A remote sensing synchronous inversion method for lake water quality and topography under watershed-shoreline co-constraint

Abstract

This invention relates to a method for synchronous remote sensing inversion of lake water quality and topography under watershed-shoreline co-constraints. It uses lake meteorological characteristics, coastal and offshore topographic features, Landsat-8 band reflectance data, and spectral index data as inputs, and synchronous measured underwater topographic data and measured water quality data as outputs to train an XGBoost model. The optimal model is selected as the inversion model for underwater topographic and water quality data, and the spatiotemporal distribution of lake underwater topographic and water quality data is inverted using the estimation model. This invention overcomes the limitation of traditional remote sensing inversion, which only focuses on the spectral response within the lake, by integrating watershed topography, meteorological elements, and lake shoreline shape characteristics. It supports synchronous remote sensing estimation of lake water quality and topography under conditions of insufficient data or difficulty in field investigation (such as plateau lakes), providing reliable technical support and reference for the coordinated management of lake water environment and water resources.

Description

Technical Field

[0001] This invention belongs to the field of environmental remote sensing, specifically involving a method for synchronous remote sensing inversion of lake water quality and topography under the co-constraint of watershed and shoreline. Background Technology

[0002] Remote sensing technology has become an important tool for monitoring lake water environment, enabling high-precision quantitative extraction of key parameters such as chlorophyll a, suspended solids, colored dissolved organic matter (CDOM), and transparency. Early studies mostly established empirical relationships between remote sensing spectra and lake pigment concentrations to achieve quantitative inversion of specific lake water quality parameters. In recent years, with the rapid development of intelligent algorithms such as machine learning, many studies have used algorithms such as random forests and neural networks to train spectral bands or combinations of different bands, significantly improving the inversion accuracy and generalization ability of the models and promoting the widespread application of related technologies.

[0003] However, existing research indicates that the spatial distribution mechanisms driving lake nutrient concentrations are extremely complex. In shallow lake ecosystems, wind and wave disturbances act as a significant physical driving force, altering water mixing states and prompting significant spatiotemporal migration of algal communities. Special topographical areas such as lake bays, with their unique hydrodynamic conditions, form relatively stable algal aggregation zones. It is noteworthy that changes in the lake water environment are essentially the result of the interaction between watershed human activities and meteorological factors. Specifically, human activities such as land use change and point source emissions directly alter nutrient input fluxes, while meteorological factors such as precipitation and temperature influence water exchange processes, jointly regulating lake nutrient concentrations and their spatial distribution. Therefore, by training the model using spectral data from the lake area, considering the interaction between the watershed and lake system, and combining lakeshore characteristics with climatic factors to improve model generalization ability, a new approach can be provided for constructing a comprehensive model that combines mechanistic support and inversion performance.

[0004] In the field of lake water volume monitoring, the scarcity of measured data leads to significant uncertainty in the estimation of total lake water volume. The core of lake water volume surveys lies in measuring lake depth and underwater topography. Existing research mainly revolves around two technical approaches: direct measurement and spatial inference. The former focuses on using multi-platform sensors such as shipborne sonar and UAV lidar to accurately extract underwater elevation information; the latter aims to construct spatial inference models based on surrounding topographic features to extrapolate underwater topography in areas without measured data. Research on lake depth estimation using optical images and satellite altimetry data is still relatively limited. While remote sensing satellites have proven effective for depth mapping in intermittently flooded areas, obtaining topographic information in permanent water bodies remains challenging. To overcome these difficulties, some spatial prediction and modeling methods have been proposed to predict underwater topography by utilizing exposed topography around lakes. The relationship between topographic variables and lake depth (including lake depth and volume) has been extensively studied, supporting the development of geostatistical models for estimating lake depth and volume at the regional and even global scales. Although these estimates are generally acceptable in large-scale studies because the offset effects of underestimation and overestimation cancel each other out. However, the uncertainties and potential biases at specific lake or local scales cannot be ignored. Therefore, it is necessary to determine a specific and effective spatial modeling method to enhance the linkage response of spatial inference models "above water" (watershed topography, lake shoreline) and "below water" (water quality parameters), and to provide optimization solutions for the set of topographic variables.

[0005] In summary, conducting simultaneous remote sensing monitoring of lake water environment and water resources, and constructing a simultaneous inversion model of lake water quality and topography, is not only a scientific foundation for accurately assessing lake ecological health and supporting watershed water environment management decisions, but also an urgent need to promote the sustainable use of water resources and the construction of watershed ecological civilization. This invention integrates remote sensing spectral inversion, watershed surface information, and machine learning algorithms, breaking through the technical bottlenecks of traditional single-element monitoring, and constructing a simultaneous water quality-topography inversion framework, providing an integrated solution for multi-dimensional monitoring of lake ecosystems. Summary of the Invention

[0006] The purpose of this invention is to provide a method for synchronous remote sensing inversion of lake water quality and topography under watershed-shoreline co-constraints. The principle, process, and results of this method support synchronous remote sensing estimation of lake water quality and quantity under conditions of insufficient data or difficulty in field investigation (such as plateau lakes), and can provide a foundation for the synchronous acquisition of lake water environment and water resource information by the Landsat series satellites.

[0007] To achieve the above objectives, the present invention adopts the following solution:

[0008] A method for synchronous remote sensing inversion of lake water quality and topography under watershed-shoreline co-constraints, the method comprising:

[0009] The lake boundary is divided into several segments; for each segment, the slope, aspect, and elevation data are extracted from the distance sequence from the shore based on the DEM data, extending inland along the segment's perpendicular line; and the ratio of the segment shoreline length to the segment straight line length is calculated as the shoreline complexity for each segment of the lake boundary, which, together with the slope, aspect, and elevation data, constitutes the lake's coastal and offshore topographic feature data.

[0010] Acquire meteorological characteristic data of lakes and band reflectance data of Landsat-8 remote sensing data covering lakes; and calculate algal biomass index (ABI) and water color index (FUI) based on the band reflectance data;

[0011] The XGBoost model is trained by taking the aforementioned topographic features, meteorological features, ABI, FUI, and reflectivity data in the green, red, and near-infrared bands as inputs, and synchronous measured underwater topographic data and measured water quality data as outputs.

[0012] The optimal model is selected as the inversion model for underwater topographic data and water quality data, and the spatiotemporal distribution of lake underwater topographic data and water quality data is inverted using the inversion model.

[0013] In some embodiments of the present invention, after extracting the lake water area based on the band reflectivity data, the lake boundary is obtained based on the extracted lake water area.

[0014] In some embodiments of the present invention, the boundary of the lake is divided into several segments at equal intervals.

[0015] In some embodiments of the present invention, slope, aspect, and elevation data are extracted at equal intervals for each segment perpendicular line extending into the land.

[0016] In some embodiments of the present invention, the method further includes determining the distance of the interval: setting different interval distances for the perpendicular bisectors of the segments, and selecting the interval distance that optimizes the model accuracy as the sampling parameter for the offshore terrain data.

[0017] In some embodiments of the present invention, the lakes used for model training include lakes with different shapes, different shoreline complexities, and different water quality data coverage.

[0018] In some embodiments of the present invention, the meteorological features are daily average data obtained based on the ERA5 reanalysis dataset, including temperature, radiation, precipitation and wind speed.

[0019] In some embodiments of the present invention, the method further includes estimating the water volume of the lake by combining the lake water level and area after obtaining the inverted underwater topographic data of the lake.

[0020] In some embodiments of the present invention, the Landsat-8 remote sensing data is selected from the L2 level surface reflectance product of the Landsat-8 OLI sensor; the surface reflectance is divided by π to obtain the water-free reflectance. R rs (λ), and for R rs (λ) Outlier removal from the dataset: Values ​​with reflectance less than 0 in the band, or with green band reflectance less than 0.02 sr. -1 Or, a reflectivity greater than 0.50 sr in any visible light band. -1 The pixels are masked.

[0021] In some embodiments of the present invention, during model training, the ten-fold cross-validation method is used to evaluate the model performance by combining the mean root mean square error and mean deviation of the prediction results, and the model with the best overall accuracy is selected as the final inversion model.

[0022] This invention develops a joint inversion model for lake water quality and topography based on a watershed and shoreline co-constrained machine learning algorithm. Landsat-8 is a broadband satellite with fewer spectral bands, capturing less spectral information than the commonly used MODIS water color satellite, thus its application in lake water color inversion is limited. This invention improves the applicability of the training model by selecting lake samples of different shapes and water quality ranges during model training. In addition to visible and near-infrared bands, spectral indices indicating algal biomass and water color are introduced when selecting input features. These remote sensing spectral indices are correlated with nutrient levels and transparency in the water, which can accelerate model convergence. The input topographic feature data includes both coastal and offshore data, enabling spatial extrapolation of underwater topography. Furthermore, based on the distribution of water quality parameters and the correlation between topographic data and meteorological elements (such as extreme precipitation and wind erosion potentially altering the shape of a lake, thus affecting the geomorphology of the nearshore or lakeside zone), specific meteorological data is selected as input. Based on the characteristics of the selected lake samples and 4-dimensional lake features, a high-precision inversion model is trained.

[0023] This invention breaks through the limitations of traditional remote sensing inversion, which only focuses on the internal spectrum of lakes, by integrating watershed topography, meteorological elements, and lake shoreline shape characteristics. It supports the synchronous estimation of lake water quality and quantity by remote sensing under conditions of insufficient data or difficulty in field investigation (such as plateau lakes). The principle, process, and results of the method can provide a reference for the synchronous estimation of key elements of lake water environment and lake water resource reserves by Landsat series satellites.

[0024] It should be understood that all combinations of the foregoing concepts and the additional concepts described in more detail below may be considered part of the inventive subject matter of this disclosure, provided that such concepts do not contradict each other. Furthermore, all combinations of the claimed subject matter are considered part of the inventive subject matter of this disclosure.

[0025] The foregoing and other aspects, embodiments, and features of the teachings of the present invention will be more fully understood from the following description in conjunction with the accompanying drawings. Other additional aspects of the invention, such as features and / or beneficial effects of exemplary embodiments, will become apparent from the following description or may be learned through practice of specific embodiments according to the teachings of the present invention. Attached Figure Description

[0026] The accompanying drawings are not intended to be drawn to scale. In the drawings, every identical or nearly identical component shown in each figure can be denoted by the same reference numeral. For clarity, not every component is labeled in each figure, wherein:

[0027] Figure 1 The measured underwater topographic spatial distribution of sampled lakes (a) Lake A and (b) Lake B.

[0028] Figure 2 It is the principle of underwater topography extrapolation in lakes.

[0029] Figure 3 This is the flowchart of the XGB-LWQQ algorithm.

[0030] Figure 4 This relates to the impact of the surface data extraction interval on the algorithm's accuracy.

[0031] Figure 5 The prediction accuracy of the XGB-LWQQ algorithm based on underwater terrain data (a) Chla, (b) SD and (c) of the training set; the prediction accuracy of the XGB-LWQQ algorithm based on underwater terrain data (d) Chla, (e) SD and (f) of the validation set.

[0032] Figure 6 It is based on the XGB-LWQQ algorithm and shows how the accuracy of underwater terrain inversion varies with distance from the shore.

[0033] Figure 7 This is a comparison of the accuracy of the XGBoost algorithm with DNN, SVR, KNN, MLR, and RF algorithms.

[0034] Figure 8 The images show Landsat-8 OLI images of Lake A and Lake B, the average chlorophyll a concentration, the spatial distribution of transparency, and the results of underwater topography simulation. Detailed Implementation

[0035] To better understand the technical content of the present invention, specific embodiments are described below in conjunction with the accompanying drawings.

[0036] Various aspects of the invention are described in this disclosure with reference to the accompanying drawings, in which numerous illustrative embodiments are shown. The embodiments of this disclosure are not necessarily intended to encompass all aspects of the invention. It should be understood that the various concepts and embodiments described above, as well as those described in more detail below, can be implemented in any of many ways, as the concepts and embodiments disclosed herein are not limited to any particular implementation. Furthermore, some aspects of the invention disclosed may be used alone or in any suitable combination with other aspects of the invention disclosed.

[0037] Example 1

[0038] This embodiment takes Lake A and Lake B in a plateau province as examples to further describe the technical solution of the present invention.

[0039] The water quality data of the two lakes selected in the example cover a wide range and can well represent the optical properties of plateau lakes. Furthermore, the shapes of the two lakes include elongated, elliptical, and irregularly concave forms, ensuring the applicability of the developed model.

[0040] The model of this invention can be trained independently (each model has one output) or jointly (using XGBoost multi-output regression). In the embodiments, the model is built by independent training.

[0041] The method of this invention is based on measured chlorophyll a data, transparency, underwater topography data of lakes, and synchronized Landsat-8 OLI surface reflectance data of the study area, and obtains water-free reflectance through simple correction. R rs Using three bands of visible light R rs One near-infrared band and two remote sensing spectral indices were used as inputs, along with four watershed topographic features (slope, aspect, elevation, and shoreline complexity) and four meteorological features (temperature, radiation, precipitation, and wind speed). A lake water quality and topography inversion algorithm (XGB-LWQQ) suitable for Landsat-8 was developed using the XGBoost machine learning method. The algorithm was then applied to several lake case studies to evaluate its performance. Finally, the algorithm was applied to long-term image series to obtain the changing trends of lake water quality and topography from 2013 to 2020.

[0042] As an exemplary description, the implementation of the aforementioned method will be specifically explained below with reference to the accompanying drawings.

[0043] Step 1: Based on Landsat-8 remote sensing product data, acquire surface reflectance (SR) datasets for four bands; and collect underwater topographic data of lakes, as well as measured chlorophyll a concentration and transparency data of lakes at Landsat-8 synchronous transit time.

[0044] Using a Landsat-8 OLI sensor-based L2-level remote sensing product as the data source, the water reflectance was obtained by dividing SR by π to correct for the influence of the solar zenith angle. R rs (λ). Simultaneously, the reflectivity value of the band being less than 0, or the emissivity of the green band being less than 0.02 sr... -1 or a reflectivity greater than 0.50 sr in the visible light band -1 Outliers were removed to obtain results showing areas covered only by water.

[0045] Multibeam echo sounder was used to collect underwater topographic data of the lake, and the acquisition time, coordinate system and vertical reference surface parameters of the echo sounding data were recorded simultaneously.

[0046] Step 2: Data and algorithm structure used in the development of the XGB-LWQQ algorithm;

[0047] Based on existing lake water quality parameter sampling information, the following criteria were used to screen the data: (1) Samples covered by clouds and solar flares were removed by combining measured records and satellite imagery; (2) A ±3-hour time window was used to match the dataset; (3) A 3×3 window was retained for each sample. R rs (λ) Data with coefficients of variation all within 10%. A total of 211 sets of water quality sample data were obtained according to the above standards.

[0048] The measured water quality data of the two lakes were randomly divided into two categories: 150 groups were used for modeling and 61 groups were used for validation. The concentrations of substances in the two lakes differed significantly. Lake B had an average Chl-a of 1.09 ± 0.16 μg / L, classifying it as an oligotrophic lake, with an average transparency greater than 10 m. Lake A had an average Chl-a of 41.45 ± 21.99 μg / L, classifying it as a eutrophic lake, with a transparency of 1.32 ± 0.36 m. Furthermore, Lake A had a long and narrow shape. Figure 1 In lake (a), the shoreline is relatively straight; the water area of ​​lake B is butterfly-shaped. Figure 1 (b) The lake shore is winding and complex, with many bays.

[0049] The Normalized Water Index (NDWI) was constructed based on the green light band (561nm) and near-infrared band (865nm) of Landsat-8 OLI to extract the range of lake water areas.

[0050] The OTSU algorithm was used to determine the land-water separation threshold, extract the lake water area, and obtain the synchronous lake boundary during the image transit period. The lake boundary was then segmented into equal intervals, and the fractal dimension of each segment was calculated based on fractal theory to indicate the complexity of the lake shoreline. Slope and aspect were calculated using DEM data within the watershed. For each lake shoreline segment, the slope was extracted at equal intervals along its perpendicular bisector towards the land, based on the distance sequence from the shore. tan ( α )), slope aspect (angle) α (After normalization) and elevation values.

[0051] In this embodiment, the boundary between Lake A and Lake B is divided into 500 equally spaced segments. The boundary length is calculated using Python's `shapely.geometry` package. Based on fractal theory, the curve length between points (the length of that shoreline segment) is calculated. L shoreline ) and straight-line length (the distance between the beginning and end points of this shoreline) D shoreline The ratio of the shoreline to the lake's shoreline represents its complexity. A ratio closer to 1 indicates a straighter shoreline, while a lower ratio indicates a more tortuous shoreline. Figure 1 ).

[0052] Digital elevation model (DEM) data of the watershed surface with a spatial resolution of 30 meters was obtained from the geospatial data cloud (https: / / www.gscloud.cn / ), and slope and aspect were calculated using ArcMap. Offshore surface elevations were taken every 100 meters along the perpendicular bisector of the segmented shoreline. h 1… h k ), slope and aspect values ​​(e.g.) Figure 2 As shown, this invention uses offshore surface elevation to infer underwater topography. h '1… h ' n Together with the segmented shoreline complexity, it serves as a 4D terrain information feature for training machine learning algorithms.

[0053] Previous studies have confirmed that the algal biomass index (ABI) has a good response to trophic status indices based on Chl-a. Meanwhile, the Forel-Ule index (FUI) has been found to exhibit a power-law decay relationship with transparency, with a coefficient of determination R0. 2 The value is as high as 0.95. Since lake transparency is mainly affected by the chromaticity angle α, the FUI index is introduced as an input feature.

[0054] Considering the close spatial correlation between meteorological elements and lake water quality parameters, and the fact that extreme precipitation and wind erosion can alter lake shape, significantly impacting the geomorphology of the near-shore or lakeside zone, this example uses meteorological elements from the ERA5 global reanalysis dataset, including surface temperature, solar radiation, wind speed, and precipitation data. This dataset has a spatial resolution of 0.25° × 0.25° and a temporal resolution of 1 hour. Based on the smallest watershed unit where the lake is located, daily average values ​​of watershed temperature, solar radiation, wind speed, and precipitation for the same time period during satellite transit are statistically analyzed, forming a total of four meteorological feature values, which are used as input for model training.

[0055] The XGB-LWQQ method flow is as follows: Figure 3 As shown, terrain information features, remote sensing spectral features, and meteorological element features are input into the model for training. The XGB-LWQQ method is implemented using the XGBoost package in Python. Parameter tuning is performed during model training, including the learning rate, maximum tree depth, minimum sum of leaf node sample weights, maximum number of leaf nodes (Treenum), random sampling ratio of training samples, random sampling ratio of features, and regularization parameters. The parameter combination is optimized using a grid search method. The target parameters include the learning rate, tree depth, and minimum sum of leaf node weights. The parameter combination that meets the accuracy requirements and minimizes the model complexity index (tree depth × number of leaf nodes) is prioritized to reduce model complexity and enhance generalization ability, ultimately determining the optimal model structure parameters.

[0056] In this embodiment, the input parameters are optimized. Multiple XGBoost machine learning algorithms are constructed through permutations and combinations of the input parameters, and the one with the best accuracy is selected as the final model input. Tables 1-3 show the model inversion accuracy of each output parameter under different input parameter combinations. Due to the large number of parameter combinations, only R is retained in the tables for water quality parameter inversion. 2 Models with a value ≥0.7 were used as a comparison reference, and R-values ​​were preserved for underwater topographic inversion. 2 Models with a precision ≥0.5. 200 samples were selected from the total sample data for model training during input parameter selection. Tables 1-3 show the training set precision.

[0057] Table 1. Evaluation results of Chla inversion accuracy for different input parameters.

[0058]

[0059] In the table, N refers to the number of samples.

[0060] Table 2. Evaluation results of transparency inversion accuracy for different input parameters.

[0061]

[0062] Table 3. Evaluation results of underwater topography inversion accuracy for different input parameters

[0063]

[0064] Based on the model accuracy evaluation results, Landsat-8 OLI was ultimately selected. R rs (561) R rs (655) R rs (865) Five-dimensional remote sensing spectral features, including algal biomass index (ABI) and water color index (FUI), were combined with meteorological features and coastal and offshore topographic features as inputs for model training.

[0065] In this embodiment, the model parameters obtained by training using the above-described method are as follows: learning rate of 0.05, maximum tree depth of 4 layers, resampling rate of 0.8, maximum number of leaf nodes of 100, regularization parameter of L2 regularization coefficient of 0.01, initial maximum number of iterations of 1000, and random sampling ratio of training samples of 7:3.

[0066] This embodiment also analyzes the impact of different sampling intervals on the perpendicular bisector of the lake shore on model accuracy. Four XGBoost machine learning models were constructed by setting the sampling interval distances to 50m, 100m, 200m, and 500m. Simultaneously, to ensure model stability and independence from the training set, a ten-fold cross-validation method was used. The model performance (mean RMSE and MAPE) was evaluated using 10 randomly selected training and validation sets. The results are as follows: Figure 4 As shown, it can be seen that the model accuracy and sampling interval distance are not linearly related. The accuracy of a sampling interval of 100m is better than that of 50m, 200m and 500m. Therefore, in this embodiment, the sampling interval with the highest accuracy (100m) is selected as the surface data extraction interval.

[0067] With optimal input parameter combinations and sampling accuracy, the XGB-LWQQ algorithm demonstrates high accuracy on both training and validation datasets. Figure 5 In the training set, most data points are evenly distributed around the 1:1 line, while the accuracy of the validation set is slightly lower than that of the training set. Accuracy statistics of the water quality inversion model show that the average accuracy of the Chla and SD training sets in cross-validation is... R 2 The values ​​were 0.98 and 0.96, respectively; the RMSE was 6.21 μg / L and 0.48 m; and the MAPE was 15.78% and 32.43% (N = 182). In the validation set, the average values ​​of Chla and SD inversion results were... R 2The values ​​were 0.86 and 0.88, respectively; the RMSEs were 10.72 μg / L and 0.92 m, respectively; and the MAPEs were 28.96% and 45.97% (N = 94), respectively. In the underwater topography inversion model, the average value of the ten-fold cross-validation training set was... R 2 The mean value was 0.96, and the RMSE and MAPE were 1.56% and 0.08% respectively (N = 350); the average value on the validation set was... R 2 The accuracy was 0.60, with RMSE and MAPE of 5.26% and 0.27% respectively (N = 150). Multiple rounds of validation metrics showed that the water quality and topography inversion models had weak dependence on the training set and good anti-interference performance, indicating that the inversion models have good generalization ability.

[0068] However, due to the diverse causes of lakes, complex geological structures, and limited spectral signal penetration, the inversion of underwater topography using terrestrial surface features has a high inversion accuracy in lakes with regular shapes and regular basin topographic variations (RMSE of 5.44m for Lake A), but a relatively low inversion accuracy in lakes with irregular shapes or greater depths (RMSE of 8.30m for Lake B).

[0069] This embodiment further extracts underwater topographic inversion values ​​at different distances from the shore and compares them with the measured underwater topographic results. With increasing distance from the shore, RMSE increases from 5.15m to 7.24m, MAPE increases from 0.26% to 0.32%, and the error increases by 0.06% per kilometer. Figure 6 Therefore, the accuracy of this underwater topography inversion is slightly lower in areas farther from the lake shore than in areas closer to the lake shore.

[0070] This embodiment further compares the accuracy results of the XGBoost algorithm with several other common intelligent algorithms. Figure 7 The algorithms used include Deep Neural Networks (DNNs), Support Vector Machines (SVRs), K-Nearest Neighbors (KNNs), Multiple Linear Regression (MLRs), and Random Forests (RFs). Due to limitations in the number of training samples, DNN models exhibit relatively poor accuracy, with overfitting and an RMSE close to 50m. SVRs and KNNs show significantly better accuracy than DNNs, with scatter plots evenly distributed on both sides of the 1:1 line. The RMSE and MAPE of Multiple Linear Regression, Random Forests, and XGBoost algorithms are close; however, XGBoost has a smaller bias (BIAS) of only -0.51. Therefore, the XGB-LWQQ algorithm of this invention outperforms other mainstream intelligent algorithms.

[0071] Step 3: Apply the lake water quality and topography inversion algorithm to the lake remote sensing data to obtain the spatial distribution of lake chlorophyll a, transparency and underwater topography.

[0072] Figure 8 The XGB-LWQQ inversion results for Lake A and Lake B are presented. The spatial distribution of chlorophyll a in the lakes obtained by the XGB-LWQQ algorithm is consistent with the spatial distribution of RGB imagery. Taking Lake A as an example, the chlorophyll a concentration is high in the eastern part of Lake A and low in the western part. Meanwhile, the spatial distribution of transparency shows an opposite trend to the chlorophyll a concentration distribution: areas with more algae and higher chlorophyll a concentration have relatively lower transparency, while clearer, more transparent waters have relatively lower chlorophyll a concentration. The underwater topographic spatial distribution results show that the elevation of both Lake A and Lake B decreases from the shore to the center. The underwater topography changes more rapidly on the southern shore of Lake A, while the topography of the eastern bay of Lake B changes more significantly, with relatively obvious topographic undulations. These findings are consistent with the measured spatial distribution of underwater topography in the lakes, demonstrating good simulation accuracy.

[0073] While the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the invention. Those skilled in the art to which this invention pertains can make various modifications and refinements without departing from the spirit and scope of the invention.

Claims

1. A method for synchronous remote sensing inversion of lake water quality and topography under watershed-shoreline co-constraints, characterized in that, The method includes: The lake boundary is divided into several segments at equal intervals. For each segment, the slope, aspect, and elevation data are extracted from the offshore distance sequence based on the DEM data at equal intervals along the segment's perpendicular line towards the land. The ratio of the shoreline length to the straight line length of each segment of the lake boundary is calculated as the shoreline complexity, which, together with the slope, aspect, and elevation data, constitutes the lake's coastal and offshore topographic feature data. Acquire meteorological characteristic data of lakes and band reflectance data of Landsat-8 remote sensing data covering lakes; and calculate algal biomass index (ABI) and water color index (FUI) based on the band reflectance data; The XGBoost model is trained by taking the aforementioned topographic features, meteorological features, ABI, FUI, and reflectivity data in the green, red, and near-infrared bands as inputs, and synchronous measured underwater topographic data and measured water quality data as outputs. The optimal model is selected as the inversion model for underwater topographic data and water quality data, and the spatiotemporal distribution of lake underwater topographic data and water quality data is inverted using the inversion model.

2. The method according to claim 1, characterized in that, After extracting the lake's water area based on the band reflectivity data, the lake's boundary is obtained based on the extracted lake water area.

3. The method according to claim 1, characterized in that, The method also includes determining the interval distance: setting different interval distances for the perpendicular bisectors of the segments, and selecting the interval distance that optimizes the model accuracy as the sampling parameter for the offshore terrain data.

4. The method according to claim 1, characterized in that, The lakes used for model training include lakes of different shapes, shoreline complexities, and water quality data coverage.

5. The method according to claim 1, characterized in that, The meteorological features are daily average data obtained based on the ERA5 reanalysis dataset, including temperature, radiation, precipitation, and wind speed.

6. The method according to claim 1, characterized in that, The method also includes estimating the lake's water volume by combining the lake's water level and area after obtaining the retrieved underwater topographic data.

7. The method according to claim 1, characterized in that, The Landsat-8 remote sensing data used is the L2 level surface reflectance product of the Landsat-8 OLI sensor. The surface reflectance is divided by π to obtain the water-free reflectance. R rs (λ), and for R rs (λ) Outlier removal in the dataset: Pixels with a band reflectance value less than 0, or a green band reflectance less than 0.02 sr⁻¹, or any visible band reflectance greater than 0.50 sr⁻¹ are masked.

8. The method according to claim 1, characterized in that, During model training, the ten-fold cross-validation method is used to evaluate model performance by combining the mean root mean square error and mean bias of the prediction results. The model with the best overall accuracy is selected as the final inversion model.

Images