Regional sea level change analysis and prediction method and device

Abstract

The invention discloses a regional sea level change analysis and prediction method and device, and relates to the field of sea level height monitoring, and the method comprises the steps: carrying out the regional sea level change trend analysis and regional sea level change periodical analysis of an obtained sea level height data set, and obtaining a trend analysis result and a periodical analysis result; performing regional sea level climate change response analysis on the sea level height data set and the greenhouse gas content data to obtain a climate change response analysis result; performing regional sea level statistical prediction on the sea level height data set, and performing set prediction by using different prediction modes and different constraints based on a statistical prediction result to obtain a set prediction result; and based on a trend analysis result, a periodic analysis result, a climate change response analysis result and a set prediction result, predicting the regional sea level high-resolution value to obtain a future sea level height. According to the method, the change trend and periodicity of the sea level can be accurately analyzed, and future sea level change can be predicted.

Description

Technical Field

[0001] The present application relates to the technical field of sea level monitoring, and in particular to a method and device for analyzing and predicting regional sea level changes. Background Art

[0002] Sea level change is a key indicator of global climate change, with profound impacts on human society and natural ecosystems. Accurately monitoring and predicting sea level change is crucial for developing strategies to address climate change. In recent years, with advances in science and technology, sea level monitoring technology has continued to develop, accumulating a large amount of time series data. This data provides a rich source of information for analyzing and predicting sea level change. However, sea level data are often affected by multiple factors, including seasonal fluctuations, long-term trends, and random noise. Therefore, effectively extracting useful information from this data has become a critical issue.

[0003] In the current field of sea level monitoring, there is limited analysis of long-term fluctuations across a region. Furthermore, it is necessary to remove interfering noise, conduct trend and periodicity analyses, and then use multiple methods to generate predictions, involving multiple complex processes. However, for sea level-related decision-making, analyzing and predicting sea level changes is essential. Therefore, accurately analyzing sea level trends and periodicity, as well as future sea level changes, has become a pressing technical challenge in this field. Summary of the Invention

[0004] The purpose of this application is to provide a method and device for analyzing and predicting regional sea level changes, which can accurately analyze the changing trends and periodicity of sea level and predict future sea level changes.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a method for analyzing and predicting regional sea level changes, the method comprising:

[0007] Get sea level height datasets and greenhouse gas content data.

[0008] A regional sea level change trend analysis is performed on the sea level height dataset to obtain a trend analysis result.

[0009] A regional sea level change periodicity analysis is performed on the sea level height dataset to obtain a periodicity analysis result.

[0010] A regional sea level climate change response analysis is performed on the sea level height dataset and the greenhouse gas content data to obtain a climate change response analysis result.

[0011] Performing regional sea level statistical prediction on the sea level height dataset to obtain statistical prediction results.

[0012] Based on the statistical prediction results, ensemble prediction is performed using different prediction modes and different constraints to obtain ensemble prediction results; the prediction modes include: ensemble Kalman filtering, nonlinear forced singular vector assimilation, relaxation approximation assimilation and linear interpolation method; the constraints include: random optimal perturbation, climate-related singular vectors and orthogonal condition nonlinear optimal perturbation.

[0013] Based on the trend analysis results, the periodicity analysis results, the climate change response analysis results and the ensemble prediction results, a high-resolution numerical value of the regional sea level is predicted to obtain the future sea level height.

[0014] In a second aspect, the present application provides a regional sea level change analysis and prediction device, which is used to implement the above-mentioned regional sea level change analysis and prediction method, and the regional sea level change analysis and prediction device includes:

[0015] Sea level high data set acquisition module, used to acquire sea level high data set;

[0016] A regional sea level change trend analysis module is used to perform regional sea level change trend analysis on the sea level height dataset to obtain trend analysis results;

[0017] a regional sea level change periodicity analysis module, configured to perform a regional sea level change periodicity analysis on the sea level height dataset to obtain a periodicity analysis result;

[0018] A regional sea level climate change response analysis module is used to perform regional sea level climate change response analysis on the sea level high data set and the greenhouse gas content data to obtain a climate change response analysis result;

[0019] A regional sea level statistical prediction module is used to perform regional sea level statistical prediction on the sea level height dataset to obtain a statistical prediction result;

[0020] An ensemble prediction module is configured to perform ensemble prediction based on the statistical prediction results using different prediction modes and different constraints to obtain an ensemble prediction result; the prediction modes include: ensemble Kalman filtering, nonlinear forced singular vector assimilation, relaxation approximation assimilation, and linear interpolation; the constraints include: random optimal perturbation, climate-related singular vectors, and orthogonal conditional nonlinear optimal perturbation;

[0021] The regional sea level high-resolution numerical prediction module is used to predict the regional sea level high-resolution numerical value based on the trend analysis results, the periodic analysis results, the climate change response analysis results and the ensemble prediction results to obtain the future sea level height.

[0022] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0023] The present application provides a method and device for analyzing and predicting regional sea level changes. By acquiring a high sea level data set and greenhouse gas content data, performing a regional sea level change trend analysis on the high sea level data set, and obtaining a trend analysis result, it is possible to clarify the overall change direction of the regional sea level in the long term or a specific time period, that is, whether the sea level is rising, falling, or relatively stable; by performing a regional sea level change periodicity analysis on the high sea level data set, obtaining a periodicity analysis result, it is possible to reveal the periodic laws of regional sea level changes; by performing a regional sea level climate change response analysis on the high sea level data set and the greenhouse gas content data, obtaining a climate change response analysis result, it is possible to clarify the sensitivity of regional sea level changes to climate change; by analyzing the high sea level data set and the greenhouse gas content data, it is possible to obtain a climate change response analysis result. The method can make statistical predictions of regional sea levels based on the data set to obtain statistical prediction results. Based on the statistical prediction results, ensemble predictions are made using different prediction models and different constraints to obtain ensemble prediction results. The method can make a preliminary estimate of the sea level change in the short term in the future based on the statistical laws of historical data. At the same time, the ensemble prediction can better consider various possible influencing factors and uncertainty factors, thereby improving the accuracy and confidence of the prediction. The method can make a high-resolution numerical prediction of the regional sea level based on the trend analysis results, the periodicity analysis results, the climate change response analysis results and the ensemble prediction results to obtain the future sea level, thereby providing a scientific and reliable quantitative basis for relevant decision-making, helping to plan response measures in advance and reduce the risks and losses caused by sea level rise. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 A flowchart of a method for analyzing and predicting regional sea level changes provided in one embodiment of the present application.

[0026] Figure 2This is a schematic diagram of the structure of the regional sea level change analysis and prediction software provided in one embodiment of the present application.

[0027] Figure 3 This is a flowchart of a multi-source data preprocessing module provided in one embodiment of the present application.

[0028] Figure 4 A flowchart for analyzing regional sea level change trends provided in one embodiment of the present application.

[0029] Figure 5 A flowchart for analyzing regional sea level change cycles provided in one embodiment of the present application.

[0030] Figure 6 A flow chart of regional sea level response to climate change provided in one embodiment of the present application.

[0031] Figure 7 A flowchart of regional sea level statistical prediction provided in one embodiment of the present application.

[0032] Figure 8 A flowchart of regional sea level ensemble prediction provided in one embodiment of the present application.

[0033] Figure 9 A flowchart of high-resolution numerical prediction of regional sea level provided in one embodiment of the present application.

[0034] Figure 10 A schematic diagram of a multi-source data processing interface provided in one embodiment of the present application.

[0035] Figure 11 A schematic diagram of a trend analysis interface provided in one embodiment of the present application.

[0036] Figure 12 A schematic diagram of a cycle analysis interface provided in one embodiment of the present application.

[0037] Figure 13 A schematic diagram of a climate response analysis interface provided in one embodiment of the present application.

[0038] Figure 14 A schematic diagram of the regional sea level statistical prediction interface provided in one embodiment of the present application.

[0039] Figure 15 A schematic diagram of the regional sea level ensemble prediction interface provided in one embodiment of the present application.

[0040] Figure 16 A schematic diagram of a high-resolution regional sea level prediction interface provided in one embodiment of the present application.

[0041] Figure 17 A schematic diagram of the results output interface provided in one embodiment of the present application.

[0042] Figure 18 A schematic diagram of the functional modules of a regional sea level change analysis and prediction device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0044] This application first processes global multi-source data, filters out data with poor signal quality, and then performs spatiotemporal bilinear interpolation to extract data to form a data set for subsequent analysis. In the trend analysis, linear and quadratic polynomials are used for fitting to conduct trend analysis. In the periodic analysis, wavelet analysis and Fourier transform are used to perform periodic analysis of the sea level signal to demonstrate its periodicity. In the climate response analysis, the correlation coefficient method is used to determine the response relationship of sea level height to greenhouse gases. At the same time, statistical prediction, ensemble prediction and high-resolution prediction are used to perform sea level prediction analysis.

[0045] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0046] In an exemplary embodiment, Figure 1 As shown, a method for analyzing and predicting regional sea level changes is provided, and the method for analyzing and predicting regional sea level changes includes:

[0047] S1: Obtain sea level height dataset and greenhouse gas content data.

[0048] S2: performing a regional sea level change trend analysis on the sea level height dataset to obtain a trend analysis result.

[0049] S3: Performing a regional sea level change periodicity analysis on the sea level height dataset to obtain a periodicity analysis result.

[0050] S4: performing a regional sea level climate change response analysis on the sea level height dataset and the greenhouse gas content data to obtain a climate change response analysis result.

[0051] S5: Performing regional sea level statistical prediction on the sea level height dataset to obtain statistical prediction results.

[0052] S6: based on the statistical prediction result, ensemble prediction is performed using different prediction modes and different constraints to obtain an ensemble prediction result; the prediction modes include ensemble Kalman filter (EnKF), nonlinear forced singular vector assimilation, relaxation method assimilation and linear interpolation method; the constraints include random optimal disturbance, climate-related singular vector and orthogonal condition nonlinear optimal disturbance.

[0053] S7: based on the trend analysis result, the periodicity analysis result, the climate change response analysis result and the ensemble prediction result, a regional sea level high-resolution numerical prediction is performed to obtain a future sea level height.

[0054] In step S1, the regional sea level change analysis and prediction can filter and extract sea level data of any region in the world, and form a data set using spatial and temporal bilinear interpolation; in step S2, a polynomial fitting method is used for trend analysis and display; in step S3, wavelet analysis and Fourier transform are used, the former is used for denoising, and the latter is used for time domain to frequency domain conversion to facilitate the extraction of cycles; in step S4, the response of sea level change to greenhouse gases is converted into the size of the correlation coefficient between them, and the calculation of the correlation coefficient can directly reflect the response degree; in step S5, the exponential smoothing method or autoregressive moving average method (ARMA) is used for statistical prediction; in step S6, different prediction modes and multiple constraints are used to obtain the final prediction result; in step S7, a long short-term memory (LSTM) network of time series is used to predict the regional sea level interannual variation. The historical sea level time series and the factors that significantly affect the regional sea level interannual variation are used as input data, and the future sea level height is used as output. The implementation of the above steps S1 to S7 can not only accurately analyze the change trend and periodicity of the sea level, but also predict the future sea level change. These research results will provide scientific basis and technical support for coping with sea level rise, disaster prevention and mitigation, and formulating relevant policies and measures, and provide comprehensive insights and solutions for understanding and coping with sea level change, thereby contributing to the safety and sustainable development of coastal areas.

[0055] As an optional implementation, the regional sea level change analysis and prediction method further comprises:

[0056] S8: the trend analysis result, the periodicity analysis result, the climate change response analysis result, the statistical prediction result, the ensemble prediction result and the future sea level height are graphically displayed. The structure diagram of the regional sea level change analysis and prediction software is shown in Figure 2 .

[0057] As an optional implementation, in step S1, obtaining a sea level height dataset specifically includes:

[0058] S11: Acquire historical data files; the historical data files include: altimetry satellite data files, tide gauge station data files, ocean reanalysis data files and CMIP5 model data files.

[0059] S12: Verify the correctness of the file format of the historical data file to obtain verified data.

[0060] S13: Performing spatiotemporal matching on the verified data using spatiotemporal bilinear interpolation to obtain a sea level height dataset.

[0061] like Figure 3 As shown in the figure, in the multi-source data preprocessing module, data files obtained from different World Meteorological and Oceanographic Data Centers are first opened and the correctness of the file format and the integrity of the file content are verified. Then, the spatiotemporal resolution of the final dataset is determined based on the spatiotemporal resolution of the different World Meteorological and Oceanographic data. Parameters such as the time series of sea level height at a specific point or region around the world are read, and the different data files are spatiotemporally aligned using spatiotemporal bilinear interpolation to integrate them into a single dataset.

[0062] In multi-source data preprocessing, the specific steps are as follows:

[0063] (1) To integrate different data files into one dataset, we need to consider the differences in spatiotemporal resolution of different data files. We use spatiotemporal bilinear interpolation to unify the spatiotemporal resolution of different data files. First, we determine the four points closest to the interpolation point: v1, v2, v3, and v4.

[0064] (2) For the four nearest neighboring points near the interpolation point, calculate their relative distances to the interpolation point and calculate weights based on the distances. Usually, the reciprocal of the distance or other weight functions are used to calculate the weights w1, w2, w3, and w4.

[0065] (3) Use the following formula to calculate the value of the difference point V:

[0066] V=v1*w1+v2*w2+v3*w3+v4*w4 (1);

[0067] As an optional implementation, in step S2, regional sea level change trend analysis is performed on the sea level height dataset to obtain trend analysis results, specifically including:

[0068] S21: Determine the fitting order of the polynomial fitting model based on the sea level height dataset.

[0069] S22: Based on the fitting order, with time as the independent variable and sea level as the dependent variable, perform polynomial fitting to obtain trend analysis results.

[0070] like Figure 4 As shown in the figure, in the regional sea level change trend analysis, the sea level height dataset processed in step S1 is opened and the correctness of the file format and the integrity of the file content are verified. Then, the processed sea level height file is input into the polynomial fitting model. The order of fitting is selected as needed, first or second order, using time as the independent variable and sea level height as the dependent variable. The polynomial fitting is performed and the results are finally output.

[0071] In the regional sea level change trend analysis, the specific steps are as follows:

[0072] (1) Reading of basic data on sea level changes.

[0073] In this step, the basic data of sea level change obtained by the multi-source data processing module is read through a function, that is, the time series of sea level change at a certain observation point (if it is a certain area, the average is taken).

[0074] (2) Selection of polynomial fitting model.

[0075] For the sea level change time series, a polynomial of appropriate order is used as the model of the polynomial fitting model.

[0076] The data is processed by the polynomial. Generally, the order of the polynomial does not exceed three. As shown below:

[0077] x i =Tt=a0+a1(t-t0)+a2(t-t0) 2 (2);

[0078] Among them, x i is the sea level height sequence; T is the theoretical time of fitting; t is the observation time; t0 is the reference time, which is the time point from which satellite time is calculated; a0, a1 and a2 are polynomial parameters.

[0079] (3) Polynomial parameter estimation

[0080] Let the model error be v i , then the error equation can be established according to the above formula:

[0081] x i +v i =a0+a1(t i -t0)+a2(t i -t0) 2 (3);

[0082] Among them, vi is the model error, t i is x i The corresponding time series.

[0083] Parameter estimation is performed under the least squares criterion. is the estimated value of polynomial parameters a0, a1, a2, record:

[0084]

[0085] Where A is the polynomial element excluding the parameters, and L is the sea level sequence. The estimated value under the least squares criterion can be obtained:

[0086]

[0087] The multinomial fitting model is as follows:

[0088]

[0089] As an optional implementation, in step S3, performing a regional sea level change periodicity analysis on the sea level height dataset to obtain a periodicity analysis result specifically includes:

[0090] S31: performing wavelet analysis on the sea level height data set to separate wavelet signals at various levels.

[0091] S32: Perform denoising processing on the wavelet signals at each level to obtain wavelet signals with denoised signals.

[0092] S33: Perform Fourier transform on the wavelet signal from which noise has been removed to obtain a transformed signal.

[0093] S34: Determine a signal with the greatest intensity based on the transformed signals.

[0094] S35: Taking the inverse of the frequency of the signal with the largest intensity to obtain a periodicity analysis result.

[0095] like Figure 5 As shown in the figure, in the analysis of regional sea level change periodicity, the sea level height dataset processed in step S1 is opened and the file format correctness and file content integrity are verified. Then, the processed sea level height file is subjected to wavelet analysis to separate the wavelets at each level. After removing some noise, the remaining signal is the sea level height change signal with periodicity. The signal is then Fourier transformed to extract the signal with the strongest intensity, calculate its period, and finally output.

[0096] In the periodic analysis of regional sea level changes, the specific steps are as follows:

[0097] (1) First, wavelet analysis is used to process the obtained data to eliminate errors. The idea of ​​wavelet analysis is to use a scaling function to represent the original signal. As the scale level decreases, the scale becomes larger and larger, and the representation of the original signal becomes increasingly rough and fuzzy, and the difference from the original signal becomes larger and larger. The wavelet function is introduced to represent the difference between the scaling function representation part and the original signal. Finally, the scaling function + wavelet function can accurately represent the original signal. The coefficients applied to the scaling function and the wavelet function are the results of the wavelet transform. Using these coefficients, the original signal can be restored. The following is the general form of the wavelet function:

[0098]

[0099] Among them, ψ * is a continuous mother wave, a is a scale factor, b is a translation factor, and f(t) is an arbitrary function. Parameter a is used to locate the frequency, and parameter b is used to locate the time. A suitable wavelet basis is used for calculation. For this sea level change time series, the wavelet basis selected is the Dobesi wavelet. The Dobesi wavelet is based on orthogonal wavelets and has good denoising properties for discrete wavelet transforms. By controlling the movement and scaling of the wavelet basis on the original function through a and b, the wavelet function values ​​of the wavelet basis and the original signal are obtained. If the frequency is higher, the similarity between the two is higher, indicating that the signal of this type of wavelet is contained. If the frequency is smaller, it means that the signal of this type does not exist, and thus it is used as a noise filter. After multi-level wavelet decomposition, the noise-removed signal is finally obtained.

[0100] (2) After obtaining the noise-removed signal, it is necessary to perform periodic extraction. At this time, the Fourier transform is used to convert the signal in the time domain into the signal in the frequency domain. The formula is as follows:

[0101]

[0102] Among them, w represents frequency, t represents time, e -iwt Represents a complex function. The Fourier transform assumes that a periodic function contains multiple frequency components, and any function f(t) can be synthesized by adding multiple periodic functions. In a signal analyzed by wavelet analysis, although most noise has been removed by technical means, some subtle noise and the main periodic signal still remain. Converting the signal to frequency-intensity using the Fourier transform allows for intuitive separation of the noise and main signal, thereby obtaining the desired periodic band.

[0103] (3) After Fourier transform, the signal with the largest intensity is obtained. The period of the signal is calculated. The relationship between period and frequency is as follows:

[0104]

[0105] Here, P is the period and f is the frequency. The two are reciprocal to each other; the larger the frequency, the smaller the period. By extracting the strongest signal from the Fourier transform and finding its frequency, the reciprocal is its period, thus giving the period of the main signal.

[0106] As an optional implementation, in step S4, regional sea level climate change response analysis is performed on the sea level high dataset and greenhouse gas content data to obtain climate change response analysis results, specifically including:

[0107] S41: Obtain greenhouse gas content data.

[0108] S42: performing time averaging processing on the sea level height dataset and the greenhouse gas content data respectively to obtain a sea level height time signal and a greenhouse gas content time signal.

[0109] S43: performing correlation calculation on the sea level height time signal and the greenhouse gas content time signal to obtain a correlation result.

[0110] S44: Based on the correlation results, obtain climate change response analysis results.

[0111] like Figure 6 As shown in the figure, in the regional sea level climate change response analysis, the sea level height dataset processed in step S1 is opened, followed by the greenhouse gas content data file. Both files are verified for formatting and content integrity. Next, the greenhouse gas type to be analyzed is selected, and the processed sea level height file is time-averaged, averaging by month. The greenhouse gas content data for the corresponding region is also filtered and averaged monthly. The correlation between the sea level height data and the greenhouse gas content data is then calculated based on the time signal, with the magnitude of the correlation reflecting the degree of response. Finally, the data is output.

[0112] In the regional sea level response to climate change, the specific steps are as follows:

[0113] After reading the sea level data and greenhouse gas data, we process them and calculate the correlation coefficient, or linear correlation coefficient, which is an indicator of the degree of linear correlation between two random variables. The correlation coefficient calculation formula is:

[0114]

[0115] Where N is the number of greenhouse gas data after time matching, which is the same as the number of sea level height series data; X i is the ith greenhouse gas data; is the average value of N greenhouse gas data; Y i is the i-th sea level height data; is the average value of N sea level height data; the correlation coefficient r XY The value range of r is [-1,1], XY >0 indicates positive correlation, r XY <0 indicates negative correlation, |r XY | indicates the degree of correlation between variables. In particular, r XY =1 is called perfect positive correlation, r XY = -1 is called perfect negative correlation, r XY = 0 is called irrelevant. Usually | r XY When | is greater than 0.8, the two variables are considered to have a strong linear correlation.

[0116] As an optional implementation, in step S5, performing regional sea level statistical prediction on the sea level height dataset to obtain statistical prediction results specifically includes:

[0117] S51: Based on the sea level height dataset, a time series decomposition method is used to decompose the regional mean sea level time series into an interannual variation term, a seasonal term, and a disturbance term.

[0118] S52: Using an exponential smoothing method to fit the interannual variation term, to obtain a fitted interannual variation term.

[0119] S53: Using an autoregressive moving average method to fit the disturbance term to obtain a fitted disturbance term.

[0120] S54: superimposing the seasonal term onto the fitted interannual variation term and the fitted disturbance term to obtain a complete time series statistical prediction model.

[0121] S55: Obtain statistical prediction results based on the complete time series statistical prediction model.

[0122] like Figure 7 As shown in the figure, in regional sea level statistical prediction, the sea level height dataset processed in step S1 is first opened. Using a time series decomposition method, the regional mean sea level time series is decomposed into three components: an interannual variation term, a seasonal term, and a disturbance term. Based on the variation characteristics of these three time series components, exponential smoothing or an autoregressive moving average method is used to fit the interannual variation term and the disturbance term in the time series. The seasonal term is superimposed on the first two as a cyclic variation term. This establishes a statistical prediction model for this time series.

[0123] In regional sea level statistical prediction, the specific steps are as follows:

[0124] (1) Data pattern analysis.

[0125] Before performing time series decomposition, it is necessary to first analyze the regularity of regional sea level data. This includes examining the stability, trend, and seasonality of the data to ensure that an appropriate decomposition method is selected.

[0126] (2) Data preprocessing.

[0127] Before performing time series decomposition, some preprocessing of the data is required, such as processing missing values, outliers, or performing smoothing operations to ensure the quality of the data.

[0128] (3) Time series decomposition.

[0129] Decompose the regional mean sea level time series into interannual variation, seasonality, and disturbance. Classic time series decomposition methods such as STL decomposition (Seasonal-Trend decomposition using LOESS) or empirical mode decomposition (Empirical Mode Decomposition) can be used.

[0130] (4) Use exponential smoothing method to fit the interannual variation terms.

[0131] Assume that the time series is y1, y2, ..., y t ,…,α is the weighting coefficient, 0<α<1, the first exponential smoothing formula is:

[0132]

[0133] in, is the smoothed value of the first exponential in period t; is the smoothed value of the first exponential in period t-1.

[0134] Formula (11) is an improvement of the moving average formula. Expanding formula (11) in sequence, we have:

[0135]

[0136] in, is the smoothed value of the first exponential in period t-2.

[0137] Formula (12) shows that It is the weighted average of all historical data, and the weighting coefficients are α, α(1-α), α(1-α) 2 , ..., obviously:

[0138]

[0139] Since the weighting coefficient conforms to the exponential law and has the function of smoothing data, it is called exponential smoothing. Forecasting with this smoothed value is the first-order exponential smoothing method. The prediction model is:

[0140]

[0141] Right now:

[0142]

[0143] That is, the exponential smoothing value of period t is used as the predicted value of period t+1.

[0144] When performing exponential smoothing, the choice of weighting coefficient is very important. α should be selected between 0 and 1 based on the specific properties of the time series. The specific selection can generally follow the following principles:

[0145] ① If the time series does not fluctuate much and is relatively stable, α should be smaller, such as 0.1 to 0.5, to reduce the correction amplitude and enable the prediction model to include information from longer time series.

[0146] ② If the time series has a rapid and obvious tendency to change, α should be larger, such as 0.6 to 0.8, so that the prediction model is more sensitive and can quickly keep up with the changes in the data. When using the exponential smoothing method for prediction, in addition to choosing a suitable α, the initial value must also be determined. Generally, the average of the actual values ​​of the first few periods is used as the initial value.

[0147] When the time series changes show a linear trend, there will be obvious lag bias when using the first exponential smoothing method for prediction. Therefore, it must be corrected, that is, second exponential smoothing is performed, and the linear trend model is established using the law of lag bias. The calculation formula is:

[0148]

[0149] Where, is the smoothed value of the first exponential in period t, is the smoothed value of the quadratic exponential in period t, is the smoothed value of the quadratic exponential in period t-1. t}, when there is a linear trend starting from a certain period, the linear trend model can be used for prediction:

[0150]

[0151] When the change of the time series shows a quadratic curve trend, the triple exponential smoothing method is needed. The triple exponential smoothing method is based on the quadratic exponential smoothing and then performs another smoothing. Its calculation formula is:

[0152]

[0153] In the formula is the triple exponential smoothing value of period t, is the triple exponential smoothing value of period t-1.

[0154] The forecast model of the triple exponential smoothing method is:

[0155]

[0156] Where a t is the smoothed estimate at the current moment; b t is a linear growth trend; c t is the acceleration of the trend; T is the prediction step, that is, the time interval from the current moment to the future prediction.

[0157] (5) Use the autoregressive moving average method to fit the disturbance term.

[0158] ① Implicit Green's function of ARMA system:

[0159] The ARMA model is a second-order inhomogeneous difference equation expressed as:

[0160]

[0161] Where, X t is the time series value at the current moment t; X t-1 ,X t-2 is the time series value of the previous period t-1 and the previous two periods t-2; and is the regression coefficient, which indicates the current value of X t The degree of influence of the data of the previous period and the previous two periods; a t is white noise, assuming it is an independent and identically distributed random variable with mean zero; θ is the sliding mean coefficient, representing the error term a t Affected by the previous white noise a t-1 The degree of influence. Assume that the solution of the second-order nonhomogeneous difference equation is:

[0162]

[0163] Where G j is the coefficient, representing the past noise term a t-j For the current value X t impact.

[0164] For convenience, the B operator can be used:

[0165]

[0166] Substituting (23) into (22) and comparing the coefficients of the same power of B on both sides, we can obtain:

[0167]

[0168] ②Explicit expression of Green's function of ARMA system.

[0169] The ARMA model is essentially a second-order non-homogeneous difference equation. To find its solution, we must first find the general solution of its corresponding homogeneous difference equation.

[0170] The characteristic equation corresponding to the homogeneous equation is:

[0171]

[0172] The characteristic roots are:

[0173]

[0174] The general solution of the homogeneous difference equation is:

[0175]

[0176] Where λ1 is the first root characteristic; λ2 is the second root characteristic; c1 and c2 are arbitrary constants whose values ​​are uniquely determined by the initial conditions. The initial conditions here are:

[0177]

[0178] So we have:

[0179]

[0180] The solution is:

[0181]

[0182] Then the Green function of the ARMA system is:

[0183]

[0184] (6)Seasonal items are superimposed.

[0185] The decomposed seasonal terms are superimposed on the interannual variation terms and disturbance terms obtained by fitting to obtain a complete time series model.

[0186] (7) Model evaluation and adjustment.

[0187] To evaluate the established model, various statistical indicators (such as mean square error, mean absolute error, etc.) and graphs (such as residual graphs and prediction graphs) can be used to detect the model's degree of fit and prediction performance.

[0188] If necessary, parameter tuning can be performed to improve the accuracy of the model.

[0189] (8) Prediction.

[0190] Use the established model to predict future data. Single-step or multi-step predictions can be performed as needed.

[0191] (9) Model application and verification.

[0192] Apply the established model to real data and verify it. This verification can include backtesting on historical data and real-time prediction of future data.

[0193] As an optional implementation, in step S6, based on the statistical prediction results, different prediction modes and different constraints are used to perform collective prediction to obtain collective prediction results, specifically including:

[0194] S61: Initialize different prediction modes respectively to obtain initialized modes.

[0195] S62: Based on the statistical prediction results and different constraints, uncertainty measurement of initial conditions or model errors is performed on the initialized model to obtain multiple set members.

[0196] S63: Integrate the multiple set members to obtain a set prediction result.

[0197] like Figure 8 As shown in the figure, regional sea level ensemble predictions primarily draw on data processing methods from meteorological oceanography. Each model uses ensemble Kalman filtering, nonlinear forced singular vector assimilation, nudging assimilation, and linear interpolation to assimilate atmospheric and oceanic data for model initialization. Ensemble methods using random optimal perturbations, climate-related singular vectors, and orthogonal conditional nonlinear optimal perturbations are used to measure the uncertainty of initial conditions or model errors. Ensemble members are generated, ultimately providing ensemble predictions.

[0198] In the regional sea level ensemble prediction, the specific steps are as follows:

[0199] (1) Mode initialization.

[0200] Ensemble Kalman Filter:

[0201] Initialization: Initialize system state variables Initial error covariance matrix and the initial state prediction set

[0202] Data communication: time t k , respectively, carry out observation step, analysis step and prediction step, the specific process is as follows:

[0203] Observation step:

[0204] Compute observation set:

[0205]

[0206] where y is the mean of the observations, i.e. the main signal of the observation; y i is the i-th observation in the observation set; u i are the observation errors, whose mean is zero.

[0207] Compute observation error covariance matrix:

[0208]

[0209] where R u is the covariance matrix of the observation errors; u i are the observation errors; m is the number of set members.

[0210] Analysis step:

[0211] Compute Kalman gain matrix:

[0212]

[0213] where P f is the prediction error covariance matrix, indicating the degree of uncertainty of the predicted state; H is the observation matrix, describing the relationship between the observation values and the state variables; R u is the observation noise covariance matrix, indicating the influence of measurement noise; is the Kalman gain matrix, used to adjust the predicted state to correct it to the observation value.

[0214] Compute analysis set:

[0215]

[0216] where x is the state value of the prediction set member; is the state value of the analysis set member; y i is the observation value; is the projection of the predicted state into the observation space.

[0217] Compute analysis set mean:

[0218]

[0219] where x is the mean of the analysis state; m is the number of set members; is the analysis state of each set member.

[0220] Calculate the analysis error covariance matrix:

[0221]

[0222] Where, P a is the analysis error covariance matrix; is the error term that deviates from the mean.

[0223] Prediction step:

[0224] Compute the prediction set:

[0225]

[0226] Where, is the state value predicted for the next step; M(x) is the state transfer equation, which describes the state evolution of the system from the current moment to the next moment.

[0227] Compute the forecast ensemble mean:

[0228]

[0229] Where, is the mean of the prediction set.

[0230] Compute the forecast error covariance matrix:

[0231]

[0232] Where, P f is the prediction error covariance matrix.

[0233] (2) Measure the uncertainty of initial conditions or model errors.

[0234] Stochastic optimal perturbations (SOs): SOs are random perturbations introduced into a forecasting model to simulate uncertainty and random variations in the system. By analyzing the impact of SOs, we can assess the extent to which uncertainties in initial conditions or model errors affect forecast results.

[0235] Climatically relevant singular vectors (CSVs): CSVs are singular vectors that describe long-term variations in the climate system. By analyzing ensemble methods of CSVs, we can understand the impact of climate factors on forecasts and accurately assess the reliability of forecasts.

[0236] Orthogonal Conditional Nonlinear Optimal Perturbations (O-CNOP): O-CNOP is a method for evaluating nonlinear perturbations in a system, which can help understand the sources of model errors and their impact on prediction results.

[0237] (3) Generate ensemble members and finally give ensemble prediction results.

[0238] After the initialization and disturbance impact analysis described above, a diverse ensemble prediction can be obtained by combining multiple ensemble members. This can take into account the uncertainty of the model and initial conditions, improving the reliability of the prediction.

[0239] As an optional implementation, in step S7, based on the trend analysis results, the periodicity analysis results, the climate change response analysis results, and the ensemble prediction results, a high-resolution numerical prediction of the regional sea level is performed to obtain the future sea level, specifically including:

[0240] S71: Use neural network tools to build a sea level prediction model.

[0241] S72: Input the trend analysis result, the periodicity analysis result, the climate change response analysis result and the ensemble prediction result into the sea level prediction model to obtain the future sea level height.

[0242] like Figure 9 As shown, in the high-resolution numerical prediction of regional sea level, a neural network tool is used to establish a sea level prediction model. The historical sea level time series and the factors that significantly affect the interannual variation of regional sea level are used as model training data. The high-resolution numerical prediction model of regional sea level constructed by the neural network is used to solve and predict future sea level heights. It should be noted that in this application, the historical sea level time series is the time series obtained based on the ensemble prediction results.

[0243] In the high-resolution numerical prediction of regional sea level, the specific steps are as follows:

[0244] (1) Initialize network parameters:

[0245] Define the structure of the LSTM network, including the input dimension, hidden state dimension, and output dimension.

[0246] Initialize the weight matrix and bias vector, usually using random initialization or pre-training methods.

[0247] (2) Define LSTM unit:

[0248] The LSTM unit consists of an input gate, a forget gate, an output gate, and a cell state. Each gate has its own weight and bias, as well as an activation function (usually a sigmoid function).

[0249] (3) For each time step t:

[0250] a. Input sequence:

[0251] The input sequence is fed into the LSTM network. The input gate is calculated: it controls how much information of the input data of the current time step enters the cell state. The activation value of the input gate is: i t =σ(W xi x t +W hi h t-1 +W cf c t-1 +b f )

[0252] Calculate the forget gate: control how much information of the previous time step is retained in the cell state. The activation value of the forget gate: f t =σ(W xf x t +W hf h t-1 +W cf c t-1 +b f )

[0253] Calculate candidate cell state: The candidate cell state is a temporary state used to calculate the new cell state. Candidate cell state:

[0254] Among them, x t is the value of the input data (at time step), h t-1 is the hidden state of the previous time step, c t-1 is the cell state at the previous time step, W xi , W hi , W ci , W xf , W hf , W cf , W xc , W hc is the weight matrix, b i , b f , b is the bias vector, σ is the sigmoid activation function, and tanh is the hyperbolic tangent activation function.

[0255] b. Update cell status:

[0256] The cell state is updated based on the input gate, forget gate, and candidate cell state.

[0257] The specific calculation method is:

[0258] c. Calculate the output:

[0259] Compute the output gate: controls how much information in the cell state is passed to the output.

[0260] Activation value of the output gate: o t =σ(W xo x t +W ho h t-1 +W co c t-1 +b o ).

[0261] d. Hidden state:

[0262] The hidden state is the output of the LSTM network, which can also be passed to the next time step. The value of the hidden state is: h t =o t *tanh(c t ).

[0263] (4) Repeat step (3): Repeat the above steps at each time step until the entire input sequence is processed.

[0264] (5) Output: Finally, depending on the task, the output can be obtained from the LSTM network, which can be the output of the final time step or the output of each time step.

[0265] In summary, the present application provides a method for analyzing and predicting regional sea level changes. The method is based on satellite altimeter observation data such as T / P, Jason-1, Jason-2, Jason-3, and HY-2, coastal tide station observation data, ocean reanalysis data, and CMIP5 model data. On the one hand, according to the law of tidal changes and the characteristics of satellite altimeter data and tide station data, outliers are removed to form reliable tide data with a certain time length and a unified time zone. This data is used as the basic data for analyzing sea level change trends and periodic characteristics. On the one hand, ocean reanalysis data and CMIP model data with different grid resolutions are interpolated onto a standard grid with the same resolution, and the sea level and salinity data of the studied area are divided from the global data to make data products. In addition, a variety of prediction methods such as statistical prediction and multi-model ensemble prediction are used to construct a regional sea level change prediction model. Under the background of different greenhouse gas concentrations defined in the fifth and sixth assessment reports of the IPCC, future changes in regional sea level are predicted. This method can realize graphical display of relevant data products and sea level prediction results. Through the analysis of this product, we can not only accurately analyze the changing trends and periodicity of sea level, but also predict future sea level changes. These research results will provide scientific basis and technical support for responding to sea level rise, disaster prevention and mitigation, and the formulation of relevant policies and measures, and provide comprehensive insights and solutions for understanding and responding to sea level changes, thereby contributing to the safety and sustainable development of coastal areas.

[0266] To demonstrate the effectiveness of this method, an experimental demonstration was conducted on a randomly selected global sea level region. This representative region, covering 24 years from 1994 to 2017, was used to analyze sea level changes. This region lies between 125°E and 126°E, and 30°N and 31°N, near the junction of the East China Sea and the Yellow Sea. This offshore area is particularly representative for analyzing sea level trends and cycles.

[0267] Step 1: Follow Figure 3 The process, in Figure 10 In the interface, input altimetry satellite data files, tide gauge data files and other files, and enter the area to be analyzed in the longitude and latitude area. The program will then automatically verify the correctness and completeness of the file format, filter the longitude and latitude, remove unqualified data, and perform bilinear interpolation time-space matching. After processing, the extracted files will be saved as a dataset for subsequent analysis.

[0268] Step 2: The sea level data set processed in step 1 is directly read into the regional sea level trend analysis module. After that, the first or second term fitting is selected, and the program will perform parameter estimation based on the least squares method. Figure 11 The corresponding graphics and fitting information are displayed in the graphic display box and text display box.

[0269] Step 3: The sea level data set processed in step 1 is directly read into the regional sea level periodicity analysis module. The system will use multi-level wavelet analysis to analyze the wavelet. Then, the denoised wavelet is converted from time domain to frequency domain using Fourier transform. Finally, the signal with the highest intensity is extracted and the inverse of its frequency is taken as the period. Figure 12 The corresponding graphics and periodic information are displayed in the graphic display box and text display box.

[0270] Step 4. In addition to the sea level height dataset processed in Step 1, the regional sea level climate response analysis module also requires the longitude and latitude of the greenhouse gas screening area to be entered in the longitude and latitude screening area. Then, the greenhouse gas file needs to be entered. After entering, time matching processing and averaging are performed first. Then, the variance of the sea level file and the greenhouse gas file is calculated. Finally, the correlation coefficient is calculated according to the formula. Figure 13 The text display box below shows the correlation coefficient between this gas and sea level height. The larger the correlation coefficient, the higher the impact.

[0271] Step 5. Input the sea level height dataset processed in step 1. The regional sea level statistical forecast will be tested for stationarity, and then the ARMA time series will be constructed, the ARMR parameters will be estimated, and the results will be predicted. Figure 14 In the graphic display box, the blue lines represent historical data, and the orange lines represent predicted data.

[0272] Step 6: Regional sea level ensemble prediction requires input of unprocessed CMIP6 data. Different prediction modes, such as ensemble Kalman filtering, nonlinear forced singular vector assimilation, relaxed approximation assimilation and linear interpolation methods, are then used to combine random optimal perturbations, climate-related singular vectors and orthogonal conditional nonlinear optimal perturbations, and then perform ensemble predictions to predict future sea level heights. Figure 15 The prediction results are shown.

[0273] Step 7: Input the sea level data set processed in step 1. The regional sea level high-resolution numerical prediction will normalize it using the initialized network, then establish a neural network, set parameters and train it until the test network model error reaches the required output. Figure 16 The prediction results are shown.

[0274] Step 8. After all modules have been run, the results before the module call can be displayed in the results. Figure 17 Shows the result of calling one of the modules.

[0275] Based on the same inventive concept, the embodiments of the present application also provide a regional sea level change analysis and prediction device for implementing the above-mentioned regional sea level change analysis and prediction method. The implementation scheme of the problem solving provided by the system is similar to the implementation scheme described in the above method, so the specific limitations in one or more regional sea level change analysis and prediction device embodiments provided below can refer to the limitations of the regional sea level change analysis and prediction method in the above, which will not be repeated here.

[0276] In one exemplary embodiment, as shown in Figure 18 a regional sea level change analysis and prediction device is provided for implementing the above-mentioned regional sea level change analysis and prediction method, the regional sea level change analysis and prediction device comprises:

[0277] a sea level height data set acquisition module M1 for acquiring a sea level height data set.

[0278] a regional sea level change trend analysis module M2 for performing regional sea level change trend analysis on the sea level height data set to obtain a trend analysis result.

[0279] a regional sea level change periodicity analysis module M3 for performing regional sea level change periodicity analysis on the sea level height data set to obtain a periodicity analysis result.

[0280] a regional sea level climate change response analysis module M4 for performing regional sea level climate change response analysis on the sea level height data set and the greenhouse gas content data to obtain a climate change response analysis result.

[0281] a regional sea level statistical prediction module M5 for performing regional sea level statistical prediction on the sea level height data set to obtain a statistical prediction result.

[0282] a set prediction module M6 for performing set prediction based on the statistical prediction result using different prediction modes and different constraints to obtain a set prediction result; the prediction modes include set Kalman filtering, nonlinear forced singular vector assimilation, relaxation approximation method assimilation and linear interpolation method; the constraints include random optimal disturbance, climate-related singular vector and orthogonal condition nonlinear optimal disturbance.

[0283] The regional sea level high-resolution numerical prediction module M7 is used to predict the regional sea level high-resolution numerical value based on the trend analysis results, the periodic analysis results, the climate change response analysis results and the ensemble prediction results, so as to obtain the future sea level height. The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the various technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0284] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for analyzing and predicting regional sea level changes, characterized in that: The regional sea level change analysis and prediction method includes: Obtaining sea level elevation datasets and greenhouse gas content data; performing a regional sea level change trend analysis on the sea level height dataset to obtain a trend analysis result; Performing a regional sea level change periodicity analysis on the sea level height dataset to obtain a periodicity analysis result; performing a regional sea level climate change response analysis on the sea level height dataset and the greenhouse gas content data to obtain a climate change response analysis result; Performing regional sea level statistical prediction on the sea level height dataset to obtain statistical prediction results; Based on the statistical prediction results, ensemble prediction is performed using different prediction modes and different constraints to obtain ensemble prediction results; the prediction modes include: ensemble Kalman filtering, nonlinear forced singular vector assimilation, relaxation approximation assimilation, and linear interpolation method; the constraints include: random optimal perturbation, climate-related singular vectors, and orthogonal condition nonlinear optimal perturbation; Based on the trend analysis results, the periodicity analysis results, the climate change response analysis results and the ensemble prediction results, a high-resolution numerical value of the regional sea level is predicted to obtain the future sea level height.

2. The method for analyzing and predicting regional sea level changes according to claim 1, wherein: The regional sea level change analysis and prediction method further includes: The trend analysis results, the periodicity analysis results, the climate change response analysis results, the statistical prediction results, the ensemble prediction results and the future sea level height are graphically displayed.

3. The method for analyzing and predicting regional sea level changes according to claim 1, wherein: Get sea level height dataset, including: Obtain historical data files; the historical data files include: altimetry satellite data files, tide gauge data files, ocean reanalysis data files and CMIP5 model data files; Verifying the correctness of the file format of the historical data file to obtain verified data; The verified data are temporally and spatially matched using spatiotemporal bilinear interpolation to obtain a sea level height dataset.

4. The method for analyzing and predicting regional sea level changes according to claim 1, wherein: Perform regional sea level change trend analysis on the sea level height dataset to obtain trend analysis results, specifically including: Determining a fitting order of a polynomial fitting model based on the sea level height dataset; Based on the fitting order, polynomial fitting is performed with time as the independent variable and sea level as the dependent variable to obtain trend analysis results.

5. The method for analyzing and predicting regional sea level changes according to claim 1, wherein: Performing a regional sea level change periodicity analysis on the sea level height dataset to obtain periodicity analysis results, specifically including: Performing wavelet analysis on the sea level height data set to separate wavelet signals at various levels; Performing denoising on the wavelet signals at each level to obtain wavelet signals with noise removed; Performing Fourier transform on the wavelet signal after noise removal to obtain a transformed signal; Determining a signal with the greatest intensity based on the transformed signal; The reciprocal of the frequency of the signal with the largest intensity is taken to obtain a periodic analysis result.

6. The method for analyzing and predicting regional sea level changes according to claim 1, wherein: Perform regional sea level climate change response analysis on the sea level high dataset and greenhouse gas content data to obtain climate change response analysis results, specifically including: Obtaining greenhouse gas content data; performing time averaging processing on the sea level height dataset and the greenhouse gas content data respectively to obtain a sea level height time signal and a greenhouse gas content time signal; performing a correlation calculation on the sea level height time signal and the greenhouse gas content time signal to obtain a correlation result; Based on the correlation results, climate change response analysis results are obtained.

7. The method for analyzing and predicting regional sea level changes according to claim 1, wherein: Performing regional sea level statistical prediction on the sea level height dataset to obtain statistical prediction results, specifically including: Based on the sea level dataset, the regional mean sea level time series is decomposed into interannual variation term, seasonal term and disturbance term using time series decomposition method. The interannual variation term is fitted using an exponential smoothing method to obtain a fitted interannual variation term; Using an autoregressive moving average method to fit the disturbance term to obtain a fitted disturbance term; Superimposing the seasonal term on the fitted interannual variation term and the fitted disturbance term to obtain a complete time series statistical prediction model; Based on the complete time series statistical prediction model, a statistical prediction result is obtained.

8. The method for analyzing and predicting regional sea level changes according to claim 1, wherein: Based on the statistical prediction results, ensemble prediction is performed using different prediction modes and different constraints to obtain ensemble prediction results, specifically including: Initialize different prediction modes respectively to obtain initialized modes; Based on the statistical prediction results and different constraints, uncertainty measurement of initial conditions or model errors is performed on the initialized model to obtain multiple set members; The multiple set members are integrated to obtain a set prediction result.

9. The method for analyzing and predicting regional sea level changes according to claim 1, wherein: Based on the trend analysis results, the periodicity analysis results, the climate change response analysis results, and the ensemble prediction results, a high-resolution numerical prediction of the regional sea level is performed to obtain the future sea level, specifically including: Use neural network tools to build sea level prediction models; The trend analysis result, the periodicity analysis result, the climate change response analysis result and the ensemble prediction result are input into the sea level prediction model to obtain the future sea level height.

10. A device for analyzing and predicting regional sea level changes, characterized in that: The regional sea level change analysis and prediction device is used to implement the regional sea level change analysis and prediction method according to any one of claims 1 to 9, and the regional sea level change analysis and prediction device includes: Sea level high data set acquisition module, used to acquire sea level high data set; A regional sea level change trend analysis module is used to perform regional sea level change trend analysis on the sea level height dataset to obtain trend analysis results; a regional sea level change periodicity analysis module, configured to perform a regional sea level change periodicity analysis on the sea level height dataset to obtain a periodicity analysis result; A regional sea level climate change response analysis module is used to perform regional sea level climate change response analysis on the sea level high data set and the greenhouse gas content data to obtain a climate change response analysis result; A regional sea level statistical prediction module is used to perform regional sea level statistical prediction on the sea level height dataset to obtain statistical prediction results; An ensemble prediction module is configured to perform ensemble prediction based on the statistical prediction results using different prediction modes and different constraints to obtain an ensemble prediction result; the prediction modes include: ensemble Kalman filtering, nonlinear forced singular vector assimilation, relaxation approximation assimilation, and linear interpolation; the constraints include: random optimal perturbation, climate-related singular vectors, and orthogonal conditional nonlinear optimal perturbation; The regional sea level high-resolution numerical prediction module is used to predict the regional sea level high-resolution numerical value based on the trend analysis results, the periodic analysis results, the climate change response analysis results and the ensemble prediction results to obtain the future sea level height.

Images