A method and system for estimating marine heat wave return period based on extreme value analysis

Abstract

The application provides a marine heat wave return period estimation method and system based on extreme value analysis, and belongs to the technical field of marine environment monitoring. The method obtains sea surface temperature data of a sea area and performs standardization processing to obtain a standardized sea temperature spatio-temporal data set. The heat wave state of each marine grid point is identified, and a grid point heat wave information matrix is constructed. The heat wave state is regionally merged and time series associated to form a regional heat wave event object. Event-level duration features, spatial coverage features and intensity features are extracted to construct a parameter sequence. An over-threshold value method is used to extract extreme value samples, and the final over-threshold value is determined in combination with average excess value change characteristics and extreme value model stability. An extreme value model is established, the return period of a target heat wave parameter value is calculated, and is mapped to a target sea area spatial grid to form a spatially resolved marine heat wave return period risk result. The application can realize quantitative return period estimation of marine heat wave regional events, and improve the spatial expression capability and practicality of risk assessment results.

Description

Technical Field

[0001] This invention belongs to the field of marine environmental monitoring technology, specifically relating to a method and system for predicting the return period of marine heat waves based on extreme value analysis. Background Technology

[0002] As the impacts of global climate change continue to deepen, marine heat waves, a typical extreme high-temperature event in the ocean, are becoming increasingly frequent, prolonged, and severe, significantly affecting the marine ecological environment, fishery resources, and coastal production and daily life. Therefore, effectively identifying and assessing the risks of marine heat waves has gradually become an important research topic in the field of marine environmental monitoring and disaster prevention.

[0003] In existing technologies, the analysis of marine heat waves is typically based on sea surface temperature time series, using threshold determination methods to identify heat wave events, and statistically analyzing parameters such as the frequency, duration, and intensity of heat wave events to reflect the changing characteristics of marine heat waves. Some technical solutions also incorporate climate model outputs to analyze changes in marine heat waves at different time scales or under different scenarios. While these methods can characterize the general patterns of marine heat wave changes to some extent, they primarily focus on event identification and trend statistics, and remain insufficient for further quantitative expression of marine heat wave risk.

[0004] Specifically, most existing technologies are based on sea surface temperature grids or time series for analysis. The results usually reflect local temperature anomalies or general statistical characteristics, making it difficult to directly form a quantitative characterization of the risk level of marine heat wave events. Especially when facing extreme marine heat wave events with long duration, wide impact area, or high intensity, existing methods often fail to provide corresponding risk indicators such as return periods in a stable and intuitive manner. Therefore, they are unable to meet the needs of quantitative risk information in application scenarios such as marine disaster early warning, engineering defense, and regional management.

[0005] Therefore, the main problem with existing technologies is the lack of a technical means to effectively quantify key parameters of marine heat wave events and further obtain corresponding return period results. Summary of the Invention

[0006] In view of the above-mentioned problems in the prior art, the purpose of this invention is to provide a method for predicting the return period of marine heat waves based on extreme value analysis.

[0007] A method for predicting the return period of marine heat waves based on extreme value analysis includes the following steps: Obtain sea surface temperature data of the sea area to be analyzed within a preset time period, and perform standardization processing on the sea surface temperature data to unify spatial and temporal references, thereby obtaining a standardized sea surface temperature spatiotemporal dataset. Based on the standardized sea surface temperature spatiotemporal dataset, marine heat waves are identified at each marine grid point in the target sea area, and a grid point heat wave information matrix is ​​constructed. Based on the gridded heat wave information matrix, the ocean grid points that are in heat wave state at each time are spatially connected and merged, and continuous associations are established between the merged regions at adjacent times to form a set of regional heat wave event objects. Using the heat wave event objects in the region as statistical units, the event-level feature parameters of each heat wave event object in the region are extracted, and the corresponding parameter sequence is constructed. The event-level feature parameters include at least the duration, maximum coverage area and intensity parameters. The intensity parameters include at least one of the regional peak intensity, regional average intensity, regional cumulative intensity and area-weighted cumulative intensity. Extreme value samples are extracted from the parameter sequence using the over-threshold method, and the final over-threshold is determined based on the average excess value under the candidate threshold and the stability of the model parameters, forming an extreme value sample set for the corresponding parameter dimension; An extreme value model is established for the extreme value sample set, and the return period value corresponding to the target heat wave parameter value is calculated based on the extreme value model; Based on the spatial coverage of heatwave events in each region, the corresponding return period values ​​are mapped to the target sea area spatial grid to form spatially resolved marine heatwave return period results.

[0008] Preferably, the sea surface temperature data includes at least time information, spatial location information, and the corresponding sea surface temperature value; The standardization processes performed on sea surface temperature data include: Unify sea surface temperature data from different sources to a preset spatial and temporal reference. Remove invalid and missing measurement areas; Based on the unified sea surface temperature data, a standardized sea surface temperature spatiotemporal dataset can be created that can be accessed by time and spatial grid index.

[0009] Preferably, marine heat wave identification includes: A corresponding climatological threshold sequence is constructed for each ocean grid point; Based on the relationship between the sea surface temperature of the ocean grid point and the corresponding date climatological threshold, the heat wave status of the ocean grid point at each time point is determined; The gridded heat wave information matrix is ​​constructed based on the heat wave state and the intensity exceeding the threshold.

[0010] Preferably, the climatological threshold sequence is constructed based on the quantile thresholds of historical sea surface temperature samples; When the sea surface temperature at the same ocean grid point is higher than the climatological threshold for the corresponding date for a continuous preset period of time, the ocean grid point is determined to be in a state of ocean heat wave. The overthreshold intensity is the difference between the sea surface temperature at the ocean grid point and the corresponding climatological threshold for that date.

[0011] Preferably, the process of forming a set of regional heat wave event objects includes: Spatial connectivity and merging of ocean grid points that are in a heatwave state at the same time are used to form candidate heatwave regions; Establish continuous correlations based on the spatial correspondence between candidate heat wave regions at adjacent times; Candidate heatwave areas that meet the continuous association criteria will be grouped into the same heatwave event object.

[0012] Preferably, when forming candidate heat wave areas, the candidate heat wave areas are screened for effectiveness, and the screening criteria include a minimum threshold for the number of covered grid points or the covered area. When establishing a continuous association between regions, if the current candidate heat wave region and the previous candidate heat wave region meet the continuous association condition, the event number succession relationship is determined based on the overlapping area, centroid, and boundary distance.

[0013] Preferably, the final excess threshold is determined by the average excess value under the candidate threshold and the stability of the extreme value model parameters. The stability of the model parameters includes the fluctuation range of the shape parameter and the scale parameter under different candidate thresholds, which meets the preset stability condition.

[0014] Preferably, the extreme value model is a generalized Pareto distribution model or an exponential distribution model; when the same spatial grid is covered by multiple regional heat wave event objects, the regional heat wave event object with the smallest corresponding return period value is selected as the extreme heat wave risk characterization of the grid.

[0015] The second objective of this invention is to propose a marine heatwave return period prediction system based on extreme value analysis, comprising: The data processing module is used to acquire sea surface temperature data and perform standardization processing on the data to unify spatial and temporal references, remove invalid and missing areas, and form a standardized sea surface temperature spatiotemporal dataset. The heat wave identification module is used to determine the heat wave status of each ocean grid point based on standardized sea surface temperature data and historical climate thresholds, and to construct a grid point heat wave information matrix. The regional event construction module is used to spatially connect and merge heat wave grid points at the same time, and to establish continuous associations between candidate heat wave regions at adjacent times, forming a set of regional heat wave event objects. The parameter extraction module is used to extract event-level feature parameters, including duration, maximum coverage area and intensity parameters, from regional heat wave events as statistical units, and generate corresponding parameter sequences. The extreme value analysis module is used to extract extreme value samples from the parameter sequence by applying the threshold method, and to determine the final threshold by combining the average excess value change characteristics and the stability of the model parameters, and to establish an extreme value model. The return period calculation module is used to calculate the return period value corresponding to the target heat wave parameter value based on the extreme value model. The spatial mapping module is used to map the recurrence period value to a spatial grid based on the coverage of regional heat wave events. When multiple events are covered by the same grid, the regional heat wave event with the smallest corresponding recurrence period value is selected as the extreme heat wave risk representation of that grid.

[0016] The beneficial effects of this invention are: the method for predicting the return period of marine heat waves based on extreme value analysis, This invention, based on sea surface temperature data, further constructs regional heat wave event objects after completing grid-level marine heat wave identification. It then extracts parameters related to duration, coverage area, and intensity of these regional heat wave events, and establishes a quantitative relationship between heat wave parameters and return periods using over-threshold extreme value analysis. This enables a quantitative characterization of marine heat wave risk. Compared to existing technologies that primarily focus on identifying or performing general statistical analysis of marine heat wave events, this invention provides return period results corresponding to different parameter levels, making the marine heat wave analysis results more suitable for disaster early warning, risk assessment, and regional management applications. Simultaneously, this invention maps the return period results to a spatial grid of the target sea area, forming spatially resolved risk representation results, which improves the intuitiveness and practicality of identifying high-risk areas for marine heat waves. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is the average remaining lifetime map under the candidate threshold in the embodiments of the present invention; Figure 3 This is a parameter stability diagram of the extreme value model in the embodiments of the present invention; Figure 4 This is a schematic diagram of the probability distribution and theoretical fitting curve of extreme value samples of heat wave parameters in an embodiment of the present invention; Figure 5 This is a distribution diagram of the empirical return period results and the theoretical return period estimation results of heat wave parameters in the embodiments of the present invention. Detailed Implementation

[0018] Example 1 like Figure 1As shown, a method for predicting the return period of marine heat waves based on extreme value analysis is applicable to analyzing the duration, impact range, and intensity level of marine heat wave events in a target sea area based on satellite remote sensing sea surface temperature data, ocean reanalysis data, climate model output data, or a combination thereof.

[0019] This method can convert raw sea surface temperature spatiotemporal data into regional-scale marine heat wave risk quantification results, specifically including the following steps: S1. Sea Surface Temperature Data Acquisition and Standardization Processing Acquire sea surface temperature data for the area to be analyzed within a preset time period. The sea surface temperature data must include at least time information, spatial location information, and the corresponding sea surface temperature value. The data can be obtained from one or more sources, including satellite remote sensing sea surface temperature products, ocean reanalysis data, and numerical model outputs.

[0020] Since sea surface temperature data from different sources may differ in spatial resolution, temporal resolution, projection method, and missing data labeling method, the acquired sea surface temperature data is first standardized to ensure that subsequent heat wave identification and regional event tracking are performed under a unified spatiotemporal reference. Specifically, a target analysis grid and a target time series are preset, and sea surface temperature data from different sources are uniformly resampled to the target analysis grid, while time alignment is completed. Preferably, the target analysis grid adopts a regular 0.5°×0.5° latitude and longitude grid, and the target time series adopts a daily time step.

[0021] During spatial resampling, for a target location point P(x,y), the sea surface temperature value corresponding to the target location point is calculated using bilinear interpolation based on the sea surface temperature values ​​of its surrounding neighboring grid points. For data with missing data in the time dimension, the missing data markers are retained, and samples with consecutive missing data exceeding a preset number of days are removed in subsequent processing. Furthermore, a unified land-sea mask is used to remove land areas and invalid regions, thereby obtaining a standardized spatiotemporal dataset of sea surface temperature.

[0022] After the above standardization process, the sea surface temperature data obtained at each time point have a consistent spatial and temporal reference, which can be directly used for subsequent grid-level marine heat wave identification and regional heat wave event object construction.

[0023] S2, grid-level marine heat wave identification Based on the standardized sea surface temperature spatiotemporal dataset obtained in step S1, marine heat waves are identified for each ocean grid point within the target sea area. Specifically, for each grid point, a corresponding climatological threshold sequence is constructed based on the historical sea surface temperature sequence of that grid point, and this threshold sequence is used as the benchmark for heat wave determination.

[0024] In this embodiment, a relative threshold method is used to identify marine heat wave events. The 90th percentile threshold is calculated using historical sea surface temperature samples within a preset time window around the same date each year. Preferably, an 11-day sliding window is used. When the sea surface temperature at a certain grid point is higher than the 90th percentile threshold for the corresponding date for at least 5 consecutive days, it is determined that a marine heat wave event has occurred at that grid point within the corresponding time period.

[0025] For identified grid point heatwave events, record their start time, end time, duration, and intensity information. The daily exceedance intensity on day t is defined as:

[0026] in, Let be the actual sea surface temperature at that grid point on day t. This is the 90th percentile threshold for that grid point on day t. When A value greater than 0 indicates that the day is in an over-threshold state. The cumulative intensity of a single grid-point heatwave event is defined as:

[0027] in, and These represent the start and end times of the heatwave event at that grid point, respectively.

[0028] Based on the above identification results, a grid-based heatwave information matrix is ​​constructed using "time-space grid points" as the index. This matrix includes at least the heatwave status markers for each grid point at each time point and the corresponding daily exceedance intensity information. This matrix serves as the input data for subsequent construction of regional heatwave event objects.

[0029] S3, Construction of Regional Heatwave Event Objects After obtaining the gridded heat wave information matrix in step S2, the heat wave state layers at each time point are merged into regions, and continuous associations are established between adjacent time points to form regional heat wave event objects.

[0030] Specifically, for any given heatwave state layer, connected regions of the heatwave are identified according to a preset spatial adjacency rule. Preferably, an eight-neighborhood connectivity rule is used to merge spatially adjacent grid points that are all in a heatwave state into the same region. After merging, candidate regions are filtered to remove regions with fewer covered grid points than a preset minimum grid point threshold. Preferably, the minimum grid point threshold is set to 4 grid points; under a 0.5°×0.5° grid condition, this threshold can correspond to the smallest continuous heatwave region on the order of approximately 1°×1°, thereby reducing the interference of local noise points or isolated small hot spots on the identification of regional events.

[0031] Based on this, a continuous association is established between candidate regions at adjacent time points. Let the candidate region at the previous time point be... The candidate region at the current moment is When one of the following conditions is met, and The association is between adjacent time segments of heatwave event objects in the same region: (1) The overlapping area of ​​the two regions accounts for no less than 50% of the area of ​​the smaller region of the two regions; (2) The distance between their centroids is no more than 2 grid cells, and the minimum distance between their boundaries is no more than 1 grid cell.

[0032] When a region at the current moment cannot satisfy the above-mentioned continuous association condition with any region at the previous moment, it is numbered as a new regional heat wave event object; when a region at the current moment satisfies the continuous association condition with multiple regions at the previous moment, the region at the previous moment with the largest overlapping area is preferably selected as its successor object.

[0033] Through the above processing, a set of regional heat wave event objects with unique numbers is formed. Each regional heat wave event object corresponds to a start time, an end time, and the spatial coverage of each time in its life cycle.

[0034] S4. Extraction of Feature Parameters of Regional Heat Wave Events Feature parameters are extracted from the heat wave events obtained in step S3 to form input samples for subsequent extreme value analysis. These feature parameters include at least the duration, maximum coverage area, regional average intensity, regional peak intensity, regional cumulative intensity, and area-weighted cumulative intensity.

[0035] The duration is defined as the total time from the first occurrence to the end of a heatwave event in the corresponding area; the maximum coverage area is defined as the maximum coverage area of ​​the heatwave event at each moment throughout its entire lifecycle, which can be expressed as:

[0036] in, Let t be the number of heatwave grid points covered by the heatwave event object in this region at time t. This represents the area corresponding to a single grid point.

[0037] At time t, the regional average intensity is defined as the average of the daily exceedance threshold intensities of all heat wave grid points within the heat wave event object in that region, i.e.:

[0038] in, The daily exceedance threshold intensity of the heat wave event object in the region at the k-th heat wave grid point at time t.

[0039] Regional peak intensity is defined as the maximum value of the regional average intensity at all times during the lifetime of a heat wave event in that region, i.e.:

[0040] The regional cumulative intensity is uniformly defined as the time-cumulative value of the regional average intensity over the lifetime of the heat wave event object in that region, that is:

[0041] Area-weighted cumulative intensity is defined as the cumulative product of the regional average intensity and the covered area over the lifetime of a heatwave event in that region, i.e.:

[0042] in, Let be the coverage area of ​​the heat wave event in the region at time t.

[0043] After parameter extraction, the duration sequence, maximum coverage area sequence, regional peak intensity sequence, regional cumulative intensity sequence, and area-weighted cumulative intensity sequence were constructed respectively.

[0044] The above parameter sequences all use regional heat wave events as the basic statistical units, so that subsequent extreme value analysis is directly oriented towards the regional heat wave process, rather than individual grid point temperature values.

[0045] S5. Extreme value sample extraction and threshold determination For the duration sequence, maximum coverage area sequence, regional peak intensity sequence, regional cumulative intensity sequence, and area-weighted cumulative intensity sequence constructed in step S4, extreme value samples are extracted using the over-threshold method. Specifically, a candidate threshold set is set for each parameter sequence, and the final over-threshold is determined based on the excess characteristics of samples under different candidate thresholds and the changes in model parameters, thereby extracting the corresponding extreme value samples.

[0046] In one implementation, the candidate threshold set can be a set of discrete thresholds between the 80th and 95th percentiles of the corresponding parameter sequence sample values. For any candidate threshold... Extract all parameters in the sequence that are greater than the candidate threshold. The samples are used as candidate over-threshold samples, and the corresponding average over-threshold value is calculated:

[0047] in, For exceeding the threshold The number of samples, These are the sample values ​​for the corresponding parameters.

[0048] To determine the final excess threshold, the average excess value and the parameter estimation results of the corresponding extreme value model are calculated for each candidate threshold, and a comprehensive judgment is made by combining the mean remaining lifetime map and the parameter stability map.

[0049] Figure 2 This is a plot of average remaining lifetime, with the horizontal axis representing candidate thresholds. The vertical axis represents the average excess value under the corresponding candidate threshold; Figure 3 This is a parameter stability plot, with the horizontal axis representing the candidate threshold. The vertical axis represents the shape parameter ξ and the scale parameter σ of the fitted extreme value model. When the average excess value changes approximately linearly or relatively steadily with the threshold within a certain candidate threshold interval, and the shape parameter ξ and the scale parameter σ remain relatively stable within this interval, this interval is determined as a selectable threshold interval.

[0050] Under the condition of meeting the requirement for the number of samples exceeding the threshold, it is preferable to select the smaller threshold within the selectable threshold range as the final threshold, so as to balance the stability of the extreme value model fitting and the number of extreme value samples. In a specific implementation, such as Figure 2 , Figure 3 As shown, the mean remaining lifetime map and the parameter stability map together indicate that the threshold has good stability when it is in the range of 115 to 130. Therefore, the final overthreshold can be selected in this range to extract the overthreshold extreme value samples of the corresponding parameter sequence.

[0051] To ensure the independence of extreme value samples, for multiple related records exceeding the threshold from the same heat wave event object in the same region, only one sample value representing the extreme degree of the heat wave event object in that region is retained as an independent extreme value sample under the corresponding parameter dimension. Specifically, for the duration series, the sample value with the largest duration is retained; for the maximum coverage area series, the sample value with the largest maximum coverage area is retained; for the regional peak intensity series, the sample value with the largest regional peak intensity is retained; and for the regional cumulative intensity series and the area-weighted cumulative intensity series, the sample value with the largest corresponding value is retained respectively.

[0052] After the above processing, extreme value sample sets for duration, extreme value sample sets for maximum coverage, extreme value sample sets for regional peak intensity, extreme value sample sets for regional cumulative intensity, and extreme value sample sets for area-weighted cumulative intensity are formed, which are used for extreme value model fitting in step S6.

[0053] S6. Extreme value model fitting and goodness-of-fit test Extreme value models are established for the extreme value sample sets of duration, maximum coverage, regional peak intensity, regional cumulative intensity, and area-weighted cumulative intensity obtained in step S5. Preferably, a generalized Pareto distribution model is used to fit the samples exceeding the threshold; for some parameter samples with tail features close to exponential decay, an exponential distribution model can also be established as a comparison model.

[0054] In one implementation, maximum likelihood estimation is used to obtain the parameter values ​​of the candidate extreme value model. When parameter interval estimation is required, the Markov chain Monte Carlo method can also be used to sample and estimate the model parameters. For multiple candidate models, the final model is selected by combining the information criterion and the goodness-of-fit index. Preferably, the Akaike information criterion is used as the model selection criterion, and the model with the smallest Akaike information criterion value is selected as the target extreme value model for the corresponding parameter dimension.

[0055] After completing the model fitting, a goodness-of-fit test is performed on the fitting results, and a probability distribution diagram of the heat wave parameters is plotted. For example... Figure 4 As shown, Figure 4 This is a probability distribution plot using the duration parameter as an example. The horizontal axis represents the heat wave parameter value, the vertical axis represents the probability density, the bars represent the extracted extreme value sample distribution, and the fitted curve represents the theoretical probability density distribution corresponding to the established extreme value model. By comparing the consistency between the sample distribution and the theoretical distribution curve, the fitting effect of the established extreme value model on the corresponding extreme value samples of the heat wave parameter can be determined.

[0056] After the above processing, the target extreme value models and their goodness-of-fit test results corresponding to the duration, maximum coverage area, regional peak intensity, regional cumulative intensity and area-weighted cumulative intensity are obtained respectively.

[0057] S7. Recurrence Period Calculation Based on the extreme value model obtained in step S6, the return period corresponding to the target parameter value is calculated. Let the probability that a certain parameter value is exceeded within a preset time unit be... Then the return period value corresponding to this parameter value Represented as:

[0058] In one implementation, the preset time unit is years. For the empirical return period, the empirical exceedance probability can be determined based on the sorting results of the extreme value samples, and the corresponding empirical return period value can be calculated. For the theoretical return period, the theoretical exceedance probability corresponding to different parameter values ​​can be calculated based on the target extreme value model obtained in step S6, and the theoretical return period value can be further obtained.

[0059] After calculating the empirical and theoretical return periods, the extracted extreme value samples are superimposed with the theoretical estimates to construct the return period distribution map corresponding to the heat wave parameters.

[0060] like Figure 5 As shown, Figure 5 This is a return period distribution chart using the duration parameter as an example. The horizontal axis represents the return period value, and the vertical axis represents the heat wave parameter value. The discrete points in the chart represent the empirical return period results corresponding to the extracted extreme value samples. The fitted curve represents the theoretical estimate result calculated by the target extreme value model, and the intervals on both sides of the curve represent the range of the theoretical estimate result. This return period distribution chart can intuitively represent the relationship between different heat wave parameter levels and their corresponding return period values.

[0061] For a given target parameter level in practical applications, such as a regional heat wave event whose duration exceeds a set number of days, whose maximum coverage area exceeds a set range, or whose regional cumulative intensity exceeds a set threshold, the corresponding return period value can be calculated based on the established return period distribution relationship. After the above processing, the return period calculation results under different parameter dimensions such as duration, maximum coverage area, regional peak intensity, regional cumulative intensity, and area-weighted cumulative intensity are obtained, which are then used for spatial output in step S8.

[0062] S8. Spatialization mapping of return period values ​​and output of results. Based on the return period values ​​of various heat wave parameters obtained in step S7, the corresponding return period values ​​are mapped to the spatial grid of the target sea area to form a spatially resolved marine heat wave risk result.

[0063] Specifically, based on the spatial coverage of a regional heatwave event within its lifecycle, the return period value of the corresponding parameter dimension of the regional heatwave event is assigned to the grid area it covers. For the same parameter dimension, the return period value corresponding to the duration, the return period value corresponding to the maximum coverage area, the return period value corresponding to the regional peak intensity, the return period value corresponding to the regional cumulative intensity, and the return period value corresponding to the area-weighted cumulative intensity can be mapped to the corresponding coverage grid, thereby forming the spatial risk distribution result under the corresponding parameter dimension.

[0064] When a single grid is covered by multiple regional heatwave events during the analysis period, this embodiment uses the shortest return period priority principle to determine the final return period value of the grid. That is, among the multiple regional heatwave events covering the grid, the one with the smallest corresponding return period value is taken as the final return period value of the grid under the corresponding parameter dimension. Since a smaller return period value corresponds to a higher event frequency and a greater risk level, this method can reflect the highest risk state experienced by the grid throughout the entire analysis period.

[0065] After completing the spatial mapping of return period values, the following maps can be generated: duration return period distribution map, maximum coverage area return period distribution map, regional peak intensity return period distribution map, regional cumulative intensity return period distribution map, and area-weighted cumulative intensity return period distribution map. For cases requiring risk level expression, the spatial grid can be classified according to preset return period intervals to generate marine heat wave risk distribution results for different risk levels.

[0066] After the above processing, the return period calculation results obtained in step S7 can be further converted into a marine heat wave risk map with spatial location correspondence, thereby intuitively representing the differences in marine heat wave risk in different areas within the target sea area under different parameter dimensions such as duration, impact range and intensity level.

[0067] Example 2 A marine heatwave recurrence period prediction system based on extreme value analysis includes a data processing module, a heatwave identification module, a regional event construction module, a parameter extraction module, an extreme value analysis module, a recurrence period calculation module, and a spatial mapping module.

[0068] The data processing module is used to acquire sea surface temperature data of the sea area to be analyzed within a preset time period, and to perform standardization processing on the sea surface temperature data to unify spatial and temporal references. At the same time, invalid and missing areas are removed to form a standardized sea surface temperature spatiotemporal dataset that can be called by time and spatial grid index.

[0069] The heat wave identification module is used to determine the heat wave status of each ocean grid point in the target sea area at each time based on standardized sea surface temperature data and historical climate threshold sequences, and to calculate the intensity of exceeding the threshold, thereby constructing a grid point heat wave information matrix.

[0070] The regional event construction module is used to spatially connect and merge grid points that are in a heat wave state at the same time to form candidate heat wave regions. It also establishes continuous associations between candidate heat wave regions at adjacent times and groups candidate heat wave regions that meet the continuous association conditions into the same regional heat wave event object set, thereby forming regional heat wave event objects.

[0071] The parameter extraction module is used to extract event-level feature parameters, including duration, maximum coverage area and intensity parameters, from regional heat wave events as statistical units, and to construct the corresponding parameter sequences.

[0072] The extreme value analysis module is used to extract extreme value samples from event-level parameter sequences using the over-threshold method, and to determine the final over-threshold by combining the average excess value change characteristics and the stability of the extreme value model parameters, thereby establishing an extreme value model.

[0073] The return period calculation module is used to calculate the return period value corresponding to the target heat wave parameter value based on the extreme value model.

[0074] The spatial mapping module is used to map the corresponding return period value to a spatial grid based on the spatial coverage of regional heat wave events. When multiple events are covered by the same grid, the shortest return period value is taken, thus forming a spatially resolved return period result for marine heat waves.

[0075] The modules work together in sequence. The standardized sea surface temperature spatiotemporal dataset generated by the data processing module provides input to the heat wave identification module. The gridded heat wave information matrix generated by the heat wave identification module provides input to the regional event construction module. The regional event objects are used by the parameter extraction module to generate parameter sequences. The extreme value analysis module establishes an extreme value model for the parameter sequences. The return period calculation module uses the extreme value model to calculate the return period value. The spatial mapping module maps the return period value to the target sea area spatial grid, and finally forms a spatially resolved marine heat wave return period result.

[0076] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting the return period of marine heat waves based on extreme value analysis, characterized in that, Includes the following steps: Obtain sea surface temperature data of the sea area to be analyzed within a preset time period, and perform standardization processing on the sea surface temperature data to unify spatial and temporal references, thereby obtaining a standardized sea surface temperature spatiotemporal dataset. Based on the standardized sea surface temperature spatiotemporal dataset, marine heat waves are identified at each marine grid point within the target sea area, and a grid point heat wave information matrix is ​​constructed. The marine heat wave identification includes: constructing a corresponding climatological threshold sequence for each marine grid point; determining the heat wave status of the marine grid point at each time point based on the relationship between the sea surface temperature of that marine grid point and the corresponding date climatological threshold; and constructing the grid point heat wave information matrix based on the heat wave status and the intensity of exceeding the threshold. Based on the gridded heat wave information matrix, the ocean grid points that are in heat wave state at each time are spatially connected and merged, and continuous associations are established between the merged regions at adjacent times to form a set of regional heat wave event objects. Using the heat wave event objects in the region as statistical units, the event-level feature parameters of each heat wave event object in the region are extracted, and the corresponding parameter sequence is constructed. The event-level feature parameters include at least the duration, maximum coverage area and intensity parameters. The intensity parameters include at least one of the regional peak intensity, regional average intensity, regional cumulative intensity and area-weighted cumulative intensity. Extreme value samples are extracted from the parameter sequence using the over-threshold method, and the final over-threshold is determined based on the average excess value under the candidate threshold and the stability of the model parameters, forming an extreme value sample set for the corresponding parameter dimension; An extreme value model is established for the extreme value sample set, and the return period value corresponding to the target heat wave parameter value is calculated based on the extreme value model; Based on the spatial coverage of heatwave events in each region, the corresponding return period values ​​are mapped to the target sea area spatial grid to form spatially resolved marine heatwave return period results.

2. The method according to claim 1, characterized in that, The sea surface temperature data includes at least time information, spatial location information, and the corresponding sea surface temperature value; The standardization processes performed on sea surface temperature data include: Unify sea surface temperature data from different sources to a preset spatial and temporal reference. Remove invalid and missing measurement areas; Based on the unified sea surface temperature data, a standardized sea surface temperature spatiotemporal dataset can be created that can be accessed by time and spatial grid index.

3. The method according to claim 1, characterized in that, The climatological threshold sequence is constructed based on the quantile thresholds of historical sea surface temperature samples; When the sea surface temperature at the same ocean grid point is higher than the climatological threshold for the corresponding date for a continuous preset period of time, the ocean grid point is determined to be in a state of ocean heat wave. The overthreshold intensity is the difference between the sea surface temperature at the ocean grid point and the corresponding climatological threshold for that date.

4. The method according to claim 1, characterized in that, The process of forming the set of regional heat wave event objects includes: Spatial connectivity and merging of ocean grid points that are in a heatwave state at the same time are used to form candidate heatwave regions; Establish continuous correlations based on the spatial correspondence between candidate heat wave regions at adjacent times; Candidate heatwave areas that meet the continuous association criteria will be grouped into the same heatwave event object.

5. The method according to claim 4, characterized in that, When candidate heat wave areas are formed, the candidate heat wave areas are screened for effectiveness. The screening criteria include the minimum threshold of the number of covered grid points or the coverage area. When establishing a continuous association between regions, if the current candidate heat wave region and the previous candidate heat wave region meet the continuous association condition, the event number succession relationship is determined based on the overlapping area, centroid, and boundary distance.

6. The method according to claim 1, characterized in that, The final excess threshold is determined by the average excess value under the candidate threshold and the stability of the extreme value model parameters. The stability of the model parameters includes the fluctuation range of the shape parameter and the scale parameter under different candidate thresholds, which meets the preset stability condition.

7. The method according to claim 1, characterized in that, The extreme value model is a generalized Pareto distribution model or an exponential distribution model; when the same spatial grid is covered by multiple regional heat wave event objects, the regional heat wave event object with the smallest corresponding return period value is selected as the extreme heat wave risk characterization of the grid.

8. A system for predicting the return period of marine heat waves based on extreme value analysis, characterized in that, A method for predicting the return period of marine heat waves based on extreme value analysis as described in any one of claims 1-7, comprising: The data processing module is used to acquire sea surface temperature data and perform standardization processing on the data to unify spatial and temporal references, remove invalid and missing areas, and form a standardized sea surface temperature spatiotemporal dataset. The heat wave identification module is used to determine the heat wave status of each ocean grid point based on standardized sea surface temperature data and historical climate thresholds, and to construct a grid point heat wave information matrix. The regional event construction module is used to spatially connect and merge heat wave grid points at the same time, and to establish continuous associations between candidate heat wave regions at adjacent times, forming a set of regional heat wave event objects. The parameter extraction module is used to extract event-level feature parameters, including duration, maximum coverage area and intensity parameters, from regional heat wave events as statistical units, and generate corresponding parameter sequences. The extreme value analysis module is used to extract extreme value samples from the parameter sequence by applying the threshold method, and to determine the final threshold by combining the average excess value change characteristics and the stability of the model parameters, and to establish an extreme value model. The return period calculation module is used to calculate the return period value corresponding to the target heat wave parameter value based on the extreme value model. The spatial mapping module is used to map the recurrence period value to a spatial grid based on the coverage of regional heat wave events. When multiple events are covered by the same grid, the regional heat wave event with the smallest corresponding recurrence period value is selected as the extreme heat wave risk representation of that grid.

Images