A method and system for calculating dynamic surface water storage using SWOT satellite data.

Abstract

This application relates to the field of water resources survey technology, and in particular to a method and system for calculating dynamic surface water storage using SWOT satellite data. The method collects water surface height data and calculates the water surface area, performing environmentally sensitive dynamic weight allocation and multi-source fusion to obtain the water surface area of ​​the target surface water body; it then performs water level measurements; it fuses and corrects all calculated water level measurement data to obtain the corrected water surface elevation; it acquires underwater topographic data and uses a multimodal deep learning framework to perform feature-level fusion and error minimization optimization on the underwater topographic data to obtain the underwater topographic data; and finally, it calculates the water storage of the target surface water body. This invention solves the problems of existing technologies, such as single observation data sources and susceptibility to environmental interference, and features robustness and strong all-weather adaptability.

Description

Technical Field

[0001] This application relates to the field of water resources survey technology, and in particular to a method and system for calculating dynamic surface water storage using SWOT satellite data. Background Technology

[0002] Surface water is an indispensable strategic resource for supporting socio-economic development and maintaining ecosystem balance. Dynamically and accurately monitoring changes in the water storage of surface water bodies such as lakes and reservoirs is of vital importance for optimal water resource allocation, flood and drought disaster early warning, water environment management, and global climate change research.

[0003] Traditional methods for calculating dynamic surface water storage mainly rely on field measurements from ground hydrological stations and remote sensing technology from single data sources. While ground stations (such as water level gauges and flow cross sections) can provide high-precision time-series data, their spatial representativeness is limited, making it difficult to reflect the overall, highly spatially heterogeneous changes in water storage of large water bodies (such as Poyang Lake and Dongting Lake).

[0004] With the development of remote sensing technology, using satellite data for water storage estimation has become mainstream. This method typically involves three key steps: water surface area extraction, water level measurement, and underwater topography (depth sounding) data acquisition. However, existing technologies face numerous bottlenecks in these steps:

[0005] Regarding water surface area extraction: While commonly used optical remote sensing methods (such as the Normalized Difference Water Index (NDWI)) are technically mature, they are highly susceptible to interference from meteorological conditions such as clouds, fog, and rain, leading to data gaps or decreased accuracy. Furthermore, in complex terrain areas, mountain shadows and vegetation obscuring often cause misclassification or omission of water bodies. Although Synthetic Aperture Radar (SAR) technology possesses all-weather observation capabilities, its accuracy in identifying waterfront boundaries is easily affected by surface roughness (such as wind, waves, and damp soil), and data processing is relatively complex. Therefore, relying on a single remote sensing data source cannot guarantee a stable and reliable water surface area under all conditions.

[0006] In water level measurement: Traditional satellite altimeters (such as the Jason series) can measure water levels, but their spatial resolution is low, making them mainly suitable for oceans and large lakes. The new generation of SWOT (Surface Water and Ocean Topography) satellites has greatly improved observation capabilities, but their measurement signals are still affected by atmospheric delay (especially tropospheric wet delay), resulting in systematic errors. While ground-based GNSS reflectometer technology (GNSS-R / IR) can monitor water levels at fixed points and high frequencies, its accuracy is easily affected by complex environmental factors such as surface waves and multipath effects. The challenge lies in integrating the observation advantages of different platforms and eliminating their respective systematic errors to achieve centimeter-level accurate measurements.

[0007] In terms of underwater topographic data acquisition: This is the most challenging aspect of water storage estimation. Traditional shipborne sonar bathymetry (single-beam / multi-beam) offers high accuracy but is costly, time-consuming, and labor-intensive, and cannot provide comprehensive coverage of vast water areas. Water depth remote sensing inversion methods based on optical imagery are low-cost and have wide coverage, but their accuracy drops sharply in deeper or less transparent areas. Methods based on interpolation of land surface topography suffer from significant uncertainty in lake centers far from the shoreline.

[0008] In terms of water storage calculation and prediction: Existing calculation models are mostly simple grid accumulation methods based on "volume = area × water depth". This model treats the water body as a static rigid body, ignoring important physical processes such as vertical seepage, thermal expansion and contraction of water, and salinity changes. This leads to "pseudo-storage changes" in the calculation results, which cannot truly reflect hydrological and physical processes. In addition, existing methods focus more on estimating current storage and lack the ability to dynamically predict future trends.

[0009] In summary, existing surface water storage survey technologies suffer from problems such as relying on a single data source and being susceptible to environmental interference. There is an urgent need for a comprehensive calculation method that can systematically integrate multi-source data, overcome individual limitations, and take into account fine physical processes. Summary of the Invention

[0010] To address the problems of existing technologies, such as single observation data sources and susceptibility to environmental interference, this invention provides a method and system for calculating dynamic surface water storage by incorporating SWOT satellite data. This method and system are characterized by robustness and strong all-weather adaptability.

[0011] To achieve the above-mentioned objectives of this invention, the technical solution adopted is as follows:

[0012] A method for calculating dynamic surface water storage using SWOT satellite data, the method comprising the following steps:

[0013] Collect water surface height data and calculate water surface area separately; perform dynamic weight allocation and multi-source fusion of all calculated water surface area results with environmental factor sensitivity to obtain the water surface area of ​​the target surface water body;

[0014] Perform water level measurements; fuse and correct all calculated water level measurement data to obtain the corrected water surface elevation;

[0015] Underwater topographic data is acquired, and a multimodal deep learning framework is used to perform feature-level fusion and error minimization optimization on the underwater topographic data to obtain underwater topographic data.

[0016] The water storage of the target surface water body is calculated by combining the water surface area, water surface elevation, and underwater topographic data.

[0017] Preferably, water level height data is collected and water surface area is calculated separately. The specific steps are as follows:

[0018] Acquire four-band optical remote sensing images and calculate the NDWI index to distinguish between water bodies and non-water bodies:

[0019]

[0020] in For green band reflectivity, Near-infrared reflectance, segmented by threshold. The cumulative pixel area of ​​the region is obtained ;

[0021] Acquire three-band optical remote sensing images, and apply the SAM model to perform zero-sample segmentation of the RGB images to obtain water mask images, thus obtaining the segmented water regions:

[0022]

[0023] in For the segmentation function, Input three-band RGB optical remote sensing images;

[0024] right Water body index Spectral screening was performed, among which , , The reflectances of the green, red, and blue bands are respectively. If the area exceeds a preset threshold, the water area is confirmed, and the total area of ​​the mask pixels is calculated by integration. ;

[0025] Acquire synthetic aperture radar imagery based on SAR echo intensity Calculate OTSU automatic threshold ,Will The pixels are classified as water areas, and the pixel areas are accumulated to obtain... ;

[0026] Obtain water surface height data acquired by SWOT satellite Dynamically monitor water body boundaries and calculate the pixel area within the boundaries:

[0027] in Based on water level Edge detection function, This is the dynamic water height threshold.

[0028] Furthermore, considering environmental impact, the water surface area is calculated based on the aforementioned multi-source water surface observation data; the calculated water surface area results are then subjected to dynamic weight allocation and multi-source fusion sensitive to environmental factors to obtain the water surface area of ​​the target surface water body: the specific steps are as follows:

[0029] Considering the influence of clouds and fog, construct the weights corresponding to the NDWI water index method:

[0030]

[0031] in This is the cloud and fog coverage error term. NDWI specific adjustment parameters to account for sensitive cloud and fog interference;

[0032] Considering terrain influence factors, the weights corresponding to the SAM model segmentation method are constructed as follows:

[0033]

[0034] in This is the shadow interference error term. SAM-specific adjustment parameters to account for sensitive vegetation shading;

[0035] Considering meteorological influencing factors, the weights corresponding to the OTSU threshold segmentation method are constructed as follows:

[0036]

[0037] in For the rainfall interference error term, SAR-specific adjustment parameters are designed to take into account sensitive, all-weather coarse data;

[0038] Considering the impact factor of satellite acquisition error, construct the SWOT analysis for satellite weights:

[0039]

[0040] in This is the spatial resolution error term. SWOT-specific adjustment parameters are provided to consider monitoring scale and sensitivity.

[0041] The surface area of ​​the target water body was calculated. .

[0042] Furthermore, water level measurements are performed, with the following specific steps:

[0043] Acquire GNSS signal data and use GNSS-R technology to extract preliminary water level. :

[0044]

[0045] in At the speed of light, This is the time delay of the reflected signal;

[0046] Water level was extracted using GNSS-IR technology. ;

[0047]

[0048] in, For phase difference, For signal frequency, This refers to the satellite's elevation angle;

[0049] A water surface DEM was generated using a LiDAR sensor and simultaneously scanned during the transit of the SWOT satellite. The satellite data was then aligned using the ICP algorithm.

[0050]

[0051] in For SWOT point cloud, Pointing clouds for drones, Let be a rotation matrix. The translation vector is obtained from the average water level height in the aligned DEM. ;

[0052] Obtain SWOT satellite radar echo time delay Extract water level :

[0053] .

[0054] Furthermore, all the calculated water level measurement data are fused and corrected to obtain the corrected water surface elevation. The specific steps are as follows:

[0055] Constructing the dynamic coefficients corresponding to GNSS-R technology:

[0056]

[0057] in This is the wave interference error term. To standardize the adjustment parameters;

[0058] Constructing the dynamic coefficients corresponding to GNSS-IR:

[0059]

[0060] in This is the phase drift error term;

[0061] Construct the dynamic coefficients corresponding to the UAV water level gauge:

[0062]

[0063] in This is the positioning error term;

[0064] Constructing the dynamic coefficients corresponding to SWOT satellite technology:

[0065]

[0066] in This is the atmospheric delay error term;

[0067] Dynamic weight allocation and multi-source fusion are performed to obtain the water surface elevation:

[0068]

[0069] Further optimization through Kalman filtering iterative steps :

[0070]

[0071] in For variance-based Kalman gain, For the current measurement, For the predicted value, the coefficients are iteratively optimized using the least squares method to achieve closed-loop correction. As the corrected water surface elevation.

[0072] Furthermore, to obtain underwater topographic data, the specific steps are as follows:

[0073] Water depth is extracted using single / multibeam echo sounding techniques. :

[0074]

[0075] in For the speed of sound waves, To delay the echo time, a multi-beam fan-shaped scan is used to cover the underwater area;

[0076] Based on remote sensing image data Water depth was extracted using the random forest method. :

[0077]

[0078] in, For pre-trained random forest models;

[0079] Obtain land surface topographic interpolation and use the kriging method to estimate the topography. :

[0080]

[0081] in, To minimize variance weights, These are measurements from known points on the surrounding land surface. This is the global average constant.

[0082] Furthermore, a multimodal deep learning framework is used to learn underwater terrain data to obtain underwater terrain data. The specific steps are as follows:

[0083] Construct fusion weights for single / multibeam bathymetry:

[0084]

[0085] in This is the depth resolution error term. These are dedicated parameters;

[0086] Construct the fusion weights corresponding to remote sensing inversion:

[0087]

[0088] in This is the spectral noise error term. These are dedicated parameters;

[0089] Construct the fusion weights corresponding to land surface topography interpolation:

[0090]

[0091] in This is the interpolation sparsity error term calculated based on the known point density. These are dedicated parameters;

[0092] Using a multimodal deep learning framework for terrain data , , Feature fusion is performed to obtain the final bottom terrain data. ,in This represents a convolutional neural network that includes multiple convolutional and attention layers.

[0093] Furthermore, by integrating data on water surface area, water surface elevation, and underwater topography, the water storage capacity of the target surface water body is calculated. The specific steps are as follows:

[0094] Construct a four-dimensional hydrological coupling model;

[0095] water surface area Water surface elevation Underwater topographic data Input the data into a four-dimensional hydrological coupling model to calculate the water storage of the target surface water body:

[0096]

[0097] in, The seepage coefficient is... For vertical seepage gradient, The coefficient of volumetric thermal expansion is 1. Based on temperature changes from fused water temperature data, This is the salinity expansion coefficient. Based on changes in water salinity data, This is a timing adjustment item.

[0098] Furthermore, after determining the water storage, a water level-water storage model and an area-water storage model were constructed to conduct subsequent dynamic monitoring of the target surface water body. The specific steps are as follows:

[0099] The final water level at multiple time phases Final area Final terrain Seepage loss item and thermal expansion term As input sequences, construct a water level-water storage model:

[0100]

[0101] LSTM includes gating mechanisms (forget gate, input gate, output gate) to capture long-term dependencies and optimizes network weights by training with historical water storage data.

[0102] Construct an area-water storage model:

[0103]

[0104] in , , The coefficients are based on historical data and are updated in real time using SWOT satellite dynamic monitoring data.

[0105] A dynamic surface water storage survey system combining SWOT satellite data includes a water surface area extraction module, a water surface elevation extraction module, an underwater topography extraction module, and a water storage calculation module.

[0106] The water surface area extraction module is used to collect water surface height data and calculate the water surface area separately; it performs dynamic weight allocation and multi-source fusion sensitive to environmental factors on all the calculated water surface area results to obtain the water surface area of ​​the target surface water body;

[0107] The water surface elevation extraction module is used for water level measurement; it fuses and corrects all the calculated water level measurement data to obtain the corrected water surface elevation.

[0108] The underwater terrain extraction module is used to acquire underwater terrain data and uses a multimodal deep learning framework to perform feature-level fusion and error minimization optimization on the underwater terrain data to obtain underwater terrain data.

[0109] The water storage calculation module is used to calculate the water storage of the target surface water body by integrating water surface area, water surface elevation, and underwater topographic data.

[0110] The beneficial effects of this invention are as follows:

[0111] The proposed method for calculating dynamic surface water storage using SWOT satellite technology employs a dynamic weighted fusion model to effectively suppress the influence of environmental factors on optical imagery. Combining the all-weather, wide-coverage advantages of SAR and SWOT, the method achieves a water surface area extraction accuracy and reliability far exceeding any single data source, realizing high-precision water surface area calculation. Water level measurements are performed using GNSS-R, GNSS-IR, UAV water level gauges, and SWOT satellite monitoring technologies, and all calculated water level measurement data are fused and corrected to effectively control major error sources, achieving ultra-high precision water level measurement. A multimodal deep learning framework is used to perform feature-level fusion and error minimization optimization on underwater topographic data to obtain underwater topographic data.

[0112] This invention does not rely on any single technology. Through multi-source complementarity, even when some data is of poor quality due to weather conditions, the system can still output reliable results from other data sources such as SAR and SWOT, ensuring the continuity and stability of the survey work. It solves the problems of existing technologies, such as relying on a single observation data source and being susceptible to environmental interference, and also has… Attached Figure Description

[0113] Figure 1 This is a flowchart illustrating a method for calculating dynamic surface water storage using SWOT satellite data.

[0114] Figure 2 This is a block diagram of a survey system for dynamic surface water storage that incorporates SWOT satellite data. Detailed Implementation

[0115] like Figure 1 As shown, a method for calculating dynamic surface water storage using SWOT satellite data is described, the method comprising the following steps:

[0116] Collect water surface height data and calculate water surface area separately; perform dynamic weight allocation and multi-source fusion of all calculated water surface area results with environmental factor sensitivity to obtain the water surface area of ​​the target surface water body;

[0117] Perform water level measurements; fuse and correct all calculated water level measurement data to obtain the corrected water surface elevation;

[0118] Underwater topographic data is acquired, and a multimodal deep learning framework is used to perform feature-level fusion and error minimization optimization on the underwater topographic data to obtain underwater topographic data.

[0119] The water storage of the target surface water body is calculated by combining the water surface area, water surface elevation, and underwater topographic data.

[0120] In one specific embodiment, water surface height data is collected and water surface area is calculated. The specific steps are as follows:

[0121] Acquire four-band optical remote sensing images and calculate the NDWI index to distinguish between water bodies and non-water bodies:

[0122]

[0123] in For green band reflectivity, Near-infrared reflectance, segmented by threshold. The cumulative pixel area of ​​the region is obtained ;

[0124] Acquire three-band optical remote sensing images, and apply the SAM model to perform zero-sample segmentation of the RGB images to obtain water mask images, thus obtaining the segmented water regions:

[0125]

[0126] in For the segmentation function, Input three-band RGB optical remote sensing images;

[0127] right Water body index Spectral screening was performed, among which , , The reflectances of the green, red, and blue bands are respectively. If the area exceeds a preset threshold, the water area is confirmed, and the total area of ​​the mask pixels is calculated by integration. ;

[0128] Acquire synthetic aperture radar imagery based on SAR echo intensity Calculate OTSU automatic threshold ,Will The pixels are classified as water areas, and the pixel areas are accumulated to obtain... ;

[0129] Obtain water surface height data acquired by SWOT satellite Dynamically monitor water body boundaries and calculate the pixel area within the boundaries:

[0130] in Based on water level Edge detection function, This is the dynamic water height threshold.

[0131] In this embodiment, the specific process of optimizing the SAM model segmentation using the TSI terrain shading index includes calculating before SAM segmentation. ,in The slope angle is the terrain slope angle (calculated based on DEM). Solar zenith angle (obtained based on image time). The solar azimuth angle (obtained based on image metadata). The terrain azimuth is used (calculated based on the DEM), and the TSI value ranges from 0 to 1, representing from full shadow to full illumination. Then, TSI is applied as a correction factor to the SAM mask. If TSI < 0.5, the correction formula is as follows: To increase pixel weights in low-light areas, thereby reducing the impact of shadow interference error rates exceeding 30%, and to optimize... The dynamic weight fusion model is input to achieve accurate correction of water body segmentation under vegetation cover and complex terrain.

[0132] In one specific embodiment, considering environmental impact, the water surface area is calculated based on the aforementioned multi-source water surface observation data; the calculated water surface area results are then subjected to dynamic weight allocation and multi-source fusion sensitive to environmental factors to obtain the water surface area of ​​the target surface water body: the specific steps are as follows:

[0133] Considering the influence of clouds and fog, construct the weights corresponding to the NDWI water index method:

[0134]

[0135] in The cloud and fog coverage error term is calculated in real time based on satellite cloud images, with a range of 0-1. The default value is 0.2, a dedicated adjustment parameter for NDWI used when sensitive to cloud and fog interference;

[0136] Considering terrain influence factors, the weights corresponding to the SAM model segmentation method are constructed as follows:

[0137]

[0138] in To incorporate the terrain DEM calculation, a shadow interference error term in the range of 0-1 is included. The default value is 0.15, a dedicated adjustment parameter for SAM used for sensitive vegetation shading;

[0139] Considering meteorological influencing factors, the weights corresponding to the OTSU threshold segmentation method are constructed as follows:

[0140]

[0141] in This is a rainfall disturbance error term based on meteorological data, ranging from 0 to 1. The default value is 0.25, a dedicated adjustment parameter for SAR data that is sensitive to all-weather but coarse data;

[0142] Considering the impact factor of satellite acquisition error, construct the SWOT analysis for satellite weights:

[0143]

[0144] in To calculate the missed detection rate in small water bodies based on SWOT pixel size of 100-250 meters, the spatial resolution error term, ranging from 0-1, is used. The default value is 0.3, a dedicated adjustment parameter for SWOT analysis used in sensitive large-scale monitoring;

[0145] The surface area of ​​the target water body was calculated. .

[0146] In one specific embodiment, water level measurement is performed, and the specific steps are as follows:

[0147] Acquire GNSS signal data and use GNSS-R technology to extract preliminary water level. :

[0148]

[0149] in At the speed of light, This is the time delay of the reflected signal;

[0150] Water level was extracted using GNSS-IR technology. ;

[0151]

[0152] in, For phase difference, For signal frequency, This refers to the satellite's elevation angle;

[0153] A water surface DEM was generated using a LiDAR sensor and simultaneously scanned during the transit of the SWOT satellite. The satellite data was then aligned using the ICP algorithm.

[0154]

[0155] in For SWOT point cloud, Pointing clouds for drones, Let be a rotation matrix. The translation vector is obtained from the average water level height in the aligned DEM. ;

[0156] Obtain SWOT satellite radar echo time delay Extract water level :

[0157] .

[0158] In one specific embodiment, all calculated water level measurement data are fused and corrected to obtain the corrected water surface elevation. The specific steps are as follows:

[0159] Constructing the dynamic coefficients corresponding to GNSS-R technology:

[0160]

[0161] in This is a wave disturbance error term calculated in real time based on wind speed data, ranging from 0 to 1. The adjustment parameter is uniformly set to 0.1-0.5, with a default value of 0.3.

[0162] Constructing the dynamic coefficients corresponding to GNSS-IR:

[0163]

[0164] in The phase drift error term, in the range of 0-1, is calculated based on the signal SNR.

[0165] Construct the dynamic coefficients corresponding to the UAV water level gauge:

[0166]

[0167] in This is a positioning error term based on GPS / IMU fusion, ranging from 0 to 1.

[0168] Constructing the dynamic coefficients corresponding to SWOT satellite technology:

[0169]

[0170] in This is an atmospheric delay error term calculated based on a wet delay model, ranging from 0 to 1.

[0171] Dynamic weight allocation and multi-source fusion are performed to obtain the water surface elevation:

[0172]

[0173] Further optimization through Kalman filtering iterative steps :

[0174]

[0175] in For variance-based Kalman gain, For the current measurement, For the predicted value, the coefficients are iteratively optimized using the least squares method to achieve closed-loop correction. As the corrected water surface elevation.

[0176] In one specific embodiment, the steps for acquiring underwater topographic data are as follows:

[0177] Water depth is extracted using single / multibeam echo sounding techniques. :

[0178]

[0179] in For the speed of sound waves, To delay the echo time, a multi-beam fan-shaped scan is used to cover the underwater area;

[0180] Based on remote sensing image data Water depth was extracted using the random forest method. :

[0181]

[0182] in, For pre-trained random forest models;

[0183] Obtain land surface topographic interpolation and use the kriging method to estimate the topography. :

[0184]

[0185] in, To minimize variance weights, These are measurements from known points on the surrounding land surface. This is the global average constant.

[0186] In one specific embodiment, a multimodal deep learning framework is used to learn underwater terrain data to obtain underwater terrain data. The specific steps are as follows:

[0187] Construct fusion weights for single / multibeam bathymetry:

[0188]

[0189] in This is a depth resolution error term calculated based on the water depth range, in the range of 0-1. This is a dedicated parameter with a default value of 0.2.

[0190] Construct the fusion weights corresponding to remote sensing inversion:

[0191]

[0192] in This is a spectral noise error term in the range of 0-1, calculated based on image quality. This is a dedicated parameter with a default value of 0.15.

[0193] Construct the fusion weights corresponding to land surface topography interpolation:

[0194]

[0195] in This is the interpolation sparsity error term calculated based on known point density, ranging from 0 to 1. This is a dedicated parameter with a default value of 0.25.

[0196] Using a multimodal deep learning framework for terrain data , , Feature fusion is performed to obtain the final bottom terrain data. ,in This represents a convolutional neural network that includes multiple convolutional and attention layers.

[0197] In one specific embodiment, the water storage capacity of the target surface water body is calculated by integrating water surface area, water surface elevation, and underwater topographic data, specifically through the following steps:

[0198] Construct a four-dimensional hydrological coupling model;

[0199] water surface area Water surface elevation Underwater topographic data Input the data into a four-dimensional hydrological coupling model to calculate the water storage of the target surface water body:

[0200]

[0201] in, For dynamic queries based on a riverbed sediment type database, such as clay ,gravel The seepage coefficient was calculated using a soil permeability model. This refers to the vertical seepage gradient derived from the water level gradient based on Darcy's law. The default value is 0.00021 / °C, which is the standard value for the volumetric thermal expansion coefficient of water at around 20°C. Based on temperature changes from fused water temperature data, This is the salinity expansion coefficient. Based on changes in water salinity data, This is a time-series adjustment term for calculating seasonal fluctuations based on historical SWOT data.

[0202] In one specific embodiment, after determining the water storage capacity, a water level-water storage model and an area-water storage model are constructed to conduct subsequent dynamic monitoring of the target surface water body. The specific steps are as follows:

[0203] The final water level at multiple time phases Final area Final terrain Seepage loss item and thermal expansion term As input sequences, construct a water level-water storage model:

[0204]

[0205] LSTM includes gating mechanisms (forget gate, input gate, output gate) to capture long-term dependencies and optimizes network weights by training with historical water storage data.

[0206] Construct an area-water storage model:

[0207]

[0208] in , , The fitting coefficients are calculated from measured water storage samples using the least squares method based on historical data, and are updated in real time using SWOT satellite dynamic monitoring data.

[0209] Example 2

[0210] In this embodiment, a surface water storage survey was conducted on a real reservoir, using the method of this invention combined with multi-source data (such as optical remote sensing images, SAR images, SWOT satellite data, GNSS signals, UAV LiDAR, and depth sounder data). The survey was conducted during the flood season of 2024, under environmental conditions including 40% cloud cover, 30% shadow interference, and 20% rainfall interference, with a SWOT satellite resolution of approximately 150m.

[0211] First, four-band optical remote sensing images (10m resolution) were acquired, and the preliminary water surface area A1 was extracted using the NDWI water index method. The calculation formula is as follows: Where G is the green band reflectance (average 0.12) and NIR is the near-infrared band reflectance (average 0.08). By segmenting the region where NDWI>0 by thresholding, the cumulative pixel area is A1≈104.25km² (with an error of about 4.25% due to cloud and fog).

[0212] Secondly, the initial water surface area A2 was extracted from the three-band RGB imagery (5m resolution) using the SAM model segmentation method. Zero-sample segmentation of the RGB imagery was performed using the SAM model to obtain the water mask image M=SAM(I), which was further combined with water indexes. (G≈0.15, R≈0.10, B≈0.08) Spectral screening is performed. If WaterIndex>0.1, the water area is confirmed. The total area of ​​the mask pixels is calculated by integration to obtain A2≈105.21km².

[0213] Then, the SAR image (resolution 20m) was used to extract the initial water surface area A3 using the OTSU thresholding method. Based on the SAR echo intensity pixel (average -15dB), the OTSU automatic threshold T_{OTSU}≈-12dB was calculated, and pixels with pixel > T_{OTSU} were classified as water areas. The accumulated pixel area was A3≈112.20km² (all-weather but coarse due to rainfall).

[0214] Finally, the preliminary water surface area A4 was extracted from the SWOT satellite data (height resolution 10cm) using SWOT techniques. The water body boundary was dynamically monitored using water surface height data h(t) (average 4.8m), and the contour extraction formula was used. Calculate the pixel area within the boundary, where h_{threshold}=mean(h(t))-std(h(t))≈4.5m, and obtain A4≈120.06km² (15% false negative rate for large scale but small water body).

[0215] A dynamic weighted fusion model is adopted, and the weights are dynamically adjusted according to meteorological conditions: cloud and fog coverage error. =0.40 (σ1=0.2), shadow interference =0.30 (σ²=0.15), rainfall disturbance =0.20 (σ3=0.25), spatial resolution =0.15 (σ4=0.3), we get w1≈0.18, w2≈0.22, w3≈0.30, w4≈0.30. The final water surface area A_{final}=w1·A1+w2·A2+w3·A3+w4·A4≈114.27km², showing an error of 14.27%, which is better than the average of ~20% for individual methods.

[0216] Using GNSS-R technology, the time delay of the reflected signal is analyzed to be approximately τ≈33.4ns (speed of light c=3×10^8m / s), according to the formula. ≈4.77m (wave interference error 25%). GNSS-IR technology measures a phase difference Δφ≈1.2rad (frequency f=1.575GHz, elevation angle e=30°), formula... ≈4.61m (phase drift error 18%). The UAV level gauge uses LiDAR to generate a water surface DEM, aligns the SWOT point cloud using the ICP algorithm, and minimizes... The average height was obtained ≈4.81m (positioning error 8%). SWOT satellite radar echo time delay τ≈33.4ns, formula ≈4.01m (atmospheric delay error 12%).

[0217] Construct a four-source data closed-loop correction system, with dynamic coefficients using the softmax form: =0.25, =0.18, =0.08, =0.12 (κ=0.3), ≈0.20, β≈0.25, δ≈0.35, γ≈0.20. Preliminary fusion. ≈4.55m, further optimized through Kalman filtering iteration, formula (K=0.5 based on variance, h_{pred}=5.10m), the final water level height is obtained as h_{corrected}≈4.82m, with an error of 3.6%, which shows that the effects of centimeter-level waves and delays have been resolved.

[0218] Single / multi-beam depth sounding is used to transmit acoustic waves, with an echo time delay of t≈4ms, according to the formula. ≈3.21m.

[0219] Remote sensing inversion was used to predict from image data X (feature vector [reflectivity, texture]) using the random forest method, with the formula z2=RF(X)≈2.94m.

[0220] Kriging is used for interpolation based on land surface topography. The formula is... The weights λ_i are based on autocorrelation, μ=3.0, so z3≈3.33m.

[0221] A multimodal deep learning framework is used for fusion, with weights... ( =0.10, η1=0.2)≈0.35, λ2≈0.30, λ3≈0.35. The final bottom terrain data is obtained by fusing these values. The accuracy is approximately 3.19m with an error of 6.3%, demonstrating improved precision in complex environments.

[0222] Water storage was calculated using a four-dimensional hydrological coupling model. , where γ=0.10 (gravel type). q / z = 0.05 (Darcy's Law), ε = 0.00021 / °C, ΔT = 5°C (SWOT water temperature data), β = 0.0008 / psu, ΔS = 2psu (salinity sensor), ΔV_{time} = 10 (historical SWOT seasonal fluctuations). The calculated seepage loss is approximately 0.01, heat / salinity correction is approximately 0.999, V_base is approximately 114.27 × (4.82 - 3.19) ≈ 186.26, and V is approximately 195.34 million cubic meters, with an error of 2.33%. This demonstrates that the spurious change problem has been resolved.

[0223] Further, a water level-water storage model is constructed, and time-series data is processed through an LSTM neural network with 64 hidden layers and MSE loss. The input includes the final water level. Final area Final terrain Seepage loss item and thermal expansion term ,formula The area-water storage model is used to predict the trend of changes during the flood season (using multinomial regression). The 4D-HCM optimization coefficient is linked.

[0224] The results of this embodiment show that the method of the present invention significantly improves the accuracy of water storage estimation and is suitable for water resource management in complex environments.

[0225] Example 3

[0226] like Figure 2 As shown, a dynamic surface water storage survey system combining SWOT satellite data includes a water surface area extraction module, a water surface elevation extraction module, an underwater topography extraction module, and a water storage calculation module.

[0227] The water surface area extraction module is used to collect water surface height data and calculate the water surface area separately; it performs dynamic weight allocation and multi-source fusion sensitive to environmental factors on all the calculated water surface area results to obtain the water surface area of ​​the target surface water body;

[0228] The water surface elevation extraction module is used for water level measurement; it fuses and corrects all the calculated water level measurement data to obtain the corrected water surface elevation.

[0229] The underwater terrain extraction module is used to acquire underwater terrain data and uses a multimodal deep learning framework to perform feature-level fusion and error minimization optimization on the underwater terrain data to obtain underwater terrain data.

[0230] The water storage calculation module is used to calculate the water storage of the target surface water body by integrating water surface area, water surface elevation, and underwater topographic data.

[0231] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

Claims

1. A method for calculating dynamic surface water storage using SWOT satellite data, characterized in that: The method includes the following steps: Collect water surface height data and calculate water surface area separately; perform dynamic weight allocation and multi-source fusion of all calculated water surface area results with environmental factor sensitivity to obtain the water surface area of ​​the target surface water body; Perform water level measurements; fuse and correct all calculated water level measurement data to obtain the corrected water surface elevation; Underwater topographic data is acquired, and a multimodal deep learning framework is used to perform feature-level fusion and error minimization optimization on the underwater topographic data to obtain underwater topographic data. Based on the combined data of water surface area, water surface elevation, and underwater topography, the water storage of the target surface water body is calculated. The specific steps for collecting water surface height data and calculating the water surface area are as follows: Acquire four-band optical remote sensing images and calculate the NDWI index to distinguish between water bodies and non-water bodies: in For green band reflectivity, Near-infrared reflectance, segmented by threshold. The cumulative pixel area of ​​the region is obtained ; Acquire three-band optical remote sensing images, and apply the SAM model to perform zero-sample segmentation of the RGB images to obtain water mask images, thus obtaining the segmented water regions: in For the segmentation function, Input three-band RGB optical remote sensing images; right Water body index Spectral screening was performed, among which , , The reflectances of the green, red, and blue bands are respectively. If the area exceeds a preset threshold, the water area is confirmed, and the total area of ​​the mask pixels is calculated by integration. ; Acquire synthetic aperture radar imagery based on SAR echo intensity Calculate OTSU automatic threshold ,Will The pixels are classified as water areas, and the pixel areas are accumulated to obtain... ; Obtain water surface height data acquired by SWOT satellite Dynamically monitor water body boundaries and calculate the pixel area within the boundaries: in Based on water level Edge detection function, This refers to the dynamic water body height threshold. The specific steps for calculating the water surface area of ​​the target surface water body, considering environmental impact and based on multi-source water surface observation data, are as follows: [Details of the steps are missing from the provided text.] Considering the influence of clouds and fog, construct the weights corresponding to the NDWI water index method: in This is the cloud and fog coverage error term. NDWI specific adjustment parameters to account for sensitive cloud and fog interference; Considering terrain influence factors, the weights corresponding to the SAM model segmentation method are constructed as follows: in This is the shadow interference error term. SAM-specific adjustment parameters to account for sensitive vegetation shading; Considering meteorological influencing factors, the weights corresponding to the OTSU threshold segmentation method are constructed as follows: in For the rainfall interference error term, SAR-specific adjustment parameters are designed to take into account sensitive, all-weather coarse data; Considering the impact factor of satellite acquisition error, construct the SWOT analysis for satellite weights: in This is the spatial resolution error term. SWOT-specific adjustment parameters are provided to consider monitoring scale and sensitivity. The surface area of ​​the target water body was calculated as follows: .

2. The method for calculating dynamic surface water storage using SWOT satellite data as described in claim 1, characterized in that: The specific steps for measuring water level are as follows: Acquire GNSS signal data and use GNSS-R technology to extract preliminary water level. : ， in At the speed of light, This is the time delay of the reflected signal; Water level was extracted using GNSS-IR technology. ; in, For phase difference, For signal frequency, This refers to the satellite's elevation angle; A water surface DEM was generated using a LiDAR sensor and simultaneously scanned during the transit of the SWOT satellite. The satellite data was then aligned using the ICP algorithm. in For SWOT point cloud, Pointing clouds for drones, Let be a rotation matrix. The translation vector is obtained from the average water level height in the aligned DEM. ; Obtain SWOT satellite radar echo time delay Extract water level : 。 3. The method for calculating dynamic surface water storage using SWOT satellite data as described in claim 2, characterized in that: The specific steps for fusing and correcting all calculated water level measurement data to obtain the corrected water surface elevation are as follows: Constructing the dynamic coefficients corresponding to GNSS-R technology: in This is the wave interference error term. To standardize the adjustment parameters; Constructing the dynamic coefficients corresponding to GNSS-IR: in This is the phase drift error term; Construct the dynamic coefficients corresponding to the UAV water level gauge: in This is the positioning error term; Constructing the dynamic coefficients corresponding to SWOT satellite technology: in This is the atmospheric delay error term; Dynamic weight allocation and multi-source fusion are performed to obtain the water surface elevation: Further optimization through Kalman filtering iterative steps : in For variance-based Kalman gain, For the current measurement, For the predicted value, the coefficients are iteratively optimized using the least squares method to achieve closed-loop correction. As the corrected water surface elevation.

4. The method for calculating dynamic surface water storage using SWOT satellite data as described in claim 1, characterized in that: The specific steps for obtaining underwater topographic data are as follows: Water depth is extracted using single / multibeam echo sounding techniques. : in For the speed of sound waves, To delay the echo time, a multi-beam fan-shaped scan is used to cover the underwater area; Based on remote sensing image data Water depth was extracted using the random forest method. : in, For pre-trained random forest models; Obtain land surface topographic interpolation and use the kriging method to estimate the topography. : in, To minimize variance weights, These are measurements from known points on the surrounding land surface. This is the global average constant.

5. The method for calculating dynamic surface water storage using SWOT satellite data as described in claim 4, characterized in that: The underwater topography data is obtained by learning underwater topography data using a multimodal deep learning framework. The specific steps are as follows: Construct fusion weights for single / multibeam bathymetry: in This is the depth resolution error term. These are dedicated parameters; Construct the fusion weights corresponding to remote sensing inversion: in This is the spectral noise error term. These are dedicated parameters; Construct the fusion weights corresponding to land surface topography interpolation: in This is the interpolation sparsity error term calculated based on the known point density. These are dedicated parameters; Using a multimodal deep learning framework for terrain data , , Feature fusion is performed to obtain the final bottom terrain data. ,in This represents a convolutional neural network that includes multiple convolutional and attention layers.

6. The method for calculating dynamic surface water storage using SWOT satellite data as described in claim 1, characterized in that: Based on comprehensive data on water surface area, water surface elevation, and underwater topography, calculate and determine the water storage of the target surface water body. The specific steps are as follows: Construct a four-dimensional hydrological coupling model; water surface area Water surface elevation Underwater topographic data Input the data into a four-dimensional hydrological coupling model to calculate the water storage of the target surface water body: in, The seepage coefficient is... For vertical seepage gradient, The coefficient of volumetric thermal expansion is 1. Based on temperature changes from fused water temperature data, This is the salinity expansion coefficient. Based on changes in water salinity data, This is a timing adjustment item.

7. The method for calculating dynamic surface water storage using SWOT satellite data as described in claim 6, characterized in that: After determining the water storage, a water level-water storage model and an area-water storage model were constructed for subsequent dynamic monitoring of the target surface water body. The specific steps are as follows: The final water level at multiple time phases Final area Final terrain Seepage loss item and thermal expansion term As input sequences, construct a water level-water storage model: LSTM includes gating mechanisms: forget gate, input gate, and output gate, to capture long-term dependencies and optimize network weights by training with historical water storage data. Construct an area-water storage model: in , , The coefficients are based on historical data and are updated in real time using SWOT satellite dynamic monitoring data.

8. A survey system for dynamic surface water storage combined with SWOT satellite data, employing the calculation method for dynamic surface water storage combined with SWOT satellite data as described in any one of claims 1-7, characterized in that: It includes modules for extracting water surface area, extracting water surface elevation, extracting underwater topography, and calculating water storage capacity. The water surface area extraction module is used to collect water surface height data and calculate the water surface area separately; it performs dynamic weight allocation and multi-source fusion sensitive to environmental factors on all the calculated water surface area results to obtain the water surface area of ​​the target surface water body; The water surface elevation extraction module is used for water level measurement; it fuses and corrects all the calculated water level measurement data to obtain the corrected water surface elevation. The underwater terrain extraction module is used to acquire underwater terrain data and uses a multimodal deep learning framework to perform feature-level fusion and error minimization optimization on the underwater terrain data to obtain underwater terrain data. The water storage calculation module is used to calculate the water storage of the target surface water body by integrating water surface area, water surface elevation, and underwater topographic data.

Images