An independent regional meteorological element forecast error assessment method based on multidimensional deviation

Abstract

The present invention discloses a method for evaluating the forecast error of meteorological elements in independent regions based on multidimensional deviation. In the same coordinate grid, the forecast error of meteorological elements in the forecast field is evaluated by respectively calculating the distance error and axial error of meteorological element monomers in the forecast field relative to meteorological element monomers in the observation field. The meteorological element monomer refers to a spatially continuous closed area formed by scalar meteorological elements in the weather system. The present invention uses the concept of graphical similarity probability to evaluate the forecast error of meteorological elements. Compared with the "double penalty" problem caused by traditional point-to-point inspection and right-wrong inspection, the present invention uses normalized meteorological element value-area curves for error analysis, which can more reasonably and objectively reflect the overall accuracy of meteorological elements.

Description

Technical Field

[0001] The invention belongs to the technical field of intelligent weather forecasting, and in particular relates to an evaluation method for forecast errors based on distribution structure characteristics of meteorological elements. Background Art

[0002] With the development of numerical forecasting technology and the improvement of computing power, numerical models have the ability to forecast smaller-scale weather systems. The quantitative calculation of weather forecast errors has always been a hot topic in many studies. A reasonable error definition can not only objectively reflect the forecast effect, but also be used in artificial intelligence models to guide artificial intelligence models to learn in an objective and correct direction.

[0003] At present, the main factor in forecast error assessment is precipitation. Domestic scientific research and business personnel still use traditional classification test methods based on target hit rate as the core for precipitation forecast verification and evaluation, such as TS scoring. These test methods mainly consider the point-to-point comparison of precipitation intensity and use the presence / absence judgment to obtain an overall evaluation of precipitation. For high-resolution precipitation forecasts, even if the structure and intensity of the rain belt are consistent with the actual situation, due to the slight deviation of the rain belt position, it may still produce excessive false alarm rate and missed alarm rate, resulting in a low final forecast score. This type of traditional statistical test method causes precipitation forecasts to suffer "double penalties" caused by small differences in space and time in high-resolution model verification and evaluation. It is difficult to obtain sufficient evaluation information and cannot objectively reflect the true forecasting ability of precipitation.

[0004] To solve this problem, researchers have developed a variety of spatial verification methods. The main spatial verification methods currently include neighborhood methods, scale decomposition methods, and spatial verification methods for object attributes. Spatial verification methods for object attributes mainly focus on analyzing the location, shape, intensity and other attribute information of the object. The most representative one is the Method for Object-based Diagnostic Evaluation (MODE) method. The MODE method determines the degree of matching between objects by calculating the similarity characteristics of the test objects. This method uses ellipse fitting technology to fit the test objects into multiple independent ellipse areas, and then uses the centroid distance, area ratio, direction angle, overlap ratio and other attributes of the ellipse as separate factors to calculate the overall spatial similarity characteristics. In the MODE method, the test area is fitted with an ellipse, which is affected by multiple factors such as the smoothing radius and the filtering threshold. The evaluation conclusions obtained by different smoothing radii lack consistency. In addition, the MODE method does not ultimately summarize the individual error factors of the test object into a factor that can be intuitively understood by non-professional users. The neighborhood space verification method focuses on adjusting the forecast and observation data to a larger scale by upscaling, reducing the accidental information of high-resolution data by spatial smoothing or statistical probability distribution, and thus measuring the similarity characteristics between forecast and observation. From a theoretical perspective, this method is not suitable for the evaluation of high-resolution forecast results. Summary of the invention

[0005] Purpose of the invention: In view of the above-mentioned existing problems and shortcomings, the purpose of the present invention is to provide an independent regional meteorological element forecast error assessment method based on multidimensional deviation, and to use the concept of graphic similarity probability to assess the meteorological element forecast error. Compared with the "double penalty" problem caused by traditional point-to-point inspection and right-wrong inspection, the present invention uses the normalized meteorological element value-area curve for error analysis, which can more reasonably and objectively reflect the overall accuracy of meteorological elements.

[0006] Technical solution: To achieve the above invention purpose, the present invention adopts the following technical solution: a method for evaluating the forecast error of meteorological elements in independent regions based on multidimensional deviation, in the same coordinate grid, by respectively calculating the forecast field meteorological element monomer d (f) Relative observation field meteorological element monomer d (o) The distance error and axial error of the forecast field meteorological element are used to evaluate the forecast error of the forecast field meteorological element. The meteorological element monomer refers to the spatially continuous closed area formed by the scalar meteorological elements in the weather system. Further, the forecast field meteorological element monomer d (f) Relative observation field meteorological element monomer d (o) The calculation process of distance error and axial error includes the following steps:

[0007] Step S1, calculate the forecast field meteorological element monomer d (f) Relative observation field meteorological element monomer d (o) The center of gravity of the forecast field meteorological element monomer is d (f) The horizontal and vertical coordinates of

[0008]

[0009] In the formula, x and y represent the horizontal and vertical coordinates of each point in the forecast meteorological element monomer in the coordinate system, respectively. f (xy) represents the value of the meteorological element at each point in the meteorological element monomer of the forecast field. Indicates the forecast field meteorological element monomer d (f) Integrate all points in W x (d (f) ) represents the meteorological element monomer d of the forecast field (f) The horizontal coordinate of the center of gravity, W y (d (f) ) represents the meteorological element monomer d of the forecast field (f) The ordinate of the center of gravity;

[0010] For the observed meteorological element monomer, the observed meteorological element monomer d is calculated by equations (3) and (4) respectively. (o) The horizontal and vertical coordinates of

[0011]

[0012] In the formula, x and y represent the horizontal and vertical coordinates of each point of the meteorological element monomer in the observation field in the coordinate system, respectively. o (x, y) represents the value of the target element at each point of the meteorological element monomer in the observation field. Represents the observation field meteorological element monomer d (o) Integrate all points in W x (d (o) ) represents the meteorological element monomer d of the observation field (o) The horizontal coordinate of the center of gravity, W y (d (o) ) represents the meteorological element monomer d of the observation field (o) The ordinate of the center of gravity;

[0013] Step S2, calculation of distance error Derr

[0014] The normalized center distance of the forecast field meteorological element relative to the observation field meteorological element is calculated by formula (5) and used as the distance error Derr between the two.

[0015]

[0016] In the formula, Dis represents the characteristic scale of the weather system;

[0017] Step S3, calculation of axial error Oerr

[0018] The forecast meteorological element monomer is translated so that its center of gravity coincides with the center of the forecast monomer, and then a polar coordinate system is established with the center as the origin. The rotation angle is calculated according to formula (6): After that, the overlapping area of ​​the forecast meteorological element monomer and the observed meteorological element monomer, Ψ,

[0019]

[0020] In the formula, Indicates the clockwise rotation angle of the meteorological element monomer in the forecast field The new forecast field meteorological element monomer after that;

[0021] The clockwise rotation angle is calculated by the error function of formula (7): The error between the new forecast meteorological element monomer and the observed meteorological element monomer

[0022]

[0023] d (f) \ψ represents the area remaining after deducting the overlapping area from the forecast meteorological element monomer. The meteorological elements in this area are all empty report errors, which is expressed as d (o) \ψ represents the area remaining after deducting the overlapping area from the observed meteorological element monomer. The meteorological elements in this area are all omission errors, which is expressed as

[0024] when The rotation angle corresponding to the minimum value As the axial error, it is normalized by formula (8) to obtain the normalized axial error Oerr,

[0025]

[0026] Step S4, calculation of comprehensive error Err

[0027] Calculate the comprehensive error Err of the forecast meteorological element monomer relative to the observed meteorological element monomer, Err = Derr + Oerr. Further, the meteorological element represents rainfall, radar reflectivity, temperature, visibility, and wind speed.

[0028] Furthermore, the characteristic scale Dis of the weather system in step S2 is divided according to the classification of weather systems in meteorology: when it belongs to a tornado or a small-sized convective system, the corresponding characteristic scale Dis is 10 km; when it belongs to a squall line or a super convective cell, the corresponding characteristic scale Dis is 100 km; when it belongs to a front, a cyclone, or a typhoon, the corresponding characteristic scale Dis is 1000 km.

[0029] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0030] (1) The present invention abstracts the problem of meteorological element forecast error correction into graphics and mathematical models, which is more in line with people's intuitive feelings when judging errors, and is convenient for using advanced mathematical and graphical methods to solve the problem.

[0031] (2) Compared with the prior art which adopts the fitting ellipse method and only considers the spatial distribution of elements, and inevitably introduces the defect of error, the present invention uses the original shape of the monomer for comparison, and uses the significance of the center of gravity, while considering the influence of the element value and the element distribution graph, and can more reasonably and objectively reflect the overall accuracy of the meteorological elements.

[0032] (3) In the present invention, the distance error and axial error of the monomer are normalized, and different types of errors can be directly added together, thus solving the problem that different types of errors cannot be added together. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a schematic diagram of the distance error between the forecast field meteorological element monomer and the observed meteorological element monomer according to the present invention;

[0034] Figure 2 It is a schematic diagram of the axial error between the forecast field meteorological element monomer and the observed meteorological element monomer according to the present invention;

[0035] Figure 3 This is a schematic diagram of the distribution of meteorological monomers in the meteorological element field of the weather system described in the present invention. DETAILED DESCRIPTION

[0036] The present invention is further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0037] The target element monomer in the forecast field is defined as d (f) , the target element monomer in the observation field is defined as d (o) , the error evaluation problem is transformed into finding d in the same coordinate grid. (f) and d(o) In the present invention, the overall error is decomposed into two monomers: the distance error and the axial error.

[0038] Step 1: Calculate the single meteorological element d of the forecast field respectively (f) and observed meteorological element monomer d (o) The center of gravity contains the numerical size and spatial coordinate information of the target element.

[0039] For the forecast field meteorological element monomer d (f) For example,

[0040]

[0041] In formulas (1) and (2), x and y represent the horizontal and vertical coordinates of each point in the forecast meteorological element monomer in the coordinate system, respectively. f (x, y) represents the value of the target element at each point in the forecast meteorological element monomer. Indicates the forecast field meteorological element monomer d (f) Integrate all points in W x (d (f) ) represents the forecast field meteorological element monomer d (f) The horizontal coordinate of the center of gravity, W y (d (f) ) represents the forecast field meteorological element monomer d (f) The vertical coordinate of the center of gravity.

[0042] For the observed meteorological element monomer d (o) For example,

[0043]

[0044] In formulas (3) and (4), x and y represent the horizontal and vertical coordinates of each point of the observed meteorological element in the coordinate system, respectively. o (x, y) represents the value of the target element at each point of the observed meteorological element monomer. Indicates the observed meteorological element monomer d (o) Integrate all points in W x (d (o) ) represents the observed meteorological element monomer d (o) The horizontal coordinate of the center of gravity, W y (d (o) ) represents the observed meteorological element monomer d (o) The vertical coordinate of the center of gravity.

[0045] Step 2: Calculate the distance error between the two monomers. The distance error is defined as the distance between the centroids of the two monomers divided by the characteristic scale of the weather system, thus obtaining the normalized distance error:

[0046]

[0047] (5) In the formula, Dis represents the characteristic scale of the weather system. According to the classification and definition of weather systems in meteorology, the characteristic scale of tornadoes and small-scale convective systems is 10 km, the characteristic scale of weather systems corresponding to squall lines and super convective cells is 100 km, and the characteristic scale of weather systems corresponding to fronts, cyclones, and typhoons is 1000 km. Derr is the distance error after homogenization.

[0048] Step 3: Calculate the axial error of the two monomers. (f) and observed meteorological element monomer d (o) The center of gravity of the forecast field coincides with that of the forecast field, and a polar coordinate system with the center of gravity as the origin is established. (f) The target element distribution is represented by R f (ρ, θ), observed meteorological element monomer d (o) The target element distribution is represented by R o (ρ, θ), where (ρ, θ) represents the polar diameter and polar angle in the polar coordinate system. (o) As a benchmark, let the forecast field meteorological element monomer d (f) Angle of rotation clockwise around the center of gravity The rotated forecast meteorological element monomer and the observed meteorological element monomer d (o) To achieve this, the overlap region is defined

[0049]

[0050] (6) Represents the forecast field meteorological element monomer d (f) Rotate counterclockwise around its center of gravity The new forecast meteorological element monomer is obtained. The target element distribution is expressed as Define the error function:

[0051] where d (f) \ψ represents the forecast field meteorological element monomer d (f) After deducting the overlapping part Ψ, the remaining small area contains all target elements that are empty report errors, which can be expressed as d (o) \ψ represents the observed meteorological element monomer d (o) After deducting the overlap Ψ, the remaining small area, the target element in this area is the underreporting error:

[0052] The value ranges from 1 to 360. The minimum, corresponding to is the minimum axial error, and normalized to obtain the normalized axial error Oerr:

[0053]

[0054] Step 4: The comprehensive error is defined as follows:

[0055] Err=Derr+Oerr (9).

[0056] Taking the precipitation distribution in a certain place in Jiangsu Province in a certain period of time as an example, according to the aforementioned method of the present invention, it is calculated that the corresponding geographical coordinates of the centroid of the precipitation observation field are (185.925km, 122.574km), and the corresponding geographical coordinates of the centroid of the precipitation forecast field are (178.155km, 89.624km). The angle between the axes of the two is 14°. This weather system belongs to a mesoscale weather system, and the corresponding characteristic scale is 100km. The centroid distance error Derr, axis error Oerr and comprehensive error Err of the observation field and the forecast field are calculated as shown below,

[0057] The center of gravity distance error is

[0058] The axis error is Oerr = 14 / 180 × π = 0.2443

[0059] The comprehensive error is Derr+Oerr=0.5738.

[0060] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical solutions and technical concepts of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for evaluating the forecast error of independent regional meteorological elements based on multidimensional deviation, characterized by: In the same coordinate grid, the forecast field meteorological element monomer d (f) Relative observation field meteorological element monomer d (o) The distance error and axial error of the forecast field are used to evaluate the forecast error of the meteorological elements. The meteorological element monomer refers to the spatially continuous closed area formed by scalar meteorological elements in the weather system. The forecast field meteorological element monomer d (f) Relative observation field meteorological element monomer d (o) The calculation process of distance error and axial error includes the following steps: Step S1, calculate the forecast field meteorological element monomer d (f) Relative observation field meteorological element monomer d (o) Center of gravity For the forecast field meteorological element monomer, the forecast field meteorological element monomer d is calculated by equations (1) and (2) respectively. (f) The horizontal and vertical coordinates of In the formula, x and y represent the horizontal and vertical coordinates of each point in the forecast meteorological element monomer in the coordinate system, respectively. f (x, y) represents the value of the meteorological element at each point in the meteorological element monomer of the forecast field. Indicates the forecast field meteorological element monomer d (f) Integrate all points in W x (d (f) ) represents the forecast field meteorological element monomer d (f) The horizontal coordinate of the center of gravity, W y (d (f) ) represents the forecast field meteorological element monomer d (f) The ordinate of the center of gravity; For the observed meteorological element monomer, the observed meteorological element monomer d is calculated by equations (3) and (4) respectively. (o) The horizontal and vertical coordinates of In the formula, x and y represent the horizontal and vertical coordinates of each point of the meteorological element monomer in the observation field in the coordinate system, respectively. o (x, y) represents the value of the target element at each point of the meteorological element monomer in the observation field. Represents the observation field meteorological element monomer d (o) Integrate all points in W x (d (o) ) represents the meteorological element monomer d of the observation field (o) The horizontal coordinate of the center of gravity, W y (d (o) ) represents the meteorological element monomer d of the observation field (o) The ordinate of the center of gravity; Step S2, calculation of distance error Derr The normalized center distance of the forecast field meteorological element relative to the observation field meteorological element is calculated by formula (5) and used as the distance error Derr between the two. In the formula, Dis represents the characteristic scale of the weather system; Step S3, calculation of axial error Oerr The forecast meteorological element monomer is translated so that its center of gravity coincides with the center of the observed meteorological element monomer, and then a polar coordinate system is established with the center as the origin. The rotation angle is calculated according to formula (6): After that, the overlapping area of ​​the forecast meteorological element monomer and the observed meteorological element monomer, Ψ, In the formula, Indicates the clockwise rotation angle of the meteorological element monomer in the forecast field The new forecast field meteorological element monomer after that; The clockwise rotation angle is calculated by the error function of formula (7): The error between the new forecast meteorological element monomer and the observed meteorological element monomer d (f) \ψ represents the area remaining after deducting the overlapping area from the forecast meteorological element monomer. The meteorological elements in this area are all empty report errors, which is expressed as in Represents polar coordinates The forecast value of meteorological elements at the location, Represents polar coordinates Observed values ​​of meteorological elements at the location, Represents polar coordinates Meteorological element forecast value at the location; d (o) \ψ represents the area remaining after deducting the overlapping area from the observed meteorological element monomer. The meteorological elements in this area are all omission errors, which is expressed as when The rotation angle corresponding to the minimum value As the axial error, it is normalized by formula (8) to obtain the normalized axial error Oerr, Step S4, calculation of comprehensive error Err Calculate the comprehensive error Err of the forecast meteorological element monomer relative to the observed meteorological element monomer, Err = Derr + Oerr.

2. The independent regional meteorological element forecast error assessment method based on multidimensional deviation according to claim 1 is characterized by: The meteorological elements represent rainfall, radar reflectivity, temperature, visibility, and wind speed.

3. The independent regional meteorological element forecast error assessment method based on multidimensional deviation according to claim 1 is characterized by: The characteristic scale Dis of the weather system in step S2 is divided according to the classification of weather systems in meteorology: when it belongs to a tornado or a small-sized convective system, the corresponding characteristic scale Dis is 10 km; when it belongs to a squall line or a super convective cell, the corresponding characteristic scale Dis is 100 km; when it belongs to a front, a cyclone, or a typhoon, the corresponding characteristic scale Dis is 1000 km.

Images