KR102389141B1 - Very-short-term forecast method using burgers’equation and semi-lagrangian, recording medium and device for performing the method - Google Patents

Abstract

A very-short-term forecast method using a Burgers equation and a semi-Lagrangian comprises the steps of: calculating an initial motion vector using a variational echo tracking (VET) algorithm which takes a two-dimensional radar precipitation grid field of consecutive time periods as an input value; generating a motion vector at each prediction time by using the Burgers equation with the calculated vector as an initial value; and generating a prediction field by substituting the updated motion vector generated at each prediction time into a posterior semi-Lagrangian at each prediction time. Accordingly, the motion vector is calculated for each prediction field using the Burgers equation, and the calculated motion vector is substituted into a MAPLE semi-Lagrangian at each prediction time, thereby improving prediction accuracy.

Description

Ultra-short-term precipitation forecasting method using Burgers equation and anti-Lagrangian method, recording medium and apparatus for performing the same

The present invention relates to an ultra-short-term precipitation forecasting method using the Burgers equation and the anti-Lagrangian method, and to a recording medium and apparatus for performing the same, and more particularly, calculating a motion vector for each prediction field using the Burgers equation, It relates to a technique for improving prediction accuracy by substituting a calculated vector into the anti-Lagrangian technique for each prediction time.

Radar-based precipitation ultra-short-term forecasts can be obtained by Lagrangian extrapolation, which is mainly based on advection. This method can produce results with high spatiotemporal resolution and uses a simpler algorithm compared to the Numerical Weather Predictions (NWP) model.

A representative algorithm using Lagrangian extrapolation is MAPLE (McGill Algorithm for Precipitation Nowcasting by Lagrangian Extrapolation) developed at McGill University. The extrapolation method of MAPLE was developed by German and Zawadzki (2002), and later papers on various resolutions and prediction areas were published and verified.

MAPLE uses the VET (variational echo tracking) algorithm that minimizes the cost function to calculate the motion vector to calculate a single fixed vector and uses this fixed vector to create a prediction field. A disadvantage of this method is that it does not fit well when the vertical source-sink term or the temporal variability of the motion vector field dominates over advection.

Precipitation prediction studies considering growth and decrease have been attempted by several research teams, but they did not yield successful results. In addition, in order to calculate the movement vector moving every time in order to consider the time variability, it is necessary to solve a complex partial differential equation (PDE) in consideration of numerous environmental variables, such as precise initial and boundary value conditions, and ambient pressure conditions. there is a difficulty

KR 10-2006847 B1 KR 10-2013-0049521 A US 7,062,066 B2

Germann, U. and Zawadzki, I., December 2002: Scale-dependence of the predictability of precipitation from continental radar images. Part I: Description of the methodology. Monthly, Weather Review, 130(12), 2859-2873.

Accordingly, it is an object of the present invention to provide an ultra-short-term precipitation forecasting method using the Burgers equation and the anti-Lagrangian method.

Another object of the present invention is to provide a recording medium in which a computer program for performing an ultra-short-term precipitation forecasting method using the Bergers equation and the anti-Lagrangian method is recorded.

Another object of the present invention is to provide an apparatus for performing an ultra-short-term precipitation forecasting method using the Bergers equation and the anti-Lagrangian method.

The ultra-short-term precipitation forecasting method using the Burgers equation and the anti-Lagrangian method according to an embodiment for realizing the object of the present invention is a VET algorithm (variational Calculating an initial motion vector using echo tracking); generating a motion vector for each prediction time using a Burgers equation using the calculated initial motion vector as an initial value; and generating a prediction field by substituting the updated motion vector for each generated prediction time for each prediction time into a backward semi-Lagrangian technique.

In an embodiment of the present invention, the calculating of the initial motion vector may include: receiving two or three consecutive two-dimensional radar precipitation grid fields as input values; and inputting the received two-dimensional radar precipitation grid field into the cost function to minimize the precipitation field.

In an embodiment of the present invention, the step of minimizing the precipitation field may include performing a complex gradient with an arbitrary initial guess as an input.

In an embodiment of the present invention, the initial guess may use VET algorithm data from a previous time or a vector consisting of 1 may be used as the initial guess.

In an embodiment of the present invention, the step of generating the motion vector for each prediction time includes: repeatedly generating motion vectors of the x-direction velocity u and the y-direction velocity v at preset time intervals using the Burgers equation; and downsizing the iteratively generated motion vector to the size of the prediction field using linear interpolation.

In an embodiment of the present invention, the step of repeatedly generating the movement vectors of the x-direction velocity u and the y-direction velocity v is the Navier-Stokes equation, which does not consider the pressure term in the Burgers equation. can be used to approximate a vector.

In an embodiment of the present invention, the Burgers equation may use a preset diffusion coefficient.

In a computer-readable storage medium according to an embodiment for realizing another object of the present invention, a computer program for performing the ultra-short-term precipitation forecasting method using the Bergers equation and the anti-Lagrangian method is recorded.

The ultra-short-term precipitation forecasting apparatus using the Burgers equation and the anti-Lagrangian method according to an embodiment for realizing another object of the present invention is a VET algorithm using a two-dimensional radar precipitation grid field of a continuous time period as an input value. an initial vector calculating unit that calculates an initial motion vector using (variational echo tracking); a motion vector calculator for generating a motion vector for each prediction time using a Burgers equation using the vector calculated from the initial vector calculator as an initial value; and a prediction field generator generating a prediction field by substituting the updated motion vector for each prediction time generated by the motion vector calculation unit for each prediction time into a backward semi-Lagrangian technique.

In an embodiment of the present invention, the initial vector calculator may minimize the precipitation field by using the two-dimensional radar precipitation grid field of two or three consecutive time zones as an input value of the cost function.

In an embodiment of the present invention, the initial vector calculator performs a complex gradient with an arbitrary initial guess as an input, and the initial guess uses VET algorithm data from a previous time or Alternatively, a vector of 1s can be used as the initial guess.

In an embodiment of the present invention, the motion vector calculating unit repeatedly generates motion vectors of the x-direction speed u and the y-direction speed v at preset time intervals using the Burgers equation, and linearly converts the repeatedly generated motion vector It can be downsized to the size of the prediction field using interpolation.

In an embodiment of the present invention, the motion vector calculator approximates the vector using a Burgers equation that does not consider a pressure term in Navier-Stokes equations, and the Burgers equation is a preset diffusion coefficients can be used.

According to the ultra-short-term precipitation forecasting method using the Burgers equation and the anti-Lagrangian method, the short-term prediction accuracy can be improved by updating the movement vector every hour using the Burgers equation. In addition, by using the anti-Lagrangian method without directly solving the advection-diffusion equation, the algorithm can be greatly simplified and the accuracy of the prediction field generation result can be increased.

1 is a block diagram of an ultra-short-term precipitation forecasting apparatus using the Bergers equation and the anti-Lagrangian method according to an embodiment of the present invention and a schematic diagram of an overall algorithm.
FIG. 2 is a schematic diagram illustrating a VET algorithm of the initial vector calculator of FIG. 1 .
FIG. 3 is a schematic diagram illustrating a motion vector generation algorithm of the motion vector generator of FIG. 1 .
4 is a diagram showing a velocity vector generated from real examples and the VET algorithm.
5 is a diagram showing a motion vector calculated by the VET algorithm at each time and a vector created by the Bergers equation.
6 is a view showing an observation field at the same time, a prediction field generated from the prior art and an algorithm of the present invention.
7 is a view showing a comparison of category evaluations of prediction fields generated from the prior art and the algorithm of the present invention by classifying actual cases into convective rainfall and stratified rainfall.
8 is a flowchart of an ultra-short-term precipitation forecasting method using the Burgers equation and the anti-Lagrangian method according to an embodiment of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS [0012] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS [0010] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS [0010] Reference is made to the accompanying drawings, which show by way of illustration specific embodiments in which the present invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the present invention. It should be understood that the various embodiments of the present invention are different but need not be mutually exclusive. For example, certain shapes, structures, and characteristics described herein with respect to one embodiment may be implemented in other embodiments without departing from the spirit and scope of the invention. In addition, it should be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the present invention. Accordingly, the following detailed description is not intended to be taken in a limiting sense, and the scope of the present invention, if properly described, is limited only by the appended claims, along with all scope equivalents to those claimed. Like reference numerals in the drawings refer to the same or similar functions throughout the various aspects.

Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the drawings.

1 is a block diagram of an ultra-short-term precipitation forecasting apparatus using the Bergers equation and the anti-Lagrangian method according to an embodiment of the present invention and a schematic diagram of an overall algorithm.

The ultra-short-term precipitation forecasting device (10, hereinafter) using the Burgers equation and the anti-Lagrangian method according to the present invention proposes a method of updating a vector every hour to improve the existing MAPLE algorithm. To improve the vector at each prediction time, the vector is generated by solving the Bergers equation.

Unlike conventional MAPLE, we propose a short-term prediction model that generates an improved prediction field by substituting the updated vector using the vector calculated by the motion vector calculation unit to the anti-Lagrangian algorithm, which is the prediction field generation unit of the existing MAPLE algorithm.

Referring to FIG. 1 , an apparatus 10 according to the present invention includes an initial vector calculator 110 , a motion vector calculator 130 , and a prediction field generator 150 .

In the device 10 of the present invention, software (application) for performing ultra-short-term precipitation forecasting using the Burgers equation and the anti-Lagrangian method may be installed and executed, and the initial vector calculation unit 110, the motion vector calculation The configuration of the unit 130 and the prediction field generator 150 can be controlled by software for performing the ultra-short-term precipitation forecast using the Burgers equation and the anti-Lagrangian method executed in the apparatus 10. there is.

The device 10 may be a separate terminal or may be a part of a module of the terminal. In addition, the configuration of the initial vector calculator 110 , the motion vector calculator 130 , and the prediction field generator 150 may be formed of an integrated module or one or more modules. However, on the contrary, each configuration may be formed of a separate module.

The device 10 may be movable or stationary. The apparatus 10 may be in the form of a server or an engine, and may be a device, an application, a terminal, a user equipment (UE), a mobile station (MS), or a wireless device. (wireless device), may be called other terms such as a handheld device (handheld device).

The device 10 may execute or manufacture various software based on an operating system (OS), that is, the system. The operating system is a system program for software to use the hardware of the device, and is a mobile computer operating system such as Android OS, iOS, Windows Mobile OS, Bada OS, Symbian OS, Blackberry OS and Windows series, Linux series, Unix series, It can include all computer operating systems such as MAC, AIX, and HP-UX.

The initial vector calculating unit 110 calculates an initial motion vector by using a variational echo tracking (VET) algorithm using a two-dimensional radar precipitation grid field of successive time zones as an input value.

In order to calculate the motion vector every time, an accurate initial motion vector is required. In the present invention, the same VET algorithm as MAPLE is used to calculate the initial motion vector. The advantage of the VET algorithm is that it can generate a vector field even in a clear area without precipitation in the precipitation field to generate a full vector field. The created vector field is a two-dimensional motion vector, V(t0, x', y')=(u(t0, x', y'), v(t0, x', y')).

The required input is a two-dimensional radar precipitation grid in two or three consecutive time zones. 1 shows an algorithm flow diagram of an initial vector calculator (or a differential echo tracking technique).

Referring to FIG. 1 , for example, if the size of the entire precipitation image consists of an n×n grid, two or three n×n matrices are input as input values. However, the calculated data is mainly m × m, which calculates a vector field with a size much smaller than the image grid size (n × n), and is used by interpolating the image grid size n × n in the future prediction field generation part. For example, the number of grids of the input field tested in the present invention is n=1248, and the number of vector grids is m=25.

Referring to FIG. 2 , the precipitation field at time t0 and the preceding time (t0-βt) is input into a cost function that minimizes. In this case, the cost function is as shown in Equation 1 below.

[Equation 1]

과

는 각각 다음의 수학식 2 및 수학식 3과 같다.here,

class

is the following Equation 2 and Equation 3, respectively.

[Equation 2]

[Equation 3]

는 두 공간 변수 x와 y의 모든 이차 편미분의 합을 의미한다.

(수학식 2) 항은

와

에서 라그랑지안을 만족하는 벡터를 찾기 위한 항이고,

(수학식 3) 항은 벡터를 단지 매끄럽게하기 위한 벌칙 함수(penalty function)라고 할 수 있다. here,

is the sum of all quadratic partial derivatives of the two spatial variables x and y.

(Equation 2) The term is

Wow

is a term to find a vector that satisfies the Lagrangian in

The term (Equation 3) can be referred to as a penalty function for just smoothing a vector.

When performing the minimization process, for example, a complex gradient method is used, and in this method, an arbitrary initial guess must be input. At this time, if there is VET data from the previous time, the initial value is used. The time of the minimization process may vary depending on the selection of the initial value.

The partial differentiation calculation method is as follows.

등의 일차 공간 미분은 다음과 같은 중앙유한차분법(centered FDM) 등을 이용할 수 있다.The motion vector fields u and v are given as mxm square matrices,

For the first-order spatial differentiation of , etc., the following centered FDM method may be used.

는 가로방향 i번째, 세로방향 j 번째 격자를 의미하고, h는 격자 해상도로, 붙어있는 두 격자 사이의 거리로 ‘km’ 나 ‘m’ 의 단위로 표현할 수 있고, 레이더 강수자료의 거리 단위와 일치시켜 사용하여야 한다. 2차 미분도 마찬가지로 다음과 같이 중앙유한차분법을 이용한다.here

denotes the horizontal i-th and vertical j-th grids, h is the grid resolution, and the distance between two adjacent grids can be expressed in units of 'km' or 'm', and the distance unit of radar precipitation data and must be used in harmony. Similarly, the second derivative uses the central finite difference method as follows.

The motion vector calculator 130 generates a motion vector for each predicted time using the Burgers equation using the vector calculated from the initial vector calculator 110 as an initial value.

Referring to FIG. 3 , the existing MAPLE method can be improved through the motion vector calculator 130 . The motion vector calculation unit 130 generates a vector according to time by numerically solving the m × m vector equation independently of the prediction field generation unit 150 .

An equation for generating a vector according to time is a two-dimensional Burgess equation as shown in Equation 4 below.

[Equation 4]

In fact, in order to obtain an accurate motion vector, it is necessary to solve a complex equation that considers an accurate initial vector and more environmental variables, but this is not possible in a short-term prediction system. Therefore, in the Navier-Stokes equations, the vector is approximated using the Burgers equation without considering the pressure term, and the vector calculated from VET is used as the initial value.

Referring to FIG. 3 , the iteration algorithm in the motion vector calculating unit 130 is shown. For example, assuming that the time interval (βt) for solving the equation for time is 1 minute and the prediction field is calculated every 10 minutes for a total of 3 hours, the total number of iterations to obtain the speed u in the x direction and the speed v in the y direction 3 hours / 1 minute = 180 times.

Since the predicted field is calculated every 10 minutes, a 10-minute interval vector matched with the predicted field size is generated using interpolation after 180 iterations. In order to increase the accuracy of the motion vector calculation, it is possible to prevent unnecessary time and memory waste by calculating in a small time step (eg, 1 minute), interpolating only the required prediction time, and using it in the prediction field generator 150 .

3, (x', y') is a lattice point in m × m, and (x, y) is a lattice point in n × n.

In Equation 4, the diffusion coefficient 's' has an effect of smoothing the vector as the value increases. For example, s=0.2 was fixed and used. To calculate vector values after t0, well-known RK4 can be used to solve numerical ordinary differential equations through loops.

The calculation method of the ordinary differential equation is as follows.

이 주어진, 다음과 같이 일반적인 상미분 방정식의 해를 수치적으로 구하고자 할 때 잘 알려진 Runge-Kutta order 4(RK4) 방법을 이용하여 다음 시간 스텝의

의 값을 구할 수 있다.initial value

Given this, when you want to numerically find the solution of a general ordinary differential equation as follows, you can use the well-known Runge-Kutta order 4 (RK4) method to

value can be obtained.

<RK4 Iteration>

The prediction field generation unit 150 generates a prediction field by substituting the motion vector updated for each prediction time generated by the motion vector calculation unit 130 to the backward anti-Lagrangian technique for each prediction time.

The method of generating a prediction field in the algorithm of the present invention is almost similar to that of MAPLE. However, in MAPLE, a prediction field is generated through a loop until a desired time using a fixed motion vector created in VET, but the algorithm of the present invention uses a motion vector that changes every hour.

The main method of the conventional MAPLE is to find a motion vector using the VET algorithm, and to generate a predicted precipitation field by substituting the motion vector into the anti-Lagrangian method. The anti-Lagrangian post method is expressed in Equation 5 below.

[Equation 5]

, 위치

에서의 예측장은 시각

에서의 예측장의 위치

에서의 값으로 근사된다는 의미이다. 여기서, 초기벡터

는 VET 알고리즘으로부터 얻어지며, 총 예측시간이 L이고 예측생성 간격이

이면, 반복횟수는

이다. 반복문을 통해 원하는 시간까지 예측장을 생성할 수 있다. 여기에 사용된

는 L시간 동안 고정된 값이 사용된다.That is, what Equation 5 means is

, location

The forecast field in

the location of the forecast field in

This means that it is approximated to the value in . Here, the initial vector

is obtained from the VET algorithm, where the total prediction time is L and the prediction generation interval is

If so, the number of iterations is

am. The prediction field can be generated up to the desired time through the loop. used here

A fixed value is used for L time.

The advantage of MAPLE is that the advection equation is calculated using the backward anti-Lagrangian method, which is faster and more stable than solving the advection equation using the forward method. Therefore, the present invention uses the backward anti-Lagrangian method, but improves the MAPLE algorithm, which is an existing ultra-short prediction algorithm, as shown in Equation 6 below by using the vector updated for each time generated by the motion vector calculator 130 can do.

[Equation 6]

은 이동벡터 산출부(130)에서 예측시각마다 산출된, 시각

에서의 이동벡터이다. 즉, 후방 반-라그랑지안 방법에서 고정된 하나의 벡터가 아닌 예측시간마다 업데이트된 새로운 벡터에 의해 예측장이 생성된다. here,

is calculated for each predicted time by the motion vector calculating unit 130, the time

is the motion vector in . That is, in the backward anti-Lagrangian method, a prediction field is generated by a new vector updated every prediction time, not by one fixed vector.

To verify the MAPLE improvement method using motion vectors, six cases in the summer of 2012 and summer of 2014 were selected as shown in FIG. 4 .

의 관계식으로 반사도를 강우강도로 나타낸 값이다. 영상의 격자 크기는 전체 1248 x 1248로, 한 격자의 거리는 0.25 km를 나타낸다. Referring to FIG. 4 , a 312 (km) x 312 (km) area is shown as three radar synthesis areas in the southeastern sea of Korea and the inland region,

It is a value expressing reflectivity as rainfall intensity with the relational expression of . The grid size of the image is 1248 x 1248 in total, and the distance of one grid is 0.25 km.

The number of grids in red in FIG. 4 is 1000 x 1000. A vector is created using this red area, and accuracy evaluation is also performed within the red area. Here, each observed precipitation field was calculated at 2.5-minute intervals, and vectors were calculated from two observation stations at 2.5-minute intervals using the VET algorithm, and the predicted field was calculated every 5 minutes for up to 3 hours, and the results were tested.

In order to compare and verify the quantitative accuracy of the prediction field, a categorical accuracy test method is used (German and Zawadzki (2002)). The category accuracy test uses the rainfall presence/absence division table of the rainfall intensity category accuracy test in Table 1 below. It is defined as Equation 7 below.

RObs ≥ r0 RObs < r0 RF ≥r0 a(hits) c(false alarms) RF < r0 b (misses) d(correct negative)

[Equation 7]

이다. 표 1에서 RObs는 관측장의 강우강도, RF는 예측장의 강우강도 값으로, 문턱값 r0를 기준으로 각 조건을 만족하는 격자수를 각각 a(hits), b(misses), c(false alarms), d(correct negative)로 나타낸다. 모두 0과 1사이의 값이며, FAR은 값이 작을수록, POD, CSI, ETS는 값이 클수록 정확한 예측장을 의미한다. here,

am. In Table 1, RObs is the rainfall intensity of the observation site and RF is the rainfall intensity value of the forecasting field. It is denoted by d (correct negative). All values are between 0 and 1, and the smaller the FAR value, the larger the POD, CSI, and ETS values mean more accurate prediction fields.

5 shows motion vectors calculated from VET at 3 o'clock, 4 o'clock, and 5 o'clock on June 30, 2012. In MAPLE, prediction is made using only one motion vector created at 3 o'clock. However, if you look at the motion vector created in VET (the first row), you can see that the motion vector changes for each time.

The second row shows the motion vector generated by the motion vector calculating unit 130 of the present invention for the 3 o'clock motion vector generated in the VET. In FIG. 5, it can be seen that the motion vector generated from the Burgers equation using only the motion vector at 3 o'clock shows a similar pattern to the motion vector of the VET calculated using the observation field at each time.

In the motion vector calculating unit 130, the Burgers equation is a 25 x 25 matrix, and is vectored using a 2.5-minute interval loop. A 25 x 25 vector calculated every 2.5 minutes has a size of 1248 x 1248 and is used by interpolation/extrapolation to a size of 1248 x 1248, which is the entire image size, for substituting the prediction field generation algorithm as in Equation 8 below.

[Equation 8]

6 shows the results of MAPLE generated at the same time as the observation field at one time interval for three hours from 19:00 on June 30, 2012 and the prediction field of the present invention algorithm updated by the Bergers equation.

Referring to FIG. 6 , the predicted field calculated by the improved MAPLE algorithm is different from the MAPLE result, and looks more similar to the actual observation field.

In Fig. 6, the first row is an observation field at one time interval for three hours from 19:00 (LST) on June 30, 2012, the second row is a prediction field made from MAPLE at the same time, and the third row is the algorithm of the present invention It is a prediction field made from

In addition, the case in FIG. 4 was classified into convective rainfall and stratified rainfall and compared with MAPLE. The categories were evaluated by dividing them into two types and averaging them. As a result, in the case of this method, the verification result was much better in the convective case.

Fig. 7(a) is a view comparing the average category accuracy evaluation values in the case of stratified rainfall, and Fig. 7(b) is a comparison result in the case of convective rainfall. In the case of stratiform type, there was no clear difference between the results of MAPLE and the improved MAPLE of the present invention, but it can be seen that the POD is slightly higher in the results of the present invention.

In the case of convective rainfall in Fig. 7(b), it can be seen that the POD, CSI, and ETS of the present method using a moving vector are much higher.

In conclusion, when a 3-hour forecast field was generated using 6 cases by the method presented in the present invention, the accuracy evaluation results were similar or slightly improved in the case of the stratified rainfall case, and in the case of the convective rainfall case, the category accuracy evaluation showed much higher scores in

8 is a flowchart of an ultra-short-term precipitation forecasting method using the Burgers equation and the anti-Lagrangian method according to an embodiment of the present invention.

The ultra-short-term precipitation forecasting method using the Burgers equation and the anti-Lagrangian method according to the present embodiment may proceed in substantially the same configuration as the apparatus 10 of FIG. 1 . Accordingly, the same components as those of the device 10 of FIG. 1 are given the same reference numerals, and repeated descriptions are omitted.

In addition, the ultra-short-term precipitation forecasting method using the Burgers equation and the anti-Lagrangian method according to the present embodiment may be executed by software (application) for performing the ultra-short-term precipitation forecasting using the Burgers equation and the anti-Lagrangian method.

The present invention proposes a method of updating a vector every time in order to improve the existing MAPLE algorithm. To improve the vector at each prediction time, the vector is generated by solving the Bergers equation.

Unlike conventional MAPLE, we propose a short-term prediction model that generates an improved prediction field by substituting the updated vector using the vector calculated by the motion vector calculation unit to the anti-Lagrangian algorithm, which is the prediction field generation unit of the existing MAPLE algorithm.

Referring to FIG. 8 , the ultra-short-term precipitation forecasting method using the Burgers equation and the anti-Lagrangian method according to the present embodiment uses a VET algorithm (variational echo tracking) using a two-dimensional radar precipitation grid field of a continuous time period as an input value. to calculate the initial motion vector (step S10).

In the step of calculating the initial motion vector (step S10), the two-dimensional radar precipitation grid field of two or three consecutive time zones is taken as an input value, and the received two-dimensional radar precipitation grid field is inputted into the cost function to precipitation. Go through the process of minimizing the intestine.

Here, in the step of minimizing the precipitation field, a complex gradient may be performed with an arbitrary initial guess as an input, and the initial guess may be obtained using VET algorithm data from a previous time or Alternatively, a vector of 1s can be used as an initial guess.

A motion vector is generated for each prediction time using the Burgers equation using the calculated vector as an initial value (step S30).

The step of generating the motion vector for each predicted time (step S30) is to repeatedly generate the motion vectors of the x-direction velocity u and the y-direction velocity v at preset time intervals using the Burgers equation, and repeatedly generated The motion vector is downsized to the size of the prediction field using linear interpolation.

In this case, the vector can be approximated using the Burgers equation without taking the pressure term into account in the Navier-Stokes equations. In addition, the Burgers equation may use a preset diffusion coefficient.

A prediction field is generated by substituting the updated motion vector for each generated prediction time for each prediction time in the backward anti-Lagrangian technique (step S50).

The present invention is an anti-Lagrangian method using a backward method as in MAPLE instead of solving a time-dependent differential equation so that the prior art method is suitable for practical use. That is, the difference from MAPLE is that the short-term prediction is improved by updating the motion vector every hour using the Bergers equation. is much simplified.

Such an ultra-short-term precipitation forecasting method using the Burgers equation and the anti-Lagrangian method may be implemented as an application or implemented in the form of program instructions that may be executed through various computer components and recorded in a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, etc. alone or in combination.

The program instructions recorded on the computer readable recording medium are specially designed and configured for the present invention, and may be known and used by those skilled in the art of computer software.

Examples of the computer-readable recording medium include a hard disk, a magnetic medium such as a floppy disk and a magnetic tape, an optical recording medium such as a CD-ROM and DVD, and a magneto-optical medium such as a floppy disk. media), and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like.

Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware device may be configured to operate as one or more software modules for carrying out the processing according to the present invention, and vice versa.

Although the above has been described with reference to the embodiments, those skilled in the art can variously modify and change the present invention within the scope without departing from the spirit and scope of the present invention described in the claims below You will understand.

The present invention is an algorithm that improves the accuracy of ultra-short-term prediction with a model that is lighter in scale than the existing numerical forecasting model and has higher accuracy than MAPLE.

10: Ultra-short-term precipitation forecasting device
110: initial vector calculation unit
130: motion vector calculation unit
150: prediction field generator

Claims (13)

calculating an initial motion vector using a VET algorithm (variational echo tracking) using a two-dimensional radar precipitation grid field of successive time zones as an input value;
generating a motion vector for each prediction time using a Burgers equation using the calculated initial motion vector as an initial value; and
generating a prediction field by substituting the updated motion vector for each generated prediction time for each prediction time into a backward semi-Lagrangian technique;
The step of generating a motion vector for each prediction time comprises:
repeatedly generating motion vectors of x-direction velocity u and y-direction velocity v at preset time intervals using the Burgers equation; and
Downsizing the iteratively generated motion vector to the size of the prediction field using linear interpolation.
The method of claim 1, wherein calculating the initial motion vector comprises:
receiving two or three consecutive two-dimensional radar precipitation grid fields as input values; and
Minimizing the precipitation field by inputting the received two-dimensional radar precipitation grid field into a cost function; an ultra-short-term precipitation forecasting method using the Bergers equation and the anti-Lagrangian method, including.
3. The method of claim 2, wherein minimizing the precipitation field comprises:
An ultra-short-term precipitation forecasting method using the Burgers equation and the anti-Lagrangian method, which performs a complex gradient with an initial guess as input.
4. The method of claim 3,
The initial guess value is an ultra-short-term precipitation forecasting method using VET algorithm data from a previous time or using a vector consisting of 1 as an initial guess value, using the Burgers equation and the anti-Lagrangian method.
delete The method of claim 1, wherein the step of repeatedly generating the motion vectors of the x-direction velocity u and the y-direction velocity v comprises:
An ultra-short-term precipitation forecasting method using the Burgers equation and the anti-Lagrangian method, which approximates a vector using the Burgers equation without taking the pressure term into account in the Navier-Stokes equations.
7. The method of claim 6,
The Burgers equation is an ultra-short-term precipitation forecasting method using the Burgers equation and the anti-Lagrangian method, which uses a preset diffusion coefficient.
A computer readable computer program for performing the ultra-short-term precipitation forecasting method using the Burgers equation and the anti-Lagrangian method according to any one of claims 1 to 4, 6 and 7 storage medium.
an initial vector calculation unit that calculates an initial motion vector using a VET algorithm (variational echo tracking) using a two-dimensional radar precipitation grid field of a continuous time period as an input value;
a motion vector calculator for generating a motion vector for each prediction time using a Burgers equation using the vector calculated from the initial vector calculator as an initial value; and
a prediction field generator generating a prediction field by substituting the updated motion vector for each prediction time generated by the motion vector calculation unit at each prediction time in a backward semi-Lagrangian technique;
The motion vector calculation unit,
Burger that repeatedly generates motion vectors of x-direction velocity u and y-direction velocity v at preset time intervals using the Burgers equation, and downsizing the repeatedly generated motion vector to the size of the prediction field using linear interpolation. An ultra-short-term precipitation forecasting device using the S equation and the anti-Lagrangian method.
The method of claim 9, wherein the initial vector calculator,
An ultra-short-term precipitation forecasting device using the Burgers equation and the anti-Lagrangian method that minimizes the precipitation field by using the two-dimensional radar precipitation grid field of two or three consecutive time zones as the input value of the cost function.
The method of claim 10, wherein the initial vector calculator,
Burgers, which performs a complex gradient with an initial guess as input, and uses the VET algorithm data from the previous time or a vector of 1s as the initial guess for the initial guess. An ultra-short-term precipitation forecasting device using equations and anti-Lagrangian method.
delete The method of claim 9, wherein the motion vector calculator comprises:
The vector is approximated using a Burgers equation that does not take the pressure term into account in the Navier-Stokes equations, and the Burgers equation uses a preset diffusion coefficient, the Burgers equation and the anti-Lagrangian method. An ultra-short-term precipitation forecasting device.

Images