CN117907965B - Three-dimensional radar echo proximity forecasting method for convection storm fine structure - Google Patents
Three-dimensional radar echo proximity forecasting method for convection storm fine structure Download PDFInfo
- Publication number
- CN117907965B CN117907965B CN202410308899.7A CN202410308899A CN117907965B CN 117907965 B CN117907965 B CN 117907965B CN 202410308899 A CN202410308899 A CN 202410308899A CN 117907965 B CN117907965 B CN 117907965B
- Authority
- CN
- China
- Prior art keywords
- dimensional radar
- radar echo
- dimensional
- loss function
- resolution
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000013277 forecasting method Methods 0.000 title claims abstract description 9
- 238000002310 reflectometry Methods 0.000 claims abstract description 67
- 238000004458 analytical method Methods 0.000 claims abstract description 41
- 238000000034 method Methods 0.000 claims abstract description 38
- 238000003908 quality control method Methods 0.000 claims abstract description 22
- 230000004927 fusion Effects 0.000 claims abstract description 17
- 230000006870 function Effects 0.000 claims description 55
- 238000012549 training Methods 0.000 claims description 24
- 238000004891 communication Methods 0.000 claims description 14
- 238000013528 artificial neural network Methods 0.000 claims description 8
- 238000013135 deep learning Methods 0.000 claims description 7
- 230000006855 networking Effects 0.000 claims description 5
- 230000007246 mechanism Effects 0.000 claims description 4
- 230000010365 information processing Effects 0.000 abstract description 2
- 238000004364 calculation method Methods 0.000 description 10
- 239000010410 layer Substances 0.000 description 10
- 238000004590 computer program Methods 0.000 description 5
- 238000013461 design Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 230000003287 optical effect Effects 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 238000013142 basic testing Methods 0.000 description 2
- 238000013136 deep learning model Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000013213 extrapolation Methods 0.000 description 2
- 230000005484 gravity Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000001556 precipitation Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000002356 single layer Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
- G01S13/95—Radar or analogous systems specially adapted for specific applications for meteorological use
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01W—METEOROLOGY
- G01W1/00—Meteorology
- G01W1/10—Devices for predicting weather conditions
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Physics & Mathematics (AREA)
- Environmental & Geological Engineering (AREA)
- General Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Atmospheric Sciences (AREA)
- Electromagnetism (AREA)
- Environmental Sciences (AREA)
- Ecology (AREA)
- Biodiversity & Conservation Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention relates to the technical field of weather radar information processing, and provides a three-dimensional radar echo proximity forecasting method for a convection storm fine structure, which comprises the following steps: basic reflectivity factor data of an analysis area are obtained, and a high-resolution three-dimensional radar echo data set of the analysis area is established through quality control and interpolation; according to the high-resolution three-dimensional radar echo data set, adopting SwinURNN D fusion frames to train to obtain an optimal three-dimensional radar echo proximity prediction model; and taking the closest three-dimensional radar reflectivity data sequence as an optimal three-dimensional radar echo proximity prediction model to input and output a predicted three-dimensional radar reflectivity factor sequence. The method establishes a three-dimensional radar echo data set containing multiple vertical layers as model input, and solves the problem that single-level radar echo proximity prediction in the prior art cannot reflect a convective storm structure.
Description
Technical Field
The invention relates to the technical field of weather radar information processing, in particular to a three-dimensional radar echo proximity forecasting method for a convection storm fine structure.
Background
The basic reflectivity (echo) obtained by the observation of the weather radar reflects the scale and density of precipitation particles in a meteorological target, can be used for measuring the intensity and distribution of weather phenomena, is a main way for solving the problem of strong weather early warning such as disastrous strong wind, storm, thunderstorm and the like based on 0-2 hours extrapolation (approach forecast) of the high spatial-temporal resolution radar echo, and has important application value in the fields of aviation, electric power and the like.
At present, traditional extrapolation methods represented by storm tracking, cross correlation or optical flow methods are all built on the basis of radar echo linear evolution assumption, and only the future moving direction of a convection storm can be estimated after the development and maturity of the convection storm, and nonlinear evolution such as the induction, enhancement and organization of the convection cannot be described, and the factors are the key criteria for early warning decision release in service.
With the rapid development of computer technology, deep learning technology has been gradually applied in the field of radar echo proximity prediction due to its strong nonlinear problem processing capability. The current deep learning approach prediction research is mostly based on a single-level radar echo field, but in actual conditions, the three-dimensional structures of the convection storms which cause strong weather such as disastrous storm, thunderstorm and the like to develop are different, so that the characteristics extracted by taking the single-level echo as an input deep learning model are limited, and the fine structural characteristics of the convection storms cannot be described; furthermore, in terms of loss function design, single-level deep learning radar echo proximity prediction often employs weighted loss functions for different magnitude echo fields, which is also inapplicable to multiple vertical levels of radar echo input.
Disclosure of Invention
The present invention aims to solve at least one of the technical problems existing in the related art to some extent.
Therefore, an object of the present invention is to provide a three-dimensional radar echo proximity prediction method for a fine structure of a convective storm, which solves the problem that a single-level radar echo proximity prediction in the prior art cannot reflect the structure of the convective storm.
Another object of the present invention is to provide a three-dimensional radar echo proximity prediction system for a convective storm fine structure.
In order to achieve the above purpose, the present invention provides a three-dimensional radar echo proximity prediction method for a fine structure of a convective storm, comprising the following steps:
s100, acquiring basic reflectivity factor data of an analysis area, and establishing a high-resolution three-dimensional radar echo data set of the analysis area through quality control and interpolation;
S200, training to obtain an optimal three-dimensional radar echo proximity prediction model by adopting a SwinURNN-based 3D fusion frame according to a high-resolution three-dimensional radar echo dataset;
and S300, inputting the closest three-dimensional radar reflectivity sequence data serving as an optimal three-dimensional radar echo proximity prediction model, and outputting a predicted three-dimensional radar reflectivity factor sequence.
Further, the basic reflectivity factor data of the analysis area is obtained, and a high-resolution three-dimensional radar echo data set of the analysis area is established through quality control and interpolation, and the specific method comprises the following steps:
S110, determining the spatial range, horizontal resolution and vertical resolution of an analysis area, and establishing a forecast grid;
And S120, extracting basic reflectivity factor data from the polarized weather radar in the analysis area, performing quality control, interpolating the basic reflectivity factor data to equidistant forecast grids, and establishing a high-resolution three-dimensional radar echo data set in the analysis area.
Further, the basic reflectivity factor data is used for quality control and is used for eliminating non-meteorological clutter; the basic reflectivity factor data of the quality control is interpolated to an equidistant analysis grid by a three-dimensional networking jigsaw mode.
Further, the established high-resolution three-dimensional radar echo data set is subjected to time sequence division, and in a data sequence of 40 continuous time periods for 4 hours, the first 10 frames are taken as a model training set, and the last 30 frames are taken as labels.
Further, in step S300, the nearest three-dimensional radar reflectivity data is input as an optimal three-dimensional radar echo proximity prediction model, and a predicted three-dimensional radar reflectivity factor sequence is output, specifically, real-time three-dimensional radar echo data is obtained, three-dimensional radar echo data of 10 times in the past 1 hour is substituted into the input end of the optimal three-dimensional radar echo proximity prediction model, and three-dimensional radar echo prediction results of 3 hours in the total of 30 continuous times in the future are predicted.
Furthermore, according to the high-resolution three-dimensional radar echo data set, an optimal three-dimensional radar echo proximity prediction model is obtained by training based on SwinURNN D fusion frames, and the specific method comprises the following steps:
S210, designing a three-dimensional radar echo proximity prediction model based on deep learning, wherein the three-dimensional radar echo proximity prediction model adopts a full convolution neural network UNet fused with an attention introducing mechanism and a parallel training neural network SwinRNN sub-network to generate a SwinURNN D fusion frame;
S220, designing a comprehensive loss function Comprehensively considering the balance L1 loss function and the style loss function, and three-dimensionally generating four methods for calculating the loss function, namely:；
wherein, Representing the balance L1 loss function,/>Representing style loss function,/>Representing a three-dimensional generated contrast loss function,/>Representing a vertical gradient loss function;
the balance L1 loss function The calculation formula of (2) is as follows:
；
wherein, And/>Respectively representing an observation field and a model forecasting field of the three-dimensional radar reflectivity factor; n represents the sample sum of all grid points of all pre-report times; /(I)Representing a weighting function, giving different weights to different intervals by counting the number of intervals falling in the different intervals,/>The calculation formula of (2) is as follows:
，
In the middle of For parameters of specific gravity of different threshold intervals, the larger the gamma is, the larger the weight given to a high threshold interval is, and the greater the weight is given to a high threshold intervalFor divided intervals,/>The frequencies representing different threshold intervals are specifically:
，
Wherein the method comprises the steps of For the frequency of different threshold intervals,/>Is the total frequency;
style loss function The calculation formula of (2) is as follows:
；
wherein, And/>Respectively representing style characteristics of a radar reflectivity factor observation field and a model forecasting field after background noise is removed; t and L respectively represent the time step and the number of layers of the style characteristic model EFFICIENTNET;
Specifically, the style loss function calculates the style characteristics of the radar reflectivity factor observation field and the model forecast field by inputting EFFICIENTNET respectively, and then compares the L2 loss.
Generating an fight loss functionThe calculation formula of (2) is as follows:
；
wherein, Generating an contrast loss function for three-dimensional radar echo reflectivity factors, including/>Time discriminant loss function/>The loss function is highly discriminated.
In particular, the method comprises the steps of,Expressed as:
；
the G generator is SwinURNN a 3D network, For the time discriminator, discrimination is performed in the time dimension,Is a three-dimensional radar reflectivity factor observation field,/>Is a three-dimensional radar reflectivity factor model forecasting field.
In particular, the method comprises the steps of,Expressed as:
；
the G generator is SwinURNN a 3D network, For the height discriminator, discrimination is performed in the height dimension,Is a three-dimensional radar reflectivity factor observation field,/>Is a three-dimensional radar reflectivity factor model forecasting field.
Vertical gradient loss functionThe calculation formula of (2) is as follows:
；
wherein, And/>The gradients of the three-dimensional radar reflectivity factor observation field and the model forecasting field in the vertical direction are respectively.
And S230, obtaining an optimal three-dimensional radar echo proximity prediction model through model training and super-parameter adjustment.
Another aspect of the present invention provides a three-dimensional radar echo proximity prediction system for a convective storm fine structure, including:
the data acquisition module is used for acquiring basic reflectivity factor data of the analysis area and establishing a high-resolution three-dimensional radar echo data set of the analysis area through quality control and interpolation;
The model training module is used for training to obtain an optimal three-dimensional radar echo proximity prediction model by adopting a SwinURNN D fusion frame according to the high-resolution three-dimensional radar echo data set;
and the prediction module is used for inputting the nearest three-dimensional radar reflectivity data serving as an optimal three-dimensional radar echo proximity prediction model and outputting a predicted three-dimensional radar reflectivity factor sequence.
In yet another aspect, the present invention provides a non-transitory computer readable storage medium having stored thereon computer instructions for causing a computer to perform the above-described three-dimensional radar echo proximity prediction method for a fine structure of a convective storm.
Still another aspect of the present invention provides an electronic apparatus, including: the three-dimensional radar echo proximity forecasting method for the convective storm fine structure comprises a processor, a communication interface, a memory and a communication bus, wherein the processor and the communication interface are used for the memory to complete communication with each other through the communication bus, and the processor calls logic instructions in the memory to execute the three-dimensional radar echo proximity forecasting method for the convective storm fine structure.
Compared with the prior art, the three-dimensional radar echo proximity forecasting method and system for the convection storm fine structure provided by the invention have the remarkable characteristics that: a three-dimensional radar echo data set containing multiple vertical layers is established as model input, and the problem that a single-level radar echo proximity forecast in the prior art cannot reflect a convective storm structure is solved.
The model based on the SwinURNN3D fusion frame is built by fusing Unet and SwinRNN models, the characteristics of small accumulated error and good SwinRNN consistency of Unet forecasting effect are effectively combined, and the forecasting effect of the deep learning model on different step sizes is comprehensively improved.
The method combines multiple loss functions, designs the three-dimensional generation anti-loss function on the basis of the balance L1 loss function, the style loss function and the two-dimensional generation anti-loss function in single-level radar echo prediction, avoids the problem of underestimation of high-level echo prediction caused by less high-level strong echo samples, and simultaneously designs the vertical gradient loss function to avoid the uniformity of prediction results in height caused by less high-level strong echo samples.
Drawings
Fig. 1 is a flowchart of a three-dimensional radar echo proximity prediction method facing a convective storm fine structure according to an embodiment of the invention.
Detailed Description
The technical scheme of the invention is described in detail below through the drawings, but the protection scope of the invention is not limited to the embodiments.
Examples: as shown in fig. 1, the embodiment provides a three-dimensional radar echo proximity prediction method for a fine structure of a convective storm, which includes the following steps:
s100, acquiring basic reflectivity factor data of an analysis area, and establishing a high-resolution three-dimensional radar echo data set of the analysis area through quality control and interpolation;
The specific method comprises the following steps:
s110: determining the space range, horizontal resolution and vertical resolution of the forecasting method according to application requirements, and establishing a grid area with equal distances and equal altitude as a forecasting grid;
S120: and extracting historical basic reflectivity (radar echo) data from a polarized weather radar in an analysis area, performing quality control treatment to remove non-meteorological clutter, performing three-dimensional networking jigsaw on the radar echo after quality control, interpolating to an equidistant analysis grid, wherein the grid spacing is 1 km, establishing a high-resolution three-dimensional radar echo data set in the analysis area, wherein the vertical level comprises 6 layers of 1,2, 3.5, 5, 6.5 and 8 km, the length of the data set is 2019-2022 years 4-9 months, and the data interval is 6 minutes.
S200, training to obtain an optimal three-dimensional radar echo proximity prediction model by adopting a SwinURNN-based 3D fusion frame according to a high-resolution three-dimensional radar echo dataset;
The specific method comprises the following steps:
S210, designing a three-dimensional radar echo proximity prediction model structure based on deep learning, constructing a fusion network architecture SwinURNN D combining a full convolution neural network UNet which introduces an attention mechanism and a parallel training neural network SwinRNN sub-network, wherein SwinRNN operates on the lowest scale, obtaining a prediction result of a future moment after SwinRNN, and adding the prediction result with the prediction result of Unet after up-sampling to obtain a final prediction result. Further, the data set obtained in the step S100 is subjected to time sequence division, and in a data sequence of 40 continuous time periods for 4 hours, the first 10 frames are input by a model, and the last 30 frames are labels;
s220, under SwinURNN D model frame, designing a comprehensive Loss function Loss facing to the fine structure of the convection storm, comprehensively considering a balance L1 Loss function and a style Loss function, and three-dimensionally generating four methods for calculating the Loss function, namely:
；
wherein, Representing the balance L1 loss function,/>Representing style loss function,/>Representing a three-dimensional generated contrast loss function,/>Representing a vertical gradient loss function;
the balance L1 loss function The calculation formula of (2) is as follows:
；
wherein, And/>Respectively representing an observation field and a model forecasting field of the three-dimensional radar reflectivity factor; n represents the sample sum of all grid points of all pre-report times; /(I)Representing a weighting function, giving different weights to different intervals by counting the number of intervals falling in the different intervals,/>The calculation formula of (2) is as follows:
，
In the middle of For parameters of specific gravity of different threshold intervals, the larger the gamma is, the larger the weight given to a high threshold interval is, and the greater the weight is given to a high threshold intervalFor the divided intervals, the specific ranges are 20dBz,23.5dBz,27dBz, … and 70dBz; /(I)The frequencies representing different threshold intervals are specifically:
，
Wherein the method comprises the steps of For the frequency of different threshold intervals,/>Is the total frequency;
style loss function The calculation formula of (2) is as follows:
；
wherein, And/>Respectively representing style characteristics of a radar reflectivity factor observation field and a model forecasting field after background noise is removed; t and L respectively represent the time step and the number of layers of the style characteristic model EFFICIENTNET;
Specifically, the style loss function calculates the style characteristics of the radar reflectivity factor observation field and the model forecast field by inputting EFFICIENTNET respectively, and then compares the L2 loss.
Generating an fight loss functionThe calculation formula of (2) is as follows:
；
wherein, Generating an contrast loss function for three-dimensional radar echo reflectivity factors, including/>Time discriminant loss function/>The loss function is highly discriminated.
In particular, the method comprises the steps of,Expressed as:
；
the G generator is SwinURNN a 3D network, For the time discriminator, discrimination is performed in the time dimension,Is a three-dimensional radar reflectivity factor observation field,/>Is a three-dimensional radar reflectivity factor model forecasting field.
In particular, the method comprises the steps of,Expressed as:
；
the G generator is SwinURNN a 3D network, For the height discriminator, discrimination is performed in the height dimension,Is a three-dimensional radar reflectivity factor observation field,/>Is a three-dimensional radar reflectivity factor model forecasting field.
Vertical gradient loss functionThe calculation formula of (2) is as follows:
；
wherein, And/>The gradients of the three-dimensional radar reflectivity factor observation field and the model forecasting field in the vertical direction are respectively.
And S230, obtaining an optimal three-dimensional radar echo proximity prediction model through model training and super-parameter adjustment.
S300, acquiring new real-time three-dimensional radar echo data, substituting three-dimensional radar echo data of 10 times in the past 1 hour into the input end of a three-dimensional radar echo proximity prediction model, and predicting to obtain three-dimensional radar echo prediction results of 3 hours in total of 30 continuous times in the future, so that the actual strong weather monitoring and early warning service requirements can be met.
Table 1 is basic test result data of multi-layer radar echo input, under the same condition, the input multi-layer radar echo prediction can obviously improve the prediction effect on the combined reflectivity compared with the single-layer radar echo, wherein under the threshold conditions of 30, 40 and 50dBZ, the input 6-layer radar echo is improved by 12%, 9% and 68% in1 hour TS, 8%, 4% and 500% in 2 hours TS, and the improvement is also improved to different degrees in 3 hours TS and BIAS indexes.
Table 1 basic test result data for multi-layer radar echo input:
Table 2 shows that the most TS and BIAS indexes are improved after the style and three-dimensional contrast loss and the vertical gradient loss are further increased based on the multi-layer radar echo input, and the comparison shows that the style loss and the three-dimensional contrast loss are increased (UnetRNN _sty_gan), and the TS with 50dBZ threshold is obviously improved although the TS with 30 and 40dBZ thresholds is not obvious after the vertical gradient loss for the high-layer echo is further added (SwinURNN D).
Table 2 contrast after further increasing style and three-dimensional contrast loss, vertical gradient loss, based on multilayer radar echo input:
Therefore, the invention solves the problem of poor high-level prediction effect caused by few high-level strong echo samples through the comprehensive design of the model and the loss function, not only can predict the two-dimensional spatial distribution of the convection storm, but also can characterize the three-dimensional fine structure thereof, and provides technical support for the prediction and the early warning of disaster-induced strong convection weather.
In some embodiments of the present invention, a three-dimensional radar echo proximity prediction system for a fine structure of a convective storm is provided, including:
the data acquisition module is used for acquiring basic reflectivity factor data of the analysis area and establishing a high-resolution three-dimensional radar echo data set of the analysis area through quality control and interpolation;
The model training module is used for training to obtain an optimal three-dimensional radar echo proximity prediction model by adopting a SwinURNN D fusion frame according to the high-resolution three-dimensional radar echo data set;
the prediction module is used for taking the latest three-dimensional radar reflectivity data as an optimal three-dimensional radar echo proximity prediction model and outputting a predicted three-dimensional radar reflectivity factor sequence.
In another aspect, the present invention also provides a computer program product, where the computer program product includes a computer program, where the computer program can be stored on a non-transitory computer readable storage medium, and when the computer program is executed by a processor, the computer can execute a three-dimensional radar echo proximity prediction method for a fine structure of a convective storm, and specifically includes:
s100, acquiring basic reflectivity factor data of an analysis area, and establishing a high-resolution three-dimensional radar echo data set of the analysis area through quality control and interpolation;
S200, training to obtain an optimal three-dimensional radar echo proximity prediction model by adopting a SwinURNN-based 3D fusion frame according to a high-resolution three-dimensional radar echo dataset;
and S300, inputting the closest three-dimensional radar reflectivity data serving as an optimal three-dimensional radar echo proximity prediction model, and outputting a predicted three-dimensional radar reflectivity factor sequence.
In still another aspect, the present invention further provides a non-transitory computer readable storage medium having stored thereon computer instructions for causing a computer to perform a method for three-dimensional radar echo proximity prediction for a fine structure of a convective storm, specifically comprising:
s100, acquiring basic reflectivity factor data of an analysis area, and establishing a high-resolution three-dimensional radar echo data set of the analysis area through quality control and interpolation;
S200, training to obtain an optimal three-dimensional radar echo proximity prediction model by adopting a SwinURNN-based 3D fusion frame according to a high-resolution three-dimensional radar echo dataset;
and S300, inputting the closest three-dimensional radar reflectivity data serving as an optimal three-dimensional radar echo proximity prediction model, and outputting a predicted three-dimensional radar reflectivity factor sequence.
In yet another aspect, the present invention also provides an electronic device, which may include: a processor (processor), a communication interface (Communications Interface), a memory (memory), and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus. The processor can call logic instructions in the memory to execute a three-dimensional radar echo proximity prediction method facing to a fine structure of a convective storm, and specifically comprises the following steps:
s100, acquiring basic reflectivity factor data of an analysis area, and establishing a high-resolution three-dimensional radar echo data set of the analysis area through quality control and interpolation;
S200, training to obtain an optimal three-dimensional radar echo proximity prediction model by adopting a SwinURNN-based 3D fusion frame according to a high-resolution three-dimensional radar echo dataset;
and S300, inputting the closest three-dimensional radar reflectivity data serving as an optimal three-dimensional radar echo proximity prediction model, and outputting a predicted three-dimensional radar reflectivity factor sequence.
Further, the logic instructions in the memory described above may be implemented in the form of software functional units and stored in a computer-readable storage medium when sold or used as a stand-alone product. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a usb disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
As described above, although the present invention has been shown and described with reference to certain preferred embodiments, it is not to be construed as limiting the invention itself. Various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (6)
1. A three-dimensional radar echo proximity prediction method facing a convection storm fine structure is characterized by comprising the following steps:
S100, acquiring basic reflectivity factor data of an analysis area, and establishing a high-resolution three-dimensional radar echo data set of the analysis area through quality control and interpolation; the specific method comprises the following steps:
S110, determining the spatial range, horizontal resolution and vertical resolution of an analysis area, and establishing a forecast grid;
S120, extracting basic reflectivity factor data from the polarized weather radar in the analysis area, wherein the basic reflectivity factor data is used for quality control and is used for eliminating non-meteorological clutter; the basic reflectivity factor data of the quality control are interpolated to equidistant analysis grids in a three-dimensional networking jigsaw mode, and a high-resolution three-dimensional radar echo data set in an analysis area is established;
S200, training to obtain an optimal three-dimensional radar echo proximity prediction model by adopting a SwinURNN-based 3D fusion frame according to a high-resolution three-dimensional radar echo dataset, wherein the specific method comprises the following steps of:
S210, designing a three-dimensional radar echo proximity prediction model based on deep learning, wherein the three-dimensional radar echo proximity prediction model adopts a full convolution neural network UNet fused with an attention introducing mechanism and a parallel training neural network SwinRNN sub-network to generate a SwinURNN D fusion frame;
S220, designing a comprehensive loss function Comprehensively considering the balance L1 loss function and the style loss function, and three-dimensionally generating four methods for calculating the loss function, namely:
；
wherein, Representing the balance L1 loss function,/>Representing style loss function,/>Representing a three-dimensional generated contrast loss function,/>Representing a vertical gradient loss function;
S230, obtaining an optimal three-dimensional radar echo proximity prediction model through model training and super-parameter adjustment;
and S300, inputting the closest three-dimensional radar reflectivity sequence data serving as an optimal three-dimensional radar echo proximity prediction model, and outputting a predicted three-dimensional radar reflectivity factor sequence.
2. The three-dimensional radar echo proximity prediction method for the fine structure of the storm by flow according to claim 1, wherein the established high-resolution three-dimensional radar echo data set is subjected to time sequence division, and the first 10 frames are taken as a model training set and the last 30 frames are taken as labels in a data sequence of 4 hours for 40 continuous time.
3. The method for three-dimensional radar echo proximity prediction for a fine structure of a convective storm according to claim 2, wherein in step S300, the nearest three-dimensional radar reflectivity sequence data is input as an optimal three-dimensional radar echo proximity prediction model, a predicted three-dimensional radar reflectivity factor sequence is output, specifically, real-time three-dimensional radar echo data is obtained, three-dimensional radar echo data of 10 times in the past 1 hour is substituted into an input end of the optimal three-dimensional radar echo proximity prediction model, and three-dimensional radar echo prediction results of 30 times in total 3 hours are obtained.
4. A convection storm fine structure-oriented three-dimensional radar echo proximity prediction system, comprising:
The data acquisition module is used for acquiring basic reflectivity factor data of the analysis area and establishing a high-resolution three-dimensional radar echo data set of the analysis area through quality control and interpolation; the specific method comprises the following steps:
S110, determining the spatial range, horizontal resolution and vertical resolution of an analysis area, and establishing a forecast grid;
S120, extracting basic reflectivity factor data from the polarized weather radar in the analysis area, wherein the basic reflectivity factor data is used for quality control and is used for eliminating non-meteorological clutter; the basic reflectivity factor data of the quality control are interpolated to equidistant analysis grids in a three-dimensional networking jigsaw mode, and a high-resolution three-dimensional radar echo data set in an analysis area is established;
The model training module is used for training to obtain an optimal three-dimensional radar echo proximity prediction model by adopting a SwinURNN D fusion frame according to the high-resolution three-dimensional radar echo data set; the specific method comprises the following steps:
S110, determining the spatial range, horizontal resolution and vertical resolution of an analysis area, and establishing a forecast grid;
S120, extracting basic reflectivity factor data from the polarized weather radar in the analysis area, wherein the basic reflectivity factor data is used for quality control and is used for eliminating non-meteorological clutter; the basic reflectivity factor data of the quality control are interpolated to equidistant analysis grids in a three-dimensional networking jigsaw mode, and a high-resolution three-dimensional radar echo data set in an analysis area is established;
S200, training to obtain an optimal three-dimensional radar echo proximity prediction model by adopting a SwinURNN-based 3D fusion frame according to a high-resolution three-dimensional radar echo dataset, wherein the specific method comprises the following steps of:
S210, designing a three-dimensional radar echo proximity prediction model based on deep learning, wherein the three-dimensional radar echo proximity prediction model adopts a full convolution neural network UNet fused with an attention introducing mechanism and a parallel training neural network SwinRNN sub-network to generate a SwinURNN D fusion frame;
S220, designing a comprehensive loss function Comprehensively considering the balance L1 loss function and the style loss function, and three-dimensionally generating four methods for calculating the loss function, namely:
；
wherein, Representing the balance L1 loss function,/>Representing style loss function,/>Representing a three-dimensional generated contrast loss function,/>Representing a vertical gradient loss function;
S230, obtaining an optimal three-dimensional radar echo proximity prediction model through model training and super-parameter adjustment;
And the prediction module is used for inputting the nearest three-dimensional radar reflectivity sequence data serving as an optimal three-dimensional radar echo proximity prediction model and outputting a predicted three-dimensional radar reflectivity factor sequence.
5. A non-transitory computer readable storage medium having stored thereon computer instructions for causing a computer to perform the three-dimensional radar echo proximity prediction method oriented to a convective storm fine structure as claimed in any one of claims 1-3.
6. An electronic device, comprising: the method comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus, and the processor invokes logic instructions in the memory to execute the three-dimensional radar echo proximity forecasting method facing the fine structure of the convective storm according to any one of claims 1-3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410308899.7A CN117907965B (en) | 2024-03-19 | 2024-03-19 | Three-dimensional radar echo proximity forecasting method for convection storm fine structure |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410308899.7A CN117907965B (en) | 2024-03-19 | 2024-03-19 | Three-dimensional radar echo proximity forecasting method for convection storm fine structure |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117907965A CN117907965A (en) | 2024-04-19 |
| CN117907965B true CN117907965B (en) | 2024-05-24 |
Family
ID=90693997
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410308899.7A Active CN117907965B (en) | 2024-03-19 | 2024-03-19 | Three-dimensional radar echo proximity forecasting method for convection storm fine structure |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117907965B (en) |
Citations (38)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101937078A (en) * | 2009-06-30 | 2011-01-05 | 深圳市气象局 | Nowcasting method and system of thunder cloud cluster based on boundary recognition and tracer technique |
| CN102645679A (en) * | 2012-03-13 | 2012-08-22 | 天津大学 | Mesocyclone identification method based on Doppler radar echo images |
| CN102721987A (en) * | 2012-06-12 | 2012-10-10 | 中国海洋大学 | Method for prewarning Doppler radar remote sensing strong storm |
| CN103337133A (en) * | 2013-06-14 | 2013-10-02 | 广东电网公司中山供电局 | System and method for power grid thunderstorm disaster early warning based on recognition and forecast |
| CN103529492A (en) * | 2013-09-22 | 2014-01-22 | 天津大学 | Storm body position and form prediction method based on Doppler radar reflectivity image |
| CN104992071A (en) * | 2015-07-17 | 2015-10-21 | 南京信息工程大学 | Initial disturbance method based on ensemble data assimilation technology |
| CN105046358A (en) * | 2015-07-17 | 2015-11-11 | 南京信息工程大学 | Storm scale ensemble forecast perturbation method |
| CN105759274A (en) * | 2016-04-26 | 2016-07-13 | 南京信息工程大学 | Typhoon attention area radar rainfall estimation method |
| CN106054194A (en) * | 2016-05-10 | 2016-10-26 | 南京信息工程大学 | Spaceborne radar and ground-based radar reflectivity factor data three dimensional fusion method |
| CN107037504A (en) * | 2016-11-15 | 2017-08-11 | 兰州大学 | For Severe Convective Weather Forecasting assimilation dodge the method for translation proxy radar return |
| CN107121679A (en) * | 2017-06-08 | 2017-09-01 | 湖南师范大学 | Recognition with Recurrent Neural Network predicted method and memory unit structure for Radar Echo Extrapolation |
| CN107229084A (en) * | 2017-06-08 | 2017-10-03 | 天津大学 | A kind of automatic identification, tracks and predicts contracurrent system mesh calibration method |
| CN107436987A (en) * | 2016-05-26 | 2017-12-05 | 江苏省气象台 | A kind of thermal convection storm develops the method for building up of forecast conceptual model |
| CN108256439A (en) * | 2017-12-26 | 2018-07-06 | 北京大学 | A kind of pedestrian image generation method and system based on cycle production confrontation network |
| CN108761484A (en) * | 2018-04-26 | 2018-11-06 | 江苏省气象台 | A kind of sea fog monitoring method based on Multi-sensor satellite remote sensing |
| CN109116358A (en) * | 2018-08-09 | 2019-01-01 | 成都信息工程大学 | Hail identification and occurring area forecast method based on China New Generation Weather Radar |
| CN110261857A (en) * | 2019-07-17 | 2019-09-20 | 南京信息工程大学 | A kind of weather radar spatial interpolation methods |
| CN110501760A (en) * | 2019-07-29 | 2019-11-26 | 中国气象局广州热带海洋气象研究所（广东省气象科学研究所） | A kind of hail identification and nowcasting method based on weather radar |
| CN111625993A (en) * | 2020-05-25 | 2020-09-04 | 中国水利水电科学研究院 | Small watershed surface rainfall interpolation method based on mountainous terrain and rainfall characteristic prediction |
| CN111897030A (en) * | 2020-07-17 | 2020-11-06 | 国网电力科学研究院有限公司 | Thunderstorm early warning system and method |
| CN113009490A (en) * | 2021-02-20 | 2021-06-22 | 江苏省气象台 | Radar three-dimensional wind field inversion method based on high-resolution mode dynamic constraint |
| CN113313037A (en) * | 2021-06-02 | 2021-08-27 | 郑州大学 | Method for detecting video abnormity of generation countermeasure network based on self-attention mechanism |
| CN113936142A (en) * | 2021-10-13 | 2022-01-14 | 成都信息工程大学 | Rainfall approach forecasting method and device based on deep learning |
| CN114217318A (en) * | 2021-12-09 | 2022-03-22 | 中国科学院空天信息创新研究院 | Rainfall forecasting method, device, equipment and medium based on weather radar echo image |
| CA3177585A1 (en) * | 2021-04-16 | 2022-10-16 | Strong Force Vcn Portfolio 2019, Llc | Systems, methods, kits, and apparatuses for digital product network systems and biology-based value chain networks |
| CN115201938A (en) * | 2022-07-26 | 2022-10-18 | 保定市气象局 | Strong convection weather nowcasting method and system based on thunderstorm high-pressure analysis |
| CN115512299A (en) * | 2022-09-27 | 2022-12-23 | 河南大学 | Flood early warning method of U-net variant neural network based on radar image |
| CN115508918A (en) * | 2022-08-18 | 2022-12-23 | 广东省气象探测数据中心(广东省气象技术装备中心、广东省气象科技培训中心) | Ground precipitation cooperative quality control method and system based on radar combined reflectivity |
| CN115761261A (en) * | 2022-11-27 | 2023-03-07 | 东南大学 | Short-term rainfall prediction method based on radar echo diagram extrapolation |
| CN115933008A (en) * | 2022-11-22 | 2023-04-07 | 广东电网有限责任公司广州供电局 | Strong convection weather forecast early warning method |
| CN116009121A (en) * | 2023-01-31 | 2023-04-25 | 江苏省气象台 | Lightning short-time proximity forecasting method based on deep learning |
| CN116413706A (en) * | 2022-12-13 | 2023-07-11 | 浙江大学 | Method for simultaneously establishing graph and calibrating internal reference of laser radar on mobile carrier |
| CN116416131A (en) * | 2023-02-03 | 2023-07-11 | 阿里巴巴（中国）有限公司 | Target object prediction method and device |
| CN117129997A (en) * | 2023-08-04 | 2023-11-28 | 厦门市气象台(厦门市海洋气象台、海峡气象开放实验室) | Dual-polarization radar differential phase shift rate foot identification method based on convection storm |
| CN117250620A (en) * | 2023-08-16 | 2023-12-19 | 中国水利水电科学研究院 | X-band radar proximity forecasting method based on precipitation life cycle discrimination |
| CN117520466A (en) * | 2023-11-11 | 2024-02-06 | 深圳市双银科技有限公司 | Geographic information acquisition system capable of comparing similar data |
| CN117630944A (en) * | 2023-10-27 | 2024-03-01 | 中国科学院西北生态环境资源研究院 | Combined algorithm for identifying and tracking convective storm based on radar observation data |
| CN117647812A (en) * | 2023-11-24 | 2024-03-05 | 中国船舶集团有限公司第七一六研究所 | Navigation radar-based ship travel route detection and early warning method and system |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA3177620A1 (en) * | 2021-05-06 | 2022-11-06 | Strong Force Iot Portfolio 2016, Llc | Quantum, biological, computer vision, and neural network systems for industrial internet of things |
-
2024
- 2024-03-19 CN CN202410308899.7A patent/CN117907965B/en active Active
Patent Citations (38)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101937078A (en) * | 2009-06-30 | 2011-01-05 | 深圳市气象局 | Nowcasting method and system of thunder cloud cluster based on boundary recognition and tracer technique |
| CN102645679A (en) * | 2012-03-13 | 2012-08-22 | 天津大学 | Mesocyclone identification method based on Doppler radar echo images |
| CN102721987A (en) * | 2012-06-12 | 2012-10-10 | 中国海洋大学 | Method for prewarning Doppler radar remote sensing strong storm |
| CN103337133A (en) * | 2013-06-14 | 2013-10-02 | 广东电网公司中山供电局 | System and method for power grid thunderstorm disaster early warning based on recognition and forecast |
| CN103529492A (en) * | 2013-09-22 | 2014-01-22 | 天津大学 | Storm body position and form prediction method based on Doppler radar reflectivity image |
| CN104992071A (en) * | 2015-07-17 | 2015-10-21 | 南京信息工程大学 | Initial disturbance method based on ensemble data assimilation technology |
| CN105046358A (en) * | 2015-07-17 | 2015-11-11 | 南京信息工程大学 | Storm scale ensemble forecast perturbation method |
| CN105759274A (en) * | 2016-04-26 | 2016-07-13 | 南京信息工程大学 | Typhoon attention area radar rainfall estimation method |
| CN106054194A (en) * | 2016-05-10 | 2016-10-26 | 南京信息工程大学 | Spaceborne radar and ground-based radar reflectivity factor data three dimensional fusion method |
| CN107436987A (en) * | 2016-05-26 | 2017-12-05 | 江苏省气象台 | A kind of thermal convection storm develops the method for building up of forecast conceptual model |
| CN107037504A (en) * | 2016-11-15 | 2017-08-11 | 兰州大学 | For Severe Convective Weather Forecasting assimilation dodge the method for translation proxy radar return |
| CN107121679A (en) * | 2017-06-08 | 2017-09-01 | 湖南师范大学 | Recognition with Recurrent Neural Network predicted method and memory unit structure for Radar Echo Extrapolation |
| CN107229084A (en) * | 2017-06-08 | 2017-10-03 | 天津大学 | A kind of automatic identification, tracks and predicts contracurrent system mesh calibration method |
| CN108256439A (en) * | 2017-12-26 | 2018-07-06 | 北京大学 | A kind of pedestrian image generation method and system based on cycle production confrontation network |
| CN108761484A (en) * | 2018-04-26 | 2018-11-06 | 江苏省气象台 | A kind of sea fog monitoring method based on Multi-sensor satellite remote sensing |
| CN109116358A (en) * | 2018-08-09 | 2019-01-01 | 成都信息工程大学 | Hail identification and occurring area forecast method based on China New Generation Weather Radar |
| CN110261857A (en) * | 2019-07-17 | 2019-09-20 | 南京信息工程大学 | A kind of weather radar spatial interpolation methods |
| CN110501760A (en) * | 2019-07-29 | 2019-11-26 | 中国气象局广州热带海洋气象研究所（广东省气象科学研究所） | A kind of hail identification and nowcasting method based on weather radar |
| CN111625993A (en) * | 2020-05-25 | 2020-09-04 | 中国水利水电科学研究院 | Small watershed surface rainfall interpolation method based on mountainous terrain and rainfall characteristic prediction |
| CN111897030A (en) * | 2020-07-17 | 2020-11-06 | 国网电力科学研究院有限公司 | Thunderstorm early warning system and method |
| CN113009490A (en) * | 2021-02-20 | 2021-06-22 | 江苏省气象台 | Radar three-dimensional wind field inversion method based on high-resolution mode dynamic constraint |
| CA3177585A1 (en) * | 2021-04-16 | 2022-10-16 | Strong Force Vcn Portfolio 2019, Llc | Systems, methods, kits, and apparatuses for digital product network systems and biology-based value chain networks |
| CN113313037A (en) * | 2021-06-02 | 2021-08-27 | 郑州大学 | Method for detecting video abnormity of generation countermeasure network based on self-attention mechanism |
| CN113936142A (en) * | 2021-10-13 | 2022-01-14 | 成都信息工程大学 | Rainfall approach forecasting method and device based on deep learning |
| CN114217318A (en) * | 2021-12-09 | 2022-03-22 | 中国科学院空天信息创新研究院 | Rainfall forecasting method, device, equipment and medium based on weather radar echo image |
| CN115201938A (en) * | 2022-07-26 | 2022-10-18 | 保定市气象局 | Strong convection weather nowcasting method and system based on thunderstorm high-pressure analysis |
| CN115508918A (en) * | 2022-08-18 | 2022-12-23 | 广东省气象探测数据中心(广东省气象技术装备中心、广东省气象科技培训中心) | Ground precipitation cooperative quality control method and system based on radar combined reflectivity |
| CN115512299A (en) * | 2022-09-27 | 2022-12-23 | 河南大学 | Flood early warning method of U-net variant neural network based on radar image |
| CN115933008A (en) * | 2022-11-22 | 2023-04-07 | 广东电网有限责任公司广州供电局 | Strong convection weather forecast early warning method |
| CN115761261A (en) * | 2022-11-27 | 2023-03-07 | 东南大学 | Short-term rainfall prediction method based on radar echo diagram extrapolation |
| CN116413706A (en) * | 2022-12-13 | 2023-07-11 | 浙江大学 | Method for simultaneously establishing graph and calibrating internal reference of laser radar on mobile carrier |
| CN116009121A (en) * | 2023-01-31 | 2023-04-25 | 江苏省气象台 | Lightning short-time proximity forecasting method based on deep learning |
| CN116416131A (en) * | 2023-02-03 | 2023-07-11 | 阿里巴巴（中国）有限公司 | Target object prediction method and device |
| CN117129997A (en) * | 2023-08-04 | 2023-11-28 | 厦门市气象台(厦门市海洋气象台、海峡气象开放实验室) | Dual-polarization radar differential phase shift rate foot identification method based on convection storm |
| CN117250620A (en) * | 2023-08-16 | 2023-12-19 | 中国水利水电科学研究院 | X-band radar proximity forecasting method based on precipitation life cycle discrimination |
| CN117630944A (en) * | 2023-10-27 | 2024-03-01 | 中国科学院西北生态环境资源研究院 | Combined algorithm for identifying and tracking convective storm based on radar observation data |
| CN117520466A (en) * | 2023-11-11 | 2024-02-06 | 深圳市双银科技有限公司 | Geographic information acquisition system capable of comparing similar data |
| CN117647812A (en) * | 2023-11-24 | 2024-03-05 | 中国船舶集团有限公司第七一六研究所 | Navigation radar-based ship travel route detection and early warning method and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117907965A (en) | 2024-04-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108761574B (en) | Rainfall estimation method based on multi-source information fusion | |
| Dyrrdal et al. | Bayesian hierarchical modeling of extreme hourly precipitation in Norway | |
| Duc et al. | Spatial-temporal fractions verification for high-resolution ensemble forecasts | |
| CN109738970A (en) | The method, apparatus and storage medium for realizing Lightning Warning are excavated based on lightning data | |
| CN111665575B (en) | Medium-and-long-term rainfall grading coupling forecasting method and system based on statistical power | |
| CN116595394A (en) | Training method of wind speed correction model, wind speed prediction method, wind speed prediction equipment and medium | |
| Kyznarová et al. | CELLTRACK—Convective cell tracking algorithm and its use for deriving life cycle characteristics | |
| CN113159102B (en) | Multi-time-scale photovoltaic power prediction method and system | |
| CN116739172B (en) | Method and device for ultra-short-term prediction of offshore wind power based on climbing identification | |
| Kober et al. | Aspects of short‐term probabilistic blending in different weather regimes | |
| CN114564487B (en) | Meteorological raster data updating method combining forecast prediction | |
| CN113962465A (en) | Precipitation forecasting method, equipment, device and storage medium | |
| CN114357670A (en) | Power distribution network power consumption data abnormity early warning method based on BLS and self-encoder | |
| CN116127833A (en) | Wind power prediction method, system, device and medium based on VMD and LSTM fusion model | |
| Méndez-Gordillo et al. | Wind speed forecasting using a hybrid model considering the turbulence of the airflow | |
| CN117907965B (en) | Three-dimensional radar echo proximity forecasting method for convection storm fine structure | |
| Elsner et al. | Tornado damage ratings estimated with cumulative logistic regression | |
| CN112363140A (en) | Thermodynamic constraint extrapolation objective correction method based on cyclic neural network | |
| Pierce et al. | Use of a stochastic precipitation nowcast scheme for fluvial flood forecasting and warning | |
| Plant | Statistical properties of cloud lifecycles in cloud-resolving models | |
| Fadhel et al. | Exploration of an adaptive merging scheme for optimal precipitation estimation over ungauged urban catchment | |
| CN110751335B (en) | Regional ecological quality annual scene prediction evaluation method and device | |
| Mu et al. | Prediction of North Atlantic Oscillation Index Associated with the Sea Level Pressure Using DWT‐LSTM and DWT‐ConvLSTM Networks | |
| Zelalem et al. | A Bayesian modelling approach for assessing non-stationarity in annual maximum rainfall under a changing climate | |
| CN112380778A (en) | Weather drought forecasting method based on sea temperature |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |