A method, system, terminal and storage medium for evaluating the effect of a nowcast

Abstract

The application relates to a nowcast effect evaluation method and system, a terminal and a storage medium. The method comprises the following steps: calculating the membership of radar echo data to a set threshold, calculating the probability mean in the neighborhood range of each grid point by using a spatial neighborhood probability method based on the membership, and obtaining the fuzzy probability of each grid point according to the probability mean; calculating the score of the radar echo data based on the score skill of the Gaussian type membership intensity feature (IFSS); assigning weights to the forecast echo data of each time point by using a time neighborhood adaptive weighting algorithm, calculating the probability of each grid point after the weights are assigned by using a spatial neighborhood probability method, and calculating the score skill of the radar echo data based on the time neighborhood adaptive weighting algorithm (TFSS); and calculating the score skill of the radar echo data based on the intensity feature and the space-time feature (TIFSS) according to the score results of the IFSS and the TFSS. The application can realize more objective and reasonable evaluation and test, and solve the problems of early prediction and delayed prediction.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of weather forecast verification, and particularly relates to a nowcasting effect evaluation method and system, a terminal and a storage medium. BACKGROUND

[0002] With the increasing demand for accurate prediction of disastrous extreme weather, the importance of severe convective weather prediction is increasingly prominent. Good prediction and prevention of convective weather have great significance for disaster prevention, mitigation and rescue. Due to the small space-time scale, fast evolution and high nonlinearity of dynamic and physical processes of severe convective scale weather systems, the fine prediction of severe convective weather still faces great challenges.

[0003] Forecast verification is an important part of the development of severe convective weather prediction business and technology, and its purpose is to give the consistency and difference between the prediction and the actual situation and the possible reasons. Through systematic verification, problems in the prediction technology can be found and improved and perfected, and the applicability and uncertainty of the prediction can be given, so as to improve the business prediction ability and service benefit of severe convective weather. At present, the traditional and conventional verification methods represented by risk score TS (Threat Score), critical success index CSI (Critical Success Index), hit rate POD (Probability of Detection), false alarm rate FAR (False Alarm Rate) and equitable risk score ETS (Equitable Threat Score) have been widely used in severe convective business. However, the traditional severe convective weather deterministic prediction verification method is based on station observation or eyewitness report, and the verification index is calculated through a two-dimensional contingency table. It has very high requirements for spatial consistency, and needs strict "one-to-one correspondence" between the prediction and the actual situation at the grid or station, without position deviation and intensity deviation. For grids, there is a double punishment of "prediction occurs but actual situation does not occur" and "prediction does not occur but actual situation occurs". When the target of the prediction has a certain deviation, the TS score may be very low, but even the prediction with very low TS score may contain valuable information required by users (such as forecasters), such as a prediction with large position prediction deviation but good depiction of the structural characteristics of the convective system. In addition, the conventional verification method only gives the evaluation of whether the prediction is accurate or not or the degree of accuracy, and cannot give the reason for the deviation of the prediction. For some regional targets, only a simple true or false evaluation is given, which hides some positive information in the prediction.

[0004] In order to solve the verification of high spatio-temporal resolution weather forecast, some new spatial verification techniques are emerging, the current spatial verification techniques mainly include neighborhood method, scale decomposition method (Mittermaier et al, 2010; Casati et al, 2004, 2010) and spatial verification method for object properties, however, these methods cannot effectively solve the problems of time lead and lag or intensity error of the prediction results. SUMMARY

[0005] The application provides a nowcasting effect evaluation method, system, terminal and storage medium, and aims to at least solve one of the above technical problems in the prior art to some extent.

[0006] In order to solve the above problems, the application provides the following technical solutions:

[0007] A nowcasting effect evaluation method, comprising:

[0008] The Gaussian membership function is used to calculate the membership of the radar echo data to the set threshold, and based on the membership, the spatial neighborhood probability method is used to calculate the probability mean of each grid neighborhood in the radar echo data, and the probability mean is used as the fuzzy probability of each grid.

[0009] According to the fuzzy probability, the score of the radar echo data based on the Gaussian membership intensity feature is calculated.

[0010] The time neighborhood adaptive weighting algorithm is used to assign weights to the prediction echo data of each time, and the spatial neighborhood probability method with the Gaussian membership is used to calculate the probability of each grid after the weights are assigned.

[0011] According to the probability of each grid, the score of the radar echo data based on the time neighborhood adaptive weighting algorithm is calculated.

[0012] According to the IFSS and TFSS score results, the score of the radar echo data based on the intensity feature and the spatio-temporal feature is calculated.

[0013] The technical solutions adopted by the embodiments of the application further include that the Gaussian membership function used to calculate the membership of the radar echo data to the set threshold is specifically:

[0014] The Gaussian membership function is:

[0015]

[0016] wherein x is the difference between the intensity of the radar echo data and the set threshold value; c is the variance; for the radar echo data whose intensity is greater than or equal to the set threshold value, the membership degree is equal to 1, and for the radar echo data whose intensity is less than the set threshold value, the membership degree is:

[0017]

[0018] wherein P I is the spatial neighborhood probability considering the membership intensity feature, i and j are the grid coordinates in the neighborhood, T is the set threshold value, and Ref represents the echo data.

[0019] The technical scheme adopted by the embodiment of the application further includes that the spatial neighborhood probability method is used to calculate the probability mean value of each grid point neighborhood range in the radar echo data, and the probability mean value is taken as the fuzzy probability of each grid point.

[0020] The neighborhood probability of each grid point in the radar echo data is calculated by using a circular neighborhood window, N r = πr 2 is the number of grid points in the circular neighborhood range, and r is the neighborhood radius.

[0021] The probability mean value of all grid points in the circular neighborhood range is calculated as the fuzzy probability of the current grid point.

[0022] The technical scheme adopted by the embodiment of the application further includes that the score IFSS of the radar echo data based on the Gaussian-type membership intensity feature is calculated according to the fuzzy probability, and the score IFSS is specifically:

[0023]

[0024] wherein M and N are the height and width of the radar echo data, and are the neighborhood probabilities of the forecast echo data and the real-time echo data under the Gaussian-type membership intensity feature, respectively.

[0025] The technical scheme adopted by the embodiment of the application further includes that the time neighborhood adaptive weighting algorithm is used to assign weights to each time point of the forecast echo data, and the weights are specifically:

[0026] In the time neighborhood window, the root mean square error (RMSE) is used as the similarity measurement between the forecast echo data and the real-time echo data at each time point, and the weights of the forecast echo data at each time point are assigned according to the inverse of the RMSE.

[0027] The inverse of the RMSE is calculated according to the following formula:

[0028]

[0029] where t is time neighborhood, Pred t and Obs are respectively the forecast echo data and live echo data at the t time;

[0030] The weight calculation formula of the forecast echo data at each time is:

[0031] w t = IRMSE t / ∑IRMSE t

[0032] The technical scheme adopted by the embodiment of the application further includes that the probability of each grid point after the weight distribution calculated by the spatial neighborhood probability method with the added Gaussian membership is specifically:

[0033]

[0034] where N t is a time neighborhood window, P pred (i,j) is a spatial neighborhood probability considering the membership intensity feature.

[0035] The technical scheme adopted by the embodiment of the application further includes that the score TFSS of the radar echo data based on the time neighborhood adaptive weighting algorithm is calculated based on the probability of each grid point and is specifically:

[0036] If the weight of the current time is higher than that of the previous and next times, the score is calculated only by the forecast result of the current time, otherwise, the score is calculated according to the time neighborhood adaptive weighting algorithm; the score TFSS of the radar echo data based on the time neighborhood adaptive weighting algorithm is:

[0037]

[0038] where, is a neighborhood probability obtained by time neighborhood window adaptive weighting, P obs is an original spatial neighborhood probability.

[0039] The technical scheme adopted by the embodiment of the application further includes that the score TIFSS of the radar echo data based on the intensity feature and the space-time feature is calculated according to the IFSS and the TFSS score results and is specifically:

[0040]

[0041] Another technical scheme adopted by the embodiment of the application is a nowcasting effect evaluation system, which includes:

[0042] The membership calculation module is configured to calculate the membership of radar echo data to a set threshold by using a Gaussian membership function, calculate the probability mean in a neighborhood range of each grid point by using a spatial neighborhood probability method based on the membership, and obtain the fuzzy probability of each grid point according to the probability mean.

[0043] The first scoring module is configured to calculate the score skill score (IFSS) of the radar echo data based on the Gaussian membership intensity feature according to the fuzzy probability.

[0044] The weight distribution module is configured to distribute weights for the forecast echo data of each time point by using a time neighborhood adaptive weighting algorithm, and calculate the probability of each grid point after the weight distribution by using a spatial neighborhood probability method with the Gaussian membership.

[0045] The second scoring module is configured to calculate the score skill score (TFSS) of the radar echo data based on the time neighborhood adaptive weighting algorithm based on the probability of each grid point.

[0046] The third scoring module is configured to calculate the score skill score (TIFSS) of the radar echo data based on the intensity feature and the space-time feature according to the IFSS and the TFSS score results.

[0047] Another technical solution adopted by the embodiment of the application is a terminal, which comprises a processor and a memory coupled with the processor, wherein

[0048] The memory stores program instructions for implementing the nowcasting effect evaluation method.

[0049] The processor is configured to execute the program instructions stored in the memory to control the nowcasting effect evaluation.

[0050] Another technical solution adopted by the embodiment of the application is a storage medium storing program instructions executable by a processor, wherein the program instructions are configured to execute the nowcasting effect evaluation method.

[0051] Compared with the prior art, the nowcasting effect evaluation method, system, terminal and storage medium of the embodiment of the application have the following beneficial effects: the nowcasting effect evaluation method, system, terminal and storage medium of the embodiment of the application weaken the strict spatial matching requirement by introducing the spatial neighborhood scheme, and can achieve more objective and reasonable evaluation and verification. The nowcasting effect evaluation method, system, terminal and storage medium of the embodiment of the application can give more useful information and increase the tolerance of the intensity in the actual verification by introducing the Gaussian membership function in fuzzy logic to mine the value of the forecast grid point below the threshold. The nowcasting effect evaluation method, system, terminal and storage medium of the embodiment of the application can make the result of the current forecast have the time feature by introducing the time neighborhood to adaptively weight the forecast results of certain time points before and after, and solve the deficiencies of the early forecast and the delayed forecast. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 is a flow chart of the nowcast effect evaluation method of the embodiment of the present application;

[0053] Figure 2 is a schematic diagram of the Gaussian membership function of the embodiment of the present application under different c values and the difference between the grid echo intensity and the threshold value;

[0054] Figure 3 is a schematic diagram of the grid probability field of the embodiment of the present application;

[0055] Figure 4 is a schematic diagram of the neighborhood probability method of the embodiment of the present application;

[0056] Figure 5 is a schematic diagram of the structure of the nowcast effect evaluation system of the embodiment of the present application;

[0057] Figure 6 is a schematic diagram of the terminal structure of the embodiment of the present application;

[0058] Figure 7 is a schematic diagram of the structure of the storage medium of the embodiment of the present application. DETAILED DESCRIPTION

[0059] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0060] Please refer to Figure 1 is a flow chart of the nowcast effect evaluation method of the embodiment of the present application. The nowcast effect evaluation method of the embodiment of the present application includes the following steps:

[0061] S10: Obtain radar echo data within a certain time;

[0062] In this step, the obtained radar echo data is 11 pieces of CAPPI data of the new generation S-band Doppler radar reflectivity factor covering the Guangdong region, including 20 times of prediction echo data and real-time echo data within 2 hours every 6 minutes (the time interval is 6 minutes), and the specific time and time parameters can be set according to actual operation. In order to reflect the main evolution characteristics of the short-time strong convective radar echo, the sea level height of the radar echo data selected by the embodiment of the present application is 2.5 kilometers, the horizontal spatial resolution is 0.01°x0.01°, and the horizontal grid number is 700x900.

[0063] S20: calculating the membership of the radar echo data to the set threshold value using a Gaussian membership function, calculating the probability mean of the neighborhood range of each grid point in the radar echo data based on the membership of the radar echo data to the set threshold value, and taking the probability mean as the fuzzy probability of each grid point;

[0064] In this step, the embodiment of the application converts the hard division of 0 / 1 into soft division of fuzzy logic by introducing the membership of the fuzzy logic system, converts the radar echo data less than the set threshold value into the membership of the threshold value, so that the radar echo data less than the set threshold value also has a certain reference value. The specific conversion method is as follows:

[0065] The Gaussian function is defined as:

[0066]

[0067] The Gaussian function is a bell-shaped curve, in which the parameter a controls the amplitude of the function, the parameter b controls the horizontal position of the bell-shaped curve, i.e. the mean value, and c reflects the width of the bell-shaped curve, i.e. the variance. In the embodiment of the application, the maximum membership of the Gaussian membership function is 1, and the horizontal position is set at 0, i.e. a = 1 and b = 0, so the Gaussian membership function is:

[0068]

[0069] where x is the difference between the intensity of the radar echo data and the set threshold value. For the radar echo data whose intensity is greater than or equal to the set threshold value, the membership (i.e. the probability) is equal to 1, and for the radar echo data less than the set threshold value, the membership is determined by the Gaussian membership function:

[0070]

[0071] In the formula, P I is the spatial neighborhood probability considering the membership intensity characteristics, T is the set threshold value, and c is the only variable of the Gaussian membership function, which plays a role in attenuation adjustment of the forecast score. The greater the value of c, the wider the bell-shaped curve, indicating that the tolerance for the echo intensity below the threshold value is greater, i.e. when the magnitude of the forecast and the actual situation differs greatly, the score may not differ greatly; the smaller the value of c, the narrower the bell-shaped curve, indicating that the tolerance for the echo intensity below the threshold value is smaller, i.e. when the magnitude of the forecast and the actual situation increases, the score will rapidly attenuate.

[0072] Specifically, as shown in Figure 2 , it is a schematic diagram of the change of the Gaussian membership function with the difference between the grid echo intensity and the threshold value under different values of c. As shown in Figure 2It can be seen that when the difference between the grid echo intensity and the threshold is greater than 0, the probability is 1, that is, the 0 / 1 type binary in the spatial neighborhood probability method; when the difference is less than 0, the probability is a decimal less than 1 and greater than 0, and the closer the difference is to 0, the closer the probability is to 1, and the greater the difference, the closer the probability is to 0. When c = 5, the curve is very steep, meaning that the tolerance for the difference is smaller, and when c = 20, the curve is relatively moderate, meaning that the tolerance for the difference is relatively high. For example, when c = 5, the membership (probability) of the difference of 5 is 0.61, and when c = 20, the membership (probability) is 0.97. In actual application, when the difference is 5 dBz, the threshold is set to 20 dBz, and the predicted intensity is 15 dBz, the reference value may not be large; and when the threshold is set to 50 dBz, the predicted intensity is 45 dBz, and the reference value at this time is greater than that when the threshold is 20 dBz. Therefore, by setting different thresholds, the scoring conditions under different intensity levels can be simulated.

[0073] The 0 / 1 type binary probability in the spatial neighborhood probability method is converted into a fuzzy probability by the fuzzy logic system of the Gaussian type membership function, and then the probability mean in the neighborhood range of each grid point is calculated by using the spatial neighborhood probability method, so as to obtain the fuzzy probability of each grid point. At this time, the grid probability field is as shown in Figure 3 , wherein the probability of the center grid point is 80.81%.

[0074] Further, the probability mean in the neighborhood range of each grid point is calculated by using the spatial neighborhood probability method, as shown in Figure 4 , which is a neighborhood probability method schematic diagram of the embodiment of the present application. The embodiment of the present application calculates the neighborhood probability of each grid point in the radar echo data by using a circular neighborhood window, wherein N r = πr 2 is the number of grid points in the circular neighborhood range, r is the neighborhood radius, and i and j are the grid point coordinates in the neighborhood. In the figure, the neighborhood radius r takes 3.5 times the grid point distance as an example. In the figure, the grid point value exceeding the threshold is 1, which is represented by a dark background, and the value not exceeding the threshold is 0. It can be seen that there are 37 effective grid points in the neighborhood range, and 17 grid points exceed the threshold, so the probability of the center grid point under the threshold is 45.95%. The above processing is performed on each grid point on the predicted echo data and the live echo data, so that the probability of each grid point exceeding the threshold can be obtained. After the processing, the number of grid points of the predicted echo data and the live echo data remains unchanged, but the value of each grid point is no longer the reflectivity or 1 / 0 of the grid point, but a probability value containing the spatial neighborhood echo information. Subsequent grid point to grid point calculation contains the spatial distribution characteristics in a certain range of the grid point.

[0075] S30: Intensity-based score (IFSS) of radar echo data based on Gaussian membership intensity features, calculated according to fuzzy probability.

[0076] In this step, the IFSS (Indicative Skill Score) formula based on the Gaussian membership strength feature is as follows:

[0077]

[0078] Where M and N are the height and width of the radar echo data, and These represent the neighborhood probabilities of the predicted echo data and the actual echo data under the Gaussian membership intensity characteristics, respectively.

[0079] S40: The time neighborhood adaptive weighting algorithm is used to assign different weights to the predicted echo data at different times, and the spatial neighborhood probability method with Gaussian membership degree is used to calculate the probability of each grid point.

[0080] In this step, the 6-minute time resolution of the severe convective weather radar echo forecast is relatively high. In severe convective weather systems, the movement of rain clouds exhibits strong nonlinear characteristics, which may lead to temporal mismatches in radar echo nowcasting. That is, the forecast score at the current moment may not be high, but the forecast echo data from the previous or next moment may be more closely matched with the current observed echo data in terms of intensity and extent. The previous moment is considered an early forecast, and the next moment is a delayed forecast. Considering the high temporal resolution, the forecast echo data from several consecutive moments may have significant similarities to the current observed echo data. Therefore, different weights can be assigned to the forecast echo data at different moments, with higher weights for similar moments, thus more reasonably calculating the forecast score for the current moment. Therefore, in this embodiment, the root mean square error (RMSE) is used as a similarity measure between the forecast echo data and the observed echo data at different moments within the time neighborhood window. Then, different weights are assigned to the forecast echo data at each moment according to the reciprocal of the RMSE.

[0081] Specifically, the formula for calculating the reciprocal of RMSE is:

[0082]

[0083] Where M and N are the height and width of the radar echo image, i and j are the grid coordinates, and t is the time neighborhood, taking values ​​of 0, ±1, ±2, ±3, etc. t Obs and tb represent the predicted echo data and the actual echo data at time t, respectively.

[0084] The weighting formula for the forecast echo data at each time point is as follows:

[0085] w t = IRMSE t / ∑IRMSE t (6)

[0086] After the weight assignment of the forecast echo data at different times, the grid probability is calculated by using the spatial neighborhood probability method with the Gaussian membership:

[0087]

[0088] where N t is a time neighborhood window, t is a time neighborhood, and takes values of 0, ±1, ±2, ±3, etc. When t takes the value of ±2, N t represents a time neighborhood window of 5 time points, i.e. 2 time points before and after the current time point.

[0089] Based on the above, the embodiments of the present application give a certain weight to the forecast echo data at the previous and next time points by using the time neighborhood adaptive weighting algorithm, and then give the weighted average result to the forecast echo data at the current time point. If the weight of the forecast echo data at the current time point is higher than that at the previous and next time points, it indicates that the forecast at the current time point is the most accurate, and in this case, the fusion of the forecast results at the previous and next time points will weaken the accuracy of the forecast at the current time point. Conversely, if the weight of the forecast echo data at the current time point is lower than that at the previous and next time points, it indicates that the forecast at the current time point is an early forecast or a delayed forecast, and in this case, the fusion of the forecast results at the previous and next time points will be more accurate.

[0090] In actual applications, other similarity measurement indicators such as correlation coefficients can also be used as a replacement of RMSE.

[0091] S50: calculating a score skill score TFSS (Time-based FSS) of radar echo data based on a time neighborhood adaptive weighting algorithm;

[0092] In this embodiment of the present application, the score skill score of the adaptive weighting algorithm is constrained, specifically: if the weight at the current time point is higher than that at the previous and next time points, only the forecast result at the current time point is used to calculate the score, and otherwise, the time window score is calculated according to the adaptive weighting. The score skill score TFSS based on the time neighborhood adaptive weighting algorithm is defined as follows:

[0093]

[0094] where P is a neighborhood probability obtained by time neighborhood window adaptive weighting, and P obs is an original spatial neighborhood probability.

[0095] S60: calculating a score skill score TIFSS (Time-Intersity-based FSS, TIFSS) of the radar echo data based on the intensity feature and the space-time feature according to the IFSS and TFSS score results;

[0096] In this step, the score skill score TIFSS is defined as follows:

[0097]

[0098] Since the live echo data does not need to be time-neighborhood window weighted, therefore represents the live echo neighborhood probability by introducing the Gaussian membership intensity feature, represents the neighborhood probability after adaptive weighting of the time neighborhood window on the basis of introducing the Gaussian membership intensity feature.

[0099] It can be understood that the embodiments of the present application only take radar echo nowcasting as an example, and in other embodiments of the present application, it is also applicable to numerical mode prediction, rainfall prediction and other application scenarios similar to radar echo nowcasting.

[0100] Based on the above, the nowcasting effect evaluation method of the embodiments of the present application weakens the strict spatial matching requirement by introducing the spatial neighborhood scheme, can realize more objective and reasonable evaluation and testing. By introducing the Gaussian membership function in fuzzy logic to mine the value of the forecast grid below the threshold, more useful information can be given, and the tolerance of the intensity in the actual testing is increased. By introducing the time neighborhood, the prediction results of a certain time before and after are adaptively weighted, so that the result of the current prediction has the time feature, and the deficiencies of the advance prediction and the delayed prediction are solved.

[0101] Please refer to Figure 5 , which is a structure schematic diagram of the nowcasting effect evaluation system of the embodiments of the present application. The nowcasting effect evaluation system 40 of the embodiments of the present application comprises:

[0102] The membership calculation module 41 is used to calculate the membership of the radar echo data to the set threshold by using the Gaussian membership function, calculate the probability mean in the neighborhood range of each grid point in the radar echo data based on the membership, and take the probability mean as the fuzzy probability of each grid point; the radar echo data comprises a plurality of prediction echo data and live echo data of a set time;

[0103] The first scoring module 42 is used to calculate the score skill score IFSS of the radar echo data based on the Gaussian membership intensity feature according to the fuzzy probability;

[0104] The weight distribution module 43 is configured to assign weights to the forecasted echo data of each time point by using a time neighborhood adaptive weighting algorithm, and to calculate the probability of each grid point after the weight distribution by using a spatial neighborhood probability method with a Gaussian membership;

[0105] The second scoring module 44 is configured to calculate a score technique score TFSS of the radar echo data based on the time neighborhood adaptive weighting algorithm based on the probability of each grid point.

[0106] The third scoring module 45 is configured to calculate a score technique score TIFSS of the radar echo data based on the intensity feature and the space-time feature according to the IFSS and the TFSS score results.

[0107] Referring to Figure 6 , a terminal structure diagram of an embodiment of the present application is shown. The terminal 50 includes a processor 51 and a memory 52 coupled to the processor 51.

[0108] The memory 52 stores program instructions for implementing the above-mentioned near-forecast effect evaluation method.

[0109] The processor 51 is configured to execute the program instructions stored in the memory 52 to control the near-forecast effect evaluation.

[0110] The processor 51 can also be referred to as a CPU (Central Processing Unit). The processor 51 can be an integrated circuit chip with signal processing capability. The processor 51 can also be a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application-Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0111] Referring to Figure 7 , a structure diagram of a storage medium of an embodiment of the present application is shown. The storage medium of the embodiment of the present application stores a program file 61 capable of implementing all the above-mentioned methods, wherein the program file 61 can be stored in the above-mentioned storage medium in the form of a software product, including a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes a U disk, a mobile hard disk, a ROM (Read-Only Memory), a RAM (Random Access Memory), a magnetic disk or an optical disk, and various media capable of storing program codes, or a terminal device such as a computer, a server, a mobile phone, a tablet, etc.

[0112] The foregoing description of the disclosed embodiments enables a person skilled in the art to make or use the application. Modifications of these embodiments will occur to persons of skill in the art, and, while certain embodiments within the scope of the application are submitted as examples, it is the intent of the inventors that any equivalents of the subject matter described herein be included in the scope of the application. The application is not limited to the described embodiments but rather to the appended claims.

Claims

1. A method for evaluating the effect of nowcasting, characterized by, The method comprises the following steps: The radar echo data is calculated by using a Gaussian membership function to calculate the membership of a set threshold, based on which a spatial neighborhood probability method is used to calculate the probability mean of each grid neighborhood in the radar echo data, and the probability mean is taken as the fuzzy probability of each grid; The radar echo data is calculated based on the Gaussian membership intensity feature score IFSS according to the fuzzy probability; A time neighborhood adaptive weighting algorithm is used to assign weights to the forecast echo data at each time, and a spatial neighborhood probability method with added Gaussian membership is used to calculate the probability of each grid after the weights are assigned; The radar echo data is calculated based on the time neighborhood adaptive weighting algorithm score TFSS based on the probability of each grid; The radar echo data is calculated based on the intensity feature and space-time feature score TIFSS according to the IFSS and TFSS score results.

2. The nowcast effect evaluation method according to claim 1, characterized by, The radar echo data is calculated by using a Gaussian membership function to calculate the membership of a set threshold, and the specific steps are as follows: The Gaussian membership function is as follows: Wherein, x is the difference between the intensity of the radar echo data and the set threshold; c is the variance; for the radar echo data with intensity greater than or equal to the set threshold, the membership is equal to 1, and for the radar echo data with intensity less than the set threshold, the membership is: where P I is the spatial neighborhood probability considering the intensity feature of membership, i and j are the coordinates of the grid points in the neighborhood, T is a set threshold, and Ref represents the echo data. The spatial neighborhood probability method is used to calculate the probability mean of each grid neighborhood in the radar echo data, and the probability mean is taken as the fuzzy probability of each grid, and the specific steps are as follows: calculating a neighborhood probability for each grid point in the radar return data using a circular neighborhood window, N r = πr 2 where N is the number of grid points within the circular neighborhood range and r is the neighborhood radius. The probability mean of all grids in the circular neighborhood range is calculated as the fuzzy probability of the current grid.

3. The nowcast effect evaluation method according to claim 2, characterized by, The radar echo data is calculated based on the Gaussian membership intensity feature score IFSS according to the fuzzy probability, and the specific steps are as follows: wherein M and N are the height and width of the radar echo data, and are the neighborhood probabilities of the forecast echo data and the live echo data under the Gaussian membership intensity feature, respectively.

4. The nowcast effect evaluation method according to claim 3, characterized by, A time neighborhood adaptive weighting algorithm is used to assign weights to the forecast echo data at each time, and the specific steps are as follows: In the time neighborhood window, the root mean square error RMSE is used as the similarity measure between the forecast echo data and the actual echo data at each time, and the inverse of the RMSE is used to assign weights to the forecast echo data at each time; The calculation formula of the inverse of the RMSE is as follows: where t is the time neighborhood, Pred t and Obs are the predicted and observed echo data at time t, respectively. The calculation formula of the weight of the forecast echo data at each time is as follows: w t = IRMSE t / ∑IRMSE t .

5. The nowcast effect evaluation method according to claim 4, characterized by, The spatial neighborhood probability method with added Gaussian membership is used to calculate the probability of each grid after the weights are assigned, and the specific steps are as follows: where N t is the time neighborhood window, P pred (i,j) is the spatial neighborhood probability considering the membership intensity feature.

6. The nowcast effect evaluation method according to claim 5, characterized by, The radar echo data is calculated based on the time neighborhood adaptive weighting algorithm score TFSS based on the probability of each grid, and the specific steps are as follows: If the weight of the current time is higher than that of the previous and next times, only the forecast result of the current time is used to calculate the score, otherwise, the score is calculated according to the time neighborhood adaptive weighting algorithm; the score of the time neighborhood adaptive weighting algorithm score TFSS is as follows: wherein, P is the neighborhood probability obtained by time neighborhood window adaptive weighting obs P is the original space neighborhood probability.

7. The nowcast effect evaluation method according to claim 6, characterized by, The radar echo data is calculated based on the intensity feature and space-time feature score TIFSS according to the IFSS and TFSS score results, and the specific steps are as follows:

8. A nowcast impact assessment system, characterized by, The method comprises the following steps: The membership calculation module is configured to calculate the membership of radar echo data to a set threshold by using a Gaussian membership function, and calculate the probability mean of a neighborhood range of each grid point in the radar echo data based on the membership, and take the probability mean as the fuzzy probability of each grid point; The radar echo data includes a plurality of forecast echo data and real-time echo data at a plurality of time points; The first scoring module is configured to calculate the score skill score (IFSS) of the radar echo data based on the Gaussian membership intensity feature according to the fuzzy probability; The weight distribution module is configured to assign a weight to each time point of the forecast echo data by using a time neighborhood adaptive weighting algorithm, and calculate the probability of each grid point after the weight distribution by using a spatial neighborhood probability method with the Gaussian membership; The second scoring module is configured to calculate the score skill score (TFSS) of the radar echo data based on the time neighborhood adaptive weighting algorithm based on the probability of each grid point; The third scoring module is configured to calculate the score skill score (TIFSS) of the radar echo data based on the intensity feature and the space-time feature according to the IFSS and TFSS score results.

9. A terminal, characterized by comprising: The terminal includes a processor and a memory coupled to the processor, wherein The memory stores program instructions for implementing the nowcast effect evaluation method of any one of claims 1-7; The processor is configured to execute the program instructions stored in the memory to control the nowcast effect evaluation.

10. A storage medium, characterized by The memory stores processor-executable program instructions for executing the nowcast effect evaluation method of any one of claims 1-7.

Images