Method and device for evaluating influence of tropical cyclone on marine environment

Abstract

The embodiment of the invention discloses a method and device for evaluating the influence of tropical cyclones on the marine environment. The method comprises the following steps: continuously acquiring marine environment parameters of a tropical cyclone influenced sea area, and constructing a multi-dimensional parameter data set based on the marine environment parameters; derivation features are calculated based on the multi-dimensional parameter data set; taking the cyclone strength in the vortex category and the tropical cyclone data as classification features, taking data except the vortex category and the cyclone strength in the multi-dimensional parameter data set and the derivative features as continuous features, and constructing input data according to the classification features and the continuous features; and outputting an environment evaluation result by the trained multi-parameter coupling model based on the input data. In the embodiment, the classification features and the continuous features can be input into the multi-parameter coupling model at the same time, then the dominant factors are determined by quantifying importance ranking, the environment evaluation result is generated according to the dominant factors to predict the environment evolution in a period of time in the future, and the influence of the tropical cyclone on the marine environment is more accurately and quantitatively predicted.

Description

TECHNICAL FIELD

[0001] Embodiments of the specification belong to the field of marine environment monitoring, and particularly relate to a method and device for evaluating the influence of a tropical cyclone on a marine environment. BACKGROUND

[0002] As one of the most destructive extreme weather events, the strong wind stress of a tropical cyclone can induce strong mixing and upwelling in the upper ocean, thereby affecting the marine environment. Accurate evaluation of this influence is of great significance for understanding the mechanisms of the marine ecosystem responding to extreme weather and warning of ecological disasters. However, the existing evaluation methods generally directly rely on short-term satellite observations or a single numerical model, ignoring the regulation of eddies on the vertical ocean response, resulting in insufficient evaluation of long-term effects and difficulty in quantifying the impact of cyclones, which leads to low evaluation accuracy. SUMMARY

[0003] Embodiments of the present disclosure provide a method and device for evaluating the influence of a tropical cyclone on a marine environment, aiming to solve one or more of the above problems and other potential problems.

[0004] According to a first aspect of the present disclosure, a method for evaluating the influence of a tropical cyclone on a marine environment is provided. The method includes continuously obtaining marine environment parameters of a sea area affected by the tropical cyclone, constructing a multi-dimensional parameter dataset based on the marine environment parameters, the marine environment parameters including tropical cyclone data, sea surface temperature, chlorophyll-a concentration, sea surface height, wind stress curl, and ocean vertical profile data; calculating derived features based on the multi-dimensional parameter dataset, the derived features including sea surface temperature gradient strength calculated based on the sea surface temperature and eddy category calculated based on the sea surface height; taking the eddy category and the cyclone intensity in the tropical cyclone data as classification features, taking the data in the multi-dimensional parameter dataset and the derived features except the eddy category and the cyclone intensity as continuous features, and constructing input data from the classification features and the continuous features; outputting an environmental evaluation result based on the input data by a trained multi-parameter coupling model, the multi-parameter coupling model including a random forest module and an XGBoost module, the random forest module being used to determine a dominant factor in the continuous features according to the importance of the continuous features in contributing to marine productivity, and the XGBoost module being used to generate the environmental evaluation result according to the dominant factor, the environmental evaluation result including a change value of the dominant factor within a future preset time length and corresponding warning information of the change value.

[0005] According to a second aspect of the present disclosure, an evaluation device for the influence of a tropical cyclone on a marine environment is provided. The device comprises a parameter acquisition module configured to continuously acquire marine environment parameters of a sea area affected by a tropical cyclone, and to construct a multi-dimensional parameter dataset based on the marine environment parameters. The marine environment parameters include tropical cyclone data, sea surface temperature, chlorophyll-a concentration, sea surface height, wind stress curl, and marine vertical profile data. The device further comprises a feature extraction module configured to calculate derived features based on the multi-dimensional parameter dataset. The derived features include sea surface temperature gradient strength calculated based on the sea surface temperature and eddy category calculated based on the sea surface height. The device further comprises an input data construction module configured to construct input data from classification features and continuous features. The classification features include the eddy category and the cyclone intensity in the tropical cyclone data, and the continuous features include data in the multi-dimensional parameter dataset and the derived features except for the eddy category and the cyclone intensity. The device further comprises a prediction module configured to output an environment evaluation result based on the input data from a trained multi-parameter coupling model. The multi-parameter coupling model includes a random forest module and an XGBoost module. The random forest module is used to determine a dominant factor in the continuous features according to the importance of the contribution of the continuous features to marine productivity. The XGBoost module is used to generate the environment evaluation result according to the dominant factor. The environment evaluation result includes a change value of the dominant factor within a preset time period in the future and corresponding warning information.

[0006] According to a third aspect of the present disclosure, an electronic device is provided. The electronic device comprises one or more processors, and a memory associated with the one or more processors. The memory is configured to store program instructions. When the program instructions are read and executed by the one or more processors, the method according to the first aspect is performed.

[0007] According to a fourth aspect of the present disclosure, a computer program product is provided. The computer program product comprises a computer program. When the computer program is executed by a processor, the method according to the first aspect is implemented.

[0008] The scheme provided by the embodiments of the present disclosure can input classification features and continuous features into a multi-parameter coupling model according to multi-dimensional marine environment parameters, combined with calculated marine temperature gradient strength and eddy category. Then, the dominant factor is determined by quantifying the importance, and the environment evolution in the future is predicted according to the environment evaluation result generated based on the dominant factor, so that the influence of the tropical cyclone on the marine environment is more accurately quantified and predicted. BRIEF DESCRIPTION OF DRAWINGS

[0009] The above and other features, advantages, and aspects of the embodiments of the present disclosure will become more apparent by describing in detail the following embodiments with reference to the attached drawings. In the drawings, the same or similar reference numerals refer to the same or similar elements, and:

[0010] Figure 1A flowchart of a method for evaluating the impact of a tropical cyclone on a marine environment according to some embodiments of the present disclosure is shown.

[0011] Figure 2 A structural diagram of an apparatus for evaluating the impact of a tropical cyclone on a marine environment according to some embodiments of the present disclosure is shown.

[0012] Figure 3 A schematic block diagram of an electronic device according to some embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0013] For the purposes of the present application, the term "about" means that the following

[0014] The terms "comprises", "comprising", "includes", "including", "has", "having" and their conjugates, as used herein, are intended to cover the situation where literal goods or materials are included together with non-literal goods or materials or vice versa. For example, the process, method, system, product, or apparatus that comprises a list of steps or units are not necessarily limited to those steps or units which are recited but can include other steps or units that are not expressly listed or can include steps or units that are fundamental to the process, method, system, product, or apparatus in question. Depending on the context, the word "if" as used herein can be interpreted to mean "when" or "upon" or "in response to determining" or "in response to detecting".

[0015] Figure 1 A flowchart of a method 100 for evaluating the impact of a tropical cyclone on a marine environment according to some embodiments of the present disclosure is shown. The method 100 can be performed by a terminal, which can include but is not limited to a mobile phone, a tablet computer, a desktop computer, a server, etc. As shown in the method 100, step 102 can continuously acquire marine environment parameters of a sea area affected by a tropical cyclone, and construct a multi-dimensional parameter dataset based on the marine environment parameters. The marine environment parameters include tropical cyclone data, sea surface temperature, chlorophyll-a concentration, sea surface height, wind stress curl, and marine vertical profile data. Figure 1

[0016] ​In this embodiment, tropical cyclone data can be obtained through real-time tropical cyclone archives from the Regional and Mesoscale Meteorology Branch (RMMB). Sea surface temperature (SST), chlorophyll a concentration, sea surface height, ocean vertical profile data, wind stress curl, surface wind field, and mixing layer depth can be obtained through marine color biogeochemical reanalysis product data services, such as the Copernicus Marine Environment Monitoring Service (CMEMS). Net heat flux can be obtained from the OFES dataset of the Asia Pacific Data Research Center. The ocean vertical profile data includes daily three-dimensional (longitude, latitude, depth) field data of seawater temperature, nutrient concentration, dissolved inorganic carbon concentration, pH, etc. Wind stress curl and surface wind field are used to analyze wind forcing and eddy dynamics, while net heat flux is used to quantify the vertical mixing process and air-sea heat exchange in the upper ocean. Data such as sea surface temperature, chlorophyll a concentration, and sea level height have a certain spatial resolution, allowing them to be represented as grid cells to reflect data differences across different ocean regions. By continuously acquiring these marine environmental parameters, these data can be synchronized temporally and spatially, enabling unified interpolation or aggregation of all data onto a daily timescale and a common spatial grid (e.g., ...). By aligning time and space, a unified and consistent dataset of multidimensional parameters can be obtained, which is a multidimensional parameter dataset.

[0017] Furthermore, multidimensional parameter datasets can be preprocessed before subsequent processing. Preprocessing can include, for example, data cleaning, which involves cleaning the raw data to remove obvious outliers caused by sensor malfunctions or transmission errors. Spatiotemporal interpolation methods (such as nearest neighbor interpolation or optimal interpolation) can also be used to fill in missing sea surface temperature or chlorophyll a data in satellite remote sensing data due to cloud cover. For time-series data such as wind stress curl, high-frequency noise can be removed using moving average or low-pass filtering algorithms, retaining signals related to the cyclone scale.

[0018] In method 100, step 104 can calculate derived features based on a multidimensional parameter dataset, including the sea surface temperature gradient intensity calculated based on sea surface temperature and the eddy category calculated based on sea surface height.

[0019] In this embodiment, based on the multidimensional parameter dataset, a Sobel-type convolutional filter can be used, for example, to calculate the gradient of sea surface temperature in the east-west direction at each grid point. and gradient in the north-south direction Then according to The gradient modulus, i.e., the sea surface temperature gradient strength, is calculated. In addition, the eddy category can be identified and classified according to the sea surface height, i.e., a corresponding physical threshold is set according to historical experience, and the eddy category is determined by the continuous comparison result of the sea surface height and the threshold. As an example, the area with a sea surface height continuously lower than 0.05 m can be identified as a cyclonic eddy (representing a sea water divergence and upwelling area), the area with a sea surface height continuously higher than 0.8 m can be identified as an anticyclonic eddy (representing a sea water convergence and sinking area), and the like.

[0020] In the method 100, step 106 can take the eddy category and the cyclone intensity in the tropical cyclone data as classification features, take the data in the multi-dimensional parameter data set and the derived features except the eddy category and the cyclone intensity as continuous features, and construct input data from the classification features and the continuous features.

[0021] In the embodiment, the multi-dimensional parameter data set and the derived features can be integrated to construct input data for subsequent input of a multi-parameter coupling model. The input data will be divided into classification features and continuous features. The classification features (such as cyclone intensity, eddy category, etc.) represent discrete dynamic states or category attributes. After independent coding (for example, one-hot coding), the classification features are input as classification features, which can avoid the model from misinterpreting the unordered relationship between categories as a continuous numerical relationship, and better represent the differences in ocean response under different eddy dynamic backgrounds. The continuous features retain the fine numerical information of the original observation, and provide a basis for quantifying physical processes and biochemical responses for the model.

[0022] In the method 100, step 108 can output an environmental assessment result based on the input data by using a trained multi-parameter coupling model. The multi-parameter coupling model includes a random forest module and an XGBoost module. The random forest module is used to determine a dominant factor in the continuous features according to the importance of the continuous features in contributing to ocean productivity. The XGBoost module is used to generate an environmental assessment result according to the dominant factor. The environmental assessment result includes a change value of the dominant factor within a future preset time length and corresponding warning information of the change value.

[0023] In this embodiment, the random forest module can select a feature to make the impurity of the split child node decrease the most compared to the parent node at each node split of each decision tree according to the input data. By aggregating the impurity reduction values caused by each feature at all node splits in all decision trees in the entire random forest, the average impurity reduction value can be calculated by dividing the aggregated impurity reduction value by the number of decision trees, and the average impurity reduction value can be used as the importance score of the feature to determine the importance ranking. The features with scores higher than a score threshold or rankings higher than a ranking threshold will be used as the dominant factors with greater contribution. Then, the XGBoost module generates a series of simple decision trees in a gradient boosting manner, i.e., in a serial manner, each new tree is used to correct the collective errors of all previous trees, and the environmental assessment results are generated according to the feature data corresponding to the dominant factors. The warning information can be determined by querying the conditions satisfied by the change value. For example, when the model detects that the wind stress curl along the coast of the Arabian Peninsula is significantly enhanced and the mixed layer depth is rapidly deepened, the historical simulation data can be combined to identify the potential trend of nutrient salt upwelling. If the predicted chlorophyll a concentration will exceed the abnormal threshold (e.g. ), the warning information containing “primary productivity outbreak” or “harmful algal bloom” is generated.

[0024] In other implementable manners, the size of the contribution of each environmental variable to the marine response can also be determined by the Pearson correlation coefficient and principal component analysis, and then the importance ranking can be constructed according to the contribution size, so that the random forest module can directly determine the dominant factors according to the importance ranking.

[0025] In addition, geographic information systems can also be used to convert various data into intuitive visual interfaces. For example, the spatial range of sea surface temperature cooling and the spatial response of chlorophyll a can be rendered in real time by a spatial distribution map, the vertical migration of nitrate and pH within 0-100 meters can be displayed by a three-dimensional profile, and the acidification risk of the sea area under the influence of a cyclone can be displayed by color coding (e.g., red represents high risk).

[0026] In an implementable manner, the method further comprises:

[0027] determining and marking the strong upwelling area and the cooling core area according to the comparison between the sea surface temperature gradient strength and the preset gradient threshold value;

[0028] verifying the vortex category based on the wind stress curl.

[0029] In this embodiment, the strong upwelling area and the cooling core area are determined and marked according to the comparison between the sea surface temperature gradient strength and the gradient threshold value (e.g. The comparison of the two can be used to accurately locate the strong upwelling area and the cooling core area induced by the cyclone, to monitor the cooling effect of the cyclone on the ocean in real time, and to evaluate the strength and range of the upwelling, and can also be used as a dimensional data in the input data for more accurate prediction. The wind stress curl can be used to verify the identified vortex category. For example, an area with a wind stress curl greater than The area can be determined as a strong positive vortex (cyclonic vortex) forcing area, and the area is compared with the area determined as a cyclonic vortex according to the sea surface height. The direct correlation between the vortex generation and the wind forcing can be determined. Therefore, the vortex category recognition is considered accurate only when the spatial coincidence of the two is higher than a preset threshold, otherwise the vortex category will be re-identified.

[0030] In an implementation manner, the method further comprises:

[0031] Based on the historical detection data, a training set is determined, the training set comprising input data samples constructed by the historical classification features and the historical continuous features, and environment assessment result samples;

[0032] Based on the input data samples, a predicted environment assessment result is generated by an initial model;

[0033] The initial model is subjected to at least one round of model training with the environment assessment result samples as a supervision signal, to obtain a multi-parameter coupled model.

[0034] In this embodiment, according to the historical detection data, the historical classification features and the historical continuous features collected in history can be obtained, as well as the feature change values at the detection time nodes of these features after a preset time length (24-48 hours), and the warning information labeled by the artificial for these change values, so as to construct the input data samples and the environment assessment result samples, and further construct the training set. In the process of training the initial model using the training set, the generator of the model can generate a predicted environment assessment result based on the input data samples. According to the comparison of the environment assessment result samples and the predicted environment assessment result, a generator loss can be obtained. The predicted environment assessment result is subjected to loss judgment using the generator loss, which can be judged using a comparison loss function (such as a cross-entropy loss function with label smoothing). The generator loss can be a large value, and then the generator loss can be back-propagated to the generator to guide the optimization of the parameters of the generator, to realize one round of supervised training of the generator. Such a training process can be iteratively executed round by round until the generator can generate a more accurate predicted environment assessment result, i.e., the loss value calculated by the loss function is smaller. After the training is completed, the multi-parameter coupled model can output the environment assessment result.

[0035] Specifically, during the model training process, a grid search method can be used to optimize the learning rate (learning_rate), tree depth (max_depth) of XGBoost, and the number of decision trees (n_estimators) of the random forest to minimize the prediction mean square error. K-fold cross validation method can also be used to verify the model.

[0036] In an implementation manner, the method further includes:

[0037] In the training set, the corresponding cyclone stage of each input data sample is labeled, and the training weight of the input data sample in the active stage of the cyclone is increased. The cyclone stage includes a pre-cyclone stage, an active cyclone stage, and a post-cyclone dissipation stage.

[0038] In this embodiment, the influence of the cyclone can be divided into three stages. The pre-cyclone stage represents the undisturbed background ocean state, the active cyclone stage represents the core influence period from development to dissipation of the cyclone, and the post-cyclone dissipation stage represents the stage of ocean environment recovery and sustained response. Under different stages, the values of each feature in the input data also differ, so it is necessary to distinguish by stage to improve training accuracy. For example, by comparing the changes of sea temperature cooling area, chlorophyll burst area, vortex position and range in the three stages, it can be found that, for example, significant cooling and chlorophyll increase appear on the right side of the cyclone path in the active period, and these features gradually weaken in the recovery period. And by analyzing the vertical structure changes of temperature, nutrient salt, dissolved inorganic carbon, and pH at 0-200 meters water depth, it can be found that, for example, the mixed layer deepens and the subsurface nutrient salt upwells in the active period, and the vertical gradient gradually rebuilds in the recovery period. According to these differences, it can also be verified whether the amplitude of the quantitative change value of each feature is reasonable. In addition, since the active stage of the cyclone is the core period of prediction, higher training weight will be assigned to the data of this stage to ensure that the model focuses on training the data of the active stage of the cyclone, which can better predict the future feature change value within the preset time period based on the data of this stage.

[0039] In an implementation manner, the method further includes:

[0040] In the historical detection data, the daily parameter values corresponding to the key areas on the cyclone path are extracted day by day to calculate the change trend of the daily parameter values based on smoothing processing, and to calculate the peak lag period between different parameters based on the cross-correlation function.

[0041] In this embodiment, daily parameter values ​​for key regions along the cyclone path (e.g., the cyclone vortex center, right-side upwelling region, etc.) can be extracted daily from historical monitoring data. This yields the time series of the corresponding parameters, which can then be smoothed (e.g., by moving average) to identify multi-day trends in parameter variation (e.g., gradual recovery of sea surface temperature). Furthermore, cross-correlation functions can be used to calculate the peak lag time between different parameter sequences (e.g., chlorophyll concentration peaks 36–48 hours after nutrient (nitrate) upwelling, or sea surface temperature experiences its strongest drop within 24 hours after wind stress peak). This data can reveal the evolution of parameters of vortices of different properties along the cyclone path over time. When training the model, this data can be added to the training data to enable the trained model to more accurately predict parameter changes over a future period.

[0042] Figure 2 A schematic diagram of the structure of a tropical cyclone impact assessment apparatus 200 according to some embodiments of this disclosure is shown. The various embodiments in this specification are described in a progressive manner, with reference to each other for similar or identical parts. Each embodiment focuses on describing the differences from other embodiments. In particular, the apparatus embodiments are basically similar to the method embodiments, so the description is relatively simple, and relevant parts can be referred to the description of the method embodiments. Figure 2 As shown, the device 200 includes a parameter acquisition module 201, configured to continuously acquire marine environmental parameters of the sea area affected by tropical cyclones, and construct a multidimensional parameter dataset based on the marine environmental parameters, including tropical cyclone data, sea surface temperature, chlorophyll a concentration, sea surface height, wind stress curl, and ocean vertical profile data; a feature extraction module 202, configured to calculate derived features based on the multidimensional parameter dataset, including sea surface temperature gradient intensity calculated based on sea surface temperature and vortex category calculated based on sea surface height; and an input data construction module 203, configured to use vortex category and cyclone intensity from the tropical cyclone data as classification features, and to construct the multidimensional parameter dataset. Data in the parameter dataset and derived features, excluding vortex category and cyclone intensity, are treated as continuous features. Input data is constructed from categorical features and continuous features. Prediction module 204 is configured to output environmental assessment results based on the input data and a trained multi-parameter coupling model. The multi-parameter coupling model includes a random forest module and an XGBoost module. The random forest module is used to determine the dominant factor among the continuous features according to the importance of their contribution to marine productivity. The XGBoost module is used to generate environmental assessment results based on the dominant factor. The environmental assessment results include the change value of the dominant factor within a preset future time period and the warning information corresponding to the change value.

[0043] In an implementation, the tropical cyclone data includes life cycle, moving path, center position, intensity and moving speed of the cyclone, the ocean vertical profile data includes sea water temperature, nutrient salt concentration, dissolved inorganic carbon concentration and pH value, and the ocean environmental parameters further include surface wind field, mixed layer depth and net heat flux.

[0044] In an implementation, the feature extraction module 202 is further configured to calculate gradients of the sea surface temperature in the east-west direction and the north-south direction at each grid point, to calculate a sea surface temperature gradient strength according to the two gradients, and to determine a vortex category of each identified region based on a comparison result between the sea surface height and a preset height threshold.

[0045] In an implementation, the feature extraction module 202 is further configured to determine and label a strong upwelling region and a cooling core region according to a comparison between the sea surface temperature gradient strength and a preset gradient threshold, and to verify the vortex category based on wind stress curl.

[0046] In an implementation, the apparatus further includes a model training module configured to determine a training set based on historical detection data, the training set including input data samples constructed from historical classification features and historical continuous features, and environmental evaluation result samples, to generate predicted environmental evaluation results from an initial model based on the input data samples, and to perform at least one round of model training on the initial model with the environmental evaluation result samples as supervisory signals to obtain the multi-parameter coupled model.

[0047] In an implementation, the model training module is further configured to label a cyclone phase corresponding to each input data sample in the training set, and to increase a training weight of input data samples in an active cyclone phase, the cyclone phase including a pre-cyclone phase, an active cyclone phase and a post-cyclone dissipation phase.

[0048] In an implementation, the model training module is further configured to extract daily parameter values corresponding to key regions on a cyclone path in the historical detection data day by day, to calculate a variation trend of the daily parameter values based on smoothing processing, and to calculate a peak lag period between different parameters based on a cross-correlation function.

[0049] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to the embodiments of the present specification are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in or transmitted by a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through a wired (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. integrated with one or more available media sets. The available media can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a digital versatile disc (DVD)), or a semiconductor medium (for example, a solid state disk (SSD)) and the like.

[0050] Figure 3 A block diagram of an electronic device 300 that can implement various embodiments of the present disclosure is shown. As shown, the electronic device 300 includes a processor 310, a disk drive 320, an input / output interface 330, a network interface 340, and a memory 350. The processor 310, the disk drive 320, the input / output interface 330, the network interface 340, and the memory 350 can be communicatively connected through a communication bus 360. Figure 3

[0051] The processor 310 can be implemented in the form of a general-purpose CPU, a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is configured to execute related programs to implement the technical solutions provided by the present application.

[0052] ​The memory 350 can be implemented in the form of a ROM (Read Only Memory), a RAM (Read Access Memory), a static memory, a dynamic memory device, etc. The memory 350 can store an operating system 351 for controlling the operation of the electronic device 300, a basic input / output system (BIOS) 352 for controlling the low-level operation of the electronic device 300. In addition, a web browser 353, a data storage management system 354, etc. can also be stored. In summary, when the technical solutions provided in the present application are implemented by software or firmware, the relevant program codes are stored in the memory 350 and are executed by the processor 310.

[0053] The input / output interface 330 is configured to connect an input / output module to realize information input and output. The input / output module can be configured as a component in the device (not shown in the figure) or can be externally connected to the device to provide corresponding functions. The input device can include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device can include a display, a vibrator, a prompt light, etc.

[0054] The network interface 340 is configured to connect a communication module (not shown in the figure) to realize the communication interaction between the device and other devices. The communication module can realize communication through a wired manner (such as USB, network cable, etc.) or through a wireless manner (such as mobile network, WIFI, Bluetooth, etc.).

[0055] The bus 360 includes a channel for transmitting information between various components (such as the processor 310, the disk drive 320, the input / output interface 330, the network interface 340, and the memory 350) of the device.

[0056] It should be noted that although the above device only shows the processor 310, the disk drive 320, the input / output interface 330, the network interface 340, the memory 350, the bus 360, etc., in the specific implementation process, the device can also include other components necessary for normal operation. In addition, those skilled in the art can understand that the above device can only contain the components necessary to implement the method of the present application, and does not necessarily contain all the components shown in the figure.

[0057] Program code for carrying out methods of the present disclosure can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, causes the machine to perform the functions / acts specified in the flowcharts and / or block diagrams. The program code can execute entirely on a machine, partly on a machine, as a stand-alone software package, partly on a machine and partly on a remote machine or entirely on a remote machine or server.

[0058] In the context of the present disclosure, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium would include one or more lines of a system, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. Further, while operations are depicted in a particular, sequential order, this should not be understood as requiring or implying that the operations are performed in the order shown or in sequential order, or that all illustrated operations are necessary for realizing the desired results. In certain circumstances, multitasking and parallel processing can be advantageous. Likewise, while specific implementations are discussed herein, the scope of the present disclosure is not limited to the specific details and representations herein. Certain features that are described in the context of separate embodiments can also be implemented in combination in a single embodiment. Conversely, various features that are described in the context of a single embodiment can also be implemented in multiple embodiments separately or in any suitable sub-combination.

[0059] Although the subject matter has been described in language specific to structural features and / or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as example forms of implementing the claims.

Claims

1. A method for assessing the impact of tropical cyclones on the marine environment, characterized in that, The method includes: Continuously acquire marine environmental parameters of the sea areas affected by tropical cyclones, and construct a multidimensional parameter dataset based on the marine environmental parameters, which include tropical cyclone data, sea surface temperature, chlorophyll a concentration, sea surface height, wind stress curl, and ocean vertical profile data; Derived features are calculated based on the multidimensional parameter dataset, including sea surface temperature gradient intensity calculated based on sea surface temperature and eddy category calculated based on sea surface height; The vortex category and cyclone intensity in the tropical cyclone data are used as classification features, and the data other than vortex category and cyclone intensity in the multidimensional parameter dataset and derived features are used as continuous features. The input data is constructed from the classification features and continuous features. Based on the input data, the trained multi-parameter coupled model outputs environmental assessment results. The multi-parameter coupled model includes a random forest module and an XGBoost module. The random forest module is used to determine the dominant factor among the continuous features according to the importance ranking of the contribution of continuous features to marine productivity. The XGBoost module is used to generate environmental assessment results based on the dominant factor. The environmental assessment results include the change value of the dominant factor within a preset time period in the future and the early warning information corresponding to the change value.

2. The method for assessing the impact of tropical cyclones on the marine environment according to claim 1, characterized in that, The tropical cyclone data includes the cyclone's life cycle, path, center location, intensity, and speed. The ocean vertical profile data includes seawater temperature, nutrient concentration, dissolved inorganic carbon concentration, and pH. The ocean environmental parameters also include surface wind field, mixing layer depth, and net heat flux.

3. The method for assessing the impact of tropical cyclones on the marine environment according to claim 1, characterized in that, The calculation of derived features based on the multidimensional parameter dataset includes: Calculate the sea surface temperature gradient in the east-west and north-south directions at each grid point, and calculate the sea surface temperature gradient intensity based on the two gradients. The vortex category of each identified region is determined based on the continuous comparison between sea surface height and a preset height threshold.

4. The method for assessing the impact of tropical cyclones on the marine environment according to claim 1, characterized in that, The method further includes: The strong upwelling region and the cooling core region are determined and marked by comparing the sea surface temperature gradient intensity with the preset gradient threshold. Verification of vortex categories based on wind stress curl.

5. The method for assessing the impact of tropical cyclones on the marine environment according to claim 1, characterized in that, The method further includes: Based on historical detection data, a training set is determined, which includes input data samples constructed from historical classification features and historical continuous features, as well as environmental assessment result samples. Based on the input data sample, the initial model generates predicted environmental assessment results; Using the environmental assessment result samples as supervision signals, the initial model is trained for at least one round to obtain a multi-parameter coupled model.

6. The method for assessing the impact of tropical cyclones on the marine environment according to claim 5, characterized in that, The method further includes: The cyclone phase corresponding to each input data sample is labeled in the training set, and the training weight of the input data samples in the active phase of the cyclone is increased. The cyclone phase includes the pre-cyclone phase, the active phase of the cyclone, and the post-cyclone dissipation phase.

7. The method for assessing the impact of tropical cyclones on the marine environment according to claim 5, characterized in that, The method further includes: Daily parameter values ​​corresponding to key areas along the cyclone path are extracted from historical monitoring data. The changing trend of daily parameter values ​​is calculated based on smoothing, and the peak lag period between different parameters is calculated based on cross-correlation function.

8. A device for assessing the impact of tropical cyclones on the marine environment, characterized in that, The device includes: The parameter acquisition module is configured to continuously acquire marine environmental parameters of the sea area affected by tropical cyclones, and construct a multidimensional parameter dataset based on the marine environmental parameters. The marine environmental parameters include tropical cyclone data, sea surface temperature, chlorophyll a concentration, sea surface height, wind stress curl, and ocean vertical profile data. The feature extraction module is configured to calculate derived features based on the multidimensional parameter dataset, the derived features including sea surface temperature gradient intensity calculated based on sea surface temperature and eddy category calculated based on sea surface height; The input data construction module is configured to use vortex category and cyclone intensity in tropical cyclone data as classification features, and data other than vortex category and cyclone intensity in multidimensional parameter dataset and derived features as continuous features, and construct input data from the classification features and continuous features; The prediction module is configured to output environmental assessment results from a trained multi-parameter coupled model based on the input data. The multi-parameter coupled model includes a random forest module and an XGBoost module. The random forest module is used to determine the dominant factor among the continuous features according to the importance ranking of the contribution of continuous features to marine productivity. The XGBoost module is used to generate environmental assessment results based on the dominant factor. The environmental assessment results include the change value of the dominant factor within a preset future time and the early warning information corresponding to the change value.

9. An electronic device, characterized in that, include: One or more processors, and a memory associated with the one or more processors, the memory being used to store program instructions that, when read and executed by the one or more processors, perform the steps of a method for assessing the impact of a tropical cyclone on the marine environment according to any one of claims 1-7.

10. A computer program product, characterized in that, The system includes a computer program that, when executed by a processor, implements a method for assessing the impact of tropical cyclones on the marine environment according to any one of claims 1-7.

Images