Tropical cyclone forecasting system and method

Abstract

A system and method for generating low-pressure system forecasts are provided. The method utilizes a de-cyclone process and a re-cyclone process to avoid washout associated with location divergence. During the de-cyclone process, low-pressure information is identified and extracted from multiple forecast models, thereby generating multiple discrete cyclone models and multiple associated background models. The background models are then blended and the discrete cyclone models are weighted, thereby generating a blended background model and a weighted discrete cyclone model, respectively. During the re-cyclone process, the weighted discrete cyclone model is added to the blended background model, thereby generating a forecast representing the amplitude associated with the tropical cyclone.

Description

Technical Field

[0001] Cross - Reference to Related Applications

[0001] This application claims priority based on U.S. Provisional Patent Application No. 63 / 524,509, filed on June 30, 2023, the entire content of which is incorporated herein by reference.

[0002] Technical Field

[0002] The present invention generally relates to tropical cyclone prediction. More specifically, the present invention relates to an objective method for applying a deep learning model for deterministic prediction of the complete spatio - temporal fields of wind, pressure, and rainfall associated with tropical cyclones.

Background Art

[0003] Background

[0003] Many existing weather prediction engines are built for multi - model ensemble blends, with the blend itself involving sophisticated weighting and sophisticated bias correction. Unfortunately, with existing blending techniques, large disturbances in the prediction are washed out. While perhaps satisfactory for certain prediction purposes such as temperature prediction that minimizes the root mean square error, the amplitude of the disturbance becomes important when the prediction involves extreme events such as hurricanes, cyclones, typhoons (each a "tropical cyclone"). Therefore, it would be beneficial to have blending techniques that do not wash out large disturbances.

[0004]

[0004] In the case of various existing prediction models, specific important features of individual tropical cyclone prediction models are spatially dispersed over time, and as a result, the blend leads to blurring, attenuation, or other distortions (each an "aberration") of such important features in the final blended tropical cyclone prediction model. This is particularly problematic in situations where the prediction branches, such as the exemplary multi - model prediction tracks of Hurricane Sandy shown in FIGS. 1A - 1H. Therefore, it would be beneficial to have modeling techniques that adjust for spatial divergence to avoid aberrations in the modeled tropical cyclone.

[0005]

[0005] Referring to Figure 2, an example of a distorted forecast is provided. In particular, Figure 2 shows a Global Forecast System ("GFS") forecast model showing two separate eyes, one of which is shifted 200 miles west from the other. As can be seen, the overlapping eyes are the result of blending two models with different predicted locations and / or trajectories for a particular tropical cyclone. In addition to showing two eyes, the effect of blending these two modeled cyclones into a single forecast model is a broadened, weakened, and distorted wind field that does not accurately reflect the modeled wind field associated with either model. This significant distortion of the wind field in this blend has a direct impact on the forcing in wave models employing the blended forecast. Therefore, it would be beneficial to have modeling techniques that eliminate or otherwise reduce the distortion to the associated wind field.

[0006]

[0006] As the forecast period lengthens, the spatial divergence of the forecast model tends to increase, so existing blending techniques have limited value for forecasts beyond several days or hours. However, preparation for tropical cyclones often requires prior notification of the level of danger and the expected trajectory of the cyclone. Therefore, it would be beneficial to have a blending technique that facilitates long-term forecasting of tropical cyclones.

[0007]

[0007] Existing forecasting systems rely on certain important information provided by existing software, such as the Tropical Cyclone Forecasting Automated System (hereinafter referred to as "ATCF" respectively). Unfortunately, precipitation is not a parameter described in any way in the ATCF. For example, there is no summary or reduced dimension of this parameter assigned to a particular tropical cyclone. Therefore, precipitation must be extracted from the entire grid of the forecasting model guidance. The challenge then is to identify tropical cyclones within these grids. Insofar as existing systems can extract precipitation from the entire grid of the forecasting model guidance, such extraction is time-consuming and unreliable. Therefore, it would be beneficial to have a system and method for obtaining tropical cyclone precipitation information quickly and reliably. [Overview of the Initiative] [Means for solving the problem]

[0008] overview

[0008] The present invention includes a system and method for blending conventional prediction models. The system and method utilize a special modeling technique that is more suitable for predicting tropical cyclones compared to the modeling techniques of existing systems. In particular, the blending technique of the present invention does not wash away large disturbances associated with tropical cyclones.

[0009]

[0009] In some embodiments, the present invention adjusts spatial divergence and thereby avoids associated distortion of the modeled tropical cyclone. In some such embodiments, the cyclone element is removed from the model before blending and added back to the model after blending.

[0010]

[0010] Referring to Figure 3, a high-level flowchart of one embodiment of the present invention is provided. In the provided flowchart, each of the extracted cyclone model and background model is generated for each of the two guidance models. The background model is then blended to form a blended decyclone model, and the extracted cyclone model is weighted to generate a weighted cyclone model. Finally, the weighted cyclone model is combined with the blended decyclone model to form a blended recyclone model. Thus, the present invention eliminates or otherwise reduces distortions such as strains on the wind field.

[0011]

[0011] In some embodiments, the present invention includes a system and method for rapidly and reliably obtaining precipitation information. In some such embodiments, the present invention utilizes one or more neural networks, such as U-Net, to detect and process information. In some embodiments, the neural network is used to detect information about a specific field, such as a precipitation field or a wind field. In some embodiments, the neural network is trained on millions of images of the reanalyzed field, for example, by labeling images using an objective tropical cyclone database. The neural network is then used to identify tropical cyclones in each model incorporated into a forecasting engine blend, such as a global forecasting engine blend. The tropical cyclones found in each model forecast are extracted, grouped, and weighted according to a particular method of the present invention, for example, by comparing the location of each extracted cyclone against a trajectory forecast related to the present invention. Each extracted cyclone is then recentered onto the trajectory forecast before being blended with other extracted cyclones. The application of neural networks to generate deterministic tropical cyclone forecasts provides unparalleled advantages in both the impact of tropical cyclones and the accuracy and realism of tropical cyclones within the forecasting engine blend.

[0012]

[0012] The foregoing and other objects are intended to illustrate the present invention and are not intended to be limiting. Many possible embodiments of the present invention can be made, which will be readily apparent from the following specification and the accompanying drawings, which include portions thereof. Various features and subcombinations of the present invention can be utilized without reference to other features and subcombinations. Other objects and advantages of the present invention will be apparent from the following description, embodiments of the present invention, and various features thereof, which will be interpreted in relation to the accompanying drawings described as examples and embodiments.

[0013] Simple explanation

[0013] Embodiments of the present invention will be described in detail below with reference to the attached drawings. [Brief explanation of the drawing]

[0014] [Figure 1A]

[0014] This is an image of past multi-model prediction trajectories. [Figure 1B]

[0014] This is an image of past multi-model prediction trajectories. [Figure 1C]

[0014] This is an image of past multi-model prediction trajectories. [Figure 1D]

[0014] This is an image of past multi-model prediction trajectories. [Figure 1E]

[0014] This is an image of past multi-model prediction trajectories. [Figure 1F]

[0014] This is an image of past multi-model prediction trajectories. [Figure 1G]

[0014] This is an image of past multi-model prediction trajectories. [Figure 1H]

[0014] This is an image of past multi-model prediction trajectories. [Figure 2]

[0015] This is an image of a predictive model showing two distinct eyes. [Figure 3]

[0016] This is a high-level flowchart of one embodiment of the present invention. [Figure 4]

[0017] It is a simplified graphical representation 40 of the radial wind and the vorticity of a tropical depression. [Figure 5]

[0018] It is an image of a single model prediction showing the modeled wind field. [Figure 6]

[0019] It is an image of a single model prediction of FIG. 5 showing the vorticity field. [Figure 7]

[0020] It is an image showing the vorticity field of FIG. 6 after applying the window function of the present invention. [Figure 8]

[0021] It is an image showing the discrete wind field of a tropical depression. [Figure 9]

[0022] It is an image showing the background wind field of the present invention. [Figure 10]

[0023] It is an image of a graphical polar coordinate representation of wind speeds at three radii for each of the four quadrants of a tropical depression. [Figure 11]

[0024] It is an image of a graphical polar coordinate representation of wind speeds at various radii for each of the four quadrants of a tropical depression. [Figure 12]

[0025] It is an image of a graphical Cartesian representation of the wind speed of a tropical depression. [Figure 13]

[0026] It is an image of the wind speed of FIG. 12 centered at a certain latitude and longitude. [Figure 14]

[0027] It is an image showing the representative prediction ability of the present invention in contrast to the prediction ability of existing public predictions. [Figure 15]

[0028] It is an image of a graphical representation of the wind speed and pressure field. [Figure 16]

[0029] It is an image of a graphical representation of a neural network related to a specific embodiment of the present invention. [Figure 17]

[0030] It is a figure including a plurality of images, each image including a graphical representation of the precipitation rate field.

MODE FOR CARRYING OUT THE INVENTION

[0015]

[0031] The drawings do not limit the present invention to the specific embodiments disclosed and described herein. The drawings are not necessarily to scale and instead focus on illustrating the principles of the present invention.

[0016] Detailed explanation

[0032] Where necessary, detailed embodiments of the present invention are disclosed herein, but it should be understood that the disclosed embodiments are merely illustrative of the principles of the present invention, which can be embodied in various forms. Accordingly, the specific structural and functional details disclosed herein should not be construed as restrictive, but merely as a basis for the claims and as a representative basis for teaching those skilled in the art to use the present invention in various ways in substantially any appropriately detailed structure.

[0017]

[0033] Referring to Figure 4, a simplified graphical representation of a tropical cyclone 40 includes a first curve 42 and a second curve 44 representing radial wind force and vorticity, respectively. More specifically, the first curve 42 represents the radial wind speed of the tropical cyclone, which increases linearly from zero at the center ("eye") to a maximum value at a given radius ("eye wall") 45, and then asymptotically returns to zero beyond the eye wall 45. The second curve 44 represents vorticity, which is an indicator of the amount of rotation generated by horizontal wind shear at a given point. As shown in Figure 4, the second curve 44 is a step function defined by a constant peak value inside the eye wall 45 and a zero value beyond the eye wall 45. By leveraging this fundamental understanding, the tropical cyclone wind model can be effectively removed from each forecast model before the forecast models are blended (decycloning).

[0018]

[0034] Decyclone doesn't simply use an eraser to model a grid. Instead, it leaves a physically realistic background wind field. Winds from tropical cyclones extend radially outward, sometimes for hundreds of miles, superimposed on the background flow, making it difficult to directly unravel the background flow from the radial winds of the tropical cyclone. Vorticity does not suffer from the same problem, as will be explained in more detail below.

[0019]

[0035] Referring to Figure 5, which shows a single model GFS forecast with a modeled wind field, it is observed that the tropical cyclone affects the portion of the modeled wind field located within the first region 52. The model also shows a boundary element 54 representing a cold front boundary. Since the boundary element 54 is located outside the first region 52, the tropical cyclone is expected to have minimal direct influence on the cold front boundary. Therefore, the decycloning process removes the influence of the tropical cyclone on the modeled wind field while leaving the model of the cold front boundary unchanged.

[0020]

[0036] Referring to Figure 6, the vorticity calculated from the wind field in Figure 5 is shown. As can be seen, the vorticity in the first region 52 is densely concentrated within the core of the modeled tropical cyclone, thereby facilitating the identification of vorticity associated with the modeled tropical cyclone. In some embodiments, identifying vorticity involves finding the maximum vorticity and applying a cosine tapered window radially outward. Referring to Figure 7, this window leaves only vorticity in the vicinity of the tropical cyclone.

[0021]

[0037] Referring to Figure 8, the discrete wind field of a tropical cyclone can be determined by transforming the vortex field into global spectral space. In some such embodiments, transforming the vortex field into global spectral space involves integrating in spectral space and then transforming back to generate the discrete wind field. Referring to Figure 9, the background wind field can be determined by subtracting the discrete wind field from the modeled wind field, thereby completing the decyclonic process for a particular model. In some embodiments, the presence of multiple modern tropical cyclones requires performing the decyclonic process for each tropical cyclone, thereby generating the background wind field from which each cyclone has been removed. In some such embodiments, each discrete wind field is subtracted simultaneously from the modeled wind field. In other embodiments, one or more intermediate wind fields are generated before subtracting the final discrete wind field to obtain the background wind field. In some embodiments, each model is associated with one or more relevant time periods related to the tropical cyclone.

[0022]

[0038] The decyclonic process is applied to each of several models to prepare them for blending. After blending the decyclonic models, the resulting blended model ("blended decyclonic model") is ready to be re-cyclonic with the cyclonic elements added back ("re-cyclonic"). In some embodiments, "re-cyclonic" involves generating a deterministic tropical cyclone forecast ("discrete cyclonic model") that has all the spatiotemporal parameters that can be superimposed on the official forecast grid of the blended decyclonic model.

[0023]

[0039] In some embodiments, the generation of discrete cyclone models begins by reading ATCF model guidance information about the tropical cyclone, such as information related to the cyclone's trajectory, intensity, and size. Depending on the cyclone, the volume of the model ensemble can be very large, and in some cases, it may include up to 120 different models (each a “guidance model”). Trajectory predictions are first generated using a weighted blend of at least some of the available guidance models in the ensemble. The weights are pre-seeded using a random forest deep learning technique that examines the performance of historical models, such as historical models of cyclones over the past five years. These optimized pre-seeded weights are then updated at each lead time for a given cyclone, based on agreement with official public forecasts, such as those from the National Hurricane Center (NHC) or the Joint Typhoon Warning Center (JTWC). The weighted model predicted trajectories are then blended together for all available time steps, allowing the cyclone forecasts of the present invention to be extended over longer periods, such as seven days. This methodology generates forecasts that do not contradict official public forecasts, but adds greater solutions, information, and value to what has been available in tropical cyclone forecasting to date.

[0024]

[0040] In some embodiments, the present invention utilizes maximum sustained winds, minimum central pressure, and radii of 34 knots, 50 knots, and 64 knots in each of the four quadrants of the tropical cyclone. In some embodiments of the present invention, at least some of this information is used to generate optimal deterministic predictions.

[0025]

[0041] As discussed in more detail above, in addition to determining the wind field, some models of the present invention include the total spatial field of mean sea level pressure and precipitation. In some embodiments, pressure parameters are important for driving the wind and cyclone surge models. In some such embodiments, the wind and cyclone surge models are combined with other information, such as the wind field, to create a complete tropical cyclone forecasting package.

[0026]

[0042] Model guidance available in ATCF messages is also conveniently available in the archive. Therefore, some embodiments of the present invention utilize machine learning models trained on the performance of prediction models for past tropical cyclones. In some embodiments, a random forest is trained on five years of tropical cyclones using all archived prediction models available in ATCF. In some embodiments, the training is validated using secondary information sources such as the International Best Track Archive for Climate Stewardship (IBTrACS). The objective of the random forest is not to make predictions themselves, but rather to select a prediction model that is expected to perform best in modeling the current tropical cyclone. Inputs include the location and basin of the current tropical cyclone, velocity and direction, and intensity. The results of the random forest are stochastic on a scale of 0 to 1 and are then used for initial seeding of the weights of each prediction model.

[0027]

[0043] In some embodiments, the construction of a specific field is achieved in parallel with removing tropical cyclone attributes from each of the models contributing to the prediction engine blend, thereby leaving background conditions. In the extracted cyclone shadow, the precipitation field is virtually zero, but the wind field may have a significant background component. In some embodiments, the center of the tropical cyclone is identified by a cyclone tracking algorithm, and a cosine tapered window is applied to the 10m vorticity and divergence fields around its center. The window is then removed, and the remaining wind is obtained by integration using Fourier decomposition. These resulting winds represent the background flow.

[0028]

[0044] In some embodiments, extracted cyclone-related information, such as information related to a specific field, is superimposed on top of the blended background model, thereby reintroducing the tropical cyclone. Some embodiments of the present invention are configured to reduce sharp edges in the model. In some embodiments, some vorticity is added back during the decycloning process to prevent sharp edges between the cyclone field and the predicted blend and to retain features arising from extratropical interactions. In some such embodiments, the vorticity added to the cyclone prediction center over a given effective time is proportional to the amount of vorticity initially removed and tapers away from the center. The result is a cyclone wind field that blends into the background flow behind the tropical cyclone field, as expected beyond the outer radius of the tropical cyclone storm area.

[0029]

[0045] In certain examples of the present invention, the first forecast point for a tropical cyclone is generated using a weighted blend with initial weightings. In some such examples, confirmation forecasts, such as those from the NHC, JTWC, or other government agencies and / or public forecasts, are then used, as in observations in data assimilation. The weights of each forecast model are shifted higher or lower based on how close the model trajectory of the forecast model is to the trajectory of the confirmation forecast, rewarding models that are closer and penalizing models that are further away. Thus, the forecasts of the present invention are consistent with the confirmation forecasts. In some embodiments, the present invention applies the same weighted blend to wind radius and higher temporal resolution information provided in the ATCF. This is particularly important in situations of rapid intensification and landfall.

[0030]

[0046] Referring to Figure 10, some embodiments of the present invention utilize wind speed radius information available in four quadrants (northeast, southeast, southwest, and northwest) for three thresholds (34, 50, and 64 knots). In some such embodiments, each of these 12 data points (101-112) is used to form a complete wind field, starting with interpolation in polar coordinates.

[0031]

[0047] Referring to Figure 11, some embodiments of the present invention add a 13th point 113 to represent the maximum sustained wind. In some embodiments, the azimuth angle is determined by a sine fit of the maximum available wind radius threshold. In some such embodiments, the maximum wind is located where the wind radius is maximum in the sine fit. In some embodiments, the radius of the maximum wind is determined by a vorticity constraint. In some embodiments, the remainder of the interpolated points are ramped up to the maximum wind using cosine smoothing with a width determined by the amplitude of its initial sine fit.

[0032]

[0048] Referring to Figure 12, in some embodiments, the wind field is then interpolated across a high-resolution grid of polar coordinates and subsequently transformed into Cartesian coordinates, where the wind direction is determined at a predetermined convergence angle by an implied cyclone rotation. Referring to Figure 13, a final step in some embodiments of the present invention includes transforming the wind field in Cartesian coordinates into a forecasting engine grid. Referring to Figure 13, tropical cyclone information can be centered using latitude and longitude information.

[0033]

[0049] Referring to Figure 14, some embodiments of the present invention, such as the embodiment identified in Figure 14 as "DTN's Best Tropical Forecast," provide information consistent with confirmed forecasts, such as confirmed forecasts identified in Figure 14 as "Official Public Forecast," while adding significant value to confirmed forecasts by providing trajectories and fully formed wind fields with time-level resolution for periods exceeding the period from which period information can be obtained from confirmed forecasts (e.g., 7 days).

[0034]

[0050] Some embodiments of the present invention include a fully gridded wind field. In some such embodiments, sea level pressure (SLP) is added. SLP can be an important parameter for driving wave and cyclonic surge models. In some embodiments, SLP predictions are developed using a gradient wind balance approximation for atmospheric horizontal force balance. The gradient wind balance employs a geometric wind balance of Coriolis and pressure gradient forces, setting it equal to centrifugal force. In some embodiments, such information is an important addition for strong curvature flows, such as those found in tropical cyclones. By solving for pressure gradient forces or geoelectric potential gradient, the following is obtained:

number

[0035]

[0051] In Equation 1, V is the wind speed, R is the wind radius, and f is the Coriolis parameter. The right side of Equation 1 is the geoelectric potential gradient to the right of the wind. For circulating flow, the right side is simplified to the radial gradient. Integrating Equation 1 with respect to R and setting the baseline constant (R=0) equal to the minimum central pressure (an output prediction parameter from multiple ATCF models) yields the total pressure field in polar coordinates. Referring to Figure 15, the information is converted to Cartesian coordinates and then added to the prediction engine grid.

[0036]

[0052] In some embodiments, the present invention utilizes machine learning techniques to acquire information such as precipitation information. In some such embodiments, the machine learning technique utilizes a neural network, such as a U-net, in which a decoder network is followed by an encoder network. Referring to Figure 16, the encoder downsamples using a series of convolutional neural networks and pooling to identify features in the input array. The decoder uses a transposed convolutional network with upsampling to identify where those features reside in the input. The result is an array the same size as the input, containing the probability of each pixel or grid cell being classified. This result is called semantic segmentation and provides a high level of granularity for understanding the image. Prior to this level, there are classifications where objects in the image are identified, localizations where boxes are drawn around those objects, and detections where multiple objects in the image are localized. In some embodiments, the neural network requires only a very small number of annotated images. In some embodiments, data augmentation with elastic transformations is used to significantly increase the training sample size. In some embodiments, each pixel related to a tropical cyclone is labeled.

[0037]

[0053] In some embodiments, the neural network is trained with extensive background information, such as by training on five-year reanalysis hourly precipitation rate images within a 15°×15° box, where 30% of the images contain tropical cyclones and 70% of the images are used as null samples. In some embodiments, the tropical cyclones used for training are identified using the IBTrACS database. Referring to Figure 17, a specific example utilizes six layer depths, 37 epochs, and a focus loss function. In the example represented by Figure 17, six different precipitation forecasts are taken up by the neural network, and an output segmentation array is returned with TC and non-TC labels.

[0038]

[0054] Continuing to refer to Figure 17, which shows the field of hourly precipitation rates associated with tropical cyclones, advanced blending techniques can be used to generate definitive tropical cyclone precipitation forecasts. First, each model is assigned a weight determined by the distance in the valid time between the extracted cyclone center and the confirmed forecast cyclone model center. The closer these forecast centers are, the greater the weight given to the model. More specifically, the weights have an inverse exponential relationship with the distance between the centers. In some embodiments, the tropical cyclone precipitation shield may appear substantially different for each model depending on its proximity to extratropical features such as the frontal boundary or jet stream, and this weighting technique helps ensure that the resulting precipitation image also coincides with the trajectory. In some embodiments, new weights are calculated for each valid time.

[0039]

[0055] In some embodiments, precipitation images from each model are recentered to the cyclone center. In some such embodiments, each grid point labeled as a tropical cyclone is repositioned relative to the extracted cyclone center based on its position relative to the confirmed forecast cyclone model center. Finally, the weighted and recentered images are blended for the official precipitation forecast. Referring to Figure 18, the present invention facilitates the preservation of the amplitude and other features of the tropical cyclone (represented by the right side of Figure 18) better than blends available using prior art systems and methods (represented by the left side of Figure 18).

[0040]

[0056] Various embodiments of the present invention utilize, encompass, or are facilitated by computer programs, devices, systems, methods, etc. Various embodiments of computer programs, devices, systems, and methods are implemented in hardware, software, firmware, or combinations thereof, such as by using a central management system (e.g., TCS or other central computer control system), which broadly includes server devices, computing devices, communication networks, etc. Various embodiments of server devices include computing devices that provide access to one or more common computing resources, such as internet services, email services, and data transfer services. In some embodiments, the server device also provides access to a database that stores information and data, including, but not limited to, system user information (e.g., project ID, account number, etc.) or other information and data necessary and / or desirable for implementing the computer programs, devices, systems, and methods of the present invention.

[0041]

[0057] Various embodiments of server devices and computing devices include any device, component, or apparatus having processing elements and associated memory elements. In some embodiments, the processing elements implement an operating system and, in some such embodiments, are capable of executing computer programs, also commonly known as instructions, commands, software code, executable data, applications (apps), etc. In some embodiments, the processing elements include processors, microprocessors, microcontrollers, field-programmable gate arrays, etc., or combinations thereof. In some embodiments, the memory elements are capable of storing or holding computer programs and, in some such embodiments, also store binary data, typically including data, text, databases, graphics, audio, video, combinations thereof, etc. In some embodiments, the memory elements are also known as "computer-readable storage media" and, in some such embodiments, include random-access memory (RAM), read-only memory (ROM), flash drive memory, floppy disks, hard disk drives, optical storage media such as compact discs (CDs or CD-ROMs), digital video discs (DVDs), Blu-ray®, etc., or combinations thereof. In addition to these memory elements, in some embodiments, the server device further includes a file store comprising multiple hard drives, network-attached storage, or a separate storage network.

[0042]

[0058] Various embodiments of computing devices include, in particular, mobile communication devices (including wireless devices), workstations, desktop computers, laptop computers, palmtop computers, tablet computers, portable digital assistants (PDAs), smartphones, wearable devices, or combinations thereof. Various embodiments of computing devices also include voice communication devices such as mobile phones or landline phones. In some preferred embodiments, the computing device has an electronic display capable of displaying visual graphics, images, text, etc. In certain embodiments, the computer program of the present invention facilitates interaction and communication via a graphical user interface (GUI) displayed via the electronic display. The GUI allows the user to interact with the electronic display by touching or pointing to a display area to provide information to the user control interface, which will be described in more detail below. In additional preferred embodiments, the computing device includes optical devices such as digital cameras, video cameras, and optical scanners, so that the computing device can capture, store, and transmit digital images and / or videos, barcodes, or other identifying information.

[0043]

[0059] In some embodiments, the computing device includes a user control interface that enables one or more users to share information and commands with the computing device or server device. In some embodiments, the user interface facilitates interaction via the GUI described above, or in other embodiments, includes one or more functional inputs such as buttons, keyboards, switches, scroll wheels, voice recognition elements such as microphones, and pointing devices such as mice, touchpads, tracking balls, and styluses. Embodiments of the user control interface also include a speaker for providing audible commands and feedback. Furthermore, embodiments of the user control interface enable users and / or other computing devices to interface remotely with the computing device. Includes wired or wireless data transfer elements such as communication components, removable memory, data transceivers, and / or transmitters.

[0044]

[0060] In various embodiments, the communication network may be wired, wireless, and / or a combination thereof, and may include servers, routers, switches, wireless receivers, and transmitters, as well as conductive cables or optical cables. In various embodiments, the communication network may also include local, metro, or wide-area networks, as well as the Internet or other cloud networks. Furthermore, some embodiments of the communication network may include cellular or mobile phone networks, as well as fixed-line telephone networks, public switched telephone networks, fiber optic networks, and the like.

[0045]

[0061] Various embodiments of both server devices and computing devices are connected to a communication network. In some embodiments, a server device communicates with other server devices or computing devices via the communication network. Similarly, in some embodiments, a computing device communicates with other computing devices or server devices via the communication network. In various embodiments, the connection to the communication network is wired, wireless, and / or a combination thereof. Therefore, server devices and computing devices will include appropriate components for establishing wired or wireless connections.

[0046]

[0062] Various embodiments of the computer program relating to the present invention are executed on a computing device. In other embodiments, the computer program relating to the invention is executed on one or more server devices. In addition, in some embodiments, a first part of the program, code, or instructions is executed on a first server device or a first computing device, and a second part of the program, code, or instructions is executed on a second server device or a second computing device. In some embodiments, other parts of the program, code, or instructions are also executed on other server devices. For example, in some embodiments, information is stored in a memory element associated with a server device, and as a result, the information is remotely accessible to a user of the computer program via one or more computing devices. Alternatively, in other embodiments, the information is stored directly in a memory element associated with one or more computing devices of the user. In additional embodiments of the present invention, some of the information is stored on a server device, while other parts are stored on one or more computing devices. In some embodiments, various actions and calculations described herein as being performed by or using a computer program will actually be performed by one or more computers, processors, or computing devices and / or other computing devices such as server devices, and will be understood to be performing parts of the computer program independently or in cooperation with each other.

[0047]

[0063] Various embodiments of the present invention are accessible via electronic resources such as applications, mobile "apps," or websites. In certain embodiments, a portion of the computer program is embodied as a standalone program downloadable to a user's computing device, or as a web-accessible program accessible by a user's computing device over a network. In some embodiments of a standalone program, a downloadable version of the computer program is stored, at least partially, on a server device. The user downloads at least a portion of the computer program onto their computing device over a network. After the computer program is downloaded, it is installed on the computing device in an executable format. In some embodiments of a web-accessible computer program, the user simply accesses the computer program using their computing device over a network (e.g., the Internet).

[0048]

[0064] In this specification, any reference to “one embodiment” or “embodiment” means that the features being referred to are included in at least one embodiment of the Art. Different references to “one embodiment” or “embodiment” in this specification do not necessarily refer to the same embodiment, and are not mutually exclusive unless otherwise stated and / or are readily apparent to those skilled in the art from the description. For example, features, structures, operations, etc., described in one embodiment may, but are not necessarily, be included in other embodiments. Therefore, various embodiments of the Art include various combinations and / or integrations of the embodiments described herein.

[0049]

[0065] In the foregoing description, certain terms are used for brevity, clarity, and understanding; however, such terms are used for illustrative purposes and are intended to be interpreted broadly, so that no unnecessary limitations beyond the requirements of the prior art should be implied therefrom. Furthermore, the description and examples of the present invention are illustrative, and the scope of the invention is not limited to the exact details shown or described.

[0050]

[0066] The above detailed description of the present invention is provided by reference to exemplary embodiments, illustrating the best mode intended for carrying out the invention, but it will be understood that certain changes, modifications, or variations may be made in embodying the invention as described above. Furthermore, in its configuration, anything other than that specifically described herein can be achieved by those skilled in the art without departing from the spirit and scope of the invention, and such changes, modifications, or variations should be considered within the overall scope of the invention. Accordingly, it is intended to encompass any and all changes, modifications, variations, or equivalents that fall within the true spirit and scope of the invention and the underlying principles disclosed and claimed herein. Accordingly, the scope of the invention is intended to be limited only by the appended claims, and all matters included in the above description and shown in the appended drawings should be construed as illustrative and not restrictive.

[0051]

[0067] While the features, discoveries, and principles of the present invention, the methods and uses of the present invention, the structural features, and the advantageous, novel, and useful results obtained have been described here, novel and useful structures, devices, elements, arrangements, components, and combinations are described in the appended claims.

[0052]

[0068] Furthermore, it should be understood that the following claims are intended to encompass all the general and specific features of the invention described herein, as well as all descriptions of the scope of the invention that may lie between them in terms of language.

Claims

1. A method for blending multiple predictive models, The method involves generating multiple discrete cyclone models, wherein each discrete cyclone model is generated from the respective prediction models of the multiple prediction models. Each discrete cyclone model is extracted from its respective prediction model, and thereby a background model is generated for each prediction model. Blending the aforementioned background models to generate a blended background model, Adding a weighted cyclone model to the blended background model, wherein the weighted cyclone model is generated using information from at least one of the plurality of discrete cyclone models, Methods that include...

2. The method according to claim 1, wherein each prediction model includes a modeled wind field, each discrete cyclone model includes a discrete wind field, and each background model includes a background wind field determined by subtracting the discrete wind field from the modeled wind field.

3. The ability to generate each discrete cyclone model is Calculating the vorticity of the modeled wind field, To find the maximum vorticity associated with the first tropical depression, A cosine tapered window is applied radially outward from the maximum vorticity, thereby separating the vorticity near the tropical cyclone from any other potential vorticities associated with each of the modeled wind fields. The separated vorticity is converted into a global spectral space, The method according to claim 2, including the method described in claim 2.

4. Identify the geographical center of each extracted discrete cyclone, thereby identifying the extracted center of each background model. The first vorticity quantity is added to each background model, wherein the added vorticity is centered around the respective extraction center of each background model. It further includes, Each first vorticity is proportional to the vorticity of its respective extracted discrete cyclone model. The method according to claim 3, wherein the added vorticity for each background model tapers as it moves away from each extraction center.

5. The method according to claim 3, further comprising generating the weighted cyclone model using information from each of the plurality of discrete cyclone models, wherein the values ​​of each discrete cyclone model are organized based on the position of each value relative to the center of each discrete cyclone model, and the weights of each discrete cyclone model are determined based on the distance of the center of each discrete cyclone model from the center of the weighted cyclone model.

6. The method according to claim 5, wherein the weights of each discrete cyclone model are determined using an inverse exponential relationship between the distance of the center of each discrete cyclone model from the center of the weighted cyclone model and the distance of the center of each discrete cyclone model from the center of the weighted cyclone model.

7. The method according to claim 6, wherein the weight is 1 for each discrete cyclone model that is concentric with the weighted cyclone model.

8. The method according to claim 7, wherein the center of the weighted cyclone model is determined using information from the confirmed forecast cyclone model.

9. The method of claim 8, further comprising using a neural network to identify low-pressure information for each prediction model, and extracting each discrete low-pressure model from each prediction model, comprising extracting the identified low-pressure information from each prediction model.

10. The method according to claim 9, wherein the identified low-pressure information is precipitation information.

11. The method according to claim 1, further comprising using a neural network to identify low-pressure information for each prediction model, extracting each discrete low-pressure model from each prediction model, comprising extracting the identified low-pressure information from each prediction model, wherein the identified low-pressure information is precipitation information.

12. A method for generating predictions of tropical cyclones, Using a neural network to identify low-pressure information for each of multiple guidance models, Extract the low-pressure information from each guidance model, thereby generating a background model for each guidance model. Processing the background model in order to generate a blended background model, Adding weighted low-pressure information to the first region of the blended background model, Methods that include...

13. The method according to claim 12, wherein the low-pressure information includes at least one of precipitation information, low-pressure surge information, and wind information.

14. The method according to claim 13, wherein the low-pressure information includes precipitation information, and the precipitation is substantially zero in the first region of the blended background model before the weighted low-pressure information is added to the blended background model.

15. The method according to claim 14, wherein the low-pressure information includes low-pressure surge information, and the low-pressure surge information is substantially zero in the first region of the blended background model before the weighted low-pressure information is added to the blended background model.

16. The method according to claim 15, wherein the low-pressure information includes wind information, and the wind information is not equal to zero in the first region of the blended background model before the weighted low-pressure information is added to the blended background model.

17. A system for generating tropical cyclone forecasts, A decyclonic module for identifying and extracting low-pressure information from multiple prediction models, thereby generating multiple background models, A blend module for blending the aforementioned background models and thereby generating a blended background model, A re-cyclone module for adding a weighted cyclone model to the blended background model, A system equipped with these features.

18. The system according to claim 17, comprising a weighting module for generating a weighted cyclone model using information from each of the plurality of discrete cyclone models, wherein the values ​​of each discrete cyclone model are organized based on the position of each value relative to the center of each discrete cyclone model, and the weight of each discrete cyclone model is determined based on the distance of the center of each discrete cyclone model from the center of the weighted cyclone model.

19. The system according to claim 18, wherein the weights of each discrete cyclone model are determined using an inverse exponential relationship between the distance of the center of each discrete cyclone model from the center of the weighted cyclone model and the distance of the center of each discrete cyclone model from the center of the weighted cyclone model.

20. The system according to claim 19, wherein the weight is 1 for each discrete cyclone model that is concentric with the weighted cyclone model.

Images