Marine chlorophyll spatial-temporal variation and trend prediction method and system

Abstract

The invention relates to the technical field of data processing, and provides a marine chlorophyll a spatial-temporal variation and trend prediction method and system, and the method comprises the following steps: obtaining wind speed and sea surface temperature data for a to-be-predicted region; for the wind speed data, combining the strength of Ekman transportation and the angle of a coastline to quantify coastwise upwelling, and obtaining an upwelling index calculated based on the wind speed; using a DINEOF method to reconstruct missing data in the sea surface temperature data, and performing interpolation processing on the sea surface temperature data; and predicting the upwelling index and the sea surface temperature data subjected to interpolation processing by adopting a Chl-a long-term trend prediction model which is obtained based on correlation analysis and is related to the sea surface temperature and the wind speed, so as to obtain the concentration and the change trend of the sea chlorophyll a. According to the method, the precision of wind speed data quantification processing is improved, and meanwhile, missing data are processed, so that the accuracy of the concentration and the change trend of marine chlorophyll a can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field related to data processing, and in particular to a method and system for predicting the spatiotemporal variation and trend of marine chlorophyll a. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Chlorophyll a concentration (abbreviated as Chl-a) is a key biological indicator of phytoplankton biomass and ocean productivity, and is essential for monitoring the health of marine ecosystems. Studying the temporal and spatial variation and trends of Chl-a values ​​in the ocean is of great significance for predicting the health of marine ecosystems, achieving sustainable fishery management, and early warning of harmful algal blooms.

[0004] Existing Chl-a value prediction methods mainly rely on satellite remote sensing data and model prediction methods to evaluate the biomass of phytoplankton and the productivity of marine ecosystems by obtaining the Chl-a value of the ocean area. These methods usually use indicators such as sea surface temperature (abbreviated as SST) and monsoon wind force, combined with satellite image data to predict Chl-a values. However, existing methods have problems in the following aspects: (1) Limited data integrity: For example, satellite data in some areas of the ocean are easily affected by adverse conditions such as sunlight reflection and cloud cover, resulting in incomplete or missing data. These data missing problems directly affect the prediction accuracy of Chl-a values, especially during periods of large fluctuations in environmental factors such as wind speed and upwelling, the accuracy of the prediction results decreases; (2) Although the existing prediction methods take monsoon wind force into account, they lack quantitative processing of wind speed and fail to accurately reflect the impact of wind speed on Chl-a values. This results in insufficient response of the prediction model to fluctuations in phytoplankton biomass, further limiting the accuracy of Chl-a value prediction. Summary of the invention

[0005] In order to solve the above problems, the present invention proposes a method and system for predicting the spatiotemporal variation and trend of marine chlorophyll a, which improves the accuracy of quantification processing of wind speed data. The quantified data can accurately represent the wind stress value. At the same time, the missing data is processed, thereby improving the accuracy of marine chlorophyll a concentration and changing trend.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: One or more embodiments provide a method for predicting the spatiotemporal variation and trend of marine chlorophyll a, comprising the following steps: Obtain wind speed and sea surface temperature data for the area to be predicted; Based on the wind speed data, the intensity of Ekman transport and the angle of the coastline are combined to quantify the coastal upwelling, and the upwelling index based on wind speed calculation is obtained; Use the DINEOF method to reconstruct missing data in the sea surface temperature data and interpolate the sea surface temperature data; The upwelling index and interpolated sea surface temperature data are used to predict the Chl-a long-term trend prediction model related to sea surface temperature and wind speed based on correlation analysis to obtain the marine chlorophyll a concentration and its changing trend.

[0007] One or more embodiments provide a system for predicting the spatiotemporal variation and trend of marine chlorophyll a, including: Data acquisition module: configured to acquire wind speed and sea surface temperature data for the area to be predicted; Quantification module: configured to quantify coastal upwelling based on wind speed data, the strength of Ekman transport and the angle of the coastline, and obtain an upwelling index based on wind speed calculation; Interpolation processing module: configured to use the DINEOF method to reconstruct the missing data in the sea surface temperature data and perform interpolation processing on the sea surface temperature data; Prediction module: It is configured to use the upwelling index and the interpolated sea surface temperature data to predict the Chl-a long-term trend prediction model related to sea surface temperature and wind speed based on correlation analysis to obtain the marine chlorophyll a concentration and its changing trend.

[0008] One or more embodiments provide a marine chlorophyll a spatiotemporal variation and trend prediction system, including a data acquisition terminal and a processor; Data collection devices, including a satellite system and wind speed sensors installed at each station, are used to collect sea surface temperature data and wind speed data respectively; The processor is configured to execute the above-mentioned method for predicting the spatiotemporal variation and trend of marine chlorophyll a.

[0009] Compared with the prior art, the present invention has the following beneficial effects: The present invention effectively solves the impact of missing data on the prediction of chlorophyll a concentration by quantifying wind speed data and interpolating and supplementing sea surface temperature data. Compared with the prior art, this method has achieved significant improvements in the following aspects: first, the DINEOF method is used to reconstruct sea surface temperature data, so that missing data is supplemented and the integrity of the data is improved; second, the Ekman transport model is used to accurately quantify wind speed, which improves the accuracy of the impact of wind speed on upwelling, thereby improving the applicability of the prediction model in different seasons and environments, and significantly enhancing the response ability to phytoplankton biomass fluctuations; finally, through the correlation analysis based on sea surface temperature and wind speed, the accurate prediction of the trend of chlorophyll a concentration changes is achieved, which helps to improve the accuracy and reliability of chlorophyll a concentration prediction.

[0010] The advantages of the present invention and additional advantages will be described in detail in the following specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their description are used to explain the present invention but do not constitute a limitation of the present invention.

[0012] Figure 1 is a flow chart of the prediction method of embodiment 1 of the present invention; Figure 2 is a comparison diagram of the reconstruction after logarithmic transformation and the original Chl-a data in the model construction example of Example 1 of the present invention; Figure 3 is a comparison diagram of reconstructed and original SST data in the model construction example of Example 1 of the present invention; Figure 4 is a monthly variation graph of MODIS Chl-a reconstructed at different longitude positions in the model construction example of Example 1 of the present invention; Figure 5 is a monthly variation graph of MODIS Chl-a reconstructed at different dimensional positions in the model construction example of Example 1 of the present invention; Figure 6 is a comparison chart of the monthly average time series of coastal upwelling index from 2001 to 2019 at different stations in the model construction example of Example 1 of the present invention; Figure 7 is a comparison chart of the changing trend lines of the time series of Chl-a_A and SST_A at four stations from January 2001 to December 2019 in the model construction example of Example 1 of the present invention; Figure 8This is a comparison chart of the changing trend lines of the time series of Chl-a_A and Wind_A at four stations from January 2001 to December 2019 in the model construction example of Example 1 of the present invention. DETAILED DESCRIPTION

[0013] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0014] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0015] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof. It should be noted that, in the absence of conflict, the various embodiments of the present invention and the features in the embodiments can be combined with each other. The embodiments will be described in detail below in conjunction with the accompanying drawings.

[0016] Example 1 In the technical solutions disclosed in one or more embodiments, Figures 1 to 8 As shown, a method for predicting the spatiotemporal variation and trend of marine chlorophyll a comprises the following steps: Step 1: Obtain wind speed and sea surface temperature data for the area to be predicted; Step 2: Based on the wind speed data, the intensity of Ekman transport and the angle of the coastline are combined to quantify the coastal upwelling and obtain an upwelling index based on wind speed calculation; Step 3: Use the DINEOF method to reconstruct the missing data in the sea surface temperature data and interpolate the sea surface temperature data; Step 4: The upwelling index and the interpolated sea surface temperature data are predicted using the Chl-a long-term trend prediction model related to sea surface temperature and wind speed obtained based on correlation analysis to obtain the ocean chlorophyll a concentration and its changing trend.

[0017] In this embodiment, SST data is the abbreviation of sea surface temperature data, and sea surface temperature is abbreviated as SST.

[0018] In this embodiment, wind speed and sea surface temperature data are used to predict the concentration of marine chlorophyll a. First, the wind speed is quantified by the Ekman transport model to extract the components reflecting the upwelling. This method combines the direction and intensity of the wind, as well as its relationship with the coastline to quantify the contribution of wind speed to coastal upwelling, thereby more accurately reflecting the impact of upwelling on chlorophyll a concentration. At the same time, the DINEOF method is used to interpolate the missing data, so that the data integrity is improved and the impact of incomplete satellite data on model accuracy is reduced. On this basis, the prediction model obtained by the SST and wind speed correlation analysis is used to process the pre-processed data, so that the prediction results can more accurately reflect the spatiotemporal variation and trend changes of marine chlorophyll a.

[0019] Due to the influence of multiple factors such as wind speed and sea surface temperature, the chlorophyll a concentration in different regions and time periods will show significant differences. The method of this embodiment can predict the spatial distribution differences of chlorophyll a concentration in different sea areas and the changes in chlorophyll a concentration in different time periods, thereby realizing the prediction of the spatiotemporal variation and trend of marine chlorophyll a.

[0020] In step 1, the wind speed data of the predicted area within a set time period can be obtained through the weather forecast center; the sea surface temperature (SST) data in the same time period can be obtained through satellites; and the trend of Chl-a value changes in future time periods can be predicted.

[0021] In step 2, the method for calculating the upwelling index based on wind speed includes the following steps: Step 21: dynamically adjust the drag coefficient according to the acquired wind speed data, and calculate the Ekman transport component for each month within the set period, where: (1) (2) (3) in, u Indicates the wind speed of the east wind or the west wind. Set the east wind to a positive value and the west wind to a negative value. v Indicates the wind speed of the south wind or north wind, setting the north wind as a positive value and the south wind as a negative value; Indicates air density, the value is 1.22kgm-3; represents the density of water, which is 1025kgm-3. In addition, c is the resistance coefficient, f is the Coriolis parameter.

[0022] The calculated Ekman transport component ET x and E.T. yIt is used to generate monthly graphs of Ekman transport. During the model construction process, the trend change distribution graph of Chl-a values ​​is based on monthly data, so the wind speed data here is calculated based on monthly data.

[0023] Step 22, combining the relationship between the intensity and direction of Ekman transport and the coastline to quantify coastal upwelling, specifically: calculating the upwelling index according to the obtained Ekman transport component and the angle between the coastline of the predicted area and the equator; taking the upwelling index as an influencing factor of the spatiotemporal variation of Chl-a, the larger the upwelling index, the larger the Chl-a value; The upwelling index formula is as follows: (4) in, It is the angle between the coastline and the equator.

[0024] Upwelling Index Positive values ​​correspond to favorable upwelling conditions, while negative values ​​correspond to unfavorable upwelling conditions; In the above scheme, the value of the resistance coefficient c parameter is different from that in the prior art. In the prior art, the resistance coefficient c is determined by a fixed empirical value method, such as: some studies use c=1.3×10 -3 , while other studies have used higher c values, c = 2.6 × 10 -3 The fixed empirical value will cause the calculated wind stress value to be biased, too large or too small, which will lead to the deviation of the upwelling index calculated based on the wind speed, and the prediction results will be biased when making subsequent predictions; In this embodiment, a fixed constant is not used as the drag coefficient c. This embodiment dynamically adjusts the drag coefficient c based on the actual value of the detected wind speed, as shown in formula (4). The drag coefficient calculated by this formula is related to the magnitude of the wind speed monitored in real time, and dynamic adjustment is achieved. At low wind speeds, the drag coefficient c is higher, while at high wind speeds it tends to be stable. This makes the calculation results of the Ekman transport and upwelling index more accurate, thereby improving the accuracy of the prediction of the marine chlorophyll a concentration.

[0025] Optionally, the angle between the coastline of the area to be predicted and the equator is determined by geometric measurement; The “coast angle” formed between the coastline and true north at each coastal station is measured using geometric tools and incorporated into the CUI calculation. represents the angle perpendicular to the seaward unit vector relative to the mean position of the coastline; In step 3, the data interpolation empirical orthogonal function method (hereinafter referred to as DINEOF method) is used to interpolate the sea surface temperature, which includes the following steps: Step 31, constructing the acquired sea surface temperature within the set time period into a three-dimensional matrix (y×x×t); Where y and x represent the latitude and longitude of each image, respectively, and t is the total number of images, ensuring that (y×x)>t; Step 32: Subtract the mean values ​​of the spatial and temporal dimensions from each value in the matrix, and initialize the missing data points to zero; Step 33, using iterative singular value decomposition and using a randomly selected set proportion of valid data for cross validation to determine the best empirical orthogonal function (EOF) mode; Preferably, a set proportion of valid data is randomly selected, and the set proportion may be 3%; Step 331: for the matrix output from step 32, singular value decomposition is used to decompose the matrix into three matrices, the main mode is extracted, and then the data is adjusted according to the decomposition result, such as removing noise, reconstructing data, etc.; the decomposition is repeated until the result converges, and a new EOF mode is extracted; The result converges when the EOF pattern no longer changes significantly. The preliminary decomposition results are used to adjust the data, such as removing outliers or reconstructing missing values. Step 332: Randomly select a set proportion of valid data as a validation set, and use the remaining data for training and cross-validation; Repeat random sampling and verification multiple times to ensure the consistency of EOF patterns in different data subsets; Step 333, optimal mode selection: According to the result of cross-validation, the EOF mode with the largest explained variance and the strongest robustness is selected as the optimal empirical orthogonal function (EOF) mode.

[0026] Step 34: Use the optimal EOF mode to reconstruct the sea surface temperature data set to obtain the interpolated sea surface temperature data.

[0027] In step 4, the long-term trend prediction model of Chl-a related to sea surface temperature and wind speed is specifically as follows: based on the historical data of Chl-a values, a trend change distribution diagram of the Chl-a values ​​is obtained by a trend analysis method; the abscissa of the trend change distribution diagram of the Chl-a values ​​is the time point, and the ordinate is the specific value; at the same time, the change information or numerical information of the corresponding related factors is added to the trend change distribution diagram of the Chl-a values; Establish an association between the data of relevant factors of Chl-a value changes and the trend change distribution map of Chl-a value. Specifically: for each time point, add the value of specific relevant factors; for example, at a certain time point on the distribution map, directly add the corresponding wind speed value and sea surface temperature value to the corresponding point on the distribution map. When forecasting, find the corresponding point on the trend change distribution map for the wind speed and sea surface temperature data within the current set time, so as to roughly predict the change trend of Chl-a value in the future period on the trend change distribution map; A further technical solution is a method for predicting the long-term trend of Chl-a related to sea surface temperature and wind speed based on correlation analysis, comprising the following steps: Step S1, for the area to be predicted, obtaining a satellite remote sensing data set including Chl-a value and SST data, and a reanalysis data set including wind speed data; Correlation calculations were performed on the data in the dataset to identify the relevant factors for the changes in Chl-a values, including wind speed and SST data; Step S2, quantifying the coastal upwelling by combining the wind speed data with the strength and direction of the Ekman transport and the relationship with the coastline, and obtaining an upwelling index calculated based on the wind speed; Step S3, using the DINEOF method to reconstruct the Chl-a values ​​in the satellite remote sensing data set and the missing data in the SST data set; Step S4, obtaining a trend change distribution diagram of the Chl-a value based on a trend analysis method for the reconstructed Chl-a value data; Step S5, establishing a correlation relationship between the relevant factor data of the Chl-a value change and the trend change distribution diagram of the Chl-a value, and obtaining a Chl-a long-term trend prediction model related to the sea surface temperature and wind speed.

[0028] A specific implementation method is to establish a correlation relationship between the data of the relevant factors of the change of the Chl-a value and the trend change distribution diagram of the Chl-a value. Specifically, for each time point, the value of the specific relevant factor is added; In step S1, the satellite remote sensing data is the three-level monthly composite data of Chl-a and SST of MODIS-Terra (hereinafter referred to as MODIS), with a spatial resolution of 4 kilometers and a data time span from 2001 to 2019. The published data set can be used. The reanalysis data can be a public data set released by a weather forecast center: ERA-Interim data set, which contains weekly wind speed data from 2001 to 2019, with a spatial resolution of 0.12×0.12 square meters and an observation height of 10 meters above the surface.

[0029] Step S1 also includes a method for preprocessing chlorophyll a concentration value (Chl-a value) and sea surface temperature (SST) in the satellite remote sensing data set, the steps are as follows; Step S11, according to the cloud coverage rate, remove the data whose cloud coverage rate exceeds the set value; Specifically, for Chl-a values ​​and SST data, satellite images with cloud coverage exceeding a set value are eliminated to reduce data interference caused by atmospheric factors such as cloud cover and solar flare, thereby ensuring data accuracy.

[0030] Among them, the setting value of cloud coverage can be set to be greater than 95%; Step S12, performing logarithmic transformation on the Chl-a value so that the data conforms to the normal distribution assumption; In this embodiment, in order to meet the normal distribution assumption, the Chl-a value is logarithmically transformed to meet the data requirements of the DINEOF interpolation method in order to solve the problem of a large range of Chl-a values; Step S13, subtract the long-term monthly average values ​​of all years from the monthly time series of the Chl-a value and SST data of each month to obtain the monthly abnormal values ​​of the Chl-a value and SST data (Chl-a_A and SST_A), thereby eliminating the influence of seasonal cycles; Specifically, a monthly time series refers to the Chl-a value or SST data series of a certain pixel in the same month of each year; for example, the data of January 2020, January 2021, and January 2022 constitute a monthly time series of "January"; The long-term monthly average refers to the average of all Chl-a values ​​or SST data in a certain month over many years. The value obtained is the "long-term monthly average", which reflects the normal level of data in a certain month over many years.

[0031] Deducting the long-term monthly average means subtracting the long-term monthly average from the observed value of a month in a certain year to obtain the abnormal value of this month, that is, the monthly abnormal value. This abnormal value represents the difference between the actual observed value and the normal level (long-term monthly average).

[0032] The above operation is performed for each pixel, and the obtained differences (outliers) are recorded as Chl-a_A and SST_A, respectively, indicating the fluctuations of Chl-a and SST relative to the long-term average.

[0033] Correlation analysis: The quantitative relationship between Chl-a_A and SST_A anomalies was obtained by calculating the correlation coefficient of the Chl-a_A and SST_A time series for each pixel; Step S2, quantifying the coastal upwelling by combining the wind speed data with the strength and direction of the Ekman transport and the relationship with the coastline, and obtaining an upwelling index calculated based on the wind speed; To represent monthly climate patterns, these weekly wind speed data are averaged into 12 months per year, spanning from January 2001 to December 2019. For example, the weekly data contained in January 2002 are averaged to obtain the monthly wind speed data for that month. The monthly wind speed data for each month of each year are used to calculate the corresponding monthly Ekman transport, and then the corresponding monthly upwelling index is calculated. The Ekman transport and upwelling index data for a certain month are averaged across all years to obtain the monthly Ekman transport and monthly upwelling index data for 12 months, respectively.

[0034] According to formulas (1) and (2), the Ekman transport component for each month during the study period was calculated; The calculated Ekman transport component ET x and E.T. y , used to generate monthly plots of Ekman transport to visually demonstrate its changes over time.

[0035] The long-term monthly average value of all years is deducted from the monthly time series of wind speed in each month to obtain the monthly abnormal value of wind speed data (Wind_A); the calculation method of the monthly time series of wind speed is the same as that in step S13, which will not be repeated here; Correlation analysis was performed on the wind speed data: the quantitative relationship between the Chl-a_A and Wind_A anomalies was obtained by calculating the correlation coefficient of the Chl-a_A and Wind_A time series of each pixel; Among them, time series data is the corresponding data plus time information, arranged into sequence data according to time, such as Chl-a_A time series data, which is the value of Chl-a_A plus the corresponding time, and arranged into time series data; Furthermore, the DINEOF method is used to reconstruct the Chl-a values ​​in the dataset and the missing data in the SST dataset, including the following steps: Step S31, construct each data set into a three-dimensional matrix (y×x×t); Where y and x represent the latitude and longitude of each image, t is the total number of images, and y×x>t; In this embodiment, the Chl-a data are processed using natural logarithms to prevent negative values ​​from appearing during the reconstruction process, while the original SST data are used directly without being transformed.

[0036] Step S32: for each element in the matrix, the mean values ​​in the spatial and temporal dimensions are subtracted from the matrix, and the missing data points are initialized to zero to minimize the deviation in the initial guess; Step S33, using iterative singular value decomposition and cross-validation using randomly selected 3% valid data to determine the best empirical orthogonal function (EOF) mode; Step S34: Reconstruct the entire data set using the optimal EOF mode.

[0037] Step S4, obtaining a trend change distribution diagram of the Chl-a value based on the reconstructed Chl-a value data based on a trend analysis method; Furthermore, the trend analysis method includes Man-Kendall test and Sen's slope method. Based on the Sen's slope method, the trend change of the data in the set time period is obtained. The statistical significance of the Sen's slope is tested by Man-Kendall to obtain the trend change distribution diagram, and the SST data and wind speed data of the corresponding time are associated with the Chl-a value according to the correlation coefficient. When the model is obtained and applied for prediction, the current SST data and wind speed data are detected, and the Chl-a value is predicted based on the constructed trend change distribution map and the correlation between relevant factors (SST data and wind speed data) and Chl-a value; Among them, the Man-Kendall test and Sen's slope method were used to obtain the trend change distribution diagram of the Chl-a value. The specific process is as follows: (1) Man-Kendall test: According to the time sequence, the data to be analyzed is constructed as time series data; Taking the Chl-a value as an example, the Chl-a value is associated with the corresponding time to obtain the time series data of the Chl-a value; Calculate the difference between each data point in the time series and the subsequent data point, and record the sign of these differences (positive, negative, or zero); The number of symbols of each data point is cumulatively summed to calculate the statistic S; The calculation formula of statistic S is as follows: (5) in, represents the annual average value of Chl-a value in year p, represents the annual average value of Chl-a value in year q, p is greater than q, and n represents the total number of samples; (6) To determine whether there is an upward or downward trend, we can consider the positive and negative values ​​of the S statistic. If S>0, it indicates an upward trend, which means that later observations are usually greater than earlier observations. On the contrary, if S<0, it indicates a downward trend; the mean of S is 0, that is, E(s)=0, and the variance of S is calculated by the following formula: (7) Where n is the total number of data points, m is the number of groups with the same value in the dataset, and lk is the number of data points in the kth group with the same value.

[0038] The Z transformation is used below to assess the significance of the trend; (8) (1) Sen's slope calculation: Q ij =(y j -y i ) / (x j -x i )(9) Q=median(Q ij )(10) Among them, Q ij is the slope, xi represents the i-th month, xj represents the j-th month, and j>i. i and j represents Chl-a_A of the i-th month and the j-th month respectively. Q represents all Q ij of the median.

[0039] The method for evaluating the significance of the change trend of Sen's slope is the formula (7) and (8) in the above-mentioned Man-Kendall test method; The Sen's slope value combined with the Man-Kendall test is used to determine the changing trend of chlorophyll a in a set time period (such as 20 years), that is, increase, decrease or unchanged; a changing trend graph is constructed to integrate wind speed data and SST data at the corresponding time, that is, a one-to-one correspondence is established between the wind speed data and ocean temperature data at the corresponding time point and the corresponding Chl-a value, so as to determine the interval of the changing trend based on the detected wind speed data and ocean temperature data, thereby predicting the changing trend of chlorophyll a in the future.

[0040] Taking a certain sea area as an example, the following explains the process of building a long-term trend prediction model for Chl-a related to sea surface temperature and wind speed based on data from 2001 to 2019.

[0041] For the convenience of research, the study area of ​​the example sea area is divided into 21 stations distributed in seven sub-areas. Z1 to Z7 represent the 7 sub-areas respectively, and S1 to S3 represent the stations in each area respectively.

[0042] (1) DINEOF reconstruction and validation of satellite Chl-a and SST datasets; The log-transformed MODIS Chl-a monthly data and raw SST data from 2001 to 2019 were reconstructed using the DINEOF technique. The reconstruction statistics are shown in Table 1, where the missing data rates for Chl-a and SST are 24.67% and 1.26%, respectively. This highlights the key role of adopting DINEOF in reconstructing Chl-a data, as its missing data rate is significantly higher. In addition, the means of the input and output data are almost exactly the same in terms of Chl-a (input is -0.42 and output is -0.428) and SST (both input and output are 27.27). Similarly, the standard deviations of the input and output data are also very close in terms of Chl-a (input is 0.41 and output is 0.405) and SST (input is 1.96 and output is 1.958). These similarities indicate that the distributions of the reconstructed Chl-a and SST data match the raw data very well, indicating a high accuracy of the reconstruction.

[0043] Table 1 DINEOF calculation statistics

[0044] To evaluate the quality of the reconstructed data, we selected a Chl-a image reconstructed in August and a SST image reconstructed in June and compared them with the original SST and Chl-a images. The original images showed a large number of spatial gaps, especially in the Chl-a data. In contrast, the reconstructed images are more continuous and present a more coherent spatial distribution.

[0045] The reconstructed Chl-a and SST data were further cross-validated using the aforementioned method. Figure 2 It is a density map comparing the reconstruction of the Chl-a data after logarithmic transformation and the original Chl-a data (MODIS data), and the original data is MODIS data; Figure 3It is a density map comparing the reconstructed SST data with the original SST data. The solid black line indicates the 1:1 line. The density map shows the comparison between the reconstructed data and the original Chl-a and SST data. The reconstruction results of both show a strong correlation with the original data, which is reflected in the good indicators: slope (Chl-a is 0.86, SST is 0.95), R² (Chl-a is 0.84, SST is 0.96), deviation (Chl-a is 0.002, SST is 0.04) and root mean square error (Chl-a is 0.16, SST is 1.52). In addition, the density map represents the number of data points in each 4 km × 4 km grid, showing a trend of gradually approaching the 1:1 line. This shows that the data reconstructed using the DINEOF method are both accurate and reliable.

[0046] (2) Seasonal variability of Chl-a in the sample sea area; Based on the reconstructed data, the monthly climate averages of Chl-a between 2001 and 2019 were generated. Figure 4 and Figure 5 The Hovmöller plot shows the monthly time series of Chl-a along latitudes 17°N, 21°N, and 25°N and longitudes 61°E, 64°E, and 67°E. Figure 4 and Figure 5 All graphs in the table consistently reflect the known seasonal cycle in the sample area. Chl-a concentrations peak in summer, have a secondary peak in winter, and reach minimum levels in transition months. Spatially, the highest concentrations occur near the western and northern coastlines of the sample area. This seasonal cycle is mainly driven by the southwest monsoon in summer and the northeast monsoon in winter.

[0047] In the latitudinal profile, Chl-a exhibits two annual peaks: a major peak in summer and a minor peak in winter. At 17°N, close to the equator, Chl-a concentrations remain low throughout the year. However, two distinct peaks are still observed, one in summer (August and September) and the other in winter (February and March). With increasing latitude to 21°N, Chl-a concentrations increase significantly in both seasons, with the summer peak occurring between July and September and the winter peak occurring between February and March. Further north at 25°N, in the coastal areas that form the northern boundary, Chl-a levels remain high and eutrophic for most of the time, with distinct peaks in both summer (August to October) and winter (February to March). In addition, Chl-a concentrations gradually increase with increasing longitude. Overall, it is strongly influenced by the monsoon dynamics, affecting the waters in coastal areas and open areas.

[0048] Similarly, Chl-a also shows two seasonal peaks in the longitude profile, except in the northeastern region. At 61°E, near the west coast, Chl-a levels remain high throughout the year, especially in the northern region. Two seasonal peaks are significant, occurring in summer (July to September) and winter (February to March). At 64°E, although the temporal and spatial patterns of Chl-a are similar, the overall concentration is slightly lower. Further east to 67°E, distinct temporal and spatial distribution patterns emerge. Specifically, between 21°N and 23°N, Chl-a shows two peaks in summer (August) and winter (February), while between 23°N and 24°N, Chl-a increases significantly and remains high throughout the year. These spatial variations highlight the complex interaction between monsoon forces and the unique ocean characteristics of different regions.

[0049] To further investigate the upwelling dynamics, the CUI was calculated using wind vectors from eight coastal stations (stations 1 to 8) located in the sample sea area. The time series of monthly CUI is shown in Figure 2. Figure 6 As shown, Figure 6 This is a comparison of the monthly average time series of different stations of the coastal upwelling index from 2001 to 2019: In the example area, June and August are the southwest monsoon period. The Chl-a concentrations in the eight areas affected by the monsoon are higher, indicating that upwelling is the key driving factor for Chl-a changes in these areas during the southwest monsoon. Figure 6 One curve corresponds to the detection data of one station, and eight coastal stations correspond to eight curves.

[0050] (3) Long-term trend changes of Chl-a related to SST and wind speed; To analyze the long-term trends of Chl-a and SST, the Sen's slope of each pixel was calculated using the yearly and monthly anomaly data. Positive and negative values ​​indicate increasing and decreasing trends, respectively, and zero indicates no trend. The statistical significance of the Sen's slope was assessed by the Mann-Kendall (MK) test, with significant trends marked as 1 and non-significant trends marked as 0. Non-significant Sen's slope values ​​are shaded and represented by white areas. The Sen's slope values ​​of Chl-a_A are in the statistically significant area tested by the MK test, and most of the values ​​in the entire sample area are negative, indicating that the Chl-a level is on a downward trend.

[0051] Among the 21 selected stations, significant Sen's slope values ​​were found only at four stations: Z3-S1 represents the sea area near the S1 station in the Z3 sub-region, Z5-S2 represents the sea area near the S2 station in the Z5 sub-region, Z5-S3 represents the sea area near the S3 station in the Z5 sub-region, and Z7-S2 represents the sea area near the S2 station in the Z7 sub-region. In addition, the Sen's slope values ​​of SST_A and Wind_A were calculated for these stations because they are two key factors affecting the variation of Chl_a. The calculation method of Wind_A is the same as that of Chl-a_A and SST_A. Chl-a_A showed a downward trend at all four stations. The trend of SST_A was significantly positive for Z3-S1, Z5-S2, and Z5-S3, but there was no significant trend for Z7-S2. For Wind_A, the trend was significantly positive for Z5-S2 and Z7-S2, while there was no significant trend for Z3-S1 and Z5-S3. These results indicate the complex interactions among Chl-a, SST, and wind speed in different regions.

[0052] To evaluate the impact of SST and wind speed on the long-term trend of Chl-a, the correlation coefficient (r) between Chl-a_A and SST_A, and between Chl-a_A and Wind_A for each pixel was calculated. The r values ​​between Chl-a_A and SST_A are mainly negative in the entire sample sea area, indicating a negative correlation between the two. In contrast, the r values ​​between Chl-a_A and Wind_A are mostly positive, showing a positive correlation.

[0053] Since significant Chl-a_A trends were detected at four stations: Z3-S1, Z5-S2, Z5-S3, and Z7-S2, the time series of Chl-a_A, SST_A, and Wind_A at these locations were extracted to further analyze the temporal variations and the correlations between Chl-a_A and SST_A and between Chl-a_A and Wind_A.

[0054] Figure 7 (a) is a comparison chart of the time series of Chl-a_A and SST_A at the Z3-S1 site; Figure 7 Middle (b) is a comparison chart of the time series of Chl-a_A and SST_A at the Z5-S2 site; Figure 7 Middle (c) is a comparison chart of the time series of Chl-a_A and SST_A at station Z5-S3; Figure 7 (d) is a comparison of the time series trend lines of Chl-a_A and SST_A at the Z7-S2 site. Regarding the coordinates, the left and right ordinates in the sub-graphs represent different meanings, as shown in the figure. The blue and red lines represent the trend lines of Chl-a_A and SST_A, respectively. Similarly, Figure 8The time series of Chl-a_A and Wind_A at the four stations in the same period are also shown. Figure 8 (a) is a comparison chart of the time series of Chl-a_A and Wind_A at station Z3-S1; Figure 8 Middle (b) is a comparison chart of the changing trend lines of the time series of Chl-a_A and Wind_A at the Z5-S2 site; Figure 8 Middle (c) is a comparison chart of the trend lines of the time series of Chl-a_A and Wind_A at station Z5-S3; Figure 8 Middle (d) is a comparison of the changing trend lines of the time series of Chl-a_A and Wind_A at the Z7-S2 site; the blue and red lines represent the trend lines of Chl-a_A and Wind_A, respectively.

[0055] Table 2 shows a detailed statistical summary of these correlations, covering the entire study period, southwest monsoon season, northeast monsoon season, and transition months (before and after the southwest monsoon); Table 2 shows the statistical significance of the correlation values ​​between Chl-a_A and SST_A, and between Chl-a_A and Wind_A, during the entire study period, southwest monsoon season, northeast monsoon season, and transition months at the four sites from 2001 to 2019. A value of 1 indicates a significant correlation, and 0 indicates no significance. The symbols “+” and “–” indicate positive and negative correlations, respectively.

[0056] Table 2 Correlation statistics

[0057] In Z3-S1, no significant correlations between Chl-a_A and SST_A or Wind_A were observed for any time period. In Z5-S2, two significant correlations between Chl-a_A and Wind_A were found: one was a positive correlation throughout the study period and the other was a positive correlation during the northeast monsoon period. In Z5-S3, three significant correlations between Chl-a_A and SST_A were identified: one was during the entire study period, another was during the northeast monsoon period, and the last was during the transition months. In Z7-S2, two significant correlations between Chl-a_A and Wind_A were found: one was during the entire study period and the other was during the southwest monsoon period. These results highlight the regional differences in the relationships between Chl-a_A and SST_A and between Chl-a_A and Wind_A.

[0058] In this embodiment, the model construction process is explained by taking the above sea area as an example. The above method can be used to reconstruct any sea area in the ocean to obtain a prediction model, which is within the protection range.

[0059] Example 2 Based on Example 1, this embodiment provides a marine chlorophyll a spatiotemporal variation and trend prediction system, including: Data acquisition module: configured to acquire wind speed and sea surface temperature data for the area to be predicted; Quantification module: configured to quantify coastal upwelling based on wind speed data, the strength of Ekman transport and the angle of the coastline, and obtain an upwelling index based on wind speed calculation; Interpolation processing module: configured to use the DINEOF method to reconstruct the missing data in the sea surface temperature data and perform interpolation processing on the sea surface temperature data; Prediction module: It is configured to use the upwelling index and the interpolated sea surface temperature data to predict the Chl-a long-term trend prediction model related to sea surface temperature and wind speed based on correlation analysis to obtain the marine chlorophyll a concentration and its changing trend.

[0060] Furthermore, the quantization module includes: A first calculation module: configured to dynamically adjust the drag coefficient according to the acquired wind speed data, and calculate the Ekman transport component for each month within a set period; The second calculation module is configured to calculate the upwelling index according to the obtained Ekman transport component and the angle between the coastline of the area to be predicted and the equator.

[0061] It should be noted here that the various modules in this embodiment correspond one-to-one to the various steps in Example 1, and the specific implementation process is the same, which will not be repeated here.

[0062] Example 3 Based on Example 1, this embodiment provides a marine chlorophyll a spatiotemporal variation and trend prediction system, including: a data acquisition terminal and a processor; Data collection devices, including a satellite system and wind speed sensors installed at each station, are used to collect sea surface temperature data and wind speed data respectively; The processor is configured to execute the method for predicting the spatiotemporal variation and trend of marine chlorophyll a described in Example 1.

[0063] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for predicting the spatiotemporal variation and trend of marine chlorophyll a, characterized in that: The steps include: Obtain wind speed and sea surface temperature data for the area to be predicted; Based on the wind speed data, the intensity of Ekman transport and the angle of the coastline are combined to quantify the coastal upwelling, and the upwelling index based on wind speed calculation is obtained; Use the DINEOF method to reconstruct missing data in the sea surface temperature data and interpolate the sea surface temperature data; The upwelling index and interpolated sea surface temperature data are used to predict the Chl-a long-term trend prediction model related to sea surface temperature and wind speed based on correlation analysis to obtain the marine chlorophyll a concentration and its changing trend.

2. A method for predicting the spatiotemporal variation and trend of marine chlorophyll a according to claim 1, characterized in that: The calculation method of the upwelling index includes the following steps: The drag coefficient is dynamically adjusted according to the acquired wind speed data, and the Ekman transport component is calculated for each month within the set period; The upwelling index is calculated based on the obtained Ekman transport component and the angle between the coastline of the predicted area and the equator.

3. A method for predicting the spatiotemporal variation and trend of marine chlorophyll a according to claim 1, characterized in that: The interpolation process of sea surface temperature data includes the following steps: The sea surface temperature data acquired within a set time period is constructed into a three-dimensional matrix; Subtract the mean value of the spatial and temporal dimensions from each value in the three-dimensional matrix and initialize the missing data points to zero; Iterative singular value decomposition and cross-validation using a randomly selected set proportion of valid data were used to determine the best empirical orthogonal function model; The sea surface temperature dataset is reconstructed using the optimal EOF mode to obtain the interpolated sea surface temperature data.

4. A method for predicting the spatiotemporal variation and trend of marine chlorophyll a according to claim 1, characterized in that: Based on correlation analysis, a method for obtaining a long-term trend prediction model of Chl-a related to sea surface temperature and wind speed includes the following steps: For the area to be predicted, obtain satellite remote sensing data sets including Chl-a values ​​and sea surface temperature data, as well as reanalysis data sets including wind speed data; perform correlation calculations on the data in the data sets to identify the relevant factors of Chl-a value changes, including wind speed and sea surface temperature data; The wind speed data is combined with the strength and direction of Ekman transport and the relationship with the coastline to quantify the coastal upwelling, and the upwelling index based on wind speed calculation is obtained; Use the DINEOF method to reconstruct the Chl-a values ​​in satellite remote sensing datasets and the missing data in sea surface temperature datasets; For the reconstructed Chl-a value data, a trend change distribution diagram of the Chl-a value is obtained based on the trend analysis method; The correlation relationship between the data of related factors of Chl-a value changes and the trend change distribution diagram of Chl-a value was established, and the long-term trend prediction model of Chl-a related to sea surface temperature and wind speed was obtained.

5. A method for predicting the spatiotemporal variation and trend of marine chlorophyll a as claimed in claim 4, characterized in that: It also includes the steps of preprocessing the chlorophyll a concentration value and sea surface temperature data in the satellite remote sensing data set, as follows: According to the cloud coverage rate, the data with cloud coverage rate exceeding the set value are eliminated; Chl-a values ​​were log-transformed to make the data conform to the normal distribution assumption; The long-term monthly averages of all years are subtracted from the monthly time series of the Chl-a value and sea surface temperature data of each month to obtain the monthly anomaly values ​​of the Chl-a value Chl-a_A and the monthly anomaly values ​​of the sea surface temperature data SST_A.

6. A method for predicting the spatiotemporal variation and trend of marine chlorophyll a as claimed in claim 5, characterized in that: The quantitative relationship between Chl-a_A and SST_A anomalies was obtained by calculating the correlation coefficient of the Chl-a_A and SST_A time series for each pixel; By calculating the correlation coefficient of the Chl-a_A and Wind_A time series for each pixel, the quantitative relationship between the Chl-a_A and Wind_A anomalies was obtained; According to the quantitative relationships obtained, the relevant factors for the change of Chl-a value are determined to include wind speed and sea surface temperature data.

7. A method for predicting the spatiotemporal variation and trend of marine chlorophyll a as claimed in claim 4, characterized in that: For the reconstructed Chl-a value data, the trend change distribution diagram of the Chl-a value is obtained based on the trend analysis method, including: Based on the Sen's slope method, the trend change of the data within the set time period is obtained. The statistical significance of the Sen's slope is tested by Man-Kendall to obtain the trend change distribution diagram.

8. A marine chlorophyll a spatiotemporal variation and trend prediction system, characterized in that: include: Data acquisition module: configured to acquire wind speed and sea surface temperature data for the area to be predicted; Quantification module: configured to quantify coastal upwelling based on wind speed data, the strength of Ekman transport and the angle of the coastline, and obtain an upwelling index based on wind speed calculation; Interpolation processing module: configured to use the DINEOF method to reconstruct the missing data in the sea surface temperature data and perform interpolation processing on the sea surface temperature data; Prediction module: It is configured to use the upwelling index and the interpolated sea surface temperature data to predict the Chl-a long-term trend prediction model related to sea surface temperature and wind speed based on correlation analysis to obtain the marine chlorophyll a concentration and its changing trend.

9. A marine chlorophyll a spatiotemporal variation and trend prediction system as claimed in claim 8, characterized in that: The quantization module includes: A first calculation module: configured to dynamically adjust the drag coefficient according to the acquired wind speed data, and calculate the Ekman transport component for each month within a set period; The second calculation module is configured to calculate the upwelling index according to the obtained Ekman transport component and the angle between the coastline of the area to be predicted and the equator.

10. A marine chlorophyll a spatiotemporal variation and trend prediction system, characterized by: Including data acquisition terminal and processor; Data collection devices, including a satellite system and wind speed sensors installed at each station, are used to collect sea surface temperature data and wind speed data respectively; A processor is configured to execute a method for predicting the spatiotemporal variation and trend of marine chlorophyll a as described in any one of claims 1 to 7.

Images