speed) and land use (construction land, urban land, rural settlements, cultivated land, forest land and grassland) data. Then, the measured water quality parameter data is used as the dependent variable, and the remote sensing image data, climate data and land use data are used as independent variables. Modeling analysis is performed based on the machine learning regression model to explore the model accuracy under different modeling rules, and the model accuracy is verified by the validation data set to obtain the best fitting model. Finally, the optimal model is applied to the validation data set to realize the spatial visualization of water quality parameters, explore the emission sources of pollutants, the direction of migration and diffusion, the scope of influence, and the characteristics of mixing with clean water, find out the ins and outs of pollutants, and provide technical support for subsequent water pollution prevention and control. The entire technical route is as follows: Figure 4 As shown, the process of the whole method is shown from another perspective.

[0077] The embodiments of the present invention can achieve the following beneficial effects:

[0078] 1. Currently, there is no related technology for inverting river water quality parameters using multi-source data. This embodiment couples river water quality parameters, remote sensing image data, climate data, and land use data to build an ETR machine learning model, thereby achieving the prediction of multiple water quality parameters under the same scenario, greatly improving the accuracy of river water quality parameter inversion;

[0079] 2. In the regression analysis of this embodiment, key features are selected through band transformation, correlation analysis and feature importance evaluation for the inversion of specific water quality parameters (ammonia nitrogen and total phosphorus). Among them, in the application of band transformation, according to the spectral characteristics of water bodies and water quality parameters, targeted band operation strategies are proposed to enhance specific information or reduce noise, thereby improving the accuracy of subsequent model analysis. In the process of model construction and secondary feature screening, the ETR model was selected, which has great advantages in water quality monitoring of rivers and lakes;

[0080] 3. This embodiment constructs a buffer zone centered on land use type, forming a two-dimensional spatial distribution related to the land use type; and based on the two-dimensional spatial distribution, determines the proportion of positive and negative samples, and screens the optimal buffer zone scale, so that the model can accurately reflect the impact of land use type on water quality parameters, further improving the accuracy of water quality prediction.

[0081] 4. The prediction methods for the spatiotemporal changes of river water quality parameters in the prior art can only predict a small area, and mainly use remote sensing monitoring methods such as drones to study the water quality of some river sections, which cannot achieve river water quality monitoring over a larger area. The embodiments of the present invention use free Sentinel-2 satellite images, combined with multi-source data, to accurately predict river water quality parameters over a large area on a spatiotemporal scale (spatial resolution: 10 meters; temporal resolution: 5 days).

[0082] Based on the above model construction method, Figure 5 FIG. 1 is a flow chart of a method for spatiotemporal dynamic monitoring of river water quality based on multi-source data provided by an embodiment of the present invention. Figure 5 As shown, the method specifically includes:

[0083] S210. Obtain satellite image data, climate data, and land use data of the river basin at any time period.

[0084] The time period may be a period when the meteorological environment is relatively stable. For example, multiple scene data within a month may be taken, when interference is relatively small and the meteorological conditions within a month are relatively stable.

[0085] S220. Determine the optimal band characteristics, land use type characteristics and meteorological characteristics at each location in the watershed based on the acquired data.

[0086] After the optimal model is determined in the above embodiment, here only the optimal characteristic variable combination at each position needs to be determined.

[0087] S230. Input all the features at each position into the water quality parameter monitoring model constructed by the method described in any of the above embodiments to obtain the two-dimensional spatial distribution of water quality parameters in the river basin during any period of time.

[0088] Similarly, the optimal characteristic variables at each location can be combined and input into the optimal model to obtain the water quality parameters at each location. These water quality parameters together constitute the two-dimensional spatial distribution of the parameter.

[0089] S240. Using the two-dimensional spatial distribution of the water quality parameters as a visualization layer, and overlaying the visualization layers of other covariates in the model to achieve spatial visualization of the river basin.

[0090] Spatial visualization can clearly show the spatial distribution of different water quality parameters. After using the S210-S240 method to predict the two-dimensional spatial distribution of water quality parameters in multiple time periods, spatial visualization can also be used to show the spatiotemporal changes of different water quality parameters, revealing the emission sources, migration and diffusion directions, and impact range of nitrogen and phosphorus in rivers.

[0091] Figure 6A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention is shown in FIG. Figure 6 As shown, the device includes a processor 60, a memory 61, an input device 62 and an output device 63; the number of processors 60 in the device can be one or more. Figure 6 A processor 60 is taken as an example; the processor 60, the memory 61, the input device 62 and the output device 63 in the device can be connected by a bus or other means. Figure 6 The example of connecting through bus is taken in the following.

[0092] The memory 61, as a computer-readable storage medium, can be used to store software programs, computer executable programs and modules, such as the method for constructing a spatiotemporal dynamic monitoring model of river water quality based on multi-source data in the embodiment of the present invention, or the program instructions / modules corresponding to the spatiotemporal dynamic monitoring method of river water quality based on multi-source data. The processor 60 executes various functional applications and data processing of the device by running the software programs, instructions and modules stored in the memory 61, that is, to realize the above-mentioned method for constructing a spatiotemporal dynamic monitoring model of river water quality based on multi-source data, or the spatiotemporal dynamic monitoring method of river water quality based on multi-source data.

[0093] The memory 61 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system and at least one application required for a function; the data storage area may store data created according to the use of the terminal, etc. In addition, the memory 61 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 61 may further include a memory remotely arranged relative to the processor 60, and these remote memories may be connected to the device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0094] The input device 62 may be used to receive input digital or character information and generate key signal input related to user settings and function control of the device. The output device 63 may include a display device such as a display screen.

[0095] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the method for constructing a spatiotemporal dynamic monitoring model of river water quality based on multi-source data, or the method for spatiotemporal dynamic monitoring of river water quality based on multi-source data, is implemented in any embodiment.

[0096] The computer storage medium of the embodiment of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0097] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, which carry computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0098] The program code embodied on the computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0099] Computer program code for performing the operation of the present invention may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as C or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a spatiotemporal dynamic monitoring model for river water quality based on multi-source data, characterized in that: include: Obtain satellite imagery data, climate data, and land use data for the river basin, as well as water quality parameters at multiple sites within the basin; By analyzing the correlation between each spectral band in the satellite image data and the water quality parameters, multiple sensitive bands of the water quality parameters are determined; adding, subtracting and ratio operations are performed on each sensitive band to obtain multiple spectral band combinations; by correlation analysis, the degree of correlation between each spectral band combination and the water quality parameters is determined; and the spectral band combinations whose correlation with the water quality parameters meets the set conditions and passes the significance test are selected as the optimal spectral characteristics of each station; According to the land use data, a buffer zone is constructed with the continuous area of ​​various land use types in the river basin as the center and extending outward, and each buffer zone represents the impact range of various land use types on the surrounding environment; the river vector data is intersected with the vector data of the buffer zone corresponding to each land use type, wherein the river vector data includes the longitude and latitude of all river grids, and the vector data of the buffer zone includes the longitude and latitude of all buffer zone grids; the intersection part of each land use type is converted into a river grid layer under each land use type, wherein the value of the intersection part in each layer is 1, and the value of the non-intersection part is 0; from the river grid layer under each land use type, the 0 / 1 value at each station position is extracted as the characteristics of each land use type at each station; Taking the optimal spectral characteristics, land use type characteristics and meteorological characteristics of each station as independent variables and the water quality parameters of each station as dependent variables, a water quality parameter monitoring model suitable for the entire basin is fitted.

2. The method according to claim 1, characterized in that The step of constructing a buffer zone centered on a continuous area of ​​each land use type according to the land use data includes: Determine multiple buffer distances used to characterize the extent of influence of land use types; Taking the continuous area of ​​any land use type as the center, buffer zones of different spatial scales are delineated according to each buffer distance; the same operation is performed on each land use type to obtain buffer zones of different spatial scales under each land use type; The optimal scale of buffer zone is determined according to the proportion of sites covered by buffer zones of different spatial scales under each land use type.

3. The method according to claim 2, characterized in that The optimal scale buffer is determined according to the proportion of sites covered by buffers of different spatial scales under each land use type, including: Calculate the proportion of the number of sites covered by buffer zones of different spatial scales under each land use type to the total number of sites; If the change range of the proportions corresponding to buffer zones of different spatial scales under the same land use type is less than a first threshold, marking the same land use type; According to the sample balance of the independent variable, it is determined whether the proportions corresponding to the buffer zones of each spatial scale under the unmarked land use type exceed the set range; If the ratio corresponding to the buffer zone of a certain spatial scale exceeds the set range, the buffer zones of the certain spatial scale under all land use types will be deleted, and the remaining buffer zones will be used as the buffer zones of the optimal scale.

4. According to the method of claim 3, the first threshold is 0.05 or 0.1, and the setting range is greater than 0.2 or less than 0.

8.

5. The method according to claim 1, characterized in that The fitting is applicable to the water quality parameter monitoring model of the entire basin, including: The extreme randomized regression tree algorithm is used to fit the water quality parameter monitoring model applicable to the entire watershed.

6. A spatiotemporal dynamic monitoring method for river water quality based on multi-source data, characterized in that: include: Obtain satellite imagery data, climate data, and land use data for river basins at any time period; Based on the acquired data, the optimal band characteristics, land use type characteristics and meteorological characteristics at each location in the basin are determined; Input all the features at each position into the water quality parameter monitoring model constructed by the method described in any one of claims 1 to 5 to obtain the two-dimensional spatial distribution of the water quality parameters of the river basin at any time period; The two-dimensional spatial distribution of the water quality parameters is used as a visualization layer, which is superimposed with the visualization layers of other covariates in the model to achieve spatial visualization of the river basin.

7. An electronic device, characterized in that: include: one or more processors; a memory for storing 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 method for constructing a spatiotemporal dynamic monitoring model of river water quality based on multi-source data as described in any one of claims 1-5, or the method for spatiotemporal dynamic monitoring of river water quality based on multi-source data as described in claim 6.

8. A computer-readable storage medium, characterized in that: A computer program is stored thereon, which, when executed by a processor, implements the method for constructing a spatiotemporal dynamic monitoring model for river water quality based on multi-source data as described in any one of claims 1 to 5, or the method for spatiotemporal dynamic monitoring of river water quality based on multi-source data as described in claim 6.

Images