Multi-model multi-boundary estuary saltwater intrusion ensemble probability prediction method and system

Abstract

The application provides a multi-model multi-boundary estuary saltwater intrusion ensemble probability prediction method and system, wherein the method is coupled based on a plurality of calibrated and verified numerical models, a plurality of runoff boundaries and a plurality of wind field boundary conditions, an ensemble prediction system of multiple source uncertainty disturbances is constructed, and saltwater intrusion evolution results under a plurality of prediction scenarios are generated. Further based on historical observation data and model performance evaluation indexes, a sub-item confidence probability weighting method of the numerical model, the runoff boundary and the wind field boundary is established, and probability distribution estimation of the ensemble prediction result is realized. The method can quantitatively represent the probability and uncertainty of the occurrence of a saltwater intrusion event in a water source, and improve the accuracy and reliability of the estuary saltwater intrusion prediction.

Description

Technical Field

[0001] This invention relates to the field of estuarine saltwater intrusion forecasting technology, specifically to an ensemble probability forecasting method and system for estuarine saltwater intrusion based on multi-model and multi-boundary-condition coupling. Background Technology

[0002] Estuaries, as the confluence of terrestrial freshwater and marine saltwater, are crucial for water resource utilization, ecological environmental protection, and socio-economic development. Saltwater intrusion processes are influenced by a combination of dynamic factors, including runoff, tides, wind fields, and topography, exhibiting significant nonlinear characteristics and multi-scale spatiotemporal variations, resulting in considerable uncertainty in their evolution. During dry seasons or under extreme weather conditions, the intensity and extent of saltwater intrusion are highly sensitive to external dynamic conditions. Once saltwater intrusion occurs, it pollutes estuarine freshwater resources, impacting domestic water use, industrial water use, and agricultural irrigation, disrupting the estuarine ecosystem balance. Therefore, accurate forecasting of estuarine saltwater intrusion is of significant engineering value for ensuring estuarine water supply security and optimizing water resource allocation.

[0003] Currently, most existing saltwater intrusion forecasting methods are based on a single numerical model and deterministic initial boundary conditions. They are difficult to simultaneously characterize model structural errors, initial field errors, and external forcing uncertainties, and cannot quantitatively provide probability information on the occurrence of saltwater intrusion. This limits the application of forecast results in risk assessment and engineering scheduling.

[0004] A search of patent literature revealed Chinese invention patent application number CN201310501628.5, which discloses a method for predicting estuarine salinity based on tides and runoff. This method selects key parameters such as tidal range and runoff to establish a statistical relationship model between salinity and tides and runoff, thereby enabling the prediction and forecasting of estuarine salinity. By fitting the functional relationship between salinity and tidal range and runoff, this method simplifies the complex saltwater intrusion process into a statistical model, thus achieving short-term forecasts of 1–7 days.

[0005] However, the aforementioned existing technologies still have the following shortcomings: First, the forecasting methods based on statistical models oversimplify the complex dynamic processes of estuaries, considering only a few factors such as tides and runoff, making it difficult to reflect the impact of wind fields, topography, and multi-scale dynamic coupling on saltwater intrusion; second, although machine learning-based methods can improve fitting accuracy, they lack clear physical mechanism constraints and have limited ability to predict extreme scenarios and extrapolation; at the same time, traditional numerical model forecasts mostly rely on a single model and deterministic initial and boundary conditions, failing to systematically consider model structure errors and uncertainties of multi-source boundary conditions; furthermore, existing methods generally rely on deterministic forecast results, lacking a quantitative expression of forecast uncertainties and failing to provide information on the probability and risk of saltwater intrusion.

[0006] In summary, given the problems of the existing technologies, researching a multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction method and system has become a critical task that urgently needs to be addressed. Summary of the Invention

[0007] To address the shortcomings of existing technologies, the purpose of this invention is to provide a multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction method and system.

[0008] A multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction method provided by the present invention includes the following steps: Step S1: Obtain static topographic data of the target estuary area, select multiple calibrated and verified numerical models, each numerical model reads static topographic data and runs independently until the salinity field is stable, generate the initial salinity field corresponding to each numerical model, and the numerical model set is composed of multiple numerical models and their corresponding initial salinity fields. Step S2: For the external dynamic factors affecting saltwater intrusion, obtain runoff forecast products and wind field forecast products from multiple sources within the forecast period, and construct runoff boundary sets and wind field boundary sets respectively. Step S3: Pair each numerical model in the numerical model set, each runoff boundary in the runoff boundary set, and each wind field boundary in the wind field boundary set one by one. Each pairing result is used as a forecast scenario. For each forecast scenario, perform numerical calculations of brine intrusion using its corresponding numerical model, runoff boundary, wind field boundary, and initial salinity field to obtain the salinity spatiotemporal distribution forecast results for each forecast scenario. Step S4: Based on the historical simulation accuracy of each numerical model, each runoff boundary, and each wind field boundary, calculate the model confidence probability, runoff confidence probability, and wind field confidence probability respectively, and then calculate the comprehensive confidence probability of each forecast scenario. Step S5: Based on the salinity spatiotemporal distribution forecast results and corresponding comprehensive confidence probabilities for each forecast scenario, perform weighted statistics on the characteristic variables of saltwater intrusion, construct a joint probability distribution table, and output the probability forecast results of saltwater intrusion.

[0009] Preferably, the selection criteria for the numerical model in step S1 are: the numerical models of saltwater intrusion in each estuary are verified using historical data, the calculation and simulation accuracy reaches the preset threshold, and each numerical model has differences in physical parameterization scheme or numerical solution method.

[0010] Preferably, the external dynamic factors in step S2 include runoff and wind field, and the runoff boundary set is constructed through the following sub-steps: Step S2.1: Obtain runoff forecast products from multiple different sources within the forecast period as the basic runoff boundary; Step S2.2: Based on historical forecast error statistics, generate positive deviation runoff boundaries and negative deviation runoff boundaries for each basic runoff boundary; Step S2.3: The basic runoff boundary, the positive deviation runoff boundary, and the negative deviation runoff boundary are combined to form a runoff boundary set.

[0011] Preferably, the wind field boundary set in step S2 includes wind field forecast products from different sources: wind field simulated and forecasted by the WRF model, wind field forecasted by the GFS global forecast system, and wind field forecasted by the Shanghai Typhoon Institute of the China Meteorological Administration.

[0012] Preferably, the total number of forecast scenarios in step S3 is equal to the product of the number of models in the numerical model set, the number of runoff boundaries in the runoff boundary set, and the number of wind field boundaries in the wind field boundary set.

[0013] Preferably, in step S4, the comprehensive confidence probability is obtained by multiplying the model confidence probability, the runoff confidence probability, and the wind field confidence probability; wherein, the model confidence probability is calculated by weighted normalization based on the model score SS, the runoff confidence probability is calculated by weighted normalization based on the relative error square RS, and the wind field confidence probability is calculated by weighted normalization based on the vector relative error square VRS.

[0014] Preferably, the formula for calculating the model score SS is:

[0015] in, n is the total number of samples that can be compared.

[0016] Indicates the first Measured salinity values ​​for each sample.

[0017] Indicates the first Simulated salinity values ​​for each sample.

[0018] This represents the average of the measured salinity values ​​for all samples.

[0019] If SS is less than or equal to the preset threshold, the corresponding model confidence probability is set to 0; otherwise, the model confidence probability of the i-th numerical model is set to 0. Calculate using the following formula:

[0020] Where N is the total number of numerical models. The preset model sharpness, For the first Model scores for each numerical model.

[0021] Preferably, the formula for calculating the vector relative error balance (VRS) is:

[0022] This represents the measured wind vector of the i-th sample; This represents the wind vector predicted for the i-th sample. This represents the total number of samples involved in the statistical analysis (i.e., the total time series length of the multi-sites used for evaluation). The confidence probability of the wind field at the i-th wind field boundary Calculate using the following formula:

[0023] Where M represents the total number of wind field boundaries, The preset third sharpness; This represents the relative error margin of the i-th wind field boundary forecast product; The formula for calculating the relative error (RS) is:

[0024] in, This represents the measured value of runoff.

[0025] This represents the predicted runoff value.

[0026] This indicates the total number of measured runoff samples included in the statistics (i.e., the length of the time series used for evaluation).

[0027] Runoff confidence probability at the i-th runoff boundary Calculate using the following formula:

[0028] Where Q represents the total number of runoff boundaries. The preset second sharpness, The table shows the relative error of the i-th boundary forecast product.

[0029] Preferably, step S5 includes: Step S5.1: Extract the characteristic variables of saltwater intrusion from the salinity spatiotemporal distribution forecast results of each forecast scenario. The characteristic variables include the number of saltwater intrusions and the total duration of saltwater intrusions. Step S5.2: Based on the number of times saltwater intrusion occurs, each forecast scenario is divided into three categories: no saltwater intrusion, one saltwater intrusion, and two or more saltwater intrusions. Step S5.3: Based on the comprehensive confidence probability of each forecast scenario, calculate the probability of occurrence of different numbers of saltwater intrusions and the corresponding total duration of saltwater intrusions; Step S5.4: Construct a joint probability distribution table containing the number of brine intrusions, their corresponding probabilities, and the total duration of the brine intrusions; Step S5.5: Based on the joint probability distribution table, output the probability of 0, 1, 2 or more brine intrusions and the total duration of brine intrusions for each category.

[0030] This invention also provides a multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction system, comprising: Module M1 acquires static topographic data of the target estuary area, selects multiple calibrated and verified numerical models, each numerical model reads static topographic data and runs independently until the salinity field is stable, generates the initial salinity field corresponding to each numerical model, and the numerical model set consists of multiple numerical models and their corresponding initial salinity fields. Module M2, targeting the external dynamic factors affecting saltwater intrusion, acquires runoff forecast products and wind field forecast products from multiple sources within the forecast period, which respectively constitute the runoff boundary set and the wind field boundary set; Module M3 pairs each numerical model in the numerical model set, each runoff boundary in the runoff boundary set, and each wind field boundary in the wind field boundary set one by one. Each pairing result is used as a forecast scenario. For each forecast scenario, the corresponding numerical model, runoff boundary, wind field boundary, and initial salinity field are used to perform numerical calculations of brine intrusion to obtain the salinity spatiotemporal distribution forecast results for each forecast scenario. Module M4 calculates the model confidence probability, runoff confidence probability, and wind field confidence probability based on the historical simulation accuracy of each numerical model, each runoff boundary, and each wind field boundary, and then calculates the comprehensive confidence probability of each forecast scenario. Module M5 performs weighted statistics on the characteristic variables of saltwater intrusion based on the salinity spatiotemporal distribution forecast results and corresponding comprehensive confidence probabilities for each forecast scenario, constructs a joint probability distribution table, and outputs the probability forecast results of saltwater intrusion.

[0031] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention solves the technical problems of traditional saltwater intrusion forecasting methods, which are difficult to comprehensively characterize multi-source uncertainties and cannot provide probabilistic forecasting results, by using a multi-model, multi-boundary ensemble probabilistic forecasting method. It realizes the transformation of saltwater intrusion process from deterministic prediction to probabilistic prediction.

[0032] 2. This invention constructs a set of numerical models by selecting multiple numerical models that have differences and have been calibrated and verified, effectively avoiding the structural errors of a single numerical model; it comprehensively covers the uncertainties of external dynamic conditions by constructing a set of multi-source runoff boundaries and a set of wind field boundaries; it obtains data on the evolution of saltwater intrusion under multiple scenarios by combining the models and boundaries to form forecast scenarios and calculating the spatiotemporal distribution of salinity; it achieves quantitative weighting of different forecast results by calculating the comprehensive confidence probability of each forecast scenario based on the historical simulation accuracy; and it can quantitatively give the probability distribution of different occurrences and total durations of saltwater intrusion by constructing a joint probability distribution table and outputting the probability forecast results.

[0033] 3. This invention improves the reliability and information integrity of saltwater intrusion forecast results, enhances the risk identification capability for extreme saltwater intrusion events, and provides quantifiable technical support for estuarine water supply security, scientific water resource allocation, water resource management, and ecological environmental protection. Attached Figure Description

[0034] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a flowchart of a multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction method in an embodiment of the present invention. Detailed Implementation

[0035] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0036] To achieve accurate prediction of the uncertainty and probability of estuarine saltwater intrusion events, this invention proposes a multi-model, multi-boundary ensemble probabilistic forecasting method and system for estuarine saltwater intrusion. The method couples multiple calibrated and validated numerical models with various runoff and wind field boundary conditions. By constructing an ensemble forecasting system with multi-source uncertainties, it generates saltwater intrusion evolution results under multiple forecast scenarios. Furthermore, based on historical observation data and model performance evaluation indicators, a weighted method for the confidence probabilities of the numerical models, runoff boundaries, and wind field boundaries is established to estimate the probability distribution of the ensemble forecast results. This method can quantitatively characterize the probability and uncertainty of saltwater intrusion events in water source areas, improving the accuracy and reliability of estuarine saltwater intrusion forecasting.

[0037] Example 1: Figure 1This is a flowchart of a multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction method in an embodiment of the present invention.

[0038] like Figure 1 As shown in the figure, this embodiment provides a multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction method, including the following steps: Step S1: Obtain static topographic data of the target estuary area, select multiple calibrated and verified numerical models, each numerical model reads static topographic data and runs independently until the salinity field is stable, generate the initial salinity field corresponding to each numerical model, and the numerical model set is composed of multiple numerical models and their corresponding initial salinity fields.

[0039] In this embodiment, multiple estuarine saltwater intrusion numerical models that have been calibrated and validated using historical data are selected, including at least two of the UnFECOM-i, ECOM-si, SCHISM, and ROMS models. The selection criteria for the numerical models are: each estuarine saltwater intrusion numerical model is validated using historical data; its simulation accuracy reaches a preset threshold; and each numerical model exhibits significant differences in physical parameterization schemes (such as bottom friction and turbulent closure schemes) or numerical solution methods to overcome structural biases inherent in a single model. Each estuarine saltwater intrusion numerical model independently reads uniform static topographic data and is driven by idealized or climatological boundary conditions (such as runoff, wind field, and tides) until the salinity field change is less than a preset threshold (i.e., the salinity field stabilizes), thereby generating its corresponding initial salinity field. This initial salinity field serves as the initial condition for subsequent ensemble forecasts. The aforementioned multiple estuarine saltwater intrusion numerical models and their corresponding initial salinity fields constitute a numerical model ensemble.

[0040] Step S2: For the two key external dynamic factors affecting saltwater intrusion, obtain runoff forecast products and wind field forecast products from multiple sources within the forecast period, which constitute the runoff boundary set and the wind field boundary set, respectively.

[0041] In this embodiment, the external dynamic factors include runoff and wind field, and the forecast period is 7 days or 28 days.

[0042] Specifically, the runoff boundary set is constructed through the following sub-steps: Step S2.1: Obtain runoff forecast products from multiple different sources within the forecast period as the basic runoff boundary, such as the forecast from the Hydrology Bureau of the Yangtze River Water Resources Commission and the forecast from the Shanghai Marine Monitoring and Forecasting Center.

[0043] Step S2.2: Based on historical forecast error statistics, positive deviation runoff boundaries and negative deviation runoff boundaries are generated for each basic runoff boundary to characterize forecast uncertainty.

[0044] Specifically, based on the statistical results of the maximum error estimation of the short-term runoff forecast samples from the Yangtze River Water Resources Commission, the historical forecast error standard deviation of each runoff forecast data is calculated by comparing the forecast runoff volume within the forecast period with the measured value during the same period. The positive deviation runoff boundary at each time point is the forecast runoff volume plus the historical forecast error standard deviation, and the negative deviation runoff boundary is the forecast runoff volume minus the historical forecast error standard deviation.

[0045] Step S2.3: The basic runoff boundary, the positive deviation runoff boundary, and the negative deviation runoff boundary are combined to form a runoff boundary set.

[0046] Furthermore, the wind field boundary set includes wind field forecast products from different sources: wind field simulations and forecasts from the WRF (Weather Research and Forecasting Model), wind field forecasts from the GFS (Global Forecasting System), and wind field forecasts from the Shanghai Typhoon Institute of the China Meteorological Administration, generating diverse wind field boundaries.

[0047] Step S3: Pair each numerical model in the numerical model set, each runoff boundary in the runoff boundary set, and each wind field boundary in the wind field boundary set one by one. Each pairing result is a forecast scenario. For each forecast scenario, use the numerical model corresponding to the forecast scenario; use the runoff boundary and wind field boundary corresponding to the forecast scenario as driving inputs; use the initial salinity field corresponding to the forecast scenario as initial conditions, and run the numerical calculation of brine intrusion to obtain the forecast results of salinity spatiotemporal distribution under each forecast scenario.

[0048] In this embodiment, ensemble forecasting estimates the probability distribution of forecast errors using different models, boundary conditions, and perturbations, thereby providing uncertainty information. Numerous studies have shown that ensemble forecasting offers higher forecast skill compared to single deterministic forecasts. Based on this, ensemble forecasting of saltwater intrusion into target estuaries (such as the Yangtze River estuary) is conducted by combining pairs of the aforementioned numerical model ensemble, runoff boundary ensemble, and wind field boundary ensemble.

[0049] Specifically, each forecast scenario includes a numerical model, a runoff boundary, and a wind field boundary. The total number of forecast scenarios = number of models × number of runoff boundaries × number of wind field boundaries.

[0050] The numerical model for estuarine brine intrusion used in this embodiment is based on the three-dimensional incompressible fluid governing equations (including the momentum equation, continuity equation, and salinity transport equation, which describe the spatiotemporal evolution of estuarine tidal currents and salinity fields), and employs non-orthogonal curves. Grid and vertical The mesh, combined with the Mellor-Yamada 2.5th order turbulence closure scheme, realistically simulates the evolution of estuarine salinity.

[0051] Specifically, The momentum equation for the direction is:

[0052] The momentum equation for the direction is:

[0053] t represents time; The vertical velocity is given in the Cartesian coordinate system. The height of the free surface; The density of seawater is determined by salinity. and temperature Calculated using the seawater state equation; For reference density, take 1000 kg / m³; The vertical velocity is given in Cartesian coordinates. The continuity equation is:

[0054] in:

[0055] in, for Vertical velocity in the coordinate system; The salinity transport equation is:

[0056] in, Salinity; This is the salinity horizontal diffusion term.

[0057] The temperature transport equation is:

[0058] in, For temperature; This is the horizontal heat diffusion term; In the above system of equations, It is the Coriolis force coefficient. It is gravitational acceleration. , , It is the Jacobian function. , and Defined as:

[0059] , It is the depth of still water. The total water depth.

[0060] Vertical coordinates On the sea surface On the seabed .

[0061] h represents water depth / seabed topographic depth, indicating the distance from the datum to the seabed, used to define the σ coordinate. The h1, h2, and h3 above represent the "metric coefficients" for coordinate transformation. , , Cartesian coordinates; = , = , = , = .

[0062] and The flow rates are respectively at and Directional components are defined as follows:

[0063] in and Let be the horizontal velocity in the Cartesian coordinate system. and For definition in and Velocity in the normal direction:

[0064] and These are the vertical turbulent viscosity coefficient and the vertical turbulent diffusion coefficient, respectively. The Mellor-Yamada 2.5th order turbulent closed-loop model was used for calculation.

[0065] Horizontal momentum diffusion term:

[0066]

[0067] Salinity and heat diffusion terms:

[0068]

[0069] in, The horizontal viscosity coefficient, The horizontal diffusion coefficient is calculated using parameterized formulas for both and its value is the same as... Consistent.

[0070] The computation process, namely the numerical solution and time progression of the model, is used to discretize the above three-dimensional control equations and solve them in parallel on a heterogeneous computing platform.

[0071] Step S4: Based on the historical simulation accuracy of each numerical model, each runoff boundary, and each wind field boundary, calculate the model confidence probability, runoff confidence probability, and wind field confidence probability respectively, and then calculate the comprehensive confidence probability of each forecast scenario.

[0072] The historical simulation accuracy is quantified by the preliminary validation performance scores of each numerical model, each runoff boundary, and each wind field boundary. To calculate these preliminary validation performance scores, historical hydrological salinity observation data for the target estuary area must be obtained. A higher preliminary validation performance score indicates higher historical simulation accuracy and a corresponding higher confidence probability. The calculation and allocation of each confidence probability follows the principle of first normalizing the individual components, then multiplying and summing the results.

[0073] Furthermore, the combined confidence probability of each forecast scenario is obtained by multiplying the model confidence probability, runoff confidence probability, and wind field confidence probability. Specifically, the confidence probability of the numerical model is calculated using exponential weighted normalization based on its previous simulation accuracy (e.g., model score SS); the runoff confidence probability is calculated using exponential weighted normalization based on the average relative error (RS); and the wind field confidence probability is calculated using exponential weighted normalization based on the average vector relative error (VRS). All component probabilities are guaranteed to sum to 1.

[0074] Specifically, step S4 includes the following sub-steps: Step S4.1: The Model Score (SS) is used as the metric to evaluate the early validation performance of each numerical model. The formula for calculating SS is:

[0075] in, n is the total number of samples that can be compared.

[0076] This represents the measured salinity value of the i-th sample.

[0077] This represents the simulated salinity value of the i-th sample.

[0078] This represents the average of the measured salinity values ​​for all samples.

[0079] First, perform an initial screening of the SS results. If the SS of the numerical model is less than or equal to a preset threshold (e.g.: If the historical simulation accuracy of the numerical model is insufficient, the model confidence probability is set to 0 and it will not participate in the final probability distribution calculation; if SS is greater than the preset threshold ( Then, the SS value of the numerical model is subjected to exponential weighted normalization, and the SS value is calculated according to the following formula. Model confidence probability of a numerical model :

[0080] Where N is the total number of numerical models participating in the set (4 in this embodiment). For the first Model scores for each numerical model The lower the preset model sharpness, the higher the model confidence probability obtained by the numerical model with a higher initial validation performance score (SS score). The larger the value, the greater the sum of the model confidence probabilities of all numerical models, which is 1.

[0081] Step S4.2: Since the wind field boundary forecast results are vectors, when evaluating the accuracy performance of the early-stage wind field boundary forecast products, the Vector Relative Error Balance (VRS) is used as the indicator to evaluate the early-stage verification performance score of each wind field boundary. The formula for calculating VRS is:

[0082] This represents the measured wind vector of the i-th sample; This represents the wind vector predicted for the i-th sample. This represents the total number of samples participating in the statistical analysis (i.e., the total time series length of the multi-site data used for evaluation).

[0083] The VRS value of each wind field boundary is subjected to exponential weighted normalization to calculate the wind field confidence probability of the i-th wind field boundary. The formula is as follows:

[0084] Where M is the total number of wind field boundaries used in the ensemble forecast (3 in this embodiment). This represents the relative error margin of the i-th wind field boundary forecast product. The third sharpness is a preset value; the smaller the value, the higher the confidence probability of the wind field obtained from the wind field boundary in the early verification performance score (VRS value). The larger the value, the greater the sum of the confidence probabilities of the wind field at all wind field boundaries, which is 1.

[0085] Step S4.3: When evaluating the performance of the early runoff boundary forecast products and the accuracy score of the extreme error probability construction, the relative error balance (RS) is used as the indicator to evaluate the early validation performance score of each runoff boundary. The formula for calculating RS is:

[0086] in, This represents the measured value of runoff.

[0087] This represents the predicted runoff value.

[0088] This indicates the total number of measured runoff samples included in the statistics (i.e., the length of the time series used for evaluation).

[0089] Then, the RS value of each runoff boundary is subjected to exponential weighted normalization to calculate the runoff confidence probability of the i-th runoff boundary. Calculate using the following formula:

[0090] Where Q is the total number of runoff boundaries used in the ensemble forecast (4 in this embodiment). The table shows the relative error of the i-th boundary forecast product. The lower the preset second sharpness, the higher the runoff confidence probability obtained from the runoff boundary with the higher the prior validation performance score (RS value). The larger the value, the greater the sum of the runoff confidence probabilities at all runoff boundaries, which is 1.

[0091] Step S4.4: Each forecast scenario corresponds to a specific set of numerical models, runoff boundaries, and wind field boundaries, and their combined confidence probability... The product of the three confidence probabilities mentioned above: .

[0092] The combined confidence probability of all forecast scenarios satisfies the normalization condition: .

[0093] Step S5: Based on the salinity spatiotemporal distribution forecast results of each forecast scenario and its corresponding comprehensive confidence probability, the characteristic variables of saltwater intrusion are weighted and statistically analyzed to construct a joint probability distribution table and output the probability forecast results of saltwater intrusion, thereby quantifying the uncertainty of the saltwater intrusion process.

[0094] Specifically, step S5 includes the following sub-steps: Step S5.1: Extract the characteristic variables of saltwater intrusion from the salinity spatiotemporal distribution forecast results of each forecast scenario. The characteristic variables include the number of saltwater intrusions and the total duration of saltwater intrusions.

[0095] The salinity spatiotemporal distribution forecast results include: the salinity spatial distribution field of each grid node in the calculation area, the three-dimensional salinity field or surface / bottom salinity field at each time step, and the salinity time series of the target control section or water source location.

[0096] Further, step S5.1 includes the following sub-steps: Step S5.1.1: Extract the salinity time series at the target water source or control section from the salinity spatiotemporal distribution forecast results; Step S5.1.2: Set a salinity threshold, perform threshold determination on the salinity time series, and identify the time period when the salinity continuously exceeds the salinity threshold; Step S5.1.3: Count the number of consecutive time intervals in which the salinity exceeds the salinity threshold, and use this number as the number of saltwater intrusions in this forecast scenario.

[0097] Step S5.1.4: For each consecutive time interval exceeding the salinity threshold, calculate its duration and sum them up to obtain the total duration of saltwater intrusion for the forecast scenario.

[0098] Step S5.2: Based on the number of saltwater intrusion occurrences, each forecast scenario is divided into the following three categories: If the number of brine intrusions is equal to 1, then it is determined that 1 brine intrusion has occurred. If the number of brine intrusions is greater than or equal to 2, it is determined as multiple (≥2) brine intrusions. If the number of brine intrusions is 0, it is determined that no brine intrusion has occurred. At the same time, the total duration of saltwater intrusion corresponding to each forecast scenario is recorded.

[0099] Step S5.3: Based on the comprehensive confidence probability of each forecast scenario, perform weighted statistics on the number of saltwater intrusions and the total duration of saltwater intrusions, and calculate the probability of saltwater intrusions of different numbers and the corresponding total duration of saltwater intrusions.

[0100] Specifically, the forecast scenarios are grouped according to the number of saltwater intrusions, and the formula for calculating the probability of occurrence for different numbers of saltwater intrusions is as follows: , Where Num represents the number of brine intrusions. The probability of Num saltwater intrusions. Let N be the combined confidence probability of Num intrusions occurring in the j-th forecast scenario.

[0101] Calculate the total duration of intrusion when Num brine intrusions occur: , in, For the occurrence Total duration of secondary brine intrusion Let be the duration of the k-th saltwater intrusion in the j-th forecast scenario.

[0102] Step S5.4: Construct a joint probability distribution table containing the number of saltwater intrusions, their corresponding probabilities, and the total duration of the saltwater intrusions.

[0103] Specifically, a joint probability distribution table is formed by considering all forecast scenarios based on three factors: the number of times saltwater intrusion occurs (including three categories: 0 times, 1 time, and 2 times or more), the probability of occurrence of this type of event for each forecast scenario, and the total duration of saltwater intrusion under this type of forecast scenario.

[0104] Step S5.5: Based on the joint probability distribution table, output the following probability prediction results.

[0105] Specifically, the probability of 0 brine intrusions, the probability of 1 brine intrusion, the probability of 2 or more brine intrusions, and the total duration of brine intrusions for each category.

[0106] For example, the probability of two or more saltwater intrusions at Qingcaosha Reservoir is 6.49%, with a total duration of 232 hours; the probability of one saltwater intrusion is 9.41%, with a duration of 126 hours; and the probability of no saltwater intrusion is 84.10%.

[0107] Example 2: This invention also provides a multi-model, multi-boundary estuarine saltwater intrusion ensemble probability forecasting system. The multi-model, multi-boundary estuarine saltwater intrusion ensemble probability forecasting system can be implemented by executing the process steps of the multi-model, multi-boundary estuarine saltwater intrusion ensemble probability forecasting method. That is, those skilled in the art can understand the multi-model, multi-boundary estuarine saltwater intrusion ensemble probability forecasting method as a preferred implementation of the multi-model, multi-boundary estuarine saltwater intrusion ensemble probability forecasting system.

[0108] Specifically, this multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction system includes: Module M1 acquires static topographic data of the target estuary area, selects multiple calibrated and verified numerical models, each numerical model reads static topographic data and runs independently until the salinity field is stable, generates the initial salinity field corresponding to each numerical model, and the numerical model set consists of multiple numerical models and their corresponding initial salinity fields. Module M2, targeting the external dynamic factors affecting saltwater intrusion, acquires runoff forecast products and wind field forecast products from multiple sources within the forecast period, which respectively constitute the runoff boundary set and the wind field boundary set; Module M3 pairs each numerical model in the numerical model set, each runoff boundary in the runoff boundary set, and each wind field boundary in the wind field boundary set one by one. Each pairing result is used as a forecast scenario. For each forecast scenario, the corresponding numerical model, runoff boundary, wind field boundary, and initial salinity field are used to perform numerical calculations of brine intrusion to obtain the salinity spatiotemporal distribution forecast results for each forecast scenario. Module M4 calculates the model confidence probability, runoff confidence probability, and wind field confidence probability based on the historical simulation accuracy of each numerical model, each runoff boundary, and each wind field boundary, and then calculates the comprehensive confidence probability of each forecast scenario. Module M5 performs weighted statistics on the characteristic variables of saltwater intrusion based on the salinity spatiotemporal distribution forecast results and corresponding comprehensive confidence probabilities for each forecast scenario, constructs a joint probability distribution table, and outputs the probability forecast results of saltwater intrusion.

[0109] In module M2, the runoff boundary set is constructed through the following submodules: Module M2.1 acquires runoff forecast products from multiple sources within the forecast period as the basic runoff boundary, such as forecasts from the Hydrology Bureau of the Yangtze River Water Resources Commission and forecasts from the Shanghai Marine Monitoring and Forecasting Center.

[0110] Module M2.2 generates positive and negative deviation runoff boundaries for each basic runoff boundary based on historical forecast error statistics to characterize forecast uncertainty.

[0111] Specifically, based on the statistical results of the maximum error estimation of the short-term runoff forecast samples from the Yangtze River Water Resources Commission, the historical forecast error standard deviation of each runoff forecast data is calculated by comparing the forecast runoff volume within the forecast period with the measured value during the same period. The positive deviation runoff boundary at each time point is the forecast runoff volume plus the historical forecast error standard deviation, and the negative deviation runoff boundary is the forecast runoff volume minus the historical forecast error standard deviation.

[0112] Module M2.3 combines the basic runoff boundary, the positive deviation runoff boundary, and the negative deviation runoff boundary to form a runoff boundary set.

[0113] Furthermore, the wind field boundary set includes wind field forecast products from different sources: wind field simulations and forecasts from the WRF (Weather Research and Forecasting Model), wind field forecasts from the GFS (Global Forecasting System), and wind field forecasts from the Shanghai Typhoon Institute of the China Meteorological Administration, generating diverse wind field boundaries.

[0114] Specifically, module M4 includes the following sub-modules: Module M4.1 uses the Model Score (SS) as the metric to evaluate the early validation performance of each numerical model. The formula for calculating SS is:

[0115] in, n is the total number of samples that can be compared.

[0116] This represents the measured salinity value of the i-th sample.

[0117] This represents the simulated salinity value of the i-th sample.

[0118] This represents the average of the measured salinity values ​​for all samples.

[0119] First, perform an initial screening of the SS results. If the SS of the numerical model is less than or equal to a preset threshold (e.g.: If the historical simulation accuracy of the numerical model is insufficient, the model confidence probability is set to 0 and it will not participate in the final probability distribution calculation; if SS is greater than the preset threshold ( Then, the SS value of the numerical model is subjected to exponential weighted normalization, and the SS value is calculated according to the following formula. Model confidence probability of a numerical model :

[0120] Where N is the total number of numerical models participating in the set (4 in this embodiment). For the first Model scores for each numerical model The lower the preset model sharpness, the higher the model confidence probability obtained by the numerical model with a higher initial validation performance score (SS score). The larger the value, the greater the sum of the model confidence probabilities of all numerical models, which is 1.

[0121] Module M4.2, because the wind field boundary forecast results are vectors, uses the Vector Relative Error Balance (VRS) as the metric for evaluating the accuracy performance of the early-stage wind field boundary forecast products. The formula for calculating VRS is:

[0122] This represents the measured wind vector of the i-th sample; This represents the wind vector predicted for the i-th sample. This represents the total number of samples participating in the statistical analysis (i.e., the total time series length of the multi-site data used for evaluation).

[0123] The VRS value of each wind field boundary is subjected to exponential weighted normalization to calculate the wind field confidence probability of the i-th wind field boundary. The formula is as follows:

[0124] Where M is the total number of wind field boundaries used in the ensemble forecast (3 in this embodiment). This represents the relative error margin of the i-th wind field boundary forecast product. The third sharpness is a preset value; the smaller the value, the higher the confidence probability of the wind field obtained from the wind field boundary in the early verification performance score (VRS value). The larger the value, the greater the sum of the confidence probabilities of the wind field at all wind field boundaries, which is 1.

[0125] Module M4.3 uses the Relative Error Balance (RS) as the metric for evaluating the accuracy of early runoff boundary forecast products and the accuracy of extreme error probability assessment. The formula for calculating RS is as follows:

[0126] in, This represents the measured value of runoff.

[0127] This represents the predicted runoff value.

[0128] This indicates the total number of measured runoff samples included in the statistics (i.e., the length of the time series used for evaluation).

[0129] Then, the RS value of each runoff boundary is subjected to exponential weighted normalization to calculate the runoff confidence probability of the i-th runoff boundary. Calculate using the following formula:

[0130] Where Q is the total number of runoff boundaries used in the ensemble forecast (4 in this embodiment). The table shows the relative error of the i-th boundary forecast product. The lower the preset second sharpness, the higher the runoff confidence probability obtained from the runoff boundary with the higher the prior validation performance score (RS value). The larger the value, the greater the sum of the runoff confidence probabilities at all runoff boundaries, which is 1.

[0131] Module M4.4, each forecast scenario corresponds to a specific set of numerical models, runoff boundaries, and wind field boundaries, and its combined confidence probability The product of the three confidence probabilities mentioned above: .

[0132] The combined confidence probability of all forecast scenarios satisfies the normalization condition: .

[0133] Specifically, module M5 includes the following sub-modules: Module M5.1 extracts characteristic variables of saltwater intrusion from the salinity spatiotemporal distribution forecast results of each forecast scenario. The characteristic variables include the number of saltwater intrusions and the total duration of saltwater intrusions.

[0134] The salinity spatiotemporal distribution forecast results include: the salinity spatial distribution field of each grid node in the calculation area, the three-dimensional salinity field or surface / bottom salinity field at each time step, and the salinity time series of the target control section or water source location.

[0135] Furthermore, module M5.1 includes the following sub-modules: Module M5.1.1 extracts the salinity time series at the target water source or control section from the salinity spatiotemporal distribution forecast results; Module M5.1.2 sets a salinity threshold, performs threshold determination on salinity time series, and identifies time periods when salinity continuously exceeds the salinity threshold; Module M5.1.3 counts the number of consecutive time intervals in which salinity exceeds the salinity threshold, and uses this count as the number of saltwater intrusions that occur in this forecast scenario.

[0136] Module M5.1.4 calculates the duration of each consecutive time interval exceeding the salinity threshold and sums them to obtain the total duration of saltwater intrusion for the forecast scenario.

[0137] Module M5.2, based on the number of saltwater intrusion occurrences, categorizes each forecast scenario into the following three types: If the number of brine intrusions is equal to 1, then it is determined that 1 brine intrusion has occurred. If the number of brine intrusions is greater than or equal to 2, it is determined as multiple (≥2) brine intrusions. If the number of brine intrusions is 0, it is determined that no brine intrusion has occurred. At the same time, the total duration of saltwater intrusion corresponding to each forecast scenario is recorded.

[0138] Module M5.3 performs weighted statistics on the number of saltwater intrusions and the total duration of saltwater intrusions based on the comprehensive confidence probability of each forecast scenario, and calculates the probability of saltwater intrusions of different numbers and the corresponding total duration of saltwater intrusions.

[0139] Specifically, the forecast scenarios are grouped according to the number of saltwater intrusions, and the formula for calculating the probability of occurrence for different numbers of saltwater intrusions is as follows: , Where Num represents the number of brine intrusions. The probability of Num saltwater intrusions. Let N be the combined confidence probability of Num intrusions occurring in the j-th forecast scenario.

[0140] Calculate the total duration of intrusion when Num brine intrusions occur: , in, For the occurrence Total duration of secondary brine intrusion Let be the duration of the k-th saltwater intrusion in the j-th forecast scenario.

[0141] Module M5.4 constructs a joint probability distribution table containing the number of saltwater intrusions, their corresponding probabilities, and the total duration of the saltwater intrusions.

[0142] Specifically, a joint probability distribution table is formed by considering all forecast scenarios based on three factors: the number of times saltwater intrusion occurs (including three categories: 0 times, 1 time, and 2 times or more), the probability of occurrence of this type of event for each forecast scenario, and the total duration of saltwater intrusion under this type of forecast scenario.

[0143] Module M5.5, based on the joint probability distribution table, outputs the following probability forecast results.

[0144] Specifically, the probability of 0 brine intrusions, the probability of 1 brine intrusion, the probability of 2 or more brine intrusions, and the total duration of brine intrusions for each category.

[0145] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0146] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction method, characterized in that, Includes the following steps: Step S1: Obtain static topographic data of the target estuary area, select multiple calibrated and verified numerical models, each numerical model reads the static topographic data and runs independently until the salinity field is stable, generate the initial salinity field corresponding to each numerical model, and the numerical model set is composed of multiple numerical models and the corresponding initial salinity fields. Step S2: For the external dynamic factors affecting saltwater intrusion, obtain runoff forecast products and wind field forecast products from multiple sources within the forecast period, and construct runoff boundary sets and wind field boundary sets respectively. Step S3: Pair each numerical model in the numerical model set, each runoff boundary in the runoff boundary set, and each wind field boundary in the wind field boundary set one by one. Each pairing result is used as a forecast scenario. For each forecast scenario, perform numerical calculations of brine intrusion using the corresponding numerical model, runoff boundary, wind field boundary, and initial salinity field to obtain the salinity spatiotemporal distribution forecast result for each forecast scenario. Step S4: Based on the historical simulation accuracy of each numerical model, each runoff boundary, and each wind field boundary, calculate the model confidence probability, runoff confidence probability, and wind field confidence probability respectively, and then calculate the comprehensive confidence probability of each forecast scenario. Step S5: Based on the salinity spatiotemporal distribution forecast results and corresponding comprehensive confidence probabilities for each forecast scenario, perform weighted statistics on the characteristic variables of saltwater intrusion, construct a joint probability distribution table, and output the probability forecast results of saltwater intrusion.

2. The multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction method according to claim 1, characterized in that, The selection criteria for the numerical models mentioned in step S1 are: the numerical models of saltwater intrusion in each estuary are verified using historical data, the calculation and simulation accuracy reaches the preset threshold, and there are differences between the numerical models in terms of physical parameterization schemes or numerical solution methods.

3. The multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction method according to claim 1, characterized in that, The external dynamic factors mentioned in step S2 include runoff and wind field, and the runoff boundary set is constructed through the following sub-steps: Step S2.1: Obtain runoff forecast products from multiple different sources within the forecast period as the basic runoff boundary; Step S2.2: Based on historical forecast error statistics, generate positive deviation runoff boundaries and negative deviation runoff boundaries for each of the basic runoff boundaries; Step S2.3: The basic runoff boundary, the positive deviation runoff boundary, and the negative deviation runoff boundary together constitute the runoff boundary set.

4. The multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction method according to claim 1, characterized in that, The wind field boundary set mentioned in step S2 includes wind field forecast products from different sources: wind field simulated and forecasted by the WRF model, wind field forecasted by the GFS global forecast system, and wind field forecasted by the Shanghai Typhoon Research Institute of the China Meteorological Administration.

5. The multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction method according to claim 1, characterized in that, The total number of forecast scenarios in step S3 is equal to the product of the number of models in the numerical model set, the number of runoff boundaries in the runoff boundary set, and the number of wind field boundaries in the wind field boundary set.

6. The multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction method according to claim 1, characterized in that, The comprehensive confidence probability mentioned in step S4 is obtained by multiplying the model confidence probability, the runoff confidence probability, and the wind field confidence probability; wherein, the model confidence probability is calculated by weighted normalization based on the model score SS, the runoff confidence probability is calculated by weighted normalization based on the relative error square RS, and the wind field confidence probability is calculated by weighted normalization based on the vector relative error square VRS.

7. The multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction method according to claim 6, characterized in that, The formula for calculating the model score SS is: in, n is the total number of samples compared; Indicates the first Measured salinity values ​​for each sample; Indicates the first Simulated salinity values ​​for each sample; This represents the average of the measured salinity values ​​for all samples. If SS is less than or equal to the preset threshold, the corresponding model confidence probability is set to 0; otherwise, the... Model confidence probability of a numerical model Calculate using the following formula: Where N is the total number of numerical models. The preset model sharpness, For the first Model scores for each numerical model.

8. The multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction method according to claim 6, characterized in that, The formula for calculating the vector relative error (VRS) is as follows: This represents the measured wind vector of the i-th sample; This represents the wind vector predicted for the i-th sample. This indicates the total number of samples participating in the statistical analysis; The confidence probability of the wind field at the i-th wind field boundary Calculate using the following formula: Where M represents the total number of wind field boundaries, The preset third sharpness; This represents the relative error margin of the i-th wind field boundary forecast product; The formula for calculating the relative error difference (RS) is as follows: in, This represents the measured runoff value; This represents the predicted runoff value; This represents the total number of measured runoff samples included in the statistics; Runoff confidence probability at the i-th runoff boundary Calculate using the following formula: Where Q represents the total number of runoff boundaries. The preset second sharpness, The table shows the relative error of the i-th boundary forecast product.

9. The multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction method according to claim 1, characterized in that, Step S5 includes: Step S5.1: Extract the characteristic variables of saltwater intrusion from the salinity spatiotemporal distribution forecast results of each forecast scenario. The characteristic variables include the number of saltwater intrusions and the total duration of saltwater intrusions. Step S5.2: Based on the number of times the saltwater intrusion occurred, each forecast scenario is divided into three categories: no saltwater intrusion, one saltwater intrusion, and two or more saltwater intrusions. Step S5.3: Based on the comprehensive confidence probability of each forecast scenario, calculate the probability of occurrence of different numbers of saltwater intrusions and the corresponding total duration of the saltwater intrusions; Step S5.4: Construct a joint probability distribution table containing the number of brine intrusions, the corresponding probability of occurrence, and the total duration of brine intrusions; Step S5.5: Based on the joint probability distribution table, output the probability of 0, 1, 2 or more brine intrusions and the total duration of brine intrusions for each category.

10. A multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction system, employing the multi-model, multi-boundary estuarine saltwater intrusion ensemble probability prediction method according to any one of claims 1-9, characterized in that, include: Module M1 acquires static topographic data of the target estuary area, selects multiple calibrated and verified numerical models, each of the numerical models reads the static topographic data and runs independently until the salinity field stabilizes, generates an initial salinity field corresponding to each numerical model, and the numerical model set is composed of multiple numerical models and the corresponding initial salinity fields. Module M2, targeting the external dynamic factors affecting saltwater intrusion, acquires runoff forecast products and wind field forecast products from multiple sources within the forecast period, which respectively constitute the runoff boundary set and the wind field boundary set; Module M3 pairs each of the numerical models in the numerical model set, each of the runoff boundaries in the runoff boundary set, and each of the wind field boundaries in the wind field boundary set one by one. Each pairing result is used as a forecast scenario. For each forecast scenario, the corresponding numerical model, the runoff boundary, the wind field boundary, and the initial salinity field are used to perform numerical calculations on brine intrusion to obtain the salinity spatiotemporal distribution forecast result for each forecast scenario. Module M4 calculates the model confidence probability, runoff confidence probability, and wind field confidence probability based on the historical simulation accuracy of each numerical model, each runoff boundary, and each wind field boundary, and then calculates the comprehensive confidence probability of each forecast scenario. Module M5 performs weighted statistics on the characteristic variables of saltwater intrusion based on the salinity spatiotemporal distribution forecast results and corresponding comprehensive confidence probabilities for each forecast scenario, constructs a joint probability distribution table, and outputs the probability forecast results of saltwater intrusion.

Images