Marine environment data standardization method and system

Abstract

The invention discloses a marine environment data standardization method and system, and relates to the field of marine environment monitoring, and the method comprises the steps: carrying out the space grid processing of the environment data of a selected target sea area, obtaining a plurality of target sea area grid division schemes, and selecting the optimal grid data corresponding to the optimal grid division scheme; performing difference analysis on the optimal gridding data to determine existence of a marine ecological environment influence factor plane distribution reference normal form, and constructing the marine ecological environment influence factor plane distribution reference normal form; constructing a seasonal change reference normal form of the influence factors of the marine ecological environment, and constructing an interannual change reference normal form of the influence factors of the marine ecological environment; and supplementing missing data of the target sea area according to the three reference normal forms, and completing standardization of the marine environment data in the target sea area. According to the method, the missing marine environment data can be complemented on the spatio-temporal scale.

Description

Technical Field

[0001] The present invention relates to the field of marine environmental monitoring, and particularly to a method and system for standardizing marine environmental data. Background Art

[0002] The monitoring, management and application of marine environmental data are the key to the prediction and prevention of marine ecological environment risks. Due to the large capital and manpower investment required for offshore ship cruising and online buoy water quality monitoring systems, the spatial coverage and monthly frequency of monitoring data are limited, data in some sea areas are seriously missing, and the existing monitoring data have inconsistent spatio-temporal granularities, making it difficult to comprehensively reflect environmental conditions such as marine eutrophication. In particular, the spatio-temporal granularities of the original monitoring data are significantly different from those of eutrophication symptoms such as red tides and hypoxia, making it difficult to support the analysis of the environmental process mechanisms of ecological events. Therefore, at present, in scientific research and environmental management such as the simulation, evaluation and prevention and control plan formulation of the eutrophication state in coastal waters, standardized reference data with consistent spatio-temporal granularities are needed as support, which is of great significance for the prediction and prevention of marine ecological environment risks.

[0003] In the prior art, Yongzhi Liu et al. proposed an improved Cressman interpolation method and applied it to the standardization of the original monitoring data of total nitrogen in the ocean. Rushui Xiao et al. proposed a method for standardizing the spatiotemporal missing data of marine nitrogen and phosphorus nutrients by combining Chebyshev Polynomial Fitting (CPF) and STL (Seasonal-Trend decomposition using LOESS, STL) time series data decomposition and reconstruction. This method first uses spatial interpolation CPF to interpolate the spatially non-uniformly distributed marine nitrogen and phosphorus nutrient monitoring survey statistical data to the same grid as the Copernicus Marine Environment Monitoring Service (CMEMS) product data (1 / 12° horizontal resolution). Then the monitoring data and CMEMS product data are standardized (Minmax normalization). Secondly, STL is used to extract the trend and seasonal signals of the standardized monitoring data and CMEMS product data. Thirdly, the seasonal decomposition of the monitoring data is replaced by the seasonal decomposition of the CMEMS product data (calculated according to the proportion of the original time series). Finally, the monitoring data is decomposed into new seasons, and the trend terms and residual terms are summarized to obtain the completed data. The completed monitoring data retains the trend and detailed fluctuation of the original monitoring data, so that it can more realistically depict the changes in nitrogen and phosphorus nutrients in the target sea area, which is convenient for subsequent analysis. The above data standardization method adopts Chebyshev polynomial fitting when completing spatial data, which realizes the conversion of scattered plane distribution data into grid data covering the entire sea area. This method belongs to statistical interpolation method, which ignores the correlation of complex biogeochemical processes in the ocean with space and time, and can only be applied to spatial data interpolation in a single way. For the standardization of time series data, the above method performs STL decomposition on the same element data from different sources, replaces the seasonal items with each other, and then re-summarizes the items to obtain the completed data. This process not only ignores the important influence of abnormal data on the trend of factor changes, but also relies heavily on exogenous data, which has great limitations in the actual application of data standardization.

[0004] In summary, how to complete the missing marine environmental data in terms of time and space is an important issue that needs to be solved urgently. Summary of the invention

[0005] The embodiments of the present invention provide a method and system for standardizing marine environmental data, which can solve the problem in the prior art that missing marine environmental data cannot be supplemented in time and space scales.

[0006] An embodiment of the present invention provides a method for standardizing marine environmental data, including the following steps: Using the station coverage rate reflecting the aggregated data after removing outliers in the geographical coverage range of the target sea area, the grain saturation rate measuring the density of monitoring stations per unit sea area, and the seasonal balance degree reflecting the distribution balance of monitoring data in different seasons within a year, screen the environmental data of the target sea area and perform grid division; Use the hierarchical clustering analysis method to group the grid-divided data by similarity, and use the radial basis function to model each group of data to obtain the plane distribution reference paradigm of marine ecological environment impact factors; Select the environmental data corresponding to the year when the seasonal balance degree reaches the set threshold; according to the distribution pattern of the selected environmental data, determine the corresponding seasonal change reference paradigm of marine ecological environment impact factors; Use the seasonal change reference paradigm of marine ecological environment impact factors to obtain the seasonal change standardized data for each year; according to the seasonal change standardized data, use the Fourier series equation to model and obtain the interannual change reference paradigm of marine ecological environment impact factors; Use the plane distribution reference paradigm of marine ecological environment impact factors, the seasonal change reference paradigm of marine ecological environment impact factors, and the interannual change reference paradigm of marine ecological environment impact factors to fill in the missing plane distribution data, seasonal change data, and interannual change data in the target sea area respectively, and complete the standardization of marine environmental data in the target sea area.

[0007] Further, the specific steps of screening the environmental data of the target sea area include: Based on the environmental data of the target sea area, use the station coverage rate, the grain saturation rate, and the seasonal balance degree as the core criteria for measuring the quality of environmental data; For the environmental data screened according to each core criterion, use the standard classification method of equal interval and quantile to divide the environmental data into different grades of high, medium-high, medium, medium-low, and low according to the data quality; Use the high-quality environmental data as the environmental data of the screened target sea area.

[0008] Further, the specific steps of performing grid division include: Design 5 target sea area grid division schemes for the high-quality environmental data according to different spatial resolutions. The grid division schemes include: 0.25°×0.25°, 0.2°×0.2°, 0.15°×0.15°, 0.1°×0.1°, and 0.05°×0.05°; Use the Kappa coefficient and relative deviation to select the optimal grid division scheme from the 5 target sea area grid division schemes.

[0009] Further, before using the hierarchical clustering analysis method to perform similarity grouping on the gridded data, the following steps are also included: According to the optimal gridded data corresponding to the optimal grid division scheme, select the plane distribution year-on-year data of the influencing factors in the target sea area in different months for one-way ANOVA; If the difference P is greater than 0.05, it is determined that there is a plane distribution reference paradigm for marine ecological environment influencing factors.

[0010] Further, the specific steps of using the hierarchical clustering analysis method to perform similarity grouping on the gridded data are as follows: In the process of hierarchical clustering analysis, the Euclidean distance measurement method and the ward sum of squared deviations clustering merging strategy are adopted. According to the "bottom-up" agglomeration method, the data is gradually merged through cyclic iteration, and a clustering structure dendrogram is formed; Obtain the similar data in the optimal gridded data corresponding to the optimal grid division scheme according to the clustering structure dendrogram, and group the similar data for similarity grouping.

[0011] Further, the specific steps of obtaining the plane distribution reference paradigm for marine ecological environment influencing factors are as follows: Each group of data after similarity grouping represents a plane distribution pattern. The radial basis function is used to model each group of data to obtain the plane distribution reference paradigm for marine ecological environment influencing factors. The formula is: ; Among them, represents the grouped data after similarity analysis, represents the center, is the coefficient term of the model, is the form of the basis function, is the number of radial basis functions, i represents the index variable, which is used to number different radial basis functions.

[0012] Further, the specific steps of determining the seasonal change reference paradigm for the corresponding marine ecological environment influencing factors according to the distribution pattern of the selected environmental data are as follows: Select the environmental data of the year with a seasonal balance degree of 100%. The environmental data has "U"-shaped and non-"U"-shaped distribution patterns; For the data with a "U"-shaped distribution pattern, use the exponential function for modeling to obtain the seasonal change reference paradigm for marine ecological environment influencing factors. The formula is: ; Among them, is the intercept, which controls the position of the curve on the x-axis; is the coefficient of the first-order term, controlling the slope of the curve; is the coefficient of the second-order term, controlling the degree of curve bending, x is the month; For data with a non-"U"-shaped distribution pattern, obtain the difference between the data with a non-"U"-shaped distribution pattern and the data with a "U"-shaped distribution pattern, and use the sine function to model the difference; ; Among them, is the amplitude, controlling the peak value of the sine wave; is the phase, controlling the horizontal translation of the sine wave; is the vertical translation, controlling the vertical offset of the sine wave.

[0013] Furthermore, the interannual change reference paradigm of the marine ecological environment impact factors has the formula: ; Among them, the constant term represents the average value of the data over the entire time range, used to describe the overall offset of the data; the frequency parameter represents the period of each filtering, the larger it is, the shorter the period, and the higher the periodicity of the curve on the abscissa; represents the harmonic order, takes 1 and 2, corresponding to the first harmonic and the second harmonic respectively; the cosine term and the sine term represent the amplitudes of the cosine wave and the sine wave with different frequencies in the data, x represents the year.

[0014] The embodiment of the present invention provides a standardization system for marine environment data, including: A data acquisition module, configured to use the station coverage rate reflecting the aggregated data excluding outliers in the geographical coverage range of the target sea area, the granularity saturation measuring the density of monitoring stations per unit sea area, and the seasonal balance degree reflecting the distribution balance of monitoring data in different seasons within a year to screen the environmental data of the target sea area and perform gridification; A planar distribution reference paradigm construction module, configured to use the hierarchical clustering analysis method to group the gridified data by similarity, and use the radial basis function to model each group of data to obtain the planar distribution reference paradigm of the marine ecological environment impact factors; A seasonal change reference paradigm construction module, configured to select the environmental data corresponding to the year when the seasonal balance degree reaches the set threshold; determine the corresponding seasonal change reference paradigm of the marine ecological environment impact factors according to the distribution pattern of the selected environmental data; The interannual change benchmark paradigm construction module is used to obtain the seasonally varying benchmarked data for each year using the seasonal change benchmark paradigm of marine ecological environment impact factors; based on the seasonally varying benchmarked data, the interannual change benchmark paradigm of marine ecological environment impact factors is obtained by modeling using the Fourier series equation; The marine environment data standardization module is used to complete the standardization of marine environment data in the target sea area by using the planar distribution benchmark paradigm of marine ecological environment impact factors, the seasonal change benchmark paradigm of marine ecological environment impact factors, and the interannual change benchmark paradigm of marine ecological environment impact factors to fill in the missing planar distribution data, seasonal change data, and interannual change data in the target sea area respectively.

[0015] The embodiments of the present invention provide a method and system for standardizing marine environment data. Compared with the prior art, the beneficial effects are as follows: The environmental data of the target sea area is screened and gridded using the station coverage, grain size saturation, and seasonal balance; based on the gridded data, the planar distribution benchmark paradigm of marine ecological environment impact factors, the seasonal change benchmark paradigm of marine ecological environment impact factors, and the interannual change benchmark paradigm of marine ecological environment impact factors are obtained; through the planar distribution benchmark paradigm of marine ecological environment impact factors, the seasonal change benchmark paradigm of marine ecological environment impact factors, and the interannual change benchmark paradigm of marine ecological environment impact factors, the regularity of the marine ecological environment impact factors in the target sea area in terms of planar distribution, seasonal change, and interannual change can be reflected, that is, the change law on the spatio-temporal scale, so as to fill in the missing sea area data on the spatio-temporal scale and standardize the data of the target sea area. Description of the Drawings

[0016] Figure 1 It is the overall model architecture of a method for standardizing marine environment data provided by the embodiments of the present invention; Figure 2 It is the backpropagation neural network structure of a method for standardizing marine environment data provided by the embodiments of the present invention; Figure 3 It is the flowchart of a method for standardizing marine environment data provided by the embodiments of the present invention; Figure 4 It is the interannual change benchmark paradigm and standardized data of DIN in the target sea area of a method for standardizing marine environment data provided by the embodiments of the present invention; Figure 5 It is the interannual change benchmark paradigm and standardized data of PO 4 -P in the target sea area of a method for standardizing marine environment data provided by the embodiments of the present invention. Detailed Embodiments

[0017] To make the above objects, features, and advantages of the present invention more apparent and understandable, the following provides a detailed description of the specific embodiments of the present invention in conjunction with the accompanying drawings. Many specific details are set forth in the following description to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0018] An embodiment of the present invention provides a method for standardizing marine environmental data, including the following steps: Step 1: Use the aggregated data reflecting the elimination of outliers to screen the environmental data of the target sea area and grid it, including the station coverage rate of the geographical coverage area of the target sea area, the granularity saturation rate measuring the density of monitoring stations per unit sea area, and the seasonal balance degree reflecting the distribution balance of monitoring data in different seasons within a year.

[0019] Step 2: Use the hierarchical clustering analysis method to group the gridded data by similarity, and use the radial basis function to model each group of data to obtain the plane distribution reference paradigm of marine ecological environment impact factors.

[0020] Step 3: Select the environmental data corresponding to the year when the seasonal balance degree reaches the set threshold; determine the corresponding seasonal change reference paradigm of marine ecological environment impact factors according to the distribution pattern of the selected environmental data.

[0021] Step 4: Use the seasonal change reference paradigm of marine ecological environment impact factors to obtain the seasonal change standardized data for each year; according to the seasonal change standardized data, use the Fourier series equation to model and obtain the interannual change reference paradigm of marine ecological environment impact factors.

[0022] Step 5: Use the plane distribution reference paradigm of marine ecological environment impact factors, the seasonal change reference paradigm of marine ecological environment impact factors, and the interannual change reference paradigm of marine ecological environment impact factors to supplement the missing plane distribution data, seasonal change data, and interannual change data of the target sea area respectively, and complete the standardization of marine environmental data in the target sea area.

[0023] I. The overall model architecture of the present invention is as Figure 1 shown, consisting of an original data processing module, an aggregated data processing module, and a reference data processing module.

[0024] 1. Original data processing module.

[0025] Due to the diversity of monitoring platforms and means, marine environmental data are scattered among multiple departments and not processed according to unified technical specifications, actually forming multiple heterogeneous, autonomous, and distributed data sources, resulting in difficulties in data sharing and common use. The present invention constructs a marine multi-source heterogeneous data fusion framework aiming to overcome the limitations of the spatial sparsity of a single data source. First, the target sea area is subjected to spatial grid processing. Then, according to the nearest neighbor matching principle, the geographical locations of multi-source heterogeneous data are aligned to the standard grid, and a standard grid independent variable matrix is constructed. Finally, at the same resolution grid scale, linear regression data fusion is used for the independent variable matrices from different sources to effectively integrate multi-source heterogeneous data, thereby obtaining more reliable influencing factor data on an ideal spatio-temporal scale. The expression is as follows: (1).

[0026] In the formula, represents the data sets from different sources, represents the independent variable matrix under the standard grid, represents the resolution of the sea area grid, represents the regression coefficient, represents the error term.

[0027] 2. Data collection and processing module.

[0028] During the data collection process, problems such as human operation errors, instrument precision limitations, or communication transmission errors are inevitable, which may lead to outliers in the monitoring data. The present invention uses Chauvenet's test for efficient identification of outliers. This method takes into account the sample size and can give an adaptive discrimination coefficient that varies with the sample size, making it particularly suitable for cases where the spatial sample distribution is uneven or the sample size is small. In the process of collecting and processing data, it is necessary to implement the annotation of data quality. Therefore, in the process of calculating data quality evaluation indicators, the present invention comprehensively considers multiple key factors of marine environmental data in the time and space dimensions. Station Coverage (SC), Granularity Saturation (GS), and Seasonal Balance (SB) are constructed as three core criteria for measuring data quality. These indicators jointly reflect the representativeness and integrity of the original collected data in the time and space scales, providing a comprehensive evaluation framework for data quality.

[0029] First, the station coverage reflects the geographical coverage of the aggregated data after outlier removal in the target sea area. Specifically, spatial interpolation of the data is performed by Kriging method, and the ratio of the interpolated data coverage area to the total area of the entire target sea area is calculated as the quantitative index of the station coverage. The expression is as follows: (2).

[0030] In the formula, is the interpolated area, is the total area of the target sea area.

[0031] Secondly, the grain saturation measures the density of monitoring stations per unit sea area. This index is determined by comparing the number of stations after outlier processing with the possible maximum number of stations, so as to reflect the spatial distribution uniformity of the data. The expression is as follows: (3).

[0032] In the formula, is the number of monitoring stations, are the sub-sea areas or bays of the target sea area, is the area of a certain sub-sea area or bay.

[0033] Finally, the seasonal balance degree focuses on the distribution balance of monitoring data in different seasons of a year. This index calculates the monitoring frequency of each season in the four seasons of the whole year and compares this frequency with the total number of the whole year to obtain the balance degree of the four-season monitoring. The expression is as follows: (4).

[0034] In the formula, is the month to which the data belongs, is the season determination function, is the number of different seasons involved in the given month.

[0035] In order to achieve a comprehensive assessment of data quality, the present invention uses a standard classification method of equal intervals and quantiles to divide the quality grade range of the original aggregated data. The data quality is divided into 5 grades, namely high, medium-high, medium, medium-low, and low, to more precisely describe different levels of data quality, as shown in Table 1. This systematic data quality management method can ensure the scientific and reasonable distinction and use of data with different quality grades. Among them, the environmental data of high quality grade is used as the research basis to seek the plane distribution benchmark paradigm of marine ecological environment impact factors, the seasonal change benchmark paradigm of marine ecological environment impact factors, and the interannual change benchmark paradigm of marine ecological environment impact factors. While the data below the high quality grade (medium-high, medium, medium-low, low) need to be filled in space, seasonally, and interannually based on the spatio-temporal distribution benchmark paradigm due to their poor reliability or missing data. In addition, the comprehensive quality grade of the original aggregated data depends on the lowest standard value among 3 sub-indicators, thus ensuring the conservativeness and reliability of the quality control results and avoiding overestimating the comprehensive data quality due to abnormally high values of a single indicator.

[0036] Table 1 Classification of the comprehensive quality grade of the original aggregated data.

[0037]

[0038] 3. Benchmark data processing module.

[0039] The spatio-temporal distribution benchmark paradigm of marine ecological environment impact factors is the basis for filling in the spatio-temporal missing data in the target sea area. Therefore, it is first necessary to judge the existence of the benchmark paradigm of the impact factor. After the present invention completes the assessment of the comprehensive data quality grade, it selects high-quality data for spatial grid processing. According to different spatial resolutions, 5 grid division schemes for the target sea area are given (0.25°×0.25°, 0.2°×0.2°, 0.15°×0.15°, 0.1°×0.1°, 0.05°×0.05°), and the optimal gridded data is determined according to the Kappa coefficient and relative deviation. Then, through the differential analysis of the optimal gridded data, the existence of the plane distribution benchmark paradigm of marine ecological environment impact factors is judged. According to the actual application requirements, the differential analysis method used is one-way analysis of variance (ANOVA). Based on this method, the differential analysis is carried out on the year-on-year data of the plane distribution of the impact factors in the target sea area in typical months (March, May, August, October) to preliminarily judge whether there is a certain law in the plane distribution of this factor. If the difference is too large, there is no benchmark paradigm. Specifically, if the difference P is greater than 0.05, there is a plane distribution benchmark paradigm of marine ecological environment impact factors; if the difference P is less than 0.05, there is no plane distribution benchmark paradigm of marine ecological environment impact factors, and there is no need to standardize the marine environmental data in the target sea area.

[0040] Among them, determining the optimal gridded data according to the kappa coefficient and relative deviation includes: obtaining the environmental data of the target sea area and the kappa coefficients and relative deviations of various target sea area grid division schemes, and according to the gap between the kappa coefficient and the relative deviation, selecting the kappa coefficient and relative deviation with the largest gap, and determining the corresponding target sea area grid division scheme as the optimal grid division scheme.

[0041] On the premise of the existence of the plane distribution reference paradigm, it can be proved that the influencing factors of the target sea area may have internal laws in typical months. Therefore, it is necessary to further explore the similarity of data distribution in different periods. The present invention uses hierarchical clustering analysis to group the year-on-year data of plane distribution within typical months. In the process of hierarchical clustering analysis, the Euclidean distance measurement method and the ward sum of squared deviations clustering merger strategy are adopted, and in accordance with the "bottom-up" agglomerative method, through cyclic iteration, the data of adjacent years with higher similarity are gradually merged, and a clustering structure pedigree diagram is formed. The final result of hierarchical clustering analysis will reveal which years have similar data distribution laws.

[0042] Each group after similarity analysis can represent a plane distribution pattern. The present invention uses a radial basis function to model and express each group of data. The expression is as follows: (5).

[0043] In the formula, represents the grouped data after similarity analysis, represents the center, is a norm operation. The present invention adopts the Euclidean norm, that is, the Euclidean distance between two points is taken. To simplify the multi-parameter input, the geographical location of the minimum value of the influencing factors of the target sea area is used as the reference coordinate, and the spatial parameters (longitude and latitude) are converted into distance parameters to construct a functional relationship with the distance as the independent variable and the influencing factor concentration value as the dependent variable. is the coefficient term of the model, is the form of the basis function, is the number of radial basis functions, i represents the index variable, which is used to number different radial basis functions.

[0044] Meanwhile, in order to test the uncertainty of the radial basis function in expressing the plane distribution reference paradigm, the present invention uses the Monte Carlo method to randomly iterate and sample the input data, and the number of sampling times is greater than 500 times, and then the relative standard deviation (RSD) is used to characterize the uncertainty caused by the model parameters.

[0045] After the benchmarking process of the planar distribution data is completed, select the annual data with a seasonal balance degree reaching 100% to establish the benchmark paradigm for the seasonal variation of marine ecological environment impact factors. Over the years, there have been two main distribution patterns for high-quality seasonal variation data: the "U" type and the non-"U" type. For the data showing a "U" type distribution, the present invention uses an exponential function for modeling and expression. The expression is as follows: (6).

[0046] In the formula, is the intercept, controlling the position of the curve on the x-axis; is the coefficient of the first-order term, controlling the slope of the curve; is the coefficient of the second-order term, controlling the degree of curve bending, x is the independent variable, representing different months.

[0047] For the data showing a non-"U" type distribution, the present invention first calculates the difference between it and the typical "U" type distribution data. Subsequently, a sine function is used to model the difference to quantify and express the potential impact of external conditions such as rainfall on the seasonal variation of the impact factors. Finally, the benchmark paradigm for the seasonal variation of non-"U" type distribution data can be expressed as the difference between the typical "U" type distribution data and the seasonal variation adjustment value The expression is as follows: (7).

[0048] In the formula, is the amplitude, controlling the peak value of the sine wave; is the phase, controlling the horizontal translation of the sine wave; is the vertical translation, controlling the vertical offset of the sine wave, x is the independent variable, representing different months.

[0049] Finally, the present invention integrates the benchmark paradigm for the seasonal variation of each year to obtain the annual average concentration value of the impact factors, and uses the Fourier series equation to model and express the benchmark paradigm for the interannual variation of marine ecological environment impact factors. The expression is as follows: (8).

[0050] In the formula, the constant term represents the average value of the data over the entire time range, which can be used to describe the overall offset of the data; the frequency parameter represents the period of each filter, the larger it is, the shorter the period, and the higher the periodicity of the curve on the abscissa; represents the harmonic order, where takes 1 and 2, corresponding to the first harmonic and the second harmonic respectively; the cosine term and sine terms represent the amplitudes of cosine and sine waves with different frequencies in the data, x where \(x\) is the independent variable representing different years.

[0051] 4. Filling in the missing spatio-temporal data.

[0052] After passing through the raw data processing module, the aggregated data processing module, and the reference data processing module, a spatio-temporal distribution reference paradigm of the influencing factors in the target sea area can be obtained. Based on this reference paradigm, the spatio-temporal data below high quality can be filled in. Among them, when filling in the planar distribution data below high quality, the key issue lies in determining the applicable planar distribution reference paradigm. Therefore, the present invention adopts a data-driven paradigm determination method, that is, using a backpropagation neural network to optimize the process of filling in the planar distribution data. On the basis of the optimal grid division, we construct a parallel BP neural network architecture, as shown in Figure 2 shown. Using high-quality planar distribution data to iteratively train the BPNN so that it has the ability to match the most suitable planar distribution reference paradigm for new incoming data. According to the selected planar distribution reference paradigm, the standardization of the spatial distribution data to be filled in is achieved through spatial fitting. And the reliability of the filled-in data is tested using the kappa coefficient and the relative standard deviation. When filling in the seasonal change data below high quality, based on the existing monthly data and combined with the minimum deviation method, first determine the applicable seasonal change reference paradigm. Then, according to the selected seasonal change reference paradigm, the standardization of the seasonal change data to be filled in is achieved through time series fitting. And the reliability of the seasonal filling result is tested using the relative deviation. Finally, after the planar distribution and seasonal change data are benchmarked, the missing inter-annual change data can be directly filled in based on the fitting curve of the inter-annual change reference paradigm.

[0053] When filling in the spatial distribution data, first determine the applicable planar distribution reference paradigm for the data to be filled in based on the backpropagation neural network. The determined planar distribution reference paradigm can reflect the distribution law of the influencing factors in different spatial positions within the target sea area. Then, according to the selected planar distribution reference paradigm, the standardization of the spatial data to be filled in is achieved through spatial fitting. When filling in the seasonal change data, based on the existing monthly data and combined with the minimum deviation method, first determine the applicable seasonal change reference paradigm. The determined seasonal change reference paradigm can reflect the change pattern of the influencing factors in different seasons within the target sea area. Then, according to the selected seasonal change reference paradigm, the standardization of the seasonal data to be filled in is achieved through time series fitting. When filling in the inter-annual change data, the inter-annual change reference paradigm can reflect the long-term change trend of the influencing factors in different years within the target sea area. Therefore, the standardization of the inter-annual data to be filled in can be directly achieved based on the fitting curve of the inter-annual change reference paradigm.

[0054] II. The key part of the marine environment data standardization method proposed by the present invention lies in: 1. The present invention conducts fusion and outlier detection on multi-source heterogeneous data in the marine environment, comprehensively considers multiple key factors of the monitoring data in the target sea area in terms of time and space dimensions, and constructs three data quality control indicators: station coverage, grain saturation, and seasonal balance. Based on high-quality data, statistical methods are used to summarize the regularities shown by the influencing factors in the target sea area in terms of planar distribution, seasonal variation, and interannual variation, and mathematical equations are used to model and express them respectively, constructing a spatio-temporal distribution benchmark paradigm for the influencing factors in the target sea area. This can not only explain the main variation laws of the influencing factors at the marine spatio-temporal scale, but also lay a foundation for filling in the missing spatio-temporal data below high quality.

[0055] 2. During the process of filling in the planar distribution data of factors below high quality in the target sea area, the standardization process of the missing planar distribution data to be filled is optimized by constructing a backpropagation neural network. Using the constructed data quality control indicators, the high-quality data of the influencing factors are filtered and screened according to the minimum value principle. Then, the high-quality planar distribution data are used as the training set to iteratively train the backpropagation neural network, enabling it to have the ability to match the most suitable planar distribution benchmark paradigm for new incoming data. Furthermore, based on the most suitable planar distribution benchmark paradigm, the missing planar distribution data below high quality are filled by spatial fitting. For the missing seasonal variation data below high quality, based on the existing monthly data and combined with the minimum deviation method, the missing seasonal variation data to be filled are filled by time series fitting. After the benchmarking of the planar distribution data and seasonal variation data of the influencing factors is completed, the missing interannual variation data can be directly filled based on the fitting curve of the interannual variation benchmark paradigm. This progressive spatio-temporal missing data standardization method can ultimately effectively improve and unify the spatio-temporal granularity of the influencing factors in the target sea area in terms of planar distribution, seasonal variation, and interannual variation.

[0056] III. The advantages of the present invention are: Marine environmental data usually has certain variation laws in the time and space dimensions. However, due to the limitations of monitoring conditions, relevant data often shows problems such as limited spatial coverage and monthly frequencies, serious data gaps in some sea areas, and inconsistent spatio-temporal granularities of existing monitoring data. Therefore, standardizing marine environmental data is crucial for aspects such as simulating, evaluating, and formulating prevention and control plans for the eutrophication status in coastal waters. The present invention comprehensively considers multiple key factors of marine environmental data in the time and space dimensions, constructs three data quality control sub-indicators: station coverage, granularity saturation, and seasonal balance, and provides a comprehensive evaluation framework for the data quality of elements in the target sea area. Then, based on high-quality data, a spatio-temporal distribution benchmark paradigm of elements in the target sea area is summarized. This benchmark paradigm provides a clear mathematical equation description of the spatio-temporal distribution characteristics of influencing elements in the target sea area and has high reliability. In addition, when missing data in the planar distribution or time-series changes of influencing elements in the target sea area are input, based on the corresponding benchmark paradigm and combined with a backpropagation neural network, the supplementation of marine environmental data can be effectively completed, the spatio-temporal coverage range of influencing elements in the target sea area can be improved, and the standardization of marine environmental data in the target sea area can be achieved. As Figure 3 shown in the flowchart of the present invention.

[0057] An embodiment of the present invention provides a system for standardizing marine environmental data, including: A data acquisition module, configured to use the station coverage of the collection data reflecting the elimination of outliers in the geographical coverage range of the target sea area, the granularity saturation measuring the density of monitoring stations per unit sea area, and the seasonal balance reflecting the distribution balance of monitoring data in different seasons within a year to screen the environmental data of the target sea area and perform gridification.

[0058] A planar distribution benchmark paradigm construction module, configured to use hierarchical clustering analysis to group the gridified data by similarity and use a radial basis function to model each group of data to obtain a planar distribution benchmark paradigm of marine ecological environment influencing elements.

[0059] A seasonal change benchmark paradigm construction module, configured to select the environmental data corresponding to the years when the seasonal balance reaches a set threshold; determine the corresponding seasonal change benchmark paradigm of marine ecological environment influencing elements according to the distribution pattern of the selected environmental data.

[0060] An inter-annual change benchmark paradigm construction module, configured to use the seasonal change benchmark paradigm of marine ecological environment influencing elements to obtain the seasonal change benchmarked data for each year; use the Fourier series equation to model according to the seasonal change benchmarked data to obtain an inter-annual change benchmark paradigm of marine ecological environment influencing elements.

[0061] The marine environment data standardization module is used to complete the standardization of marine environment data in the target sea area by filling in the missing planar distribution data, seasonal change data, and interannual change data in the target sea area using the planar distribution benchmark paradigm of marine ecological environment impact factors, the seasonal change benchmark paradigm of marine ecological environment impact factors, and the interannual change benchmark paradigm of marine ecological environment impact factors respectively.

[0062] A specific embodiment is as follows: This embodiment discloses a method for standardizing marine environment data, and the specific steps are as follows: S1. Quality assessment of ecological environment data in a certain sea area.

[0063] The monitoring stations of marine environment data are usually unevenly distributed in space. Most monitoring stations are distributed in the coastal waters according to administrative divisions, while there are much fewer monitoring stations in the central sea area. The calculation results using formulas (2)-(4) show that there is a certain similarity among the quality control sub-indicators of marine environment data over the years, all showing a trend of fluctuating in the early stage (from the early 1980s to the early 1990s), rising in the middle stage (from the mid-1990s to the mid-2000s), and stabilizing in the late stage (from the late 2000s to the early 2020s). The high-quality data of impact factors are filtered and screened according to the minimum value principle. Among them, the proportion of data below low quality is as high as 51.66% (DIN) and 57.82% (PO 4 -P), while the proportion of data above medium and high quality is only 25.17% (DIN) and 23.13% (PO 4 -P). Therefore, the proportion of low-quality marine environment data is generally high, and it is urgent to standardize the marine environment data.

[0064] S2. Standardization processing of ecological environment data in a certain sea area.

[0065] Select the high-quality marine environment planar distribution data for the four seasons of spring, summer, autumn, and winter, and sequentially apply the difference analysis and similarity analysis methods to judge and analyze the planar distribution law of the elements in the target sea area. According to the difference analysis results, the P values of the planar distribution data for the four seasons are all greater than 0.05, indicating low differences, which shows that the spatial distribution of marine environment data has seasonal stability and may follow a certain planar distribution pattern. According to the similarity analysis results, the decadal spatial distribution of impact factors has obvious clustering characteristics. Among them, the early data in the 1980s have similar planar distribution laws, the mid-term data in the 1990s have similar planar distribution laws, and the recent data in the early 21st century have similar planar distribution laws. Based on this similarity grouping, use formula (5) to construct the planar distribution benchmark paradigm of impact factors, and complete the standardization of planar distribution data in combination with the backpropagation neural network. DIN and PO4 - The relative standard deviations of the -P standardized data are 35.62% and 58.44% respectively, and the kappa coefficients are 0.75 and 0.83 respectively, indicating high reliability. Then, the seasonal change benchmark paradigm of the influencing factors is constructed using formulas (6)-(7), and the standardization of the seasonal change data is completed based on the self-small deviation method. DIN and PO 4 - The relative deviations of the -P standardized data are 24.39% and 18.81% respectively. Finally, the inter-annual change benchmark paradigm of the influencing factors is constructed using formula (8), and the standardization of the inter-annual change data is completed. The results are as Figure 4 and Figure 5 shown.

[0066] The above-described embodiments merely represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all fall within the protection scope of the present invention. Therefore, the protection scope of the present invention patent shall be subject to the appended claims.

Claims

1. A method for standardizing marine environmental data, characterized in that: The following steps are involved: The environmental data of the target sea area are screened and gridded using the station coverage of the geographical coverage of the target sea area, which reflects the collected data excluding outliers, the granularity saturation, which measures the density of monitoring stations within a unit sea area, and the seasonal balance, which reflects the distribution balance of monitoring data in different seasons within a year. The hierarchical cluster analysis method was used to group the grid data by similarity, and the radial basis function was used to model each group of data to obtain the plane distribution benchmark paradigm of the factors affecting the marine ecological environment. Select the environmental data corresponding to the year when the seasonal balance reaches the set threshold; determine the corresponding benchmark paradigm of seasonal changes of marine ecological environment influencing factors based on the distribution pattern of the selected environmental data; Use the seasonal variation benchmark paradigm of marine ecological environment influencing factors to obtain the annual seasonal variation benchmark data; Based on the seasonal variation benchmark data, use Fourier series equation modeling to obtain the interannual variation benchmark paradigm of marine ecological environment influencing factors; The benchmark paradigm for the plane distribution of factors affecting the marine ecological environment, the benchmark paradigm for seasonal changes of factors affecting the marine ecological environment and the benchmark paradigm for interannual changes of factors affecting the marine ecological environment are used respectively to fill the missing plane distribution data, seasonal change data and interannual change data of the target sea area, and complete the standardization of marine environmental data in the target sea area.

2. A method for standardizing marine environmental data according to claim 1, characterized in that: The specific steps of screening the environmental data of the target sea area include: Based on the environmental data of the target sea area, station coverage, particle size saturation and seasonal balance are used as the core standards for measuring the quality of environmental data; Based on the environmental data screened by each core standard, the environmental data are classified into different levels of data quality, including high, medium-high, medium, medium-low and low, using the standard classification method of equal intervals and quantiles; High-quality environmental data will be used as the environmental data for the screened target sea areas.

3. A method for standardizing marine environmental data as claimed in claim 2, characterized in that: The meshing process specifically includes the following steps: Five target sea area grid division schemes are designed according to different spatial resolutions using high-quality environmental data, including: 0.25°×0.25°, 0.2°×0.2°, 0.15°×0.15°, 0.1°×0.1°, and 0.05°×0.05°; The Kappa coefficient and relative deviation are used to select the optimal grid division scheme from five target sea area grid division schemes.

4. A method for standardizing marine environmental data as claimed in claim 3, characterized in that: Before the gridded data are grouped by similarity using the hierarchical clustering analysis method, the method further includes: According to the optimal gridding data corresponding to the optimal grid division scheme, the year-on-year data of the plane distribution of the influencing factors of the target sea area in different months are selected for single-factor variance analysis; If the difference P is greater than 0.05, it is determined that there is a benchmark paradigm for the plane distribution of factors affecting the marine ecological environment.

5. A method for standardizing marine environmental data as claimed in claim 3, characterized in that: The hierarchical clustering analysis method is used to group the grid data by similarity, and the specific steps include: In the process of hierarchical cluster analysis, the Euclidean distance measurement method and the Ward deviation square sum cluster merging strategy are used. According to the "bottom-up" agglomeration method, the data are gradually merged through cyclic iterations to form a cluster structure pedigree diagram; According to the clustering structure pedigree diagram, similar data in the optimal gridded data corresponding to the optimal grid division scheme is obtained, and the similar data is grouped according to similarity.

6. A method for standardizing marine environmental data according to claim 1, characterized in that: The specific steps of obtaining the planar distribution benchmark paradigm of marine ecological environment impact factors include: Each group of data in the similarity grouping represents a plane distribution pattern. The radial basis function is used to model each group of data to obtain the plane distribution benchmark paradigm of the factors affecting the marine ecological environment. The formula is: ； in, Represents the grouped data after similarity analysis, Indicates the center, is the coefficient term of the model, is the basis function form, is the number of radial basis functions, i Represents the index variable, which is used to number different radial basis functions.

7. A method for standardizing marine environmental data as claimed in claim 1, characterized in that: The method of determining the corresponding seasonal variation benchmark paradigm of the marine ecological environment influencing factors according to the distribution pattern of the selected environmental data comprises the following specific steps: Select environmental data of the year with a seasonal balance of 100%, and the environmental data has "U"-shaped and non-"U"-shaped distribution patterns; For the data of "U" type distribution pattern, the exponential function is used to model and obtain the benchmark paradigm of seasonal changes of factors affecting the marine ecological environment. The formula is: ； in, is the intercept, which controls the position of the curve on the x-axis; is the coefficient of the first-order term, which controls the slope of the curve; is the quadratic term coefficient, which controls the curvature of the curve. x for the month; For data with non-U-shaped distribution patterns, the difference between the data with non-U-shaped distribution patterns and the data with U-shaped distribution patterns is obtained, and the difference is modeled using a sine function; ； in, is the amplitude, which controls the peak value of the sine wave; is the phase, which controls the horizontal translation of the sine wave; It is the vertical translation, which controls the vertical displacement of the sine wave.

8. A method for standardizing marine environmental data as claimed in claim 1, characterized in that: The interannual variation benchmark model of the factors affecting the marine ecological environment is as follows: ； Among them, the constant term Represents the average value of the data in the entire time range, which is used to describe the overall deviation of the data; frequency parameter represents the period of each filter, The larger the period, the shorter the period, and the higher the periodicity of the curve on the abscissa; Indicates the harmonic order, Take 1 and 2, corresponding to the first and second harmonics respectively; the cosine term and the sine term Represents the amplitudes of cosine and sine waves of different frequencies in the data, x Indicates the year.

9. A standardization system for marine environmental data, characterized in that: include: The data acquisition module is used to screen and grid the environmental data of the target sea area using the station coverage of the geographical coverage of the target sea area, which reflects the collected data with outliers removed, the granularity saturation, which measures the density of monitoring stations within a unit sea area, and the seasonal balance, which reflects the distribution balance of monitoring data in different seasons within a year; The plane distribution benchmark paradigm construction module is used to group grid data by similarity using the hierarchical clustering analysis method and model each group of data using the radial basis function to obtain the plane distribution benchmark paradigm of marine ecological environment impact factors; The seasonal change benchmark paradigm construction module is used to select the environmental data corresponding to the year when the seasonal balance reaches the set threshold; according to the distribution pattern of the selected environmental data, the corresponding seasonal change benchmark paradigm of the marine ecological environment influencing factors is determined; The interannual variation benchmark paradigm construction module is used to obtain the annual seasonal variation benchmark data using the seasonal variation benchmark paradigm of marine ecological environment influencing factors; based on the seasonal variation benchmark data, the interannual variation benchmark paradigm of marine ecological environment influencing factors is obtained using Fourier series equation modeling; The marine environment data standardization module is used to use the marine ecological environment influencing factors plane distribution benchmark paradigm, marine ecological environment influencing factors seasonal change benchmark paradigm and marine ecological environment influencing factors interannual change benchmark paradigm to fill in the missing plane distribution data, seasonal change data and interannual change data of the target sea area, and complete the standardization of marine environment data in the target sea area.

Images