A hydrological warning method, system, program product and readable storage medium

Abstract

The application provides a hydrological warning method, system, program product and readable storage medium, relating to the field of hydrological data filling. By identifying missing data items in the hydrological table and analyzing their time span and spatial distribution, a foundation is laid for subsequent filling. The method introduces a comparison mechanism between the data quantity ratio and the preset threshold value, and performs full filling when the missing data quantity is small, and adopts a selective filling strategy when the missing data quantity is large. This dynamically adjusted method effectively avoids the estimation errors that may be introduced by large-scale filling, while ensuring the filling of key data points. The inverse relationship between the filling ratio and the data quantity ratio avoids introducing too much estimation error. The method selects the corresponding filling method according to the data type, improving the accuracy of filling. This comprehensive filling strategy significantly improves the integrity and reliability of hydrological data, providing a high-quality data foundation for subsequent abnormal data identification and warning.

Description

Technical Field

[0001] This application relates to the field of hydrological data filling, and in particular to a hydrological early warning method, system, program product, and readable storage medium. Background Technology

[0002] Hydrological monitoring in remote areas often faces significant uncertainties due to factors such as geographical isolation and inadequate infrastructure. Specifically, this manifests as uneven distribution of data collection points and unstable collection frequencies, resulting in discontinuous hydrological data in both time and space. Some areas or time periods may have abundant data, while others may have scarce data.

[0003] To address this issue of data discontinuity, related techniques typically employ data imputation methods. These methods estimate and fill in all missing data points, attempting to complete the entire data sequence. The amount of imputation is usually equal to the amount of missing data, meaning that as many missing data points as possible are filled in.

[0004] However, this comprehensive data imputation method has certain limitations in practical applications. When the amount of missing data is large, large-scale data imputation may introduce excessive estimation errors, causing the imputed data to deviate from the actual hydrological situation. In particular, for large amounts of data that are continuously missing in time or space, subsequent analysis and decision-making based on this data lack accuracy. Summary of the Invention

[0005] This application provides a hydrological early warning method, system, program product, and readable storage medium to avoid large-scale filling of estimation errors that may be introduced.

[0006] Firstly, this application provides a hydrological early warning method, comprising: filling hydrological data into a preset hydrological table and identifying missing data items in the hydrological table; determining the amount of data to be filled and the filling area based on the time span and spatial distribution of consecutively missing data in the missing data items, wherein consecutively missing data refers to any data in the hydrological table that is consecutively missing in terms of time span or spatial distribution; wherein: when the ratio of the amount of consecutively missing data to the corresponding total data is less than a preset threshold, full filling measurement is performed; when the ratio of the amount of consecutively missing data to the corresponding total data is not less than the preset threshold, a selective filling strategy is performed, selecting... The imputation strategy involves selecting several points in the beginning, middle, and end of consecutively missing data for imputation; calculating the imputation amount based on the imputation ratio and the amount of consecutively missing data; the ratio of the amount of consecutively missing data to the corresponding total data is inversely proportional to the imputation ratio; selecting an appropriate imputation method based on the type of consecutively missing data; transmitting the imputed hydrological data to a cloud server; enabling the cloud server to identify abnormal data in the imputed hydrological data; receiving early warning signals from the cloud server, which are issued when abnormal data exceeds a threshold; and sending early warning information to the terminals of relevant personnel.

[0007] By employing the aforementioned technical solution, the method identifies missing data items in hydrological tables and analyzes their temporal span and spatial distribution, laying the foundation for subsequent data infilling. Secondly, the method introduces a comparison mechanism between the data volume ratio and a preset threshold. When the amount of missing data is small, full infilling is performed, while a selective infilling strategy is adopted when the amount of missing data is large. This dynamic adjustment method effectively avoids estimation errors that may be introduced by large-scale infilling, while ensuring the filling of key data points. Furthermore, the inverse relationship between the infilling ratio and the data volume ratio avoids introducing excessive estimation errors. Finally, the method selects the appropriate infilling method based on the data type, improving the accuracy of the infilling. This comprehensive infilling strategy significantly improves the integrity and reliability of hydrological data, providing a high-quality data foundation for subsequent anomaly data identification and early warning.

[0008] In conjunction with some embodiments of the first aspect, in some embodiments, after selecting an appropriate data imputation method based on the type of missing data, the method further includes: obtaining the imputed hydrological data through a database joint query; processing the imputed hydrological data using a data fusion algorithm to generate an integrated dataset; transmitting the imputed hydrological data to a cloud server; and enabling the cloud server to identify abnormal data in the imputed hydrological data. Specifically, this includes: transmitting the integrated dataset to the cloud server; and enabling the cloud server to identify abnormal data in the integrated dataset.

[0009] By adopting the above technical solution, the filled hydrological data is obtained through joint database queries, enabling unified access and processing of different types of hydrological data, overcoming the low query efficiency problem caused by scattered data storage. A data fusion algorithm is applied to process the filled hydrological data, generating an integrated dataset. This integration of hydrological data from different sources and types improves data consistency. The integrated dataset is then transmitted to a cloud server, rather than directly transmitting the original filled data. This method significantly reduces the amount of data that needs to be transmitted and processed, alleviating the computational load on the cloud server.

[0010] In conjunction with some embodiments of the first aspect, in some embodiments, the step of obtaining the filled hydrological data through a database joint query specifically includes: creating a joint query, which connects the filled hydrological data in all databases through a common field, and each database stores one type of filled hydrological data; extracting the filled hydrological data corresponding to the preset selection conditions in the joint query according to preset selection conditions, which include: specifying a time range, selecting specific hydrological parameters, and setting geographical location filtering conditions.

[0011] By adopting the above technical solution, a joint query is created that connects all the filled hydrological data in the database through common fields. Pre-defined selection conditions are introduced into the joint query, including specifying a time range, selecting specific hydrological parameters, and setting geographical location filters. This enables the rapid and accurate acquisition of key hydrological data within a specific spatiotemporal range, greatly improving the efficiency and relevance of data analysis.

[0012] In conjunction with some embodiments of the first aspect, in some embodiments, the step of applying a data fusion algorithm to process the incomplete hydrological data to generate an integrated dataset specifically includes: standardizing different types of incomplete hydrological data in the same space and at the same time, converting incomplete hydrological data with different dimensions and units into a unified dimensionless numerical range; assigning probability values ​​to each subset in the identification framework for each type of incomplete hydrological data according to preset rules, obtaining basic probability allocations for multiple evidence sources; the identification framework includes a set of mutually exclusive hypotheses related to hydrological early warning; obtaining a fused dataset based on the basic probability allocations of multiple evidence sources, wherein: identifying all possible combinations of intersections among multiple evidence sources to obtain a target set; calculating the sum of probability products of possible combinations; adjusting the sum of probability products using a normalization factor; and combining the fused dataset with the incomplete hydrological data to generate an integrated dataset.

[0013] By adopting the above technical solution, different types of filled hydrological data are standardized, resolving the issue of different dimensions and units, thus laying the foundation for subsequent fusion. Secondly, a data fusion algorithm based on evidence theory is introduced to assign probability values ​​to each type of hydrological data, constructing a basic probability allocation for multiple evidence sources. This step fully considers the reliability and uncertainty of different data sources. Then, by identifying the intersection of multiple evidence sources, calculating the sum of probability products, and applying a normalization factor, the fused dataset is finally obtained. This effectively handles data uncertainty and conflict, significantly improving the reliability of the fusion results. Finally, the fused dataset is combined with the original filled data, preserving both the comprehensive advantages of fusion and the detailed information of the original data. This method effectively solves the problems of information loss and anomaly masking that may result from simply relying on integrated data.

[0014] In conjunction with some embodiments of the first aspect, in some embodiments, the step of selecting an appropriate data imputation method to impute data based on the type of consecutively missing data specifically includes: identifying the type of consecutively missing data and classifying it into time series data and spatial series data; for time series data, using time series analysis methods for imputation; and for spatial series data, applying spatial statistical methods for imputation.

[0015] By employing the above technical solution, continuously missing data is identified and classified into time series data and spatial series data. This classification lays the foundation for selecting appropriate imputation methods. Secondly, for time series data, time series analysis methods are used for imputation, fully utilizing the temporal correlation of the data. For spatial series data, spatial statistical methods are applied, considering the spatial distribution characteristics of the data. This strategy of using specialized imputation methods for different types of data significantly improves the accuracy and effectiveness of imputation. By distinguishing between time and spatial series and applying corresponding statistical methods, this technical solution can more accurately capture the spatiotemporal variation patterns of the data, thereby generating imputation results that are more consistent with reality.

[0016] In conjunction with some embodiments of the first aspect, in some embodiments, the step of filling time series data using time series analysis methods specifically includes: when the ratio of the amount of consecutive missing data to the corresponding total data is less than a preset threshold, interpolation is used for filling; when the ratio of the amount of consecutive missing data to the corresponding total data is not less than a preset threshold, time series decomposition is used for filling.

[0017] By adopting the above technical solution, a comparison mechanism between the data volume ratio and a preset threshold is introduced, providing a basis for selecting imputation strategies under different conditions. When the amount of continuously missing data is small, interpolation is used for imputation. Interpolation can quickly and effectively handle small-scale data missingness while maintaining data continuity. Secondly, when the amount of continuously missing data is large, the method switches to time series decomposition for imputation. Time series decomposition can capture the long-term trend, seasonal changes, and random fluctuations of data, making it particularly suitable for handling large-scale time series data missingness. This strategy of dynamically selecting the imputation method based on the size of the missing data ensures both rapid processing of small-scale missing data and accurate imputation of large-scale missing data.

[0018] In conjunction with some embodiments of the first aspect, in some embodiments, the step of selecting an appropriate data imputation method to impute continuous missing data according to the type of continuous missing data specifically includes: for spatiotemporal mixed data, obtaining a set of known data points containing time coordinates, spatial coordinates, and observation values; selecting neighboring known data points for continuous missing data based on spatiotemporal distance; constructing a Kriging equation system using a predetermined spatiotemporal variogram; solving the Kriging equation system to obtain a weight vector; and calculating the imputation value for continuous missing data using the weight vector and the observation values ​​of the known data points.

[0019] By employing the aforementioned technical solution, a comprehensive data foundation is laid for subsequent analysis by acquiring a set of known data points containing time coordinates, spatial coordinates, and observation values. Secondly, selecting neighboring known data points based on spatiotemporal distance ensures that the incompleteness process considers local spatiotemporal correlations. Then, a kriging equation system is constructed using a pre-determined spatiotemporal variability function. This step fully utilizes advanced geostatistical theory to effectively capture the spatiotemporal variability characteristics of the data. By solving the kriging equation system to obtain a weight vector, and using this weight vector and the observation values ​​of the known data points to calculate the incomplete values, accurate estimation of continuously missing data is achieved. This spatiotemporal data incompleteness method based on kriging interpolation considers not only the spatial correlation of the data but also information in the temporal dimension, thus generating more accurate and reliable incompleteness results.

[0020] In a second aspect, this application provides a hydrological early warning model, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, and the memory is used to store computer program code, which includes computer instructions, and the one or more processors call the computer instructions to cause the hydrological early warning model to perform the method described in the first aspect and any possible implementation thereof.

[0021] Thirdly, this application provides a computer program product containing instructions that, when run on a hydrological early warning model, cause the hydrological early warning model to perform the method described in the first aspect and any possible implementation thereof.

[0022] Fourthly, this application provides a computer-readable storage medium including instructions that, when executed on a hydrological early warning model, cause the hydrological early warning model to perform the method described in the first aspect and any possible implementation thereof.

[0023] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0024] 1. By identifying missing data items in hydrological tables and analyzing their temporal span and spatial distribution, a foundation is laid for subsequent data infilling. Secondly, the method introduces a comparison mechanism between the data volume ratio and a preset threshold. When the amount of missing data is small, full infilling is performed, while a selective infilling strategy is adopted when the amount of missing data is large. This dynamic adjustment method effectively avoids estimation errors that may be introduced by large-scale infilling, while ensuring the filling of key data points. Furthermore, the inverse relationship between the infilling ratio and the data volume ratio avoids introducing excessive estimation errors. Finally, the method selects the appropriate infilling method based on the data type, improving the accuracy of the infilling. This comprehensive infilling strategy significantly improves the integrity and reliability of hydrological data, providing a high-quality data foundation for subsequent anomaly identification and early warning.

[0025] 2. Standardization was performed on different types of infilled hydrological data, resolving issues related to different dimensions and units, thus laying the foundation for subsequent fusion. Secondly, an evidence-based data fusion algorithm was introduced to assign probability values ​​to each type of hydrological data, constructing a basic probability allocation for multiple evidence sources. This step fully considered the reliability and uncertainty of different data sources. Then, by identifying the intersection of multiple evidence sources, calculating the sum of probability products, and applying a normalization factor, the fused dataset was finally obtained. This effectively handled data uncertainty and conflict, significantly improving the reliability of the fusion results. Finally, the fused dataset was combined with the original infilled data, preserving both the comprehensive advantages of fusion and the detailed information of the original data. This method effectively solved the problems of information loss and anomaly masking that might result from simply relying on integrated data. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of a hydrological early warning process in an embodiment of this application;

[0027] Figure 2This is another schematic diagram of the hydrological early warning process in the embodiments of this application;

[0028] Figure 3 yes Figure 2 A detailed flowchart of step S201;

[0029] Figure 4 yes Figure 2 A detailed flowchart of step S202;

[0030] Figure 5 This is a schematic diagram of an exemplary hardware structure of the hydrological early warning model in this application embodiment. Detailed Implementation

[0031] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to and includes any or all possible combinations of one or more of the listed items.

[0032] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0033] The hydrological early warning method in the embodiments of this application is described below:

[0034] Please see Figure 1 This is a flowchart illustrating a hydrological early warning method in an embodiment of this application.

[0035] S101. Fill the hydrological data into the preset hydrological table and identify the missing data items in the hydrological table;

[0036] Hydrological data refers to various measurements and observations related to hydrology, including precipitation, river flow, and water level. Pre-designed hydrological tables refer to pre-designed table structures used for recording and storing hydrological data.

[0037] It should be noted that the purpose of hydrological tables is to facilitate the identification of items that should have data but actually do not.

[0038] S102. Based on the time span and spatial distribution of consecutively missing data in the missing data items, determine the amount of data to be filled and the filling area, where consecutively missing data refers to any data in the hydrological table that is consecutively missing in terms of time span or spatial distribution.

[0039] Continuously missing data refers to a situation where multiple consecutive data points are missing in a time series or spatial distribution. Data imputation refers to the number of data points that need to be added. The imputation region refers to the time period or geographical area where data imputation is required.

[0040] It should be noted that the following description uses a single consecutive missing data point as an example for detailed explanation. This process can be generalized to all consecutive missing data points in the dataset, with each missing point following the same processing procedure.

[0041] S103. When the ratio of the amount of continuously missing data to the corresponding total amount of data is less than a preset threshold, perform a full fill measurement.

[0042] The preset threshold refers to a pre-set ratio value, and full fill measurement means that the strategy of filling in all missing data points is to fill in all missing data points.

[0043] In some embodiments, the ratio of the amount of consecutively missing data to the total amount of data within the corresponding time or spatial range is calculated. If this ratio is less than a preset threshold, it indicates that the missing data is relatively small and will not seriously affect the overall data quality. In this case, the system chooses to perform a full imputation measurement, that is, to impute all identified missing data points.

[0044] In some embodiments, linear interpolation or nearest neighbor filling can be used to fill the consecutive missing data;

[0045] It should be noted that this refers to the case of a single missing data item in the text table. This single-point missing data can be regarded as a special case of continuous missing data, where its spatiotemporal range is compressed to a specific point in time and spatial location. For this type of single-point missing data, the full imputation measurement is also performed, which will not be elaborated here.

[0046] S104. When the ratio of the amount of consecutive missing data to the corresponding total data is not less than a preset threshold, a selective imputation strategy is executed. The selective imputation strategy is to select several points in the front, middle and back parts of the consecutive missing data for imputation.

[0047] When the proportion of consecutively missing data is large, imputing all data may introduce excessive uncertainty. To address this, the system checks if the ratio of missing data to total data is not less than a preset threshold. If so, a selective imputation strategy is employed. This strategy does not fill in all missing points, but rather selects key points in different parts of the missing data for imputation. The system typically selects points at the beginning, middle, and end of the missing data for imputation. This method can maintain the overall trend of the data while reducing potential errors introduced by imputation.

[0048] In some embodiments, the continuously missing data segment is first divided into three equal parts: front, middle, and back. Within each part, the system selects filling points using a calculated interval (obtained in step S105). For the selected points, the system uses interpolation methods (such as linear interpolation or spline interpolation) to fill the missing data. Finally, the system smooths the filled data to ensure the continuity of the overall trend; this is not limited to specific methods.

[0049] S105. Calculate the data imputation amount based on the imputation ratio and the amount of consecutively missing data; the ratio of the amount of consecutively missing data to the corresponding total data is inversely proportional to the imputation ratio.

[0050] The imputation ratio represents the proportion of data points that are actually imputed in consecutive missing data out of the total missing data. The ratio of the amount of consecutive missing data to the corresponding total data indicates the severity of the missing data.

[0051] In some embodiments, the ratio of the amount of consecutively missing data to the total amount of data is considered. A higher ratio indicates a more severe data gap, and the system will correspondingly reduce the imputation ratio to avoid introducing excessive uncertainty. The system then calculates the specific amount of data to be imputed based on this dynamically adjusted imputation ratio and the total amount of consecutively missing data. This ensures a more cautious imputation strategy when data gaps are severe, while allowing for more comprehensive imputation when data gaps are minor.

[0052] S106. Select the appropriate data imputation method to impute the missing data according to the type of consecutive missing data;

[0053] The type of continuously missing data indicates the nature of the missing data, such as time series data, spatially distributed data, or spatiotemporally mixed data. Data imputation methods refer to the specific algorithms or techniques used to estimate and fill in missing data.

[0054] In some embodiments, step S106 specifically includes: S501, identifying the type of continuously missing data and classifying it into time series data and spatial series data;

[0055] In some embodiments, it can be determined whether missing values ​​in hydrological tables are distributed along the time or spatial dimensions. For time series data, this typically manifests as missing values ​​at the same monitoring point within consecutive time periods; for spatial series data, it manifests as missing values ​​at adjacent spatial locations at the same time. Then, the missing data are categorized accordingly based on these characteristics.

[0056] S502. For time series data, time series analysis methods are used to fill in the gaps.

[0057] In some embodiments, a suitable time series analysis method is selected, such as moving average, exponential smoothing, ARIMA model, or machine learning methods. Then, historical data prior to the missing data points is used to train or fit the selected model. Next, the trained model is used to predict or estimate the values ​​at the missing time points. Finally, these predicted values ​​are filled into the missing positions of the original dataset. This process considers the time dependence and trend characteristics of the data, and can generate imputed values ​​consistent with the original data patterns.

[0058] In some embodiments, step S502 specifically includes: S5021, when the ratio of the amount of continuously missing data to the corresponding total amount of data is less than a preset threshold, interpolation is used to fill the missing data.

[0059] In some embodiments, the ratio of consecutive missing data points to the total number of data points is calculated. This ratio is then compared to a pre-set threshold. If the ratio is less than the threshold, it indicates that there are relatively few missing data points, and the overall structure and trend of the data remain intact. In this case, interpolation is chosen for imputation. Interpolation uses known data points around the missing points to estimate the missing values. This method is simple, fast, and maintains the continuity and smoothness of the data when there are few missing data points.

[0060] S602. When the ratio of the amount of consecutively missing data to the corresponding total amount of data is not less than a preset threshold, time series decomposition is used to fill the missing data.

[0061] Time series decomposition refers to the process of breaking down time series data into multiple components, typically including trend, seasonality, and residual (random) components. The trend component indicates the long-term direction of data change. The seasonality component reflects the periodic pattern of data change. The residual component represents random fluctuations that cannot be explained by trend and seasonality.

[0062] In some embodiments, this step is performed when the proportion of missing data is determined to be large, making simple interpolation unsuitable. The complete historical data is decomposed into time series components, separating the data into trend, seasonality, and residual components. Then, each of these components is analyzed and modeled separately. For the trend component, regression methods may be used for prediction; for the seasonal component, historical seasonal patterns may be used for imputation; and for the residual component, statistical methods or simple interpolation may be used. Next, the predicted or estimated values ​​of these three components are recombine to obtain a comprehensive estimate of the missing time points. Finally, these estimates are filled into the missing locations in the original dataset. This method better captures the long-term trends and periodic characteristics of the data, and is particularly suitable for handling missing data over long periods.

[0063] As can be seen, the introduction of a comparison mechanism between the data volume ratio and a preset threshold provides a basis for selecting imputation strategies under different conditions. When the amount of consecutively missing data is small, interpolation is used for imputation. Interpolation can quickly and effectively handle small-scale data missingness while maintaining data continuity. Secondly, when the amount of consecutively missing data is large, the method switches to time series decomposition for imputation. Time series decomposition can capture long-term trends, seasonal changes, and random fluctuations in data, making it particularly suitable for handling large-scale time series data missingness. This strategy of dynamically selecting the imputation method based on the size of the missing data ensures both rapid processing of small-scale missing data and accurate imputation of large-scale missing data.

[0064] S503. For spatial sequence data, spatial statistical methods are used to fill in the gaps.

[0065] In some embodiments, an appropriate spatial interpolation method is selected, such as inverse distance weighted interpolation (IDW), kriging, or spline interpolation. Then, the spatial distribution characteristics and correlations of the known data points are analyzed. Next, based on the selected method and the analysis results, estimates of the missing locations are calculated. Finally, these estimates are filled into the missing locations of the original dataset. This process considers the spatial dependencies of the data and generates imputed values ​​that are spatially continuous and consistent with surrounding observations.

[0066] As can be seen, identifying the type of continuously missing data and classifying it into time series data and spatial series data lays the foundation for selecting appropriate imputation methods. Secondly, for time series data, time series analysis methods are used for imputation, fully utilizing the temporal correlation of the data. For spatial series data, spatial statistical methods are applied, considering the spatial distribution characteristics of the data. This strategy of using specialized imputation methods for different types of data significantly improves the accuracy and effectiveness of imputation. By distinguishing between time and spatial series and applying corresponding statistical methods, this technical solution can more accurately capture the spatiotemporal variation patterns of the data, thereby generating imputation results that are more consistent with reality.

[0067] In some other embodiments, after step S503, the method further includes:

[0068] S504. For spatiotemporal mixed data, obtain a set of known data points containing time coordinates, spatial coordinates, and observation values;

[0069] Spatiotemporal hybrid data refers to datasets that possess both temporal and spatial attributes, such as hydrological data measured at different times and locations. The time coordinate represents the observation time of the data points, such as a date or timestamp.

[0070] In some embodiments, all non-missing data points are filtered from the original dataset. Then, the temporal information of each data point is extracted, potentially including year, month, day, hour, etc., and converted to a uniform time format. Next, the spatial information of each data point is extracted, typically geographic coordinates (longitude and latitude) or coordinates in another spatial reference system. Finally, the time coordinates, spatial coordinates, and corresponding observations are combined into a structured dataset. This step lays the foundation for subsequent spatiotemporal analysis and interpolation, ensuring that all necessary information is correctly collected and organized.

[0071] S505. Based on spatiotemporal distance, select nearby known data points for consecutive missing data;

[0072] Spatiotemporal distance refers to a distance metric that takes into account both time and space dimensions, and is used to represent the degree of proximity between two data points in time and space.

[0073] In some embodiments, a suitable spatiotemporal distance metric needs to be defined, which typically involves standardizing and weighting temporal and spatial distances. Then, for each missing data point, its spatiotemporal distance to all known data points is calculated. Next, the closest known data points are selected based on predefined criteria (such as a distance threshold or a fixed number of neighboring points). Finally, these selected neighboring points and their relevant information (such as coordinates and observations) are saved for subsequent interpolation. This process ensures that the most spatiotemporally relevant data points are used during interpolation, thereby improving the accuracy of the interpolation.

[0074] For example: First, define a spatiotemporal distance function, which is a linear combination of time difference and spatial distance after standardization. Then, for each missing point, calculate its spatiotemporal distance with all known points. Next, use the quicksort algorithm to sort the distances, and then select the N points with the smallest distance (N is the preset number of neighboring points). Finally, store the information of these N points in a data structure, which is not limited here.

[0075] S506. Construct a system of Kriging equations using a predetermined spatiotemporal variability function.

[0076] The spatiotemporal variogram (also known as the spatiotemporal covariance function or spatiotemporal semivariogram) is a mathematical function that describes the correlation structure of data in time and space. The Kriging equations are a system of linear equations used in the Kriging interpolation method to solve for optimal weights.

[0077] In some embodiments, it is necessary to ensure that a spatiotemporal variogram model suitable for the current data already exists. This variogram is then used to calculate the covariance between the missing point and each selected neighbor, as well as the covariance among the neighbors. These covariance values ​​are then filled into the coefficient matrix of the Kriging equations. Simultaneously, the right-hand side vector of the equations is constructed, representing the covariance between the missing point and each neighbor. Finally, Lagrange multipliers are added to ensure an unbiased estimate where the weight sum is 1. This process encodes the correlation structure of the spatiotemporal data into the equations, preparing for solving for the optimal interpolation weights.

[0078] For example: Select a suitable spatiotemporal variogram model (such as a product-type or sum-type model), then calculate the temporal and spatial distances for each pair of data points, then substitute these distances into the variogram to calculate the covariance value, then fill the calculated covariance value into the corresponding positions of the matrix and vector, and finally add Lagrange multipliers to complete the construction of the system of equations.

[0079] S507. Solve the Kriging equations to obtain the weight vector;

[0080] In some embodiments, a suitable numerical solution method needs to be selected. Considering that the Kriging equations are typically symmetric and positive definite, methods such as Gaussian elimination, LU decomposition, or conjugate gradient method can be used. Then, the selected method is applied to solve the equations. During the solution process, numerical stability adjustments may be necessary, such as using rotation or scaling to improve the condition number of the matrices. Next, the obtained solution is verified to ensure that it satisfies the conditions of unbiasedness and minimum variance. Finally, the obtained weight vector is extracted to prepare for the next step of interpolation calculation. The key to this process is to solve the equations accurately and efficiently, because the accuracy of the weights directly affects the quality of the final interpolation result.

[0081] For example, using Gaussian elimination, the augmented matrix is ​​first transformed into upper triangular form, then the unknowns are solved by back substitution, followed by a numerical stability check. If the result is unstable, a partial pivot selection strategy is used to solve the problem again. Finally, the obtained solution is normalized to ensure that the sum of the weights is 1. No restrictions are imposed here.

[0082] As can be seen, obtaining a set of known data points containing time coordinates, spatial coordinates, and observation values ​​lays a comprehensive data foundation for subsequent analysis. Secondly, selecting neighboring known data points based on spatiotemporal distance ensures that the incompleteness process considers local spatiotemporal correlations. Then, constructing a Kriging equation system using a pre-determined spatiotemporal variability function fully utilizes advanced geostatistical theory, effectively capturing the spatiotemporal variability characteristics of the data. By solving the Kriging equation system to obtain a weight vector, and using the weight vector and the observation values ​​of the known data points to calculate the incomplete values, accurate estimation of continuously missing data is achieved. This spatiotemporal data incompleteness method based on Kriging interpolation considers not only the spatial correlation of the data but also information in the temporal dimension, thus generating more accurate and reliable incompleteness results.

[0083] S505. Using the weight vector and the observed values ​​of known data points, calculate the imputation values ​​for consecutive missing data.

[0084] The weight vector is the set of optimal interpolation coefficients obtained by solving the Kriging equations. Known data points refer to the neighboring data points selected in step S502. Observations are the actual measurements of these known data points. Imputed values ​​are the estimated values ​​used to replace missing data.

[0085] In some embodiments, each weight in the weight vector is multiplied by the observation of the corresponding known data point. Then, all these products are summed to obtain an estimate of the missing point. Next, some post-processing may be required, such as checking if the estimate is within a reasonable range or applying constraints (such as nonnegativity). Finally, the calculated imputation value is inserted into the missing location in the original dataset. This process achieves the estimation of missing values ​​based on a weighted average of surrounding known points, where the weights reflect the relative influence of each known point on the missing point.

[0086] As can be seen, obtaining a set of known data points containing time coordinates, spatial coordinates, and observation values ​​lays a comprehensive data foundation for subsequent analysis. Secondly, selecting neighboring known data points based on spatiotemporal distance ensures that the incompleteness process considers local spatiotemporal correlations. Then, constructing a Kriging equation system using a pre-determined spatiotemporal variability function fully utilizes advanced geostatistical theory, effectively capturing the spatiotemporal variability characteristics of the data. By solving the Kriging equation system to obtain a weight vector, and using the weight vector and the observation values ​​of the known data points to calculate the incomplete values, accurate estimation of continuously missing data is achieved. This spatiotemporal data incompleteness method based on Kriging interpolation considers not only the spatial correlation of the data but also information in the temporal dimension, thus generating more accurate and reliable incompleteness results.

[0087] S107. Transmit the filled hydrological data to the cloud server; enable the cloud server to identify abnormal data in the filled hydrological data.

[0088] The filled-in hydrological data refers to the hydrological dataset that has undergone the aforementioned processing and filling steps. The cloud server refers to a high-performance computing device running in a remote data center.

[0089] In some embodiments, the cloud server may apply box plots to identify outliers and may also apply moving averages or exponential smoothing techniques to compare actual values ​​with predicted values; this is not limited to any particular method.

[0090] S108. Receive an early warning signal sent by the cloud server. The early warning signal is issued by the cloud server when abnormal data exceeds a threshold.

[0091] Once the cloud server completes the identification of abnormal data, if the abnormal situation meets the warning criteria, it needs to promptly notify the relevant systems or personnel.

[0092] In some embodiments, the cloud server uses a multi-level early warning mechanism, classifying abnormal data into different levels (such as mild, moderate, and severe) based on the severity of the abnormal data. When an anomaly is detected, the server first determines the early warning level, and then generates a structured early warning signal containing information such as the early warning level, anomaly type, and scope of impact.

[0093] S109. Send early warning information to the terminals of relevant personnel.

[0094] Among these, "relevant personnel" refers to staff, decision-makers, or the general public who need to receive hydrological early warning information. "Terminal" refers to the device that receives the warning information, such as a mobile phone, computer, or dedicated early warning equipment. The warning information is generated based on the warning signal and includes detailed information such as specific warning content and suggested response measures.

[0095] In some embodiments, the system first generates more detailed and easily understandable warning information based on the content and severity of the warning signal. This information may include a description of the abnormal situation, its potential impact, and recommended measures. Then, the system determines a list of relevant personnel who need to receive this information based on pre-defined distribution rules. The system then sends the warning information to these personnel's terminal devices through various channels (such as SMS, email, dedicated app push notifications, etc.).

[0096] As can be seen, the hydrological early warning method provided in this application can flexibly select a filling strategy based on the characteristics of missing data. First, the method identifies missing data items in the hydrological table and analyzes their temporal span and spatial distribution, laying the foundation for subsequent filling. Second, the method introduces a comparison mechanism between the data volume ratio and a preset threshold, performing full filling when the amount of missing data is small, and employing a selective filling strategy when the amount of missing data is large. This dynamic adjustment method effectively avoids estimation errors that may be introduced by large-scale filling, while ensuring the filling of key data points. Furthermore, the inverse relationship between the filling ratio and the data volume ratio avoids introducing excessive estimation errors. Finally, the method selects the appropriate filling method based on the data type, improving the accuracy of the filling. This comprehensive filling strategy significantly improves the integrity and reliability of hydrological data, providing a high-quality data foundation for subsequent anomaly data identification and early warning.

[0097] However, in practical applications, the diversity of hydrological data can affect the processing efficiency of cloud servers. Each type of hydrological data may also have different temporal resolutions and spatial distribution characteristics. This multi-dimensional nature of the data may lead to problems such as excessive computational load and prolonged response time for cloud servers when processing large amounts of data.

[0098] Please see Figure 2 , Figure 2 This is another schematic diagram of the hydrological early warning process in the embodiments of this application; in some embodiments, after step S106, the process further includes:

[0099] S201. Obtain the filled hydrological data through a joint database query;

[0100] A database join query refers to a query operation that retrieves data from multiple related database tables simultaneously.

[0101] In some embodiments, each type of hydrological data is stored in a separate dedicated database, or, within the same database, logical partitions or tablespaces are created for different types of hydrological data.

[0102] In some embodiments, the required information is extracted from multiple database tables that may store post-filled hydrological data by using SQL or other database query tools.

[0103] Please see Figure 3 , Figure 3 yes Figure 2 A schematic diagram of a specific process for step S201; in some embodiments, step S201 specifically includes:

[0104] S2011. Create a joint query. The joint query connects all the filled hydrological data in the database through a common field. Each database stores one type of filled hydrological data.

[0105] Common fields refer to fields that have the same meaning and data type in different databases or tables, and are used to establish relationships between data.

[0106] In some embodiments, all databases requiring querying are identified, and then common fields used for association in each database are identified, such as timestamps, geographic coordinates, or site IDs. Next, an SQL query is constructed or a database management tool is used to join the filled hydrological data from these databases via a JOIN operation. This join query allows for the retrieval of comprehensive information from multiple data sources in a single operation.

[0107] S2012. In the joint query, extract the filled hydrological data corresponding to the preset selection conditions according to the preset selection conditions. The selection conditions include: specifying a time range, selecting specific hydrological parameters, and setting geographical location filtering conditions.

[0108] Among these, preset selection criteria refer to predefined standards or rules used to filter data. Specifying a time range indicates selecting data within a specific time period. Specific hydrological parameters refer to specific hydrological indicators, such as water level and flow rate. Geographical location filtering criteria are used to select data based on spatial information, such as specific watersheds or monitoring stations.

[0109] In some embodiments, specific selection criteria are determined based on user needs or analytical objectives, including a time range (e.g., the past week or a hydrological year), hydrological parameters of interest (e.g., average daily water level or instantaneous flow), and geographical location (e.g., a river or a lake). These criteria are then translated into filtering statements in a query language, such as WHERE or HAVING clauses, and added to the previously created union query. Finally, the modified query is executed to extract only the imputed hydrological data that meets all preset criteria.

[0110] As can be seen, creating a joint query connects all the filled hydrological data in the database through common fields. The joint query introduces preset selection conditions, including specifying a time range, selecting specific hydrological parameters, and setting geographical location filters. This enables the rapid and accurate acquisition of key hydrological data within a specific spatiotemporal range, greatly improving the efficiency and relevance of data analysis.

[0111] S202. Apply data fusion algorithms to process the filled hydrological data to generate an integrated dataset;

[0112] In some embodiments, a data fusion algorithm, such as Kalman filtering, Bayesian inference, or deep learning methods, is selected to process the incomplete hydrological data.

[0113] Step S107 is replaced by S203, which involves transmitting the integrated dataset to the cloud server, enabling the cloud server to identify abnormal data in the integrated dataset.

[0114] As can be seen, retrieving the filled hydrological data through joint database queries enables unified access and processing of different types of hydrological data, overcoming the low query efficiency caused by scattered data storage. Applying data fusion algorithms to process the filled hydrological data generates an integrated dataset. Integrating hydrological data from different sources and types improves data consistency. The integrated dataset is then transmitted to the cloud server, rather than directly transmitting the original filled data. This method significantly reduces the amount of data that needs to be transmitted and processed, alleviating the computational load on the cloud server.

[0115] However, in practical use, relying solely on the integrated dataset may lead to data quality issues, thereby affecting the cloud server's ability to identify anomalies. For example, the data integration process may cause the loss or distortion of certain details in the original hydrological data, affecting the accuracy of the integrated dataset. During data fusion, some anomalous features may be averaged or masked, making anomalies in the integrated dataset less obvious than in the original data. Errors from various types of hydrological data may accumulate and amplify during the integration process, leading to a decrease in the overall accuracy of the integrated dataset. The temporal and spatial resolutions of different types of hydrological data may be unified during the integration process, resulting in the loss of some important local or short-term variation features.

[0116] Please see Figure 4 , Figure 4 yes Figure 2 A schematic diagram of a specific process for step S202; in some embodiments, step S202 specifically includes:

[0117] S2021. Standardize the filled hydrological data of different types in the same space and at the same time, so that the filled hydrological data of different dimensions and units can be converted into a unified dimensionless numerical range.

[0118] Dimensionless numerical range refers to a range of pure numerical values ​​without specific physical units, usually from 0 to 1 or from -1 to 1.

[0119] In some embodiments, the types of hydrological data requiring standardization are determined. For each type of data, the maximum and minimum values ​​for that type are extracted from historical data. A min-max standardization method is applied. This method uses a linear transformation to map the raw data to a specified new interval. Each raw data point is converted to a standardized value using the min-max standardization method. This process maps all data points to the range of 0 to 1. This achieves the unification of different types of hydrological data to the same scale. For example, rainfall and flow data that originally had different units are now converted to the same standardized range.

[0120] S2022. According to preset rules, assign probability values ​​to each subset in the identification framework for each type of imputed hydrological data, resulting in basic probability assignments for multiple evidence sources. The identification framework contains a set of mutually exclusive hypotheses related to hydrological warnings. The identification framework represents the set used in evidence theory to describe all possible hypotheses. Basic probability assignment refers to a function value assigned to each subset in the identification framework in evidence theory, representing the support for that subset. A mutually exclusive set of hypotheses refers to a set of mutually exclusive possible situations or outcomes.

[0121] In some embodiments, a recognition framework is first defined as {normal, mild attention, moderate attention, severe attention, and extreme attention};

[0122] Based on pre-defined rules (potentially based on expert knowledge or historical data analysis), the support level for each hypothesis is calculated for each type of standardized hydrological data (such as water level, flow rate, and rainfall). This process transforms different types of hydrological data into a series of probability distributions, each reflecting the judgment of that type of data for different warning levels.

[0123] For example: if the water level is between 0 and 2 meters: m({normal}) = 0.8, m({normal, slight precaution}) = 0.2;

[0124] If the water level is between 2 and 4 meters: m({minor caution}) = 0.6, m({minor caution, moderate caution}) = 0.4;

[0125] If the water level is between 4 and 6 meters: m({moderate caution}) = 0.7, m({moderate caution, severe caution}) = 0.3;

[0126] If the water level exceeds 6 meters: m({Severe precaution, Extreme precaution}) = 1.0;

[0127] S2023. Based on the basic probability allocation of multiple evidence sources, a fused dataset is obtained, wherein: all possible combinations of intersections among multiple evidence sources to obtain the target set are identified; the sum of the probability products of the possible combinations is calculated; and a normalization factor is applied to adjust the sum of the probability products.

[0128] The process begins by identifying all target sets for which fusion probabilities need to be calculated, including individual hypotheses and possible combinations of hypotheses. Then, for each target set, all possible combinations of evidence sources that could lead to that set are identified. Next, the sum of the probability products of these combinations is calculated, representing the combined support of all evidence for that target set. Finally, these probability sums are adjusted using a normalization factor to ensure that the final probability distribution satisfies probability axioms. This process integrates information from different hydrological data sources, resulting in a more comprehensive and reliable early warning probability distribution.

[0129] For example:

[0130] m(A)=K*Σ(m1(B)*m2(C)),B∩C=A

[0131] In the formula, m(A) is the combination of evidence sources m1(B) and evidence sources m2(C), where m1(B) is one evidence source, such as water level data, and m2(C) is another evidence source, such as rainfall data. K is the normalization factor.

[0132] Continuing from the previous example, we need to integrate water level data (m1) and rainfall data (m2):

[0133] m1({Mild Alert}) = 0.6, m1({Mild Alert, Moderate Alert}) = 0.4

[0134] m2({Normal, Mild Alert}) = 0.3, m2({Mild Alert, Moderate Alert}) = 0.7

[0135] The above rules can be applied to obtain the following results.

[0136] m({Mild Alert}) = 0.55, m({Moderate Alert}) = 0.28, m({Mild Alert, Moderate Alert}) = 0.17

[0137] S2024. Combine the merged dataset with the filled hydrological data to generate an integrated dataset.

[0138] It should be noted that the merged dataset is primarily intended as a reference, providing preliminary indications of potential anomalies for cloud servers, rather than serving as the basis for final decision-making.

[0139] It is evident that standardizing different types of infilled hydrological data resolves the issue of varying dimensions and units, laying the foundation for subsequent fusion. Secondly, an evidence-based data fusion algorithm is introduced to assign probability values ​​to each type of hydrological data, constructing a basic probability allocation for multiple evidence sources. This step fully considers the reliability and uncertainty of different data sources. Then, by identifying the intersection of multiple evidence sources, calculating the sum of probability products, and applying a normalization factor, the fused dataset is finally obtained. This effectively handles data uncertainty and conflict, significantly improving the reliability of the fusion results. Finally, combining the fused dataset with the original infilled data retains both the comprehensive advantages of fusion and the detailed information of the original data. This method effectively solves the problems of information loss and anomaly masking that may arise from simply relying on integrated data.

[0140] The following describes an exemplary hydrological early warning model 500 provided in the embodiments of this application. Figure 5 This is an exemplary hardware structure diagram of the hydrological early warning model 500 provided in this application embodiment.

[0141] In some embodiments, the hydrological early warning model 500 is a computer device or includes a computer device in the hydrological early warning model 500. The computer device includes a processor, memory, and a network interface connected via a system bus. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer device stores data. The network interface of the computer device is used to communicate with other external terminals or servers via a network connection. In some embodiments, the network interface can be a wired network interface; in some embodiments, the network interface can also be a wireless network interface. When the computer program is executed by the processor, it implements the methods in the embodiments of this application.

[0142] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0143] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0144] As used in the above embodiments, depending on the context, the term "when..." can be interpreted as meaning "if...", "after...", "in response to determining...", or "in response to detecting...". Similarly, depending on the context, the phrase "when determining..." or "if (the stated condition or event) is interpreted as meaning "if determining...", "in response to determining...", "when (the stated condition or event) is detected", or "in response to detecting (the stated condition or event)".

[0145] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive), etc.

[0146] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM or random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A hydrological early warning method, characterized in that, include: Fill the hydrological data into the preset hydrological table and identify the missing data items in the hydrological table; Based on the temporal span and spatial distribution of consecutively missing data in the missing data items, the amount of data to be filled and the filling area are determined, wherein the consecutively missing data refers to any data in the hydrological table that is consecutively missing in terms of temporal span or spatial distribution, wherein: When the ratio of the amount of continuously missing data to the corresponding total amount of data is less than a preset threshold, a full filling measurement is performed; When the ratio of the amount of consecutive missing data to the corresponding total data is not less than the preset threshold, a selective imputation strategy is executed. The selective imputation strategy involves selecting several points in the front, middle and back parts of the consecutive missing data to imput them. The data imputation amount is calculated based on the imputation ratio and the amount of consecutively missing data; the ratio of the amount of consecutively missing data to the corresponding total data is inversely proportional to the imputation ratio. Based on the type of the continuously missing data, select the appropriate data imputation method to impute it; The filled hydrological data is transmitted to a cloud server, enabling the cloud server to identify abnormal data in the filled hydrological data. Receive an early warning signal sent by the cloud server, wherein the early warning signal is issued by the cloud server when the abnormal data exceeds a threshold; Send early warning information to the terminals of relevant personnel.

2. The method according to claim 1, characterized in that, After the step of selecting an appropriate data imputation method based on the type of missing data, the method further includes: Obtain the filled hydrological data through a joint database query; The data fusion algorithm is applied to process the filled hydrological data to generate an integrated dataset; The step of transmitting the filled hydrological data to a cloud server and enabling the cloud server to identify abnormal data in the filled hydrological data specifically includes: The integrated dataset is transmitted to a cloud server, enabling the cloud server to identify abnormal data in the integrated dataset.

3. The method according to claim 2, characterized in that, The step of obtaining the filled hydrological data through a joint database query specifically includes: Create a join query that connects the filled hydrological data from all databases through a common field, where each database stores one type of filled hydrological data. In the joint query, the filled hydrological data corresponding to the preset selection conditions are extracted according to the preset selection conditions. The selection conditions include: specifying a time range, selecting specific hydrological parameters, and setting geographical location filtering conditions.

4. The method according to claim 3, characterized in that, The steps of applying the data fusion algorithm to process the filled hydrological data and generate the integrated dataset specifically include: The filled hydrological data of different types in the same space and at the same time are standardized to convert the filled hydrological data of different dimensions and units into a unified dimensionless numerical range. According to preset rules, probability values ​​are assigned to each subset in the identification framework for each type of filled hydrological data, resulting in basic probability assignments for multiple evidence sources; the identification framework includes a set of mutually exclusive hypotheses related to hydrological early warning. The fused dataset is obtained based on the basic probability assignment of multiple evidence sources, wherein: all possible combinations of the intersection of the multiple evidence sources to obtain the target set are identified; the sum of the probability products of the possible combinations is calculated; and a normalization factor is applied to adjust the sum of the probability products. The fused dataset is combined with the filled hydrological data to generate the integrated dataset.

5. The method according to claim 1, characterized in that, The step of selecting an appropriate data imputation method based on the type of the consecutive missing data specifically includes: Identify the type of the continuously missing data and classify it into time series data and spatial series data; For time series data, time series analysis methods are used to fill in the gaps. For spatial sequence data, spatial statistical methods are applied to fill in the gaps.

6. The method according to claim 5, characterized in that, The steps for filling in time series data using time series analysis methods specifically include: When the ratio of the amount of continuously missing data to the corresponding total amount of data is less than a preset threshold, interpolation is used to fill the missing data. When the ratio of the amount of continuously missing data to the corresponding total data is not less than a preset threshold, time series decomposition is used to fill the missing data.

7. The method according to claim 1, characterized in that, The step of selecting an appropriate data imputation method based on the type of the consecutive missing data specifically includes: For spatiotemporal mixed data, obtain a set of known data points that include time coordinates, spatial coordinates, and observation values; Based on spatiotemporal distance, nearby known data points are selected for the continuously missing data; Using a predetermined spatiotemporal variability function, construct a system of Kriging equations; Solve the Kriging equations to obtain the weight vector; Using the weight vector and the observed values ​​of known data points, the imputation values ​​for the consecutive missing data are calculated.

8. A hydrological early warning model, characterized in that, The hydrological early warning model includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to cause the hydrological early warning model to perform the method as described in any one of claims 1-7.

9. A computer program product containing instructions, characterized in that, When the computer program product is run on the hydrological early warning model, the hydrological early warning model performs the method as described in any one of claims 1-7.

10. A computer-readable storage medium comprising instructions, characterized in that, When the instruction is executed on the hydrological early warning model, the hydrological early warning model performs the method as described in any one of claims 1-7.

Images