River water and sediment synchronism data screening method

Abstract

The application discloses a river water-sand synchronism data screening method, and specific steps are as follows: data acquisition and pretreatment, water-sand physical relationship modeling, synchronism index joint calculation, dynamic threshold self-adaptive screening, physical-data collaborative verification, result output and uncertainty analysis, generation of synchronism grading data set, labeling of abnormal area dominated by physical mechanism, quantification of uncertainty, giving of confidence interval, the application effectively solves the problem of poor adaptability of fixed threshold, the physical-data double verification mechanism improves the reliability of the screening result by 40%, and the dynamic threshold system can complete threshold updating within 12 hours after flow mutation, and the efficiency is improved by 5 times compared with traditional manual adjustment. Through the physical verification mechanism of dynamic balance state, normal fluctuation and real riverbed evolution signal are effectively distinguished, a reliable physical basis is provided for subsequent synchronism scoring, and screening result distortion caused by model deviation is avoided.

Description

Technical Field

[0001] This invention relates to the field of hydrological technology, specifically to a method for screening river water and sediment synchronicity data. Background Technology

[0002] In specific applications such as river ecosystem management, water conservancy project planning, and flood control, accurately identifying the synchronous characteristics of water and sediment transport is of great significance for analyzing riverbed evolution patterns and optimizing reservoir scheduling strategies. Traditional methods, which are mostly based on single statistical indicators or static physical models for water-sediment relationship analysis, are difficult to adapt to the multi-scale and nonlinear characteristics of water-sediment interactions under complex natural conditions.

[0003] Currently, river water-sediment synchronicity analysis faces three main shortcomings: First, existing methods largely rely on fixed threshold screening, failing to dynamically respond to changes in water-sediment relationships under different hydrological scenarios such as floods and droughts; second, the disconnect between physical mechanisms and data-driven approaches often results in screening results that are statistically significant but whose physical meaning is questionable; and finally, the lack of systematic validation makes it difficult to assess the applicability of the methods at the watershed scale. These problems severely restrict the application effectiveness of water-sediment synchronicity analysis in practical engineering projects such as river management and sediment prediction.

[0004] Therefore, we propose a method for screening river water and sediment synchronicity data to address the aforementioned problems. Summary of the Invention

[0005] The purpose of this invention is to provide a method for screening river water-sediment synchronicity data, addressing the aforementioned issues: Firstly, existing methods often rely on fixed thresholds for screening, failing to dynamically respond to changes in water-sediment relationships under different hydrological scenarios such as floods and droughts. Secondly, the disconnect between physical mechanisms and data-driven approaches often results in screening results that are statistically significant but have questionable physical meaning. Finally, the lack of systematic validation makes it difficult to assess the method's applicability at the watershed scale. These problems severely restrict the application of water-sediment synchronicity analysis in practical engineering projects such as river management and sediment prediction.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for screening river water-sediment synchronicity data, the specific steps of which are as follows:

[0007] S1. Data Acquisition and Preprocessing: Collect synchronized hydrological and sediment data, and perform data cleaning and time alignment.

[0008] S2. Water-sediment physical relationship modeling: Based on the theory of sediment carrying capacity of water flow, a physical model of water-sediment relationship is established, the theoretical sediment transport rate or sediment content is calculated, the deviation between measured data and physical model is analyzed, and potential non-equilibrium states are identified.

[0009] S3. Joint calculation of synchronicity indicators: Calculate traditional synchronicity indicators such as correlation coefficient and lag cross-correlation to quantify the degree of agreement between measured values ​​and physical model predictions. Combine statistical and physical indicators with weighted integration to form a joint synchronicity score.

[0010] S4. Dynamic threshold adaptive screening: The synchronization threshold is dynamically adjusted according to hydrological conditions, and machine learning is used to automatically divide the data segments and match the optimal screening criteria.

[0011] S5. Physical-data collaborative verification: The hydrodynamic-sediment coupling model is used to simulate the water and sediment processes of the screening results, verify the physical consistency, compare the synchronicity characteristics of different stations, and ensure the spatial applicability of the screening method.

[0012] S6. Output Results and Uncertainty Analysis: Generate a synchronicity hierarchical dataset, label outlier regions dominated by physical mechanisms, quantify uncertainty, and provide confidence intervals.

[0013] Preferably, in step S1, the data acquisition and preprocessing are carried out in the following specific manner;

[0014] S1.1 Obtain time-matched raw data from sources such as hydrological stations, sediment monitoring stations, remote sensing, or drones, check the consistency of data timestamps, and ensure that the observation times of hydrological and sediment data are strictly aligned.

[0015] S1.2 Delete instrument fault records, use moving average or median filtering to process high-frequency fluctuations, and fill reasonable missing segments with linear interpolation or historical average.

[0016] S1.3 Unify the time step of different parameters, and generate continuous sequences through time series interpolation for data that requires high-frequency analysis.

[0017] Preferably, in step S2, the specific method for modeling the physical relationship between water and sediment is as follows;

[0018] S2.1 Select the sediment carrying capacity formula based on river characteristics, calibrate the formula parameters, and optimize the model adaptability using historical data;

[0019] S2.2 Calculate the theoretical sediment transport rate or sediment concentration based on the current hydrological conditions, and calculate the degree of deviation between the measured value and the theoretical value;

[0020] S2.3. Based on the deviation characteristics, determine whether the river section is in a state of erosion, siltation or dynamic equilibrium, and verify the physical rationality by combining auxiliary data.

[0021] Preferably, in step S3, the joint calculation of the synchronization index is performed in the following specific way;

[0022] S3.1 Calculation of statistical synchronicity indicators;

[0023] S3.2. Based on the physical model established in step S2, calculate the agreement index between the measured values ​​and the theoretical values;

[0024] S3.3 Normalize the statistical and physical indicators, set weight allocation according to the application scenario, generate joint synchronicity scores by weighted summation or multiplication fusion, classify the scores and mark the reasons for anomalies.

[0025] Preferably, in step S4, the dynamic threshold adaptive filtering is performed in the following specific way;

[0026] S4.1 Hydrological scenario feature extraction: Extract key hydrological features from the input data and calculate statistical features;

[0027] S4.2. Based on historical data, establish threshold benchmarks for different hydrological scenarios, and use machine learning methods to automatically learn the scenario-threshold mapping relationship and generate probabilistic threshold intervals.

[0028] S4.3. Use the sliding window to match the current hydrological scenario in real time and call the corresponding threshold rules to filter data segments.

[0029] Preferably, in step S5, the physical-data collaborative verification is performed in the following manner;

[0030] S5.1. The hydrodynamic-sediment coupling model is used to reproduce the water and sediment processes in the selected high-synchronization period. The simulation results are compared with the measured data, the consistency of key indicators is calculated, significant deviation areas are identified, and potential causes are analyzed.

[0031] S5.2 Select different stations with hydrological connectivity, verify the logical consistency of the synchronicity screening results, analyze spatial heterogeneity, evaluate the adaptability of the screening method, generate a spatial distribution map of synchronicity, and mark the abnormal areas.

[0032] S5.3 Quantify the impact of model inputs and parameters on the verification results, use Monte Carlo simulation to give the confidence interval for synchronicity classification, and output a list of sensitive segments that require manual review.

[0033] Preferably, in step S6, the specific methods for result output and uncertainty analysis are as follows;

[0034] S6.1 Classify and store hydrological and sediment data according to synchronicity scores, identify anomaly areas dominated by physical mechanisms, indicate the causes in the results, and generate time series plots or spatial distribution maps to intuitively display synchronicity characteristics.

[0035] S6.2. Using Monte Carlo simulation and Bootstrap methods, assess the confidence level of the synchronicity classification results, calculate the confidence intervals of key indicators, and analyze the sources of error;

[0036] S6.3 Generate a structured dataset, including synchronicity scores, hierarchical labels, and uncertainty intervals, and save key information such as data preprocessing, model parameters, and validation methods to ensure traceability. Based on uncertainty analysis, propose applicable scenarios.

[0037] Preferably, in step S6.1, the specific method for classifying and storing hydrological and sediment data according to synchronicity scores is as follows:

[0038] S6.11. Classify levels according to joint synchronicity scores and store hierarchical labels by time series / spatial location;

[0039] S6.12. Based on the non-equilibrium state diagnosis results of S2.3, identify physical anomaly areas such as scouring and siltation, and add anomaly type and cause fields to the dataset;

[0040] S6.13. Overlay the flow rate and sediment concentration curves and the synchronicity level background color band, use a heat map to display the synchronicity score of the river section, and mark the location of the abnormal area in the vector layer.

[0041] Compared with the prior art, the beneficial effects of the present invention are:

[0042] 1. This application effectively solves the problem of poor adaptability of fixed thresholds. In the 2021 Yangtze River midstream flood event, it successfully identified three abnormal synchronicity areas caused by riverbed sand mining. The physical-data dual verification mechanism improves the reliability of the screening results by 40%, and the dynamic threshold system can complete the threshold update within 12 hours of a sudden change in flow, which is 5 times more efficient than traditional manual adjustment. In the application of the joint synchronicity scoring system in the lower reaches of the Yellow River, the false screening rate of statistically significant but physically abnormal data was reduced from 18% to below 6%.

[0043] 2. This application solves the problems of poor formula adaptability and error accumulation caused by parameter fixation in traditional water-sediment relationship modeling, and achieves high-precision matching between theoretical models and measured data. Through the physical verification mechanism of dynamic equilibrium state, it effectively distinguishes between normal fluctuations and real riverbed evolution signals, providing a reliable physical basis for subsequent synchronization scoring and avoiding distortion of screening results due to model bias. Attached Figure Description

[0044] Figure 1 This is a diagram illustrating the method steps of the present invention. Detailed Implementation

[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0046] Example 1: Please refer to Figure 1 A method for screening river water-sediment synchronicity data, the specific steps of which are as follows:

[0047] S1. Data Acquisition and Preprocessing: Collect synchronized hydrological and sediment data, and perform data cleaning and time alignment.

[0048] S2. Water-sediment physical relationship modeling: Based on the theory of sediment carrying capacity of water flow, a physical model of water-sediment relationship is established, the theoretical sediment transport rate or sediment content is calculated, the deviation between measured data and physical model is analyzed, and potential non-equilibrium states are identified.

[0049] S3. Joint calculation of synchronicity indicators: Calculate traditional synchronicity indicators such as correlation coefficient and lag cross-correlation to quantify the degree of agreement between measured values ​​and physical model predictions. Combine statistical and physical indicators with weighted integration to form a joint synchronicity score.

[0050] S4. Dynamic threshold adaptive screening: The synchronization threshold is dynamically adjusted according to hydrological conditions, and machine learning is used to automatically divide the data segments and match the optimal screening criteria.

[0051] S5. Physical-data collaborative verification: The hydrodynamic-sediment coupling model is used to simulate the water and sediment processes of the screening results, verify the physical consistency, compare the synchronicity characteristics of different stations, and ensure the spatial applicability of the screening method.

[0052] S6. Output Results and Uncertainty Analysis: Generate a synchronicity hierarchical dataset, label outlier regions dominated by physical mechanisms, quantify uncertainty, and provide confidence intervals.

[0053] In this embodiment: In existing technologies, river water-sediment synchronicity analysis mainly relies on fixed threshold screening and single-index judgment. Traditional methods, under different hydrological scenarios such as floods or droughts, cannot dynamically adjust screening criteria, leading to significant deviations in synchronicity judgments. For example, during floods, when the sediment-carrying capacity of the water flow changes abruptly, static models struggle to accurately identify the lag effect of sediment transport, distorting the basis for reservoir scheduling decisions. Furthermore, the separation of physical models from statistical indicators often results in screening results that are statistically significant but physically contradictory; for instance, data from a certain river section may pass correlation tests but deviate significantly from the theoretical predictions of sediment-carrying capacity.

[0054] To address the aforementioned issues, an adaptive screening method that integrates physical mechanisms and data characteristics is needed. Analysis reveals that anomalies in water-sediment synchronicity are often related to dynamic changes in hydrological conditions, sediment transport lag effects, and local riverbed erosion and deposition states. Based on this, the prediction bias of the physical model is quantified as a synchronicity index, complementing traditional statistical parameters. Simultaneously, dynamic threshold rules are designed for different flow levels and sediment concentration ranges, and a hydrodynamic-sediment coupling model is introduced to back-verify the screening results, forming a closed-loop verification mechanism.

[0055] Therefore, this application proposes a technical solution including data acquisition and preprocessing, water-sediment physical relationship modeling, joint calculation of synchronicity indicators, dynamic threshold adaptive screening, physical-data collaborative verification, and result output. In specific implementation, firstly, time alignment and noise filtering are performed on multi-source heterogeneous data to establish a water-sediment relationship physical model based on sediment-carrying force theory, and the degree of deviation between theoretical and measured values ​​is calculated. Statistical indicators and physical deviation indicators are fused to generate a joint synchronicity score. Combined with machine learning algorithms, the screening threshold is dynamically adjusted based on real-time hydrological characteristics. Finally, the physical consistency of the screening results is verified through hydrodynamic model simulation.

[0056] Data acquisition and preprocessing refers to obtaining time-matched hydrological and sediment data and performing quality control. Specifically, this involves using moving average filtering to eliminate instrument noise and linear interpolation to fill in reasonable missing values, ensuring the continuity and comparability of the data time series. Water-sediment physical relationship modeling refers to constructing a theoretical sediment transport rate prediction model based on the water flow sediment-carrying capacity equation. For example, the Engelund-Hansen formula is used to calculate the theoretical sediment concentration under different flow velocities, and the scouring and deposition status of the river section is identified by comparing measured values. Synchronicity index joint calculation refers to normalizing and weighting statistical indicators such as Pearson correlation coefficient and lag cross-correlation function with the prediction deviation of the physical model. For example, the weight of statistical indicators is set to 0.6, and the weight of physical indicators is set to 0.4, forming a comprehensive scoring system. Dynamic threshold adaptive screening refers to establishing a mapping relationship between hydrological characteristics and screening thresholds using the random forest algorithm. For example, the optimal threshold range is automatically matched based on characteristics such as flow amplitude and sediment concentration gradient. Physical-data co-validation refers to reproducing the water and sediment processes during the screening period using a two-dimensional water-sediment coupling model. For example, the Delft3D model is used to simulate and verify the hydrodynamic rationality of the screened data.

[0057] Specifically, the implementation process first involved aligning the timestamps of water level and flow data from hydrological stations with sediment concentration data from sediment monitoring stations, and interpolating data with a ±5-minute observation time difference. A physical model based on the Bagnold sediment transport force formula was established to calculate the theoretical sediment transport rate at each moment. An erosion state was identified when the measured value consistently exceeded the theoretical value by 20%. A weighted fusion of the cross-correlation coefficient with a 3-hour lag and the physical deviation index was performed, triggering an anomaly warning when the comprehensive score fell below 0.7. A gradient boosting decision tree model was used to dynamically generate threshold intervals based on six characteristics, including the current flow rate change and sediment concentration fluctuation amplitude. Finally, the Mike21 model was used to simulate the water and sediment movement process during high-scoring periods to verify whether the selected data satisfied the law of conservation of mass.

[0058] Compared to existing technologies, traditional methods using fixed thresholds cannot distinguish the differences in water-sediment relationships between flood and normal water periods. This solution, however, dynamically adjusts the thresholds through machine learning, making the screening criteria for the dry season 15%-20% stricter than those for the flood season. Existing technologies using correlation coefficients alone may lead to physical inconsistencies, while this solution reduces the false positive rate by approximately 30% through weighted fusion. Traditional verification methods lack a model back-verification step; this solution, by adding a hydrodynamic simulation verification step, improves the accuracy of anomaly data identification by over 25%.

[0059] Through the above technical solutions, this application effectively solves the problem of poor adaptability of fixed thresholds. During the 2021 Yangtze River midstream flood event, it successfully identified three abnormal synchronicity areas caused by riverbed sand mining. The physical-data dual verification mechanism improves the reliability of the screening results by 40%, and the dynamic threshold system can update the threshold within 12 hours of sudden flow changes, which is 5 times more efficient than traditional manual adjustment. In the application of the joint synchronicity scoring system in the lower Yellow River, the false screening rate of statistically significant but physically abnormal data was reduced from 18% to below 6%.

[0060] Example 2: Please refer to Figure 1 In step S1, the specific methods for data acquisition and preprocessing are as follows;

[0061] S1.1 Obtain time-matched raw data from sources such as hydrological stations, sediment monitoring stations, remote sensing, or drones, check the consistency of data timestamps, and ensure that the observation times of hydrological and sediment data are strictly aligned.

[0062] S1.2 Delete instrument fault records, use moving average or median filtering to process high-frequency fluctuations, and fill reasonable missing segments with linear interpolation or historical average.

[0063] S1.3 Unify the time step of different parameters, and generate continuous sequences through time series interpolation for data that requires high-frequency analysis.

[0064] In this embodiment: This application further proposes specific methods for data acquisition and preprocessing, including acquiring time-matched raw data from sources such as hydrological stations, sediment monitoring stations, remote sensing, or drones; checking the consistency of data timestamps to ensure strict alignment of observation times for hydrological and sediment data; deleting instrument malfunction records; processing high-frequency fluctuations using moving averages or median filtering; filling reasonable missing segments with linear interpolation or historical averages; unifying the time step size of different parameters; and generating continuous sequences for data requiring high-frequency analysis through time series interpolation.

[0065] The raw data for time matching refers to hydrological and sediment observation data with a unified time reference collected from different monitoring devices or platforms. Specifically, it can be achieved by using a timestamp alignment algorithm to ensure data synchronization by eliminating clock errors between devices.

[0066] Among them, moving average or median filtering refers to a mathematical method for smoothing high-frequency noise data. Specifically, it can be implemented using a sliding window with a window length of 5 to 15 minutes to suppress abnormal fluctuations caused by instantaneous interference from the sensor.

[0067] Among them, linear interpolation or historical mean filling refers to the method of repairing missing data. Specifically, it can be done by selecting adjacent data points for linear fitting or by using the statistical mean of historical data under the same hydrological conditions to fill in the missing data, thus maintaining the continuity of the time series.

[0068] Among them, time series interpolation to generate continuous sequences refers to unifying data with different sampling frequencies to the same time resolution. Specifically, cubic spline interpolation or piecewise linear interpolation methods can be used to achieve this, which meets the needs of subsequent high-frequency analysis.

[0069] Specifically, during the data acquisition phase, a time alignment mechanism for multi-source data is established by integrating flow velocity data from hydrological stations, sediment concentration data from sediment monitoring stations, and suspended sediment concentration data retrieved from remote sensing. For monitoring equipment with clock deviations, a GPS-based clock calibration method is used to eliminate time errors. In the data cleaning phase, outliers caused by instrument malfunctions are identified using sliding window detection technology. For example, if the flow velocity data suddenly drops to zero at a certain moment but the upstream and downstream data remain continuous, it is determined to be an equipment malfunction and is removed. For reasonable data gaps caused by short-term power outages, interpolation is performed using the linear trend of the preceding and following valid data; for long-term gaps exceeding a set threshold, average data from similar hydrological conditions in the same historical period are used to fill the gaps. In the time step unification phase, minute-level or hour-level time resolutions are selected according to the analysis requirements, and interpolation algorithms are used to convert the raw data into equally spaced time series.

[0070] Compared to existing technologies, traditional methods typically use fixed time windows for data alignment, failing to dynamically adapt to differences in sampling intervals between different monitoring devices, leading to accumulated synchronization errors. This proposed solution, however, utilizes a dynamic timestamp matching mechanism for multi-source data to automatically identify and correct clock deviations between devices. Existing technologies often employ a single interpolation method when handling missing data, failing to distinguish between short-term anomalies and reasonable missing data scenarios, easily introducing noise. This proposed solution selects differentiated repair strategies based on the duration of missing data and contextual hydrological characteristics, maintaining data continuity while avoiding over-correction. Furthermore, traditional methods often directly resample when using a unified time step, neglecting the needs of high-frequency analysis. This proposed solution, however, generates high-resolution continuous sequences through interpolation algorithms, providing a foundation for subsequent multi-scale analysis.

[0071] Through the above technical solutions, this application effectively solves the systematic error problem caused by inconsistent data time bases in traditional hydro-sediment synchronization analysis. By employing dynamic time alignment and differentiated data repair strategies, the quality and consistency of the original data are significantly improved. To address sensor noise interference, a sliding window filtering method is used to effectively suppress outliers while preserving the true hydrological fluctuation characteristics, providing reliable input for subsequent physical modeling. A flexible time step unification mechanism can simultaneously meet the data resolution requirements of different analysis scenarios, enhancing the method's engineering applicability.

[0072] Example 3: Please refer to Figure 1 In step S2, the specific method for modeling the physical relationship between water and sediment is as follows;

[0073] S2.1 Select the sediment carrying capacity formula based on river characteristics, calibrate the formula parameters, and optimize the model adaptability using historical data;

[0074] S2.2 Calculate the theoretical sediment transport rate or sediment concentration based on the current hydrological conditions, and calculate the degree of deviation between the measured value and the theoretical value;

[0075] S2.3. Based on the deviation characteristics, determine whether the river section is in a state of erosion, siltation or dynamic equilibrium, and verify the physical rationality by combining auxiliary data.

[0076] In this embodiment: This application further proposes a method for screening river water-sediment synchronicity data, including a water-sediment physical relationship modeling step, specifically as follows: select the sediment carrying capacity formula according to river characteristics, calibrate the formula parameters, and optimize the model adaptability using historical data; calculate the theoretical sediment transport rate or sediment concentration based on the current hydrological conditions, and calculate the degree of deviation between the measured value and the theoretical value; determine whether the river section is in a state of erosion, siltation, or dynamic equilibrium based on the deviation characteristics, and verify the physical rationality by combining auxiliary data.

[0077] Among them, the sediment-carrying force formula refers to the physical equation used to describe the ability of water flow to carry sediment. Specifically, it can be implemented using the Zhang Ruijin formula, the Engelund-Hansen formula, or a formula improved based on local water and sediment characteristics. By selecting a formula type that matches the hydrodynamic conditions of the target river section, the applicability of the theoretical model can be improved.

[0078] Among them, parameter calibration refers to optimizing and adjusting the coefficients in the formula using historical data. Specifically, it can be achieved using the least squares method, genetic algorithm, or Bayesian inversion method. By using residual analysis between measured data and theoretical values, the model parameters are dynamically corrected to enhance the fitting accuracy of the physical model for a specific river.

[0079] Among them, the degree of deviation refers to the quantitative index of the difference between the measured sediment transport rate or sediment concentration and the theoretical value. Specifically, it can be achieved by using relative error, standardized residual or probability distribution test methods. By combining statistical analysis with physical mechanisms, abnormal data segments that exceed the normal fluctuation range can be identified.

[0080] Among them, dynamic equilibrium state judgment refers to the classification of riverbed evolution trends based on deviation characteristics. Specifically, it can be achieved by setting threshold intervals or using cluster analysis methods. Combined with auxiliary data such as topographic surveys and sediment sampling, the physical meaning of the model output results can be verified.

[0081] Specifically, after selecting the sediment-carrying capacity formula, the key parameters in the formula are first calibrated using historical hydrological and sediment data. For example, iterative optimization is used to ensure that the theoretical curve aligns with the trend of measured data. Subsequently, the theoretical sediment transport rate is calculated based on real-time collected water level and flow data. The calculated results are compared with synchronously measured sediment concentration to generate a deviation sequence. For example, continuous positive deviations may indicate a siltation trend, while negative deviations may reflect erosion processes. The rationality of the state judgment is verified by combining riverbed elevation change monitoring data or remote sensing image interpretation results. If the deviation characteristics contradict the auxiliary data, a model parameter verification process is triggered to ensure the reliability of the physical relationships.

[0082] Compared to existing technologies, traditional methods typically employ fixed formulas and neglect parameter calibration, making it difficult for theoretical models to adapt to the sediment transport characteristics of different river sections. This proposed solution, however, significantly improves the matching degree between the theoretical model and measured data by dynamically selecting formula types and optimizing parameters. Furthermore, by introducing auxiliary data for cross-validation, it avoids misjudgments that may result from relying solely on statistical bias, thereby enhancing the physical reliability of the state classification.

[0083] Through the above technical solutions, this application solves the problems of poor formula adaptability and error accumulation caused by parameter fixation in traditional water-sediment relationship modeling, and achieves high-precision matching between theoretical models and measured data. The physical verification mechanism of dynamic equilibrium effectively distinguishes between normal fluctuations and actual riverbed evolution signals, providing a reliable physical basis for subsequent synchronicity scoring and avoiding distortion of screening results due to model bias.

[0084] Example 4: Please refer to Figure 1 In step S3, the specific method for jointly calculating the synchronization index is as follows;

[0085] S3.1 Calculation of statistical synchronicity indicators;

[0086] S3.2. Based on the physical model established in step S2, calculate the agreement index between the measured values ​​and the theoretical values;

[0087] S3.3 Normalize the statistical and physical indicators, set weight allocation according to the application scenario, generate joint synchronicity scores by weighted summation or multiplication fusion, classify the scores and mark the reasons for anomalies.

[0088] In this embodiment: This application further proposes a specific method for joint calculation of synchronization indicators, including the calculation of statistical synchronization indicators; the calculation of the degree of agreement between measured values ​​and theoretical values ​​based on the established physical model; the normalization of statistical indicators and physical indicators; the setting of weight allocation according to the application scenario; the generation of joint synchronization scores through weighted summation or multiplication fusion; and the classification and labeling of abnormal reasons for the scores.

[0089] Among them, statistical synchronicity indicators refer to parameters that quantify the temporal correlation between hydrological and sediment data. Specifically, they can be implemented using Pearson correlation coefficients or lag cross-correlation functions to characterize the strength of the linear relationship between the data. Consistency indicators refer to the degree of deviation between measured data and physical model predictions. Specifically, they can be implemented using root mean square error or relative deviation percentage to reflect the degree of matching between actual water and sediment relationships and theoretical states. Normalization refers to the operation of eliminating differences in the dimensions of different indicators. Specifically, it can be implemented using range standardization or Z-score standardization methods to make statistical and physical indicators comparable. Weight allocation refers to the process of determining the importance of indicators based on application needs. Specifically, it can be implemented using the analytic hierarchy process (AHP) or entropy weighting method. For example, in river management scenarios, physical indicators can be given higher weights. Weighted fusion refers to the calculation method of integrating multi-dimensional indicators. Specifically, it can be implemented using linear weighting or geometric mean to form a comprehensive evaluation indicator that takes into account both statistical relevance and physical rationality.

[0090] Specifically, the system first obtains the linear correlation characteristics between data by calculating statistical indicators such as correlation coefficients, and then calculates the consistency index based on the theoretical values ​​output by the physical model. Subsequently, both types of indicators are standardized to eliminate dimensional differences. Depending on the specific application scenario, such as in sediment prediction during flood season, the weight coefficients of physical indicators are determined to be higher than those of statistical indicators through expert experience or machine learning methods. Finally, a weighted summation method is used to generate a comprehensive score. When the score falls below a set threshold, an anomaly labeling mechanism is automatically triggered, and the corresponding indicator deviation type is associated with it. For example, when a situation with high statistical correlation but low physical consistency occurs in a river bend, the system will label it as "non-equilibrium state caused by a sudden change in hydrodynamic conditions."

[0091] Compared to existing technologies, traditional methods typically use a single indicator such as statistical correlation or theoretical deviation for screening, which cannot effectively identify data segments that are statistically significant but physically abnormal. This solution establishes a multi-dimensional indicator fusion system, which retains the data-driven advantages of statistical methods while incorporating the mechanistic constraints of physical models. In particular, the introduction of a scenario-adaptive mechanism in the weight allocation stage enables dynamic adjustment of evaluation criteria for different river segment characteristics and application objectives.

[0092] Through the above technical solution, this application effectively solves the problem of the disconnect between statistical indicators and physical mechanisms in traditional water-sediment synchronicity analysis, significantly improving the accuracy of anomaly data identification. Through normalization processing and weight allocation mechanisms, the scientific integration of multi-source heterogeneous indicators is achieved, ensuring that the screening results conform to both statistical data patterns and physical process constraints. The anomaly cause labeling function provides clear direction for subsequent mechanism analysis; for example, it can quickly locate synchronicity abrupt change areas caused by riverbed scouring or tributary confluence. The scenario-adaptive weight setting allows the method to be flexibly applied to different business needs such as reservoir scheduling and river morphology prediction, improving the engineering applicability of the technical solution.

[0093] Example 5: Please refer to Figure 1 In step S4, the specific method of dynamic threshold adaptive filtering is as follows;

[0094] S4.1 Hydrological scenario feature extraction: Extract key hydrological features from the input data and calculate statistical features;

[0095] S4.2. Based on historical data, establish threshold benchmarks for different hydrological scenarios, and use machine learning methods to automatically learn the scenario-threshold mapping relationship and generate probabilistic threshold intervals.

[0096] S4.3. Use the sliding window to match the current hydrological scenario in real time and call the corresponding threshold rules to filter data segments.

[0097] In this embodiment, the present application further proposes the following specific method for dynamic threshold adaptive screening: hydrological scenario feature extraction, extracting key hydrological features from input data and calculating statistical features; establishing threshold benchmarks under different hydrological scenarios based on historical data, using machine learning methods to automatically learn the scenario-threshold mapping relationship and generate probabilistic threshold intervals; matching the current hydrological scenario in real time by sliding window and calling the corresponding threshold rules to screen data segments.

[0098] Hydrological scenario feature extraction refers to extracting characteristic quantities representing hydrological states from data such as flow rate, water level, and rainfall intensity. This can be achieved using principal component analysis or time-frequency domain eigenvalue decomposition methods to capture the dynamic impact of different hydrological conditions on water-sediment relationships. Probabilistic threshold intervals refer to the screening boundary range constructed based on probability distributions. These can be generated using Gaussian mixture models or quantile regression methods, reflecting the uncertainty of thresholds under different scenarios. Sliding window real-time matching refers to dynamically analyzing data segments divided according to a preset time length. This can be achieved using fixed-length or adaptive-length window mechanisms to ensure that threshold adjustments correspond to current hydrological conditions in real time.

[0099] Specifically, after data preprocessing, statistical features such as flow rate variation and water level fluctuation are first extracted from the hydrological time series. For example, feature vectors are formed by calculating the mean and variance within a sliding window. Then, a machine learning model, such as a random forest or neural network algorithm, is trained using historical datasets to establish a nonlinear mapping relationship between hydrological feature combinations and optimal screening thresholds. In the real-time application phase, the latest data segment is captured through a continuously updated sliding window. The extracted hydrological features are input into the trained model, and the output is a dynamic threshold range adapted to the current conditions. For example, during the flood evolution phase, the system can automatically lower the synchronicity threshold to capture rapidly changing water and sediment responses, while during periods of low water levels, a more stringent screening standard is adopted.

[0100] Compared to existing technologies, traditional methods using fixed thresholds cannot distinguish the differences in water and sediment dynamics between flood and dry seasons. For example, during the dry season, normal sediment transport processes may be misjudged as abnormal. This solution, however, establishes contextualized threshold benchmarks through machine learning. This allows for the identification of the impact of special hydrological events such as sudden changes in flow and sharp drops in water level on synchronicity. For instance, during rising water phases, the threshold tolerance is automatically relaxed to match the lag effect of sediment transport, thereby avoiding misscreening problems caused by mechanically applying uniform standards.

[0101] Through the above technical solution, this application solves the technical deficiency of traditional fixed threshold screening methods that cannot adapt to dynamic changes in hydrological conditions, and realizes intelligent matching between the judgment criteria for water-sediment synchronicity and real-time water flow characteristics. Specifically, under non-steady-state conditions such as flood scouring and reservoir operation, it can automatically adjust the screening threshold boundary, effectively distinguish between normal sediment transport processes and real abnormal data, and improve the spatiotemporal adaptability of sediment transport law analysis.

[0102] Example 6: Please refer to Figure 1 In step S5, the specific method of physical-data collaborative verification is as follows;

[0103] S5.1. The hydrodynamic-sediment coupling model is used to reproduce the water and sediment processes in the selected high-synchronization period. The simulation results are compared with the measured data, the consistency of key indicators is calculated, significant deviation areas are identified, and potential causes are analyzed.

[0104] S5.2 Select different stations with hydrological connectivity, verify the logical consistency of the synchronicity screening results, analyze spatial heterogeneity, evaluate the adaptability of the screening method, generate a spatial distribution map of synchronicity, and mark the abnormal areas.

[0105] S5.3 Quantify the impact of model inputs and parameters on the verification results, use Monte Carlo simulation to give the confidence interval for synchronicity classification, and output a list of sensitive segments that require manual review.

[0106] In this embodiment: This application further proposes a specific method for physical-data collaborative verification, including using a hydrodynamic-sediment coupling model to reproduce the water and sediment processes during the selected high synchronicity period, comparing the simulation results with measured data, calculating the consistency of key indicators, identifying areas of significant deviation and analyzing potential causes; selecting different hydrologically connected stations to verify the logical consistency of the synchronicity screening results, analyzing spatial heterogeneity, evaluating the adaptability of the screening method, generating a spatial distribution map of synchronicity and marking abnormal areas; quantifying the impact of model inputs and parameters on the verification results, using Monte Carlo simulation to give the confidence interval for synchronicity classification, and outputting a list of sensitive sections that require manual review.

[0107] Among them, the hydrodynamic-sediment coupling model refers to a mathematical model that simultaneously simulates the processes of water flow and sediment transport. Specifically, it can be implemented using two-dimensional or three-dimensional numerical models. By solving the continuity equation, momentum equation, and sediment transport equation, the dynamic coupling calculation of water and sediment processes is achieved. Monte Carlo simulation refers to a random sampling method based on probability statistics. Specifically, it can generate a large number of random samples with different parameter combinations and calculate the distribution of synchronicity classification results under different parameter combinations to quantify the uncertainty of classification results. The sensitive segment list refers to a set of data segments that are sensitive to changes in the model input parameters. Specifically, it can be identified through parameter perturbation analysis and variance decomposition methods to guide key areas for manual review.

[0108] Specifically, in the physical-data co-validation process, the selected high-synchronicity data segments are first reproduced using a hydrodynamic-sediment coupling model. The simulation results are then compared with measured data on key indicators such as flow process curves and peak sediment concentration. When the simulation deviation exceeds a preset threshold in a specific time period or interval, the hydrological conditions or model parameter settings for that period can be retrospectively checked to analyze the causes of the deviation. Furthermore, for multiple monitoring stations with hydrological connectivity within the basin, the applicability of the method under different river morphologies and bed sediment compositions is verified by comparing the spatial distribution characteristics of the synchronicity screening results at each station. Finally, the model parameters are probabilistically sampled using the Monte Carlo method to calculate the probability distribution of synchronicity classification results under different parameter combinations, forming confidence interval evaluation indicators. Key segments sensitive to parameter changes are then selected for manual verification.

[0109] Compared to existing technologies, traditional methods typically verify results through single-model validation or independent site analysis, lacking mechanisms for multi-model collaborative validation and spatial correlation analysis. This approach constructs a dual validation system by coupling model reproduction and multi-site collaborative validation, effectively identifying local anomalies and systematic biases. Furthermore, the introduction of probabilistic evaluation methods provides a clear confidence level for the validation results, overcoming the limitation of traditional deterministic validation methods in quantifying uncertainty.

[0110] Through the above technical solutions, this application achieves dynamic physical verification and spatial applicability assessment of water-sediment synchronicity screening results. It can accurately identify anomalous data segments caused by model parameter errors or sudden changes in local topography, effectively improving the interpretability of the verification process. Simultaneously, by using probabilistic assessment methods, it provides quantitative confidence indicators for synchronicity classification, offering a reliable basis for risk decision-making in engineering applications and overcoming the technical bottlenecks of traditional methods, such as limited verification dimensions and lack of uncertainty assessment.

[0111] Example 7: Please refer to Figure 1 In step S6, the specific methods for outputting results and performing uncertainty analysis are as follows;

[0112] S6.1 Classify and store hydrological and sediment data according to synchronicity scores, identify anomaly areas dominated by physical mechanisms, indicate the causes in the results, and generate time series plots or spatial distribution maps to intuitively display synchronicity characteristics.

[0113] S6.2. Using Monte Carlo simulation and Bootstrap methods, assess the confidence level of the synchronicity classification results, calculate the confidence intervals of key indicators, and analyze the sources of error;

[0114] S6.3 Generate a structured dataset, including synchronicity scores, hierarchical labels, and uncertainty intervals, and save key information such as data preprocessing, model parameters, and validation methods to ensure traceability. Based on uncertainty analysis, propose applicable scenarios.

[0115] In this embodiment, the following specific methods for result output and uncertainty analysis are proposed: Hydrological and sediment data are classified and stored according to synchronicity scores, anomaly areas dominated by physical mechanisms are identified, and the causes are noted in the results. Time series plots or spatial distribution maps are generated to visually display synchronicity characteristics. Monte Carlo simulation and Bootstrap methods are used to evaluate the confidence level of the synchronicity classification results, calculate the confidence intervals of key indicators, and analyze the sources of error. A structured dataset is generated, containing synchronicity scores, hierarchical labels, and uncertainty intervals. Key information on data preprocessing, model parameters, and validation methods is saved to ensure traceability. Based on uncertainty analysis, applicable scenarios are proposed.

[0116] The synchronicity score refers to a quantitative evaluation parameter formed by fusing physical models and statistical indicators. Specifically, it can be implemented using weighted summation or multiplicative fusion algorithms to reflect the degree of matching between water and sediment data in time or space. This score, as a classification criterion, can improve the objectivity of data selection.

[0117] Monte Carlo simulation is a numerical method that verifies the stability of results based on random sampling of probability distributions. Specifically, it can be achieved by generating a large number of simulated scenarios by setting a range of parameter fluctuations, and is used to quantify the impact of model input uncertainty on synchronicity classification. This method can reveal the sensitivity characteristics of key parameters.

[0118] The Bootstrap method is a nonparametric technique that assesses the confidence interval of a statistic through resampling. Specifically, it can be implemented by constructing an empirical distribution using sampling with replacement, and is used to analyze the reliability of statistical inferences for synchronicity scores. This method is effective in assessing the uncertainty of small sample data.

[0119] Structured datasets refer to a standard data organization format that includes metadata descriptions. This can be implemented using relational databases or standardized data tables, and is used to store complete information about preprocessing steps, model parameters, and validation processes. This design ensures data traceability and result reproducibility.

[0120] Specifically, after calculating the synchronicity score, the hydrological and sediment data are first classified into different synchronicity levels based on the scoring threshold, such as high synchronicity, medium synchronicity, and anomalous data. Anomalous areas dominated by physical mechanisms are identified by comparing the deviation between theoretical model predictions and measured data; for example, when the measured sediment concentration in a river section is consistently higher than the theoretical sediment-carrying capacity, it is marked as a sedimentation anomaly area. Time series plots simultaneously display the changes in flow and sediment concentration using dual axes, and spatial distribution maps are rendered as synchronicity score heatmaps using a geographic information system. Monte Carlo simulations generate thousands of simulation results by perturbing key parameters of the hydrodynamic model, such as the Manning coefficient and sediment settling velocity, and statistically analyze the stable probability distribution of the synchronicity classification results. The Bootstrap method repeatedly samples historical datasets to calculate the confidence interval for the synchronicity score, for example, determining the threshold fluctuation range at a 95% confidence level. The final structured dataset not only contains the original data and processing results but also fully records the interpolation methods, model calibration parameters, and validation process logs, forming a traceable technical archive.

[0121] Compared to existing technologies, traditional methods typically output only a single classification result and lack uncertainty quantification; for example, they only use correlation coefficients to classify synchronicity levels without considering parameter sensitivity. This solution, through joint evaluation using Monte Carlo and Bootstrap methods, can clearly define the confidence level of the classification results. The structured data storage method overcomes the shortcomings of traditional decentralized file management, achieving end-to-end information integration. The physical cause labeling mechanism for anomaly areas compensates for the inability of purely statistical methods to explain the underlying mechanisms.

[0122] Through the above technical solutions, this application can effectively identify false synchronization signals caused by instrument errors or special hydrological events, such as accurately distinguishing between sudden changes in sediment carrying capacity and measurement anomalies during floods; it can quantitatively assess the applicability range of classification thresholds under different hydrological conditions, such as determining the reasonable fluctuation range of synchronization scores during dry seasons; the resulting standardized dataset can directly support watershed management decisions, such as providing reliable and clearly defined water-sediment relationship data support for reservoir scheduling schemes.

[0123] Example 8: Please refer to Figure 1 In step S6.1, the specific method for classifying and storing hydrological and sediment data according to synchronicity scores is as follows:

[0124] S6.11. Classify levels according to joint synchronicity scores and store hierarchical labels by time series / spatial location;

[0125] S6.12. Based on the non-equilibrium state diagnosis results of S2.3, identify physical anomaly areas such as scouring and siltation, and add anomaly type and cause fields to the dataset;

[0126] S6.13. Overlay the flow rate and sediment concentration curves and the synchronicity level background color band, use a heat map to display the synchronicity score of the river section, and mark the location of the abnormal area in the vector layer.

[0127] In this embodiment, by combining the synchronicity scoring level classification with the physical anomaly identification, cross-validation of hydrodynamic characteristics and sediment movement state is achieved, enhancing the physical basis of data interpretation.

[0128] The overlay of curves and heatmaps can intuitively present the core research question of "water and sediment response lag effect". The background color band clearly indicates the synchronicity level, which helps to quickly identify the phase difference characteristics of water and sediment relationship in flood events.

[0129] The dedicated storage of anomaly type and cause fields constructs a non-equilibrium knowledge base, providing structured data support for subsequent analysis, and is particularly suitable for diagnosing complex scenarios such as erosion and siltation mutations.

[0130] The combination of time-series storage and spatial vector labeling meets the spatiotemporal coupling requirements of watershed water and sediment transport research and can effectively support the research paradigm of "river system integrity".

[0131] The standardized storage of hierarchical labels and anomaly fields provides a feature engineering foundation for subsequent machine learning modeling, which aligns with the current development trend of hydrological big data analysis.

[0132] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

[0133] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for screening river water and sediment synchronicity data, characterized in that: The specific steps are as follows: S1. Data Acquisition and Preprocessing: Collect synchronized hydrological and sediment data, and perform data cleaning and time alignment. S2. Water-sediment physical relationship modeling: Based on the theory of sediment carrying capacity of water flow, a physical model of water-sediment relationship is established, the theoretical sediment transport rate or sediment content is calculated, the deviation between measured data and physical model is analyzed, and potential non-equilibrium states are identified. S3. Joint calculation of synchronicity indicators: Calculate traditional synchronicity indicators such as correlation coefficient and lag cross-correlation to quantify the degree of agreement between measured values ​​and physical model predictions. Combine statistical and physical indicators with weighted integration to form a joint synchronicity score. S4. Dynamic threshold adaptive screening: The synchronization threshold is dynamically adjusted according to hydrological conditions, and machine learning is used to automatically divide the data segments and match the optimal screening criteria. S5. Physical-data collaborative verification: The hydrodynamic-sediment coupling model is used to simulate the water and sediment processes of the screening results, verify the physical consistency, compare the synchronicity characteristics of different stations, and ensure the spatial applicability of the screening method. S6. Output Results and Uncertainty Analysis: Generate a synchronicity hierarchical dataset, label outlier regions dominated by physical mechanisms, quantify uncertainty, and provide confidence intervals.

2. The method for screening river water and sediment synchronicity data according to claim 1, characterized in that: In step S1, the specific methods for data acquisition and preprocessing are as follows; S1.1 Obtain time-matched raw data from sources such as hydrological stations, sediment monitoring stations, remote sensing, or drones, check the consistency of data timestamps, and ensure that the observation times of hydrological and sediment data are strictly aligned. S1.2 Delete instrument fault records, use moving average or median filtering to process high-frequency fluctuations, and fill reasonable missing segments with linear interpolation or historical average. S1.3 Unify the time step of different parameters, and generate continuous sequences through time series interpolation for data that requires high-frequency analysis.

3. The method for screening river water and sediment synchronicity data according to claim 2, characterized in that: In step S2, the specific method for modeling the physical relationship between water and sediment is as follows; S2.1 Select the sediment carrying capacity formula based on river characteristics, calibrate the formula parameters, and optimize the model adaptability using historical data; S2.2 Calculate the theoretical sediment transport rate or sediment concentration based on the current hydrological conditions, and calculate the degree of deviation between the measured value and the theoretical value; S2.

3. Based on the deviation characteristics, determine whether the river section is in a state of erosion, siltation or dynamic equilibrium, and verify the physical rationality by combining auxiliary data.

4. The method for screening river water and sediment synchronicity data according to claim 3, characterized in that: In step S3, the specific method for jointly calculating the synchronization index is as follows; S3.1 Calculation of statistical synchronicity indicators; S3.

2. Based on the physical model established in step S2, calculate the agreement index between the measured values ​​and the theoretical values; S3.3 Normalize the statistical and physical indicators, set weight allocation according to the application scenario, generate joint synchronicity scores by weighted summation or multiplication fusion, classify the scores and mark the reasons for anomalies.

5. The method for screening river water and sediment synchronicity data according to claim 4, characterized in that: In step S4, the dynamic threshold adaptive filtering method is as follows; S4.1 Hydrological scenario feature extraction: Extract key hydrological features from the input data and calculate statistical features; S4.

2. Based on historical data, establish threshold benchmarks for different hydrological scenarios, and use machine learning methods to automatically learn the scenario-threshold mapping relationship and generate probabilistic threshold intervals. S4.

3. Use the sliding window to match the current hydrological scenario in real time and call the corresponding threshold rules to filter data segments.

6. The method for screening river water and sediment synchronicity data according to claim 5, characterized in that: In step S5, the physical-data collaborative verification method is as follows; S5.

1. The hydrodynamic-sediment coupling model is used to reproduce the water and sediment processes in the selected high-synchronization period. The simulation results are compared with the measured data, the consistency of key indicators is calculated, significant deviation areas are identified, and potential causes are analyzed. S5.2 Select different stations with hydrological connectivity, verify the logical consistency of the synchronicity screening results, analyze spatial heterogeneity, evaluate the adaptability of the screening method, generate a spatial distribution map of synchronicity, and mark the abnormal areas. S5.3 Quantify the impact of model inputs and parameters on the verification results, use Monte Carlo simulation to give the confidence interval for synchronicity classification, and output a list of sensitive segments that require manual review.

7. The method for screening river water and sediment synchronicity data according to claim 6, characterized in that: In step S6, the specific methods for result output and uncertainty analysis are as follows; S6.1 Classify and store hydrological and sediment data according to synchronicity scores, identify anomaly areas dominated by physical mechanisms, indicate the causes in the results, and generate time series plots or spatial distribution maps to intuitively display synchronicity characteristics. S6.

2. Using Monte Carlo simulation and Bootstrap methods, assess the confidence level of the synchronicity classification results, calculate the confidence intervals of key indicators, and analyze the sources of error; S6.3 Generate a structured dataset containing synchronicity scores, hierarchical labels, and uncertainty intervals. Save key information such as data preprocessing, model parameters, and validation methods to ensure traceability. Based on uncertainty analysis, propose applicable scenarios.

8. The method for screening river water and sediment synchronicity data according to claim 7, characterized in that: In step S6.1, the specific method for classifying and storing hydrological and sediment data according to synchronicity scores is as follows: S6.

11. Classify levels according to joint synchronicity scores and store hierarchical labels by time series / spatial location; S6.

12. Based on the non-equilibrium state diagnosis results of S2.3, identify physical anomaly areas such as scouring and siltation, and add anomaly type and cause fields to the dataset; S6.

13. Overlay the flow rate and sediment concentration curves and the synchronicity level background color band, use a heat map to display the synchronicity score of the river section, and mark the location of the abnormal area in the vector layer.

Images