A short-term prediction method for high tide flood in coastal areas based on probability statistics

Abstract

The application discloses a kind of short-term prediction methods of high tide flood in coastal areas based on probability statistics. Including: obtaining the historical hourly water level observation data of target coastal area tide station and preprocessing;Climate state probability distribution model is constructed, and corresponding climate state distribution parameters are obtained;Decay persistence prediction model is constructed, and the observation abnormal value of future month is calculated according to the current month observation abnormal value;The climate state distribution parameters are adjusted by the observation abnormal value of future month;The daily cumulative flood probability is calculated based on the adjusted climate state distribution parameters;Daily cumulative flood probability is calculated day by day, and the high tide flood probability sequence of coastal cities in future several months is obtained, the warning probability threshold is set, and the date higher than the warning probability threshold in high tide flood probability sequence is marked as high tide flood high risk day.The high tide flood prediction of the application is more in line with the needs of disaster risk assessment, has more information, and can provide seasonal scale prediction, leaving time for medium and long term planning and emergency preparation.

Description

Technical Field

[0001] This invention relates to the field of marine environmental forecasting and disaster prevention and mitigation technology, and more specifically to a method for short-term forecasting of high tide floods in coastal areas based on probability statistics. Background Technology

[0002] Currently, with global climate change and rising sea levels, high tide floods are occurring frequently in coastal areas of various cities, posing a serious threat to urban infrastructure, the ecological environment, and the safety of residents' lives and property. Therefore, conducting high-precision and timely high tide flood forecasting has become a key need for disaster prevention and mitigation in coastal areas.

[0003] However, current high tide flood forecasting in coastal areas mainly relies on two types of technologies: one is numerical models based on ocean dynamics. These models require a large amount of meteorological and oceanographic data as input, have complex calculation processes and high hardware costs, and are difficult to quantify the uncertainty of non-tidal factors in daily short-term forecasts, and cannot output probabilistic results, which is not conducive to decision-makers assessing risk levels; the other is statistical models based on historical data, which focus on long-term flood frequency analysis and lack the ability to accurately predict the probability of short-term floods.

[0004] Therefore, how to provide a high-precision probabilistic forecasting method that can comprehensively utilize long-term observation data, has a clear physical mechanism, and can be operated operationally is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of the above problems, the present invention proposes a short-term forecasting method for high tide floods in coastal areas based on probability statistics, which can overcome or at least partially solve the above problems.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] This invention provides a method for short-term forecasting of high tide floods in coastal areas based on probability statistics, specifically including the following steps: S1. Obtain historical hourly water level observation data from tide gauge stations in the target coastal area, and preprocess the hourly water level observation data; S2. Based on the preprocessed hourly water level observation data, a climatological probability distribution model is constructed, and the climatological distribution parameters of the corresponding target forecast month and the expected tidal range interval are calculated. S3. Construct a decay persistence prediction model, calculate the observation anomalies for future months based on the observation anomalies of the current month, and adjust the climatological distribution parameters based on the observation anomalies of the future months; S4. Calculate the daily cumulative flood probability based on the adjusted climatological distribution parameters; S5. Calculate the daily cumulative flood probability to obtain a high tide flood probability sequence for coastal cities in the next few months, set an early warning probability threshold, and mark the dates in the high tide flood probability sequence that are higher than the early warning probability threshold as high tide flood high risk days.

[0008] Furthermore, in step S1, the preprocessing of the hourly water level observation data specifically includes: Quality control was performed on the hourly water level observation data obtained from the coastal tide stations. Individual extreme outliers were checked and corrected in conjunction with weather records from the same period. Data gaps were filled by linear interpolation or by constructing a regression model using data from adjacent stations from the same period. The benchmark for hourly water level observation data after quality control was standardized. Based on the hourly water level observation data after benchmark unification, harmonic constants are extracted, which include the amplitude and lag angle of each major astronomical tide. The formula for calculating non-tidal residuals hourly is as follows:

[0009] In the formula, For hourly non-tidal residuals, The sea level height for astronomical tide forecasting, This represents the linear trend of relative sea level rise. To monitor the water level.

[0010] Furthermore, the harmonic constant is calculated using the following formula:

[0011] In the formula, Let t be the water level height at time t; l This is a sequence identifier for a time period, and is a positive integer greater than 0. The starting point for calculating tide height over a certain period of time. p, q Indicators for the main tidal constituents and accompanying tidal constituents; P The number of major tidal constituents, Q The number of accompanying tidal fractions in a given major tidal fraction; The tidal angular rate, f、u For the tidal node factor and the node correction angle, v To determine the Greenwich astronomical phase angle for tidal division, h, g The amplitude and lag angle of the tidal phase.

[0012] Furthermore, the specific process of step S2 includes: S21. The non-tidal residual data obtained after preprocessing the hourly water level observation data are grouped separately according to two methods: month and tidal range decimals. S22. Calculate the climatological mean and climatological standard deviation of the non-tidal residuals within each two-dimensional grouping determined by the month and the decile interval of the tidal range, and verify the residual distribution. S23. Use the climatological mean and climatological standard deviation that conform to a normal distribution as climatological distribution parameters. If they deviate from a normal distribution, use a log-normal distribution or a gamma distribution to fit the climatological mean and climatological standard deviation, and recalculate them.

[0013] Furthermore, the specific process of step S22 includes: S221. Extract all non-tidal residual values ​​within each two-dimensional grouping determined by the month and the decile interval of the tidal range, forming a sample subset; S222. Calculate the empirical mean and empirical standard deviation of the sample subset; S223. Introduce a global adjustment factor for tidal range dimension to optimize parameters.

[0014] Furthermore, the specific process of step S3 includes: S31. Construct a decay persistence prediction model using anomaly sequences and autocorrelation coefficients. The input of the decay persistence prediction model is the observed anomaly value of the current month, and the output is the observed anomaly value of the future month. S32. Adjust the climatological distribution parameters by using the observed outliers of the future month as the time-varying correction amount of the climatological distribution parameters.

[0015] Furthermore, the process of obtaining the abnormal sequence in step S31 includes: Based on the monthly mean sea level data from the tide gauge station, the long-term trend term is obtained by using the linear least squares method. At the same time, the average value of the monthly mean sea level data for the same month over many years is calculated to obtain the seasonal cycle term. By removing the long-term trend term and seasonal cycle term from the monthly average sea level data, the monthly average sea level anomaly sequence is obtained.

[0016] Furthermore, the process of obtaining the autocorrelation coefficient includes: The autocorrelation coefficient is obtained through calculation, and its calculation formula is as follows:

[0017] In the formula, N is the total length of the sequence. This is a monthly mean sea level anomaly sequence. The mean of the sequence; The autocorrelation coefficient is determined to be statistically significant by using a preset threshold. If the autocorrelation coefficient is greater than the preset threshold, it is considered statistically significant.

[0018] Furthermore, the specific process of step S4 includes: S41. Calculate the remaining space:

[0019] In the formula, As the initial empirical threshold, Sea level This represents the linear trend of relative sea level rise. S42. Calculate the probability of exceeding the remaining space hourly:

[0020] In the formula, F The cumulative distribution function is the function of the normal distribution. This is the adjusted climatological mean. This is the adjusted climatological standard deviation; S43. By selecting the probability of the largest hourly exceedance of the remaining space in a day, and using the autocorrelation coefficient of the hourly residuals to weight and sum the probabilities of the other 23 hours, the daily cumulative flood probability is obtained.

[0021] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a method for short-term forecasting of high tide floods in coastal areas based on probability statistics, which has the following beneficial effects: This invention acquires historical tidal gauge data and applies it to the core model, enabling rapid application to any coastal station with sufficient observation time. Furthermore, the model framework is open, allowing for easy integration with advanced domestic and international numerical, statistical, or machine learning forecasts of sea-level anomalies to further enhance forecasting capabilities. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0023] Figure 1 This is a flowchart of the short-term high tide flood statistical forecasting method provided in this embodiment of the invention; Figure 2 This is a flowchart of data preparation and preprocessing provided in the embodiments of the present invention; Figure 3 This is a flowchart for determining the high tide flood threshold provided in an embodiment of the present invention; Figure 4 A flowchart illustrating the construction of non-tidal residual climatological distribution provided in this embodiment of the invention; Figure 5This is a flowchart illustrating the improvement in residual decay persistence provided in this embodiment of the invention; Figure 6 This is a flowchart of the forecast probability calculation and product generation provided in the embodiments of the present invention; Figure 7 This is a schematic diagram of the visualization prediction results provided in an embodiment of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] This invention discloses a short-term forecasting method for high tide floods in coastal areas based on probability statistics, such as... Figure 1 As shown, the specific steps include: S1. Obtain historical hourly water level observation data from tide gauge stations in the target coastal area, and preprocess the hourly water level observation data; S2. Based on the preprocessed hourly water level observation data, a climatological probability distribution model is constructed, and the climatological distribution parameters of the corresponding target forecast month and the expected tidal range interval are calculated. S3. Construct a decay persistence prediction model, calculate the observation anomalies for future months based on the observation anomalies of the current month, and adjust the climatological distribution parameters based on the observation anomalies of the future months; S4. Calculate the daily cumulative flood probability based on the adjusted climatological distribution parameters; S5. Calculate the daily cumulative flood probability to obtain a high tide flood probability sequence for coastal cities in the next few months, set an early warning probability threshold, and mark the dates in the high tide flood probability sequence that are higher than the early warning probability threshold as high tide flood high risk days.

[0026] This invention applies historical data from tide gauge stations to the core model and can be quickly applied to any coastal station with sufficient observation time. Furthermore, the model framework is open, allowing for easy integration with advanced domestic and international numerical forecasts, statistical forecasts, or machine learning forecasts of sea-level anomalies to further improve forecasting capabilities.

[0027] The following is a detailed description of each of the above steps: Step S1, as follows Figure 2 The flowchart for data preparation and preprocessing is shown below; The raw data input is obtained by collecting long-term hourly water level observation data from major tidal gauge stations in the coastal area. The selection criteria for tide gauge stations are: mainly selecting tide gauge stations with clear high tide flood thresholds or historical inundation records, and data gap rate <5%.

[0028] The collected raw data undergoes preprocessing, which includes the following steps: (1) Automatic and manual quality control of raw data, combined with weather records of the same period, to exclude abnormal fluctuations other than extreme weather such as typhoons and cold waves, and to check and correct individual extreme abnormal values; linear interpolation is used to fill data gaps within 72 hours, and regression models are constructed to estimate and fill data gaps of more than 72 hours.

[0029] (2) Standardize the data after quality control. Since each tide station is built and operated, its observations and tide forecasts are usually based on a localized, relative calculation datum for local use, such as the "multi-year average sea level of this station" or a specific "station zero point". The local datums of different stations are different. Standardize all water level data observations to eliminate the differences in datums between different stations.

[0030] (3) Based on historical hourly water level observation data after unification of the benchmark, harmonic analysis was used to extract harmonic constants characterizing the astronomical tide variations at each station. The harmonic constants mainly include the amplitude and lag angle of each major astronomical tide. These constants are obtained by fitting the long-term water level sequence using methods such as least squares analysis or Fourier analysis, reflecting the inherent fluctuation characteristics of each tide at a specific benchmark at that station. Based on the harmonic constants of each station, the hourly astronomical tide forecast values ​​for the same period were calculated. This forecast value does not include long-term sea-level rise trends. Based on the port's tidal characteristics, a suitable astronomical tide forecasting method is used to achieve port tidal level forecasts at any given time. The harmonic forecasting method is used for deep-water ports, while the shallow-water forecasting method is used for shallow-water ports.

[0031] The harmonic constant is calculated as follows: Since the harmonic constants mainly include amplitude and lag angle, a harmonic analysis method adapted to hourly tide levels and variable time intervals of high and low tide data of different time lengths is adopted to calculate the tidal amplitude and lag angle at the station:

[0032] In the formula, Let t be the water level height at time t; l This is a sequence identifier for a time period, and is a positive integer greater than 0. The starting point for tide height during a certain period of time can be a segment of data, a year, or the entire length of the data. p, qIndicators for the main tidal constituents and accompanying tidal constituents; P The number of major tidal constituents, Q The number of accompanying tidal fractions in a given major tidal fraction; The tidal angular rate, f、u For the tidal node factor and the node correction angle, v To determine the Greenwich astronomical phase angle for tidal division, h, g The amplitude and lag angle of the tidal phase.

[0033] After obtaining the harmonic constant, based on the astronomical parameters corresponding to the future date and the angular frequency of each tide, the astronomical tide forecast value for any future time is calculated, and the tide height for any time is predicted according to the tide height expression. The formula for predicting tidal harmonics is:

[0034] In the formula: The mean sea level is the height of the mean sea level on the tidal reference surface. It is the reference surface applicable to navigation. It can also be taken as the height of the mean sea level at the zero point of the tide gauge. for The astronomical tides at certain times; The tidal angular rate; f j For the first j The intersection factor of each tidal wave, u Correct the angle for the intersection point. The initial phase angle of the Greenwich Mean Time astronomical astronomy for tidal division depends on the calculated initial time. , This is the tidal harmonic constant; The number of tidal constituents.

[0035] In shallow-water ports, especially in estuary areas, tidal waves are significantly distorted due to increased nonlinear effects, and the high-frequency oscillation component must be taken into account. To improve the accuracy of tidal forecasts, especially during high and low tides, shallow-water tidal forecasting methods are used in shallow-water ports. The formula for predicting shallow tides is:

[0036] In the formula: for The astronomical tides at certain times; To reconcile the predicted high tide; , For shallow level harmonics, and for time. function; , is the shallow water coefficient, and is a constant; This represents the number of shallow level harmonic terms.

[0037] (4) Calculation of non-tidal residuals: Non-tidal residuals reflect the influence of non-tidal factors such as wind, air pressure, and ocean circulation on water level. The formula for calculating hourly non-tidal residuals is as follows:

[0038] in In order to observe the water level, For astronomical tide forecasting, This represents the linear trend of relative sea level rise.

[0039] In this embodiment, the long-term time series hourly water level observation data collected from the tide gauge station is preferably 20 years or more; individual extreme outliers refer to outliers caused by sudden changes in water level, such as those caused by instrument malfunctions; all water level observation data are unified to the mean high tide level corresponding to the 1985 National Height Datum; astronomical tide forecasts are calculated using harmonic constants. These harmonic constants are obtained through analysis under a specific datum system, generally referred to as the theoretical depth datum; astronomical tide constituents typically refer to, for example, M2, S2, K1, O1, etc.

[0040] like Figure 3 The flowchart for determining the high tide flood threshold is shown below; (1) Collect and organize verified records of flooding events in the surrounding area of ​​the tide gauge station over the years, and accurately match the time of the events with the high-frequency water level observed at the station during the same period to form a sample set of disaster-causing water levels. Take the lower limit of the statistical data of this sample set as the initial empirical threshold for the station.

[0041] (2) For stations with limited data, select dates with clear flooding records, take the minimum value of the highest measured water level on the corresponding date as the initial threshold, and make minor adjustments based on the coastal topography around the station to ensure that the threshold matches the actual flooding situation.

[0042] (3) Using long-sequence historical water level data, the high tide flood threshold is retrospectively tested. By evaluating its hit rate and false alarm rate for historical real flooding events, the threshold is finally fine-tuned to determine the site-specific operational threshold for operational probability forecasting.

[0043] In this embodiment, the surrounding coastal topography refers to, for example, the width of the intertidal zone and the ground elevation, and the adjustment range for fine-tuning is preferably ±0.1.

[0044] Step S2, as follows Figure 4 The flowchart for constructing the non-tidal residual climatological distribution is shown below; (1) Calculate the hourly non-tidal residual data for each tide station. At the same time, it is double-grouped according to "calendar month" and "tidal range decimals" to construct its refined climatological probability distribution.

[0045] ① Grouped by "Calendar Month" (Time Dimension): All data were divided into 12 groups from January to December. This division is based on the fact that non-tidal residuals are significantly influenced by the seasonal cycles of atmospheric and oceanic processes, such as monsoon transitions, seasonal variations in sea surface temperature, and pressure system activity, resulting in stable annual patterns of variation in their statistical characteristics, such as mean and volatility. Grouping by calendar month aims to extract and characterize this inherent seasonal pattern, ensuring that probabilistic predictions for any future date are provided with a climate context consistent with that month. To guarantee the reliability of the statistics within each monthly group, the long-term historical data used to construct the climate pattern must contain a sufficient sample size (i.e., the number of hours) for each calendar month, and these samples must cover a complete climate cycle across consecutive years.

[0046] ② Grouped by "tidal range decimal places" (tidal dynamic dimension) Calculate the hourly tidal range corresponding to each non-tidal residual, for example, by taking the difference between the maximum and minimum observed water levels within a certain time window before and after that hour. Then, use the decimal method in statistics to objectively divide all historical tidal range values: sort all tidal range values ​​from smallest to largest, and find the 10th, 20th, ..., 90th percentiles as dividing points, thus dividing the entire tidal range into 10 intervals, each theoretically containing approximately 10% of the data sample. This division method ensures that the sample size within each interval is roughly balanced, rather than simply dividing the absolute value range of the tidal range equally, avoiding the instability of statistical estimation caused by large sample size disparities. The core purpose of this step is to capture the dependence of non-tidal residuals on tidal phases—at high and low tide levels, i.e., the decimal intervals of the large and small tidal ranges, the contribution of meteorological forcing such as wind and air pressure to the water level may be non-linear, and this grouping method can effectively analyze this relationship.

[0047] (2) In each group consisting of the month and the decimal place of the tidal range ( , Within this group, extract all non-tidal residual values ​​to form a sample subset. Calculate the empirical mean of this subset. Compared with the empirical standard deviation ; Introducing a global adjustment factor for tidal range dimension for parameter optimization: Calculate the global mean of the decimal interval of the tidal range over all months. and global standard deviation ; Calculate the mean adjustment factor = - Adjustment factor for standard deviation = / ; where global refers to the corresponding value of all data; Calculate the optimized climatological mean climatological standard deviation :

[0048] (3) Calculate the climatological mean of the non-tidal residuals. and climatological standard deviation The hourly non-tidal residuals can be determined. The climatological distribution is used to verify whether the residual distribution conforms to a normal distribution. If it conforms to a normal distribution, then it is directly adopted. and As distribution parameters, the probability density function (PDF) and cumulative distribution function (CDF) of the non-tidal residuals can be determined for any date (given month and expected tidal range). If the distribution deviates from a normal distribution, a log-normal or gamma distribution can be fitted, and the distribution parameters can be recalculated. and .

[0049] (4) For each group, its climatological probability distribution model is determined according to the following rules based on its sample subset: The subset is subjected to a normal distribution hypothesis test. In this embodiment, the Kolmogorov-Smirnov test is used. If the test shows that the distribution of the subset is not significantly different from the normal distribution at the preset significance level, then the climatological distribution of the group is determined to be a normal distribution.

[0050] If the test rejects the normality hypothesis, the alternative distribution fitting process is automatically initiated. Candidate models such as log-normal distribution and gamma distribution are used for fitting in turn. By comparing the values ​​of the Akaike Information Criterion (AIC), the model with the smallest AIC value is selected as the optimal fitted distribution model for that group, and its corresponding parameters are recorded.

[0051] The "distribution type identifier" and its corresponding "distribution parameters" of all the groups obtained above are constructed into a structured lookup table or lightweight database. This model serves as the core knowledge base and can respond to queries from business systems: when the month of the target forecast date and the decimal interval of the expected tidal range derived from the astronomical tide forecast are input, the model immediately returns the corresponding climatological probability distribution, providing a basis for subsequent flood probability calculations.

[0052] Step S3, as follows Figure 5 The flowchart for improving the persistence of residual decay is shown below; To enhance the predictive ability of short-term sea level anomalies (SLA) and thus optimize the timeliness and accuracy of high tide flood probability forecasts, this invention introduces a "decaying persistence" prediction model based on autocorrelation analysis of historical data. This method does not simply assume the current state remains unchanged, but rather dynamically adjusts the future climatological background based on the statistical law of the attenuation of the anomaly signal itself. The autocorrelation of monthly mean sea level (MSL) anomalies at each station is calculated. For stations with significant autocorrelation, such as those with lags of 1 month or 3 months, this persistence is incorporated into the forecast as "decaying persistence." That is, the current MSL anomaly and autocorrelation coefficient are used to adjust the climatological mean for the next 1-3 month forecast period, thereby improving medium-term forecasting techniques.

[0053] (1) For each tide station, two deterministic components are filtered out from its long-term monthly mean sea level (MSL) data: the long-term trend and the seasonal cycle. The long-term trend term was obtained by fitting using the linear least squares method; The seasonal cyclical term is obtained by calculating the average MSL for the same month over many years; After obtaining the two deterministic components, the anomaly sequence is calculated: the long-term trend and seasonal cycle terms are subtracted from the original monthly average MSL data to obtain the monthly average sea level anomaly sequence. .

[0054] (2) Calculate the autocorrelation coefficient: for each tide gauge station The sequence is calculated to determine its lag. Autocorrelation coefficient for one month The calculation formula is:

[0055] In the formula, The total length of the sequence. This is the sequence mean. This invention focuses on a lag of one month (…). The autocorrelation with a lag of 3 months is used to correspond to monthly and quarterly forecast demand.

[0056] To determine that the calculated autocorrelation does not originate from random noise, Compare with the critical value of the autocorrelation coefficient of a white noise process at a 95% confidence level. If If the autocorrelation of the lag period exceeds this threshold, it is determined that the autocorrelation of the lag period is statistically significant, and the predictable information contained therein can be used to improve forecasts.

[0057] (3) For the lag time that passes the significance test The following "decay persistence" model is used to analyze the future... Monthly SLA forecast:

[0058] in, For the current (or most recently known) month's observed outliers, The corresponding normalized autocorrelation coefficient is used as the damping factor. This model reflects the physical and statistical characteristics of the exponential decay of anomalous signals over time, compared to the simple "persistence method" (i.e., the assumption...). This can more reasonably reflect the signal attenuation process and reduce long-term forecast errors.

[0059] (4) Use the SLA information predicted by the attenuation persistence prediction model as the time-varying correction amount for the baseline climatological mean: Correcting the climatological background, for the future target forecast month, the non-tidal residual climatological mean used for probability calculation. It will be adjusted to:

[0060] In the formula, It is the original climatological mean. This is the predicted value for the attenuation persistence corresponding to the lead time. This adjustment is equivalent to treating the predicted sea-level anomaly as a systematic shift in the future non-tidal residual climate background.

[0061] This improved model is not unconditionally applicable to all sites and all lead times. Its application depends on the significance test results; for example, if a site has a significant autocorrelation only with a 1-month lag, then only the forecast within 1 month is revised; if it is also significant with a 3-month lag, then the revision can be extended to the quarterly forecast. For sites or lead times with insignificant autocorrelation, the model automatically reverts to the basic climatological model.

[0062] Step S4: Calculate the daily cumulative flood probability; (1) Calculate the remaining space : Calculate the remaining space for a specific peak moment on a future day. This value represents the water level difference from the point where a high tide flood would occur, given a known tide level and sea-level rise. The smaller the value, the higher the probability of flooding caused by non-tidal factors on that day.

[0063] (2) Calculate the hourly overthreshold probability: using the climatological distribution parameters obtained in step S3 Calculate the non-tidal residual exceedance for that hour. The probability of: , where F is the CDF of a normal distribution.

[0064] (3) Calculate the daily cumulative probability: Considering the multiple peaks within a day and the autocorrelation between hourly residuals, select the largest one in a day. Based on this, and using the autocorrelation coefficient of the hourly residuals to perform a weighted summation of the probabilities for the other 23 hours, the final daily cumulative flood probability is obtained. .

[0065] Step S5, as follows Figure 6 The forecast probability calculation and product generation are shown in the figure; The cumulative daily flood probability is calculated to obtain a high tide flood probability sequence for coastal cities over the next few months. A warning probability threshold is set, and dates in the high tide flood probability sequence that exceed the warning probability threshold are marked as high tide flood high-risk days. The forecast results can also be visualized, for example, as a "high tide flood probability calendar," providing intuitive and quantitative reference for government decision-making, public disaster prevention, and industry operations.

[0066] In this embodiment, the next few months are preferably 3 months, and the early warning probability threshold is preferably 5%.

[0067] This invention provides probabilistic forecasts, which better meet the needs of disaster risk assessment, contain more information, and can provide forecasts at seasonal scales, allowing time for medium- and long-term planning and emergency preparedness.

[0068] Example: Taking the Qinhuangdao tide gauge station in the Bohai Sea as an example; This example aims to clearly demonstrate how to apply the method to a specific station, thereby fully realizing the generation of future probability forecast products from historical data analysis.

[0069] 1. Data preparation and preprocessing (1) Data collection Hourly water level observation data from the Qinhuangdao tide gauge station were collected for 26 years, from January 1, 1997 to December 31, 2022. The data quality meets the requirement of a gap rate of less than 5%.

[0070] (2) Data quality control and benchmark unification First, automated quality control was performed on the observed water level data, and significant anomalies caused by instrument malfunctions were eliminated by combining meteorological records. For gaps shorter than 72 hours, linear interpolation was used to fill them; for longer gaps, a regression model was established using data from the nearby Tianjin station for estimation. Subsequently, all observed water level data and the original astronomical tide forecast values ​​calculated based on the harmonic constant were uniformly converted to the "Mean High Tide Level (MHHW)" under the "1985 National Height Datum". This step eliminated the systematic bias caused by the different datum levels in the station's historical data and forecast products, which is crucial for the subsequent accurate calculation of non-tidal residuals. This is a prerequisite. After conversion, a unified baseline observation water level sequence is obtained. and astronomical tide forecast sequence .

[0071] (3) Calculate key derived variables relative sea level rise trend Using monthly mean sea level data with a unified benchmark, a linear regression method was employed to calculate the linear trend from 1997 to 2022, deriving the interannual relative sea level rise rate at Qinhuangdao station and quantifying it into hourly variations. .

[0072] Non-tidal residuals According to the formula Hourly calculations yielded residual sequences reflecting the influence of meteorological and oceanic short-period processes such as wind and air pressure.

[0073] 2. Determination and calibration of high tide flood threshold at the station (1) High tide flood threshold extraction Historical disaster reports from 1997 to 2022, collected from the Qinhuangdao Flood Control and Drought Relief Headquarters and local meteorological departments, were used to identify 18 inundation events clearly attributable to tidal influence, such as "seawater intrusion" and "flooding along coastal roads." The occurrence times of these 18 events were precisely matched with the corresponding water level data at the local station, extracting 18 corresponding "disaster-causing water levels." The 5th percentile of these values ​​(i.e., one of the lowest disaster-causing water levels) was used to obtain the initial empirical threshold. .

[0074] (2) Localized calibration and verification A digital elevation model (DEM) and the elevations of major coastal roads within a 1-kilometer radius of Qinhuangdao Station were obtained. Analysis shows that the terrain in this area is relatively flat, and the standards for flood control facilities are uniform. The topographic sensitivity factor was calculated. The initial threshold is fine-tuned to 1.02. Subsequently, the threshold was retrospectively verified using water level data from 2000 to 2019 (20 years in total). Calculation results showed that the threshold achieved an 83% accuracy rate for historical inundation events, while the false alarm rate was controlled within a reasonable range of 12%. Based on this, the operational high tide flood threshold for Qinhuangdao Station was finally determined. .

[0075] 3. Construction of a non-tidal residual climatological distribution model (1) Double grouping 26 years The hourly series is grouped according to its corresponding calendar month (January-December) and the decimal interval of the tidal range. The method for dividing the tidal range into decimal intervals is as follows: all historical hourly tidal range values ​​are globally sorted, and the 10th, 20th, ..., 90th percentiles are found, thus forming 10 tidal range decimal intervals with approximately equal sample sizes.

[0076] (2) Parameter calculation and optimization For each combination of "month-tidal range decimal interval" (120 groups in total), calculate the within-group... mean and standard deviation Subsequently, a global adjustment factor for the tidal range dimension is introduced for smoothing optimization. For example, the global mean of the "5th tidal range decimal interval" data is first calculated for all years. Then calculate the adjustment factor. Ultimately, the climatological mean of this cell... Standard deviation Adjustments were made using a similar method.

[0077] (3) Determination of distribution type A normality test (Kolmogorov-Smirnov test, significance level α=0.05) was performed on the optimized samples for each group. The test revealed that over 90% of the grouped data from Qinhuangdao station passed the normality test. Therefore, the non-tidal residual climatological distribution model for Qinhuangdao station was established as a normal distribution, meaning that for any month... Tidal range decimal interval Its distribution is .

[0078] 4. Establishment of a model for the persistence of sea-level anomaly attenuation (1) Sequence calculation Remove the linear trend calculated in step 1 from the monthly average sea level data from Qinhuangdao station. With a fixed seasonal cycle, a monthly mean sea level anomaly sequence was obtained. .

[0079] (2) Autocorrelation analysis and modeling calculate The autocorrelation coefficients of the series with 1-month and 3-month lags are respectively and Both tests showed that they exceeded the 95% confidence limit for white noise processes, indicating statistical significance. Based on this, a decay persistence prediction model was established for the Qinhuangdao station. This model is used to dynamically correct the climatological background field when making 1-month and 3-month forecasts.

[0080]

[0081] 5. Model Validation and Forecast Product Generation (1) Retrospective forecast verification The period from 2018 to 2022 was used as an independent validation period. At the beginning of each month, the model was updated using data up to the end of the previous month (for the persistent portion), providing daily flood probability forecasts for the next 90 days. The Brier skill score (BSS) was used for evaluation. The results show that during the validation period, the method achieved a BSS of 0.25 for the forecast of high tide flood days at Qinhuangdao station (relative to the climate baseline forecast), demonstrating clear skill. The recall rate reached 48%, and the false alarm rate (FAR) was 9%, proving the effectiveness and practicality of the model.

[0082] (2) Operational forecast products The constructed Qinhuangdao station-specific model (including thresholds, a climatological distribution parameter library, and persistence coefficients) is integrated into the operational forecasting system. The system runs automatically monthly, outputting a three-month "Probability Outlook for High Tide Floods at Qinhuangdao Station." The product is presented in two formats: ① Data file: contains the date for each day of the next 90 days and the corresponding daily cumulative flood probability.

[0083] ② Visualized probability calendar: such as Figure 7 As shown, the calendar format is used, with the date cells transitioning from white to red, representing probabilities from low to high. For example, [the calendar format is shown here]. The dates are marked in yellow to red, with high-risk days clearly indicated as having a probability exceeding the warning threshold. This product can be directly provided to local emergency management and maritime departments to assist in decision-making.

[0084] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0085] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A short-term forecasting method for high tide floods in coastal areas based on probability statistics, characterized in that, Includes the following steps: S1. Obtain historical hourly water level observation data from tide gauge stations in the target coastal area, and preprocess the hourly water level observation data; S2. Based on the preprocessed hourly water level observation data, a climatological probability distribution model is constructed, and the climatological distribution parameters of the corresponding target forecast month and the expected tidal range interval are calculated. S3. Construct a decay persistence prediction model, calculate the observation anomalies for future months based on the observation anomalies of the current month, and adjust the climatological distribution parameters based on the observation anomalies of the future months; S4. Calculate the daily cumulative flood probability based on the adjusted climatological distribution parameters; S5. Calculate the daily cumulative flood probability to obtain a high tide flood probability sequence for coastal cities in the next few months, set an early warning probability threshold, and mark the dates in the high tide flood probability sequence that are higher than the early warning probability threshold as high tide flood high risk days.

2. The short-term forecasting method for high tide floods in coastal areas based on probability statistics as described in claim 1, characterized in that, In step S1, the preprocessing of the hourly water level observation data specifically includes: Quality control was performed on the hourly water level observation data obtained from the coastal tide stations. Individual extreme outliers were checked and corrected in conjunction with weather records from the same period. Data gaps were filled by linear interpolation or by constructing a regression model using data from adjacent stations from the same period. The benchmark for hourly water level observation data after quality control was standardized. Based on the hourly water level observation data after benchmark unification, harmonic constants are extracted, which include the amplitude and lag angle of each major astronomical tide. The formula for calculating non-tidal residuals hourly is as follows: In the formula, For hourly non-tidal residuals, The sea level height for astronomical tide forecasting, This represents the linear trend of relative sea level rise. To monitor the water level.

3. The short-term forecasting method for high tide floods in coastal areas based on probability statistics as described in claim 2, characterized in that, The harmonic constant is calculated using the following formula: In the formula, Let t be the water level height at time t; l This is a sequence identifier for a time period, and is a positive integer greater than 0. The starting point for calculating tide height over a certain period of time. p, q Indicators for the main tidal constituents and accompanying tidal constituents; P The number of major tidal constituents, Q The number of accompanying tidal fractions in a given major tidal fraction; The tidal angular rate, f、u For the tidal node factor and the node correction angle, v To determine the Greenwich astronomical phase angle for tidal division, h, g The amplitude and lag angle of the tidal phase.

4. The short-term forecasting method for high tide floods in coastal areas based on probability statistics as described in claim 1, characterized in that, The specific process of step S2 includes: S21. The non-tidal residual data obtained after preprocessing the hourly water level observation data are grouped separately according to two methods: month and tidal range decimals. S22. Calculate the climatological mean and climatological standard deviation of the non-tidal residuals within each two-dimensional grouping determined by the month and the decile interval of the tidal range, and verify the residual distribution. S23. Use the climatological mean and climatological standard deviation that conform to a normal distribution as climatological distribution parameters. If they deviate from a normal distribution, use a log-normal distribution or a gamma distribution to fit the climatological mean and climatological standard deviation, and recalculate them.

5. The short-term forecasting method for high tide floods in coastal areas based on probability statistics as described in claim 4, characterized in that, The specific process of step S22 includes: S221. Extract all non-tidal residual values ​​within each two-dimensional grouping determined by the month and the decile interval of the tidal range, forming a sample subset; S222. Calculate the empirical mean and empirical standard deviation of the sample subset; S223. Introduce a global adjustment factor for tidal range dimension to optimize parameters.

6. The short-term forecasting method for high tide floods in coastal areas based on probability statistics as described in claim 1, characterized in that, The specific process of step S3 includes: S31. Construct a decay persistence prediction model using anomaly sequences and autocorrelation coefficients. The input of the decay persistence prediction model is the observed anomaly value of the current month, and the output is the observed anomaly value of the future month. S32. Adjust the climatological distribution parameters by using the observed outliers of the future month as the time-varying correction amount of the climatological distribution parameters.

7. A short-term forecasting method for high tide floods in coastal areas based on probability statistics as described in claim 6, characterized in that, The process of obtaining the abnormal sequence in step S31 includes: Based on the monthly mean sea level data from the tide gauge station, the long-term trend term is obtained by using the linear least squares method. At the same time, the average value of the monthly mean sea level data for the same month over many years is calculated to obtain the seasonal cycle term. By removing the long-term trend term and seasonal cycle term from the monthly average sea level data, the monthly average sea level anomaly sequence is obtained.

8. A short-term forecasting method for high tide floods in coastal areas based on probability statistics as described in claim 6, characterized in that, The process of obtaining the autocorrelation coefficient includes: The autocorrelation coefficient is obtained through calculation, and its calculation formula is as follows: In the formula, N is the total length of the sequence. This is a monthly mean sea level anomaly sequence. The mean of the sequence; The autocorrelation coefficient is determined to be statistically significant by using a preset threshold. If the autocorrelation coefficient is greater than the preset threshold, it is considered statistically significant.

9. A short-term forecasting method for high tide floods in coastal areas based on probability statistics as described in claim 1, characterized in that, The specific process of step S4 includes: S41. Calculate the remaining space: In the formula, As the initial empirical threshold, Sea level This represents the linear trend of relative sea level rise. S42. Calculate the probability of exceeding the remaining space hourly: In the formula, F The cumulative distribution function is the function of the normal distribution. This is the adjusted climatological mean. This is the adjusted climatological standard deviation; S43. By selecting the probability of the largest hourly exceedance of the remaining space in a day, and using the autocorrelation coefficient of the hourly residuals to weight and sum the probabilities of the other 23 hours, the daily cumulative flood probability is obtained.

Images