Global mode near-surface element inspection method, system, equipment and medium

Abstract

The invention discloses a global mode near-surface element inspection method, system and device and a medium, and relates to the technical field of meteorological observation and forecasting, and the method comprises the steps: obtaining and processing global ground observation data and global mode forecasting data, and obtaining near-surface element observation data and near-surface element forecasting data; according to the near-surface element observation data and the near-surface element forecast data, respectively calculating inspection product parameters of each near-surface element, and generating an inspection product file; the near-surface elements comprise temperature, rainfall and wind speed, and the wind speed is synthesized by a U wind component and a V wind component; drawing an inspection result graph according to the inspection product file; the test result diagram comprises an error distribution diagram of each near-surface element and a binary classification test diagram of rainfall. According to the method and the device, global near-surface element comprehensive inspection can be realized, and the correctness of the inspection result is improved.

Description

Technical Field

[0001] The present application relates to the technical field of meteorological observation and forecasting, and in particular to a method, system, equipment and medium for verifying near-surface elements of a global model. Background Art

[0002] The World Meteorological Organization's Deterministic Numerical Prediction Verification Center began to update the unified operational verification standards for global numerical forecast models in 2010, and initially proposed the operational verification standards for isobaric surface elements. The Numerical Prediction Center completed the integration of the business system in 2013, and initially formed a unified verification platform for numerical forecast business verification and evaluation. It updated the verification and evaluation methods and technologies for isobaric surface elements in accordance with the latest verification and evaluation standards of the World Meteorological Organization (WMO). At present, eleven numerical forecast business centers, including China, have uploaded the business verification results of isobaric surface elements of global models to the unified verification platform for comparative verification. Subsequently, the European Centre for Medium-Range Weather Forecasts (ECMWF), the UK Meteorological Office, the National Centers for Environmental Prediction (NCEP) of the United States and the Japan Meteorological Agency completed the verification of precipitation forecasts of global numerical models of their respective business centers and uploaded the verification results, completing the unified comparison of precipitation verification.

[0003] As the verification of global near-surface elements has become more and more demanded, the business verification of the near-surface elements of the China Meteorological Administration Global Forecast System (CMA-GFS), which was previously only for China, cannot meet the needs of model developers and actual business applications. At the same time, the verification technology and methods also need to be gradually unified with the WMO standard. Due to the complexity of global observation processing, there is currently no complete relevant verification system in China that can verify wind, temperature and precipitation. Summary of the invention

[0004] The purpose of this application is to provide a global model near-surface element verification method, system, equipment and medium, which can realize comprehensive verification of near-surface elements for the whole world.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a global model near-surface element verification method, comprising:

[0007] Obtain and process global ground observation data and global model forecast data to obtain near-surface element observation data and near-surface element forecast data;

[0008] According to the near-surface element observation data and the near-surface element forecast data, the inspection product parameters of each near-surface element are calculated respectively to generate an inspection product file; the near-surface elements include temperature, precipitation and wind, wherein the wind speed is synthesized by the U wind component and the V wind component;

[0009] A test result map is drawn according to the test product file; the test result map includes an error distribution map of each near-ground element and a binary classification test map of precipitation.

[0010] Optionally, the global model near-surface element verification method further includes:

[0011] The inspection result graphs are summarized and uploaded to the inspection product display website for display.

[0012] Optionally, global ground observation data and global model forecast data are obtained and processed to obtain near-surface element observation data and near-surface element forecast data, including:

[0013] Acquire global ground observation data, and perform quality control on the global ground observation data based on the observation data quality control plan of the numerical forecast center to obtain near-surface element observation data;

[0014] Global model forecast data are obtained, and format conversion, decompression and near-surface element extraction are performed on the global model forecast data to obtain near-surface element forecast data.

[0015] Optionally, the observation data quality control scheme of the numerical forecast center includes climate extreme value check, inter-element consistency check, station height and station air pressure consistency check, time consistency check and continuity check; the inspection contents of the climate extreme value check include whether the wind speed, temperature, altitude, ground air pressure, sea level air pressure and humidity are within the corresponding value range; the inspection contents of the inter-element consistency check include wind direction, wind speed and temperature; the inspection contents of the station height and station air pressure consistency check include whether the station's altitude matches the ground air pressure; the inspection contents of the time consistency check include whether the rate of change of the observation element over time is within the corresponding value range; the continuity check is implemented using the standard deviation method or the adjacent observation time data difference method, the inspection content of the standard deviation method is whether the standard deviation of the observation data within a given time range is greater than a given criterion, and the inspection content of the adjacent observation time data difference method is whether the maximum value of all adjacent observation time data differences within a given time range is greater than a given criterion.

[0016] Optionally, according to the near-surface element observation data and the near-surface element forecast data, respectively calculating the inspection product parameters of each near-surface element and generating the inspection product file, including:

[0017] Selecting observation sites and actual observation conditions according to a set inspection observation station table, and extracting observation data from the near-surface element observation data;

[0018] Extracting corresponding forecast data from the near-surface element forecast data according to the actual verification time, and interpolating the forecast data to the observation site using a bilinear interpolation method; the corresponding forecast data includes grid forecast data of different forecast time effectiveness with different starting times;

[0019] Based on different observation sites, different observation conditions and different near-surface elements, the observation data and the forecast data are compared and the inspection product parameters are calculated to obtain relevant inspection results;

[0020] The relevant inspection results are written into a file named by the inspection actual time and forecast timeliness, which contains the forecast, actual situation and deviation of each observation site, to obtain an inspection product file.

[0021] Optionally, the inspection product parameters include general statistical inspection results of factors and special statistical inspection results of precipitation; the general statistical inspection results of factors include root mean square error, average error, mean absolute error, correlation coefficient, accuracy under different error thresholds, total number of effective inspection stations, number of accurate forecast stations under different error thresholds, total number of stations, number of stations with errors less than a first threshold, number of stations with errors less than a second threshold, and number of stations with errors less than a third threshold; the special statistical inspection results of precipitation include two-category inspection results of five different precipitation levels under two classification standards, number of stations with correct reports, number of stations with missed reports, number of empty reports The two-category test results include the forecast test score, fair forecast score, true skill score, handicap ratio, forecast frequency deviation, false alarm rate, missed alarm rate, empty alarm rate, forecast accuracy rate, hit rate and fairness score; wherein, for temperature, the first threshold is 1°C, the second threshold is 2°C, and the third threshold is 3°C; for wind speed, the first threshold is 1m / s, the second threshold is 2m / s, and the third threshold is 3m / s; for precipitation, the first threshold is 1mm, the second threshold is 3mm, and the third threshold is 5mm.

[0022] In a second aspect, the present application provides a global model near-surface element verification system, which is applied to the global model near-surface element verification method, including:

[0023] The data pre-processing module is used to obtain and process global ground observation data and global model forecast data to obtain near-ground element observation data and near-ground element forecast data;

[0024] A forecast verification module, used to calculate the verification product parameters of each near-surface element according to the near-surface element observation data and the near-surface element forecast data, and generate a verification product file; the near-surface elements include temperature, precipitation and wind speed, wherein the wind speed is composed of a U wind component and a V wind component;

[0025] The inspection map making module is used to draw the inspection result map according to the inspection product file; the inspection result map includes the error distribution map of each near-ground element and the binary classification inspection map of precipitation.

[0026] Optionally, the global model near-surface element verification system further includes:

[0027] The inspection product online module is used to summarize the inspection result graphs and upload them to the inspection product display website for display.

[0028] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the global model near-ground element verification method.

[0029] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the global model near-ground element verification method.

[0030] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0031] The present application provides a global model near-surface element verification method, system, equipment and medium. First, global ground observation data and global model forecast data are obtained and processed to obtain near-surface element observation data and near-surface element forecast data; secondly, according to the near-surface element observation data and near-surface element forecast data, the verification product parameters of each near-surface element are calculated respectively to generate a verification product file; then, the verification result graph is drawn according to the verification product file. The present application expands the scope of near-surface element verification from domestic to global, and expands the global near-surface elements to be verified to temperature, precipitation and wind speed, which makes the performance evaluation of the global model more scientific and reasonable, and is more convenient and quick to diagnose forecast errors occurring abroad. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0033] Figure 1 Flowchart of the global model near-surface element verification method provided for this application;

[0034] Figure 2 Flowchart of quality control of ground observation data provided for this application;

[0035] Figure 3 Module diagram of the global model near-surface element verification system provided for this application. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0037] According to the current weather forecast business and technology development needs, as well as previous research work, it is found that there are many problems with the observation of near-surface elements abroad. By drawing on international advanced inspection and quality control technologies, optimizing and improving the existing near-surface element inspection methods, processes and indicators, we will carry out global CMA-GFS global model near-surface element forecast inspection research and business applications. At present, most of the near-surface element inspections of global models in various countries around the world are only carried out for observation stations in their own country or region. In recent years, ECMWF, NCEP, the UK Meteorological Office and the Japan Meteorological Agency have carried out precipitation forecast business inspections in global near-surface elements, but have not provided inspections of other near-surface elements.

[0038] Based on this, the purpose of this application is to provide a global model near-surface element verification method, system, equipment and medium, which can realize comprehensive verification of near-surface elements for the world.

[0039] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0040] In an exemplary embodiment, the present application provides a global model near-surface element verification method, which is executed by a computer device, and can be specifically executed by a computer device such as a terminal or a server alone, or by a terminal and a server together. In the embodiment of the present application, Figure 1 As shown, the method includes the following steps.

[0041] Step 1: Obtain and process global ground observation data and global model forecast data to obtain near-surface element observation data and near-surface element forecast data.

[0042] Step 1 includes: obtaining global ground observation data, and performing quality control on the global ground observation data based on the observation data quality control plan of the numerical forecast center to obtain near-surface element observation data; obtaining global model forecast data, and performing format conversion, decompression and near-surface element extraction on the global model forecast data to obtain near-surface element forecast data.

[0043] First, use the TianQing retrieval program script run_GM.sh to download TianQing's global ground observation data from the Global Telecommunication System (GTS) to the local computer. After quality control according to the observation data quality control plan of the Numerical Prediction Center, the near-surface element observation data file rec_RSURF_yyyymmddhh_g_06.dat with an interval of 6 hours is generated. Then obtain the global model forecast data, and use the GRIB2toMICPAS program gribpost.exe to convert the GRIB2 format global model forecast data gmf.gra.yyyymmddhhh1h1h1.grb2 (where yyyymmdd is the year, month, and day, hh is the start time, and h1h1h1 is the forecast period) into MICAPS format data for verification. Then decompress the data, and then read the near-surface elements of precipitation, temperature, U wind and V wind to be extracted according to the given parameter files micapsconfig and vars_sfc_gfs. Store the extracted forecast data in a folder for each element and name it yymmddhh.h1h1h1.

[0044] There are always gross errors in meteorological observation data due to erroneous observation, transmission, reading, writing and calculation. In 1957 and 1976, the Joint Center for Numerical Forecasting of the Soviet Union conducted two investigations on the causes and frequency of gross errors, pointing out that the probability of gross errors is quite high. For example, about 20% of the total sounding reports contain at least one data with gross errors (the probability of gross errors in current observation data is less than 20%). Gross errors are erroneous values ​​that have no meteorological significance. Although the number is small, they can often cause the numerical weather forecast system to be interrupted or seriously deviate. The large differences between different assimilation systems are often caused by differences in data processing and quality control. Similarly, the distortion of forecast results is often caused by improper data quality control.

[0045] The quality control program of the observation data of the numerical prediction center includes climate extreme value check, inter-element consistency check, station height and station pressure consistency check, time consistency check and continuity check. The above checks are performed in sequence, such as Figure 2 As shown. Among them, the inspection content of the climate extreme value check includes whether the wind speed, temperature, altitude, ground pressure, sea level pressure and humidity are within the corresponding value range; the inspection content of the consistency check between the elements includes wind direction, wind speed and temperature; the inspection content of the station height and station pressure consistency check includes whether the station's altitude matches the ground pressure; the inspection content of the time consistency check includes whether the rate of change of the observation element over time is within the corresponding value range; the continuity check is implemented by the standard deviation method or the adjacent observation time data difference method. The inspection content of the standard deviation method is whether the standard deviation of the observation data within a given time range is greater than a given criterion, and the inspection content of the adjacent observation time data difference method is whether the maximum value of all adjacent observation time data differences within a given time range is greater than a given criterion.

[0046] A) Examination of climate extremes.

[0047] What to check: Wind speed, temperature, altitude, ground pressure, sea level pressure and humidity.

[0048] Inspection rules: When the inspected data exceeds the corresponding value range of the climate extreme value inspection of its elements (possibly correct, possibly wrong, wrong), the quality control code is assigned a corresponding value, and the wrong data will no longer participate in other subsequent inspection steps.

[0049] B) Check consistency between elements.

[0050] The methods for checking wind direction, wind speed and temperature consistency are the same as those for sounding.

[0051] C) Check the consistency between station altitude and station air pressure.

[0052] The checking procedure (which is included in the extreme value checking procedure) is: if (elv=>800.and.ps>=1000) iflag3=1; if (elv<800.and.ps<=850) iflag3=1; if (elv>2300.and.ps>=850) iflag3=1; if (elv<2300.and.ps<=650) iflag3=1; if (elv>3700.and.ps>=650) iflag3=1. Among them, elv represents the altitude, ps represents the ground pressure, iflag3 represents the quality control code, 0 represents available, and 1 represents an error.

[0053] D) Time consistency check and continuity check.

[0054] Temporal consistency check is to examine the observation information or observation elements from the perspective of time change rate and identify undesirable sudden changes. The method of temporal consistency check is to compare the current observation information with the observation information of adjacent time periods. If their difference exceeds the given allowable range, the current observation information is considered suspicious. The inspection criterion of temporal consistency check is related to the temporal resolution of the sample. Temporal consistency check is suitable for high temporal resolution because the correlation between adjacent samples increases with the increase of temporal resolution. Given a fixed observation platform, the difference in the observation values ​​of an element at different times is mainly caused by the change of the observation element over time.

[0055] Therefore, the temporal consistency check criterion Max(DA) is: Max(DA) = Max(δa / δt)×t. Where δa / δt represents the rate of change of the observed element over time, δa represents the change of two observations, δt represents the time difference between two observations, t represents time, Max represents the maximum value, and DA represents the change of the element. The value of Max(δa / δt) is shown in Table 1, where the Max(DA) for calculating air pressure and temperature is limited to 12hPa and 15℃ respectively.

[0056] Table 1 Maximum change rate of observation elements with altitude, latitude and time

[0057] Max(δa / δt) temperature 5℃ / 1hour Air pressure 3hPa / 1hour

[0058] The main function of the continuity check is to check whether the changes in observation parameters caused by instrument failure (such as icing, communication problems, etc.) are too small or unchanged. Compared with the time consistency check, the continuity check is to check the rationality of the time change rate of the observation elements from another perspective. There are two methods for continuity check. One is the standard deviation method, which requires that the standard deviation of the observation data within a given time range is greater than the given criterion. The second is the data difference method of adjacent observation times. It requires that the maximum value of the data difference of all adjacent observation times within a given time range is greater than the given criterion. This application adopts method 2. The inspection result acts on all data within a given time window. The continuity check criterion is 0.1. If the maximum value of the observation data difference of all adjacent observation times within a given time range is less than the given criterion, the continuity check fails.

[0059] By investigating the relevant literature and work of ECMWF, the researchers collected, processed and analyzed the arrival of global near-surface precipitation, temperature and wind data based on GTS, and formed an observation verification station table. In the process of analyzing the test results, it was found that the GTS data used showed abnormally large forecast errors at some stations outside the Chinese region, with temperature observations as high as 95 degrees, resulting in an error between the analysis and the actual situation greater than tens of degrees Celsius. In addition, especially for 24-hour precipitation observations, in a year, many GTS stations have less than half of the total samples. Therefore, in order to carry out the near-surface element verification of the Global / Regional General Numerical Weather Prediction System (GRAPES) global model, it is also necessary to verify the quality of the GTS stations and conduct quality control on the GTS stations. On this basis, the correctness of the near-surface element verification of the GRAPES global model can be guaranteed.

[0060] In addition to the above quality control scheme, the quality control threshold of temperature is selected as 14°C based on the previous sample statistics of nearly 4,000 ground observation stations worldwide from December 2019 to November 2020. In this way, there are 10 stations in the selected temperature inspection blacklist. If you want to further remove the influence of large terrain, you can also consider setting the threshold at 10°C, and there are 33 stations in the blacklist. Considering that most of the observations at the 33 stations are valid samples and the number of samples exceeds 80% of the total samples, this scheme was not adopted in the end. Since the wind at high altitude is significantly larger than that on the ground, the error of the ground wind is actually smaller than that at high altitude. According to the previous sample statistics of nearly 4,000 ground observation stations worldwide from December 2019 to November 2020, the average absolute error of wind is rarely greater than 10m / s, so the annual average absolute error of wind observation and model analysis field greater than 10m / s is selected as the quality control threshold of ground wind observation, and a ground wind inspection observation blacklist is established. There are 4 stations on the wind blacklist. The range of precipitation changes is very large around the world, ranging from no precipitation to thousands of millimeters. And precipitation is not continuous in time or space. According to previous work, the accuracy of precipitation forecast decreases as the forecast time increases, so the threshold for the final precipitation inspection observation quality control is based on the annual average absolute error of the 36-hour precipitation forecast greater than 100 mm and less than 5 valid observation precipitation samples. There are 9 stations on the blacklist selected according to this standard.

[0061] Step 2: According to the near-surface element observation data and the near-surface element forecast data, the inspection product parameters of each near-surface element are calculated respectively to generate an inspection product file. The near-surface elements include temperature, precipitation and wind speed, wherein the wind speed is synthesized by the U wind component and the V wind component.

[0062] Step 2 includes: selecting observation sites and actual observation conditions according to the set inspection observation site table, and extracting observation data from the near-surface element observation data; extracting corresponding forecast data from the near-surface element forecast data according to the inspection actual time, and interpolating the forecast data to the observation site using a bilinear interpolation method; the corresponding forecast data includes grid forecast data of different forecast time limits with different start times; based on different observation sites, different observation actual conditions and different near-surface elements, comparing the observation data and the forecast data respectively and calculating the inspection product parameters to obtain relevant inspection results; writing the relevant inspection results into a file named by the inspection actual time and forecast time limit, which contains the forecast, actual condition and deviation of each observation site, to obtain an inspection product file.

[0063] The inspection product parameters include general statistical inspection results of elements and special statistical inspection results of precipitation; the general statistical inspection results of elements include root mean square error, average error, average absolute error, correlation coefficient, accuracy under different error thresholds, total number of effective inspection stations, number of accurate forecast stations under different error thresholds, total number of stations, number of stations with errors less than the first threshold, number of stations with errors less than the second threshold, and number of stations with errors less than the third threshold; the special statistical inspection results of precipitation include two-category inspection results of five different precipitation levels under two classification standards, number of stations with correct reports, number of stations with missed reports, and number of stations with empty reports and the number of stations that correctly predicted no rain; the two-category test results include forecast test score, fair forecast score, true skill score, handicap ratio, forecast frequency deviation, false alarm rate, missed alarm rate, empty alarm rate, forecast accuracy rate, hit rate and fairness score; wherein, for temperature, the first threshold is 1°C, the second threshold is 2°C, and the third threshold is 3°C; for wind speed, the first threshold is 1m / s, the second threshold is 2m / s, and the third threshold is 3m / s; for precipitation, the first threshold is 1mm, the second threshold is 3mm, and the third threshold is 5mm.

[0064] The difficulty of global near-surface element verification lies mainly in the selection of station tables for verification and the extraction of actual conditions. In 2014, the relevant data of global SEEPS verification scores obtained from EC showed that they were quite strict in selecting precipitation verification stations. The climate station table used for SEEPS precipitation scoring changes monthly, with one station per month, and the observation stations vary slightly from month to month. February has the least stations, with only 4,650 stations, and August has the most stations, with 5,120 stations, of which 4,356 stations are in the Northern Hemisphere. That is to say, the weather in the Northern Hemisphere will determine 80% of the global model precipitation verification. In order to apply ground observations as much as possible, the climate station table with the most stations in August was selected as the station table for near-surface element forecast verification. In addition, the station table was divided into Western Europe and North Africa (25-90N, 10W-28E), Eastern Europe and West Asia (12-90N, 28-60E), North America (25-90N, 50-145W), South America (25N-60S, 30-120W), Asia (10S-90N, 60-150E), Australia and New Zealand (10-55S, 90-180E) and South Africa (35S-25N, 30W-60E) according to the WMO situation field analysis and observation verification and administrative geographical division, so as to facilitate the search for the specific area where the model works better or has problems. In order to cooperate with the inspection work of foreign cities, according to the arrangement of the National Meteorological Center, the near-surface element inspection of 48 foreign cities and 80 domestic cities was carried out.

[0065] Execute the run.sh test script located in the directory named after different elements respectively, select the observation site and the actual observation according to the proposed test observation station table of the eight global divisions, and then select the forecasts with different forecast time periods with different starting times according to the actual test time for matching. For example, if the actual observation time is yyyymmddhh, then the analysis time of 00 o'clock corresponding to the start time of hh is yymmddhh.000, and the 24-hour forecast file is yymm(dd-1)hh.024. For the temperature and wind test, a total of 11 forecast time periods are tested, including 00, 24...240 hour forecasts starting from 20 o'clock. For precipitation, a total of 9 forecast time periods are tested, including 036, 060...228 starting from 20 o'clock, and the corresponding grid forecast data are interpolated to the observation site using the bilinear interpolation method. The observed data and forecast data were compared to calculate the root mean square error, average error, mean absolute error, correlation coefficient, accuracy under different error thresholds, the total number of valid stations tested, and the number of accurate stations predicted under different error thresholds. The relevant formulas are as follows.

[0066]

[0067] Where: ME is the mean error, which represents the forecast error tendency of the global numerical forecast model (hereinafter referred to as the model); F i is the forecast value of the i-th sample of the model; i is the forecast or observation sample label, which is 1, 2, ..., n; n is the number of all test stations; O i is the actual observation value corresponding to the i-th sample forecast of the model; w i is the weight coefficient, and for site inspection, it is the reciprocal of the number of all sites, that is, 1 / n.

[0068]

[0069] Where: MAE is the mean absolute error, which indicates the magnitude of the model's forecast error. When used together with ME, it can be used to determine whether the model has a systematic error. When the two test quantities are close, if ME is positive, it means that the model forecast is too large, and if it is negative, it means that the model forecast is too small.

[0070]

[0071] Where: RMSE is the root mean square error, which indicates the size of the model error. When used together with MAE, it can be used to determine whether the model has a large error. If the values ​​are similar, it means that the model error distribution is relatively uniform. If the difference is large, it means that the model has a large error and is somewhat unstable.

[0072]

[0073] Where: r is the correlation coefficient, which represents the phase difference between the model forecast field anomaly and the observed field anomaly; F 0i is the value of the i-th sample at the initial moment of the model forecast; M F0 M is the average value of the deviation of the forecast relative to the initial forecast value in the test area; O0 is the average value of the observed relative to the model zero field anomaly in the test area; F is the average value of the forecast value; O is the average value of the observed value.

[0074]

[0075] Where: R c is the accuracy under the set error threshold; N ri For |F i -O i |<=The number of error thresholds is set. Different test elements have different thresholds, and each element can have multiple thresholds, such as temperature is 1, 2, 3℃, wind is 1, 2, 3m / s, and precipitation is 1, 3, 5mm. According to service requirements, one or more standards can be selected.

[0076] A binary classification test was also conducted for the test of discontinuous variables such as precipitation. Eleven test quantities, including TS (forecast test score), ETS (fair forecast score), PSS (true skill score), ORD (handicap ratio), BIAS (forecast frequency deviation), POFD (false alarm rate), PO (omission rate), EH (accuracy rate), NH (false alarm rate), POD (hit rate), Q (fairness score) and four quantities, including the number of stations that reported correctly, missed reports, false reports and correct forecast of no rain, were calculated for five different precipitation levels with different classification standards: light rain (>=0.1mm or 0.1-10mm), moderate rain (>=10mm or 10-25mm), heavy rain (>=25mm or 25-50mm), rainstorm (>=50mm or 50-100mm) and torrential rain (>=100mm). There were also four quantities, including the number of stations that reported correctly, missed reports, false reports and correct forecast of no rain, totaling fifteen test product parameters.

[0077] The two-category test uses a table to represent the frequency of whether a certain event is predicted and actually occurs. The four components of prediction and observation are called joint distribution, as shown in Table 2, where A represents the frequency of both prediction and actual occurrence, B represents the frequency of prediction occurring but not occurring, C represents the frequency of prediction not occurring but actually occurring, and D represents the frequency of neither prediction nor actual occurrence.

[0078] Table 2 Classification of precipitation (high temperature, etc.) inspection

[0079] Live\Forecast have none have A C none B D

[0080]

[0081] Where: DTS The forecast verification score, also known as the critical success index, defines the ratio of the predicted events to the correct forecast. It only cares about the occurrence of the predicted events and does not fully consider the events that do not occur. It is highly dependent on the climate probability, and the area with a high probability of the event will have a high score.

[0082]

[0083] Where: D PSS(TSS) The true skill score, also known as the Pirer skill score, takes into account the hit rate of precipitation forecasts and subtracts the rate of no-rain misses from it.

[0084]

[0085] Where: D PO It is the underreporting rate, which is the underreporting rate of the predicted events that actually occurred, that is, the rate at which the predicted events actually occurred but were not reported in the forecast.

[0086]

[0087] Where: D NH It is the false alarm rate, which is the ratio of false alarms in forecasted events, that is, the forecasted event occurs but does not actually occur.

[0088]

[0089] Where: D EH It is the forecast accuracy, which is the ratio of correct forecasts regardless of whether the event occurs or not. It is easy to get a high score when the probability of the predicted event occurring is high or the probability of the event not occurring is high, but when the probability of the event occurring and not occurring is the same, the score is low.

[0090]

[0091] Where: D BIAS Frequency deviation is the ratio of actual event occurrence to predicted event occurrence. A value of 1 is the best. A value greater than 1 indicates that the predicted frequency is higher than the actual frequency, with more false alarms; a value less than 1 indicates that the predicted frequency is lower than the actual frequency, with more false alarms.

[0092]

[0093] Where: D POD The hit rate is the ratio of the accurate forecast to the actual event, with 1 being the best and 0 being the minimum.

[0094]

[0095] Where: D POFD It is the ratio of the false alarm rate to the non-event rate, with 0 being the best and 1 being the worst.

[0096]

[0097] Where: D ORD It is the ratio of the product of correctly predicted events and non-events to the product of false positives and missed positives, with ∞ being the best and 0 being the worst.

[0098] D H =A (15)

[0099] D M=C (16)

[0100] D N =B (17)

[0101] D D =D (18)

[0102]

[0103] Where: D H is the number of stations reported correctly; D M is the number of missed stations; D N The number of stations with empty reports; D D The number of stations that correctly predicted no rain; D ETS is the fair forecast score, which is equivalent to the near success index minus the random probability of precipitation. It has a heavier penalty for false reports. When the random probability is very small, its score is equivalent to the critical success index. R(a) is the random probability of precipitation, defined as

[0104]

[0105] Where: D Q For fair scoring, this score considers the impact of all four classification elements on the forecast, which can maximize the expression of forecast accuracy and is not affected by sample size nor by climate probability.

[0106] The relevant test results are written into the file yyyymmddhh.hhh, which is named by the actual test time and forecast validity, and contains the forecast, actual situation and deviation of each observation station. Eleven statistical test results such as the root mean square error of a certain area at a single time are written at the end of the file. In addition, for the precipitation test, the results of fifteen test quantities of two categories under five different precipitation levels under two classification standards are added to the back. These test results can be used to summarize regional test data at any time interval such as month, season and year through the post-processing program vertglmean.exe, which is convenient for researchers to understand the model forecast performance through the changes in test data, such as the reduction of errors and the increase of accuracy.

[0107] According to the characteristics of different ground observation elements, there are slight differences in the design of test quantities. All three elements have test quantities commonly used in statistical tests: correlation coefficient, root mean square error, average error and mean absolute error. In addition, there are accuracy rates of less than 1℃, 2℃ and 3℃ for temperature, less than 1m / s, 2m / s and 3m / s for 10m wind test, and less than 1mm, 3mm and 5mm for precipitation. And 15 classified forecast verification quantities were added, providing the forecast verification score TS, fair forecast score ETS, forecast frequency deviation BIAS, false alarm rate NH, false alarm rate PO, forecast accuracy rate EH and other scores used in domestic daily inspections. It also added the real skill score PSS recommended by WMO, the concession ratio ORD used for extreme event assessment, the hit rate POD of common daily verification quantities, the false alarm rate (i.e. false alarm rate) POFD of non-events, and the statistics of the number of stations H, the number of stations M with false alarms, and the number of stations N with false alarms. It also added a fair score Q, which not only does not tend to 0 in extreme cases and does not have degradation, but also the evaluation of high-probability events is close to the forecaster's feeling, and has better consistency with the commonly used scores. In addition, the calculation of this score is almost unlikely to have no value phenomenon, and more effective forecast samples can be used. If this score is used as a rainstorm score for the public, it is relatively reasonable. From the analysis of the calculation results, the relationship between this score and the climate probability is not obvious, and it may also be suitable for the comparison of forecast levels in different regions. This evaluation result is closest to the result of the fair forecast score ETS. These statistical tests can easily reveal the forecast error margin, large error ratio, etc. of these factors, which is helpful for diagnosing model problems.

[0108] In addition, since the model products no longer generate data in MICAPS format, the forecast data for verification is converted from GRIB2 to MICAPS data using a scheduled job for daily processing. The precipitation is cumulative precipitation, and when used for verification, the latter time period is subtracted from the previous time period. The wind speed is verified by the composite of U and V winds.

[0109] Step 3: Draw a test result map based on the test product file. The test result map includes an error distribution map of each near-surface element and a two-classification test map of precipitation.

[0110] Use the two-meter temperature test result drawing script program distr_2mt.sh, the 10-meter wind test result drawing script program distr_10mw.sh, the precipitation test result drawing script program distr_rain.sh and the precipitation two-classification test result drawing script program drawrain.sh and other drawing script programs, and use GRADS to draw the error distribution maps of different near-ground elements according to the inspection product files, while the precipitation two-classification test map is drawn by PYTHON.

[0111] Step 4: Summarize the inspection result graphs and upload them to the inspection product display website for display.

[0112] The script program rain_t.sh, which summarizes the drawn test result graphs, and the script program put_rain_pic.sh, which sends the summarized test graphs to the test server, summarize the relevant test products and upload them to the test product display website for display. Based on the precipitation, temperature and wind forecast test results of the eight sub-areas developed in the early stage, the error distribution test graphs of temperature, wind and precipitation and the two-classification test graphs of precipitation are added to the test business duty webpage.

[0113] In an exemplary embodiment, the present application provides a global model near-surface element verification system to implement the above-mentioned global model near-surface element verification method. Figure 3 As shown, the system includes: a data pre-processing module, a forecast inspection module, an inspection chart making module and an inspection product online module.

[0114] Among them, the data pre-processing module is used to obtain and process global ground observation data and global model forecast data to obtain near-surface element observation data and near-surface element forecast data; the forecast verification module is used to calculate the inspection product parameters of each near-surface element according to the near-surface element observation data and the near-surface element forecast data, and generate an inspection product file; the inspection map production module is used to draw the inspection result map according to the inspection product file; the inspection product online module is used to summarize the inspection result map and upload it to the inspection product display website for display.

[0115] At present, the global model near-surface element verification system has achieved full automation from observation data and forecast data processing, verification to statistical analysis and drawing of verification results, transmission to websites and web page display, and relevant verification system operation monitoring has been established.

[0116] Compared with the prior art, this application has the following advantages:

[0117] (1) The scope of near-surface element verification has been expanded from domestic to global, which makes the performance evaluation of global models more scientific and reasonable, and is more convenient and quick to diagnose forecast errors occurring abroad.

[0118] (2) To meet the needs of economic strategy development, it can provide the forecast verification results of any relevant country’s cities, making it convenient for forecasters to revise relevant forecasts.

[0119] (3) The observation data quality control program established based on the previous inspection work effectively reduced the inspection error.

[0120] In an exemplary embodiment, the present application further provides a computer device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0121] In an exemplary embodiment, the present application further provides a computer-readable storage medium storing a computer program, which implements the steps in the above-mentioned method embodiments when executed by a processor.

[0122] In an exemplary embodiment, the present application further provides a computer program product, including a computer program, which implements the steps in the above-mentioned method embodiments when executed by a processor.

[0123] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant legal provisions.

[0124] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0125] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0126] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0127] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A global model near-surface element verification method, characterized in that: include: Obtain and process global ground observation data and global model forecast data to obtain near-surface element observation data and near-surface element forecast data; Calculate the inspection product parameters of each near-surface element according to the near-surface element observation data and the near-surface element forecast data, and generate an inspection product file; The near-surface elements include temperature, precipitation and wind speed, wherein the wind speed is composed of a U wind component and a V wind component; A test result map is drawn according to the test product file; the test result map includes an error distribution map of each near-ground element and a binary classification test map of precipitation.

2. The global model near-surface element verification method according to claim 1, characterized in that: Also includes: The inspection result graphs are summarized and uploaded to the inspection product display website for display.

3. The global model near-surface element verification method according to claim 1, characterized in that: Obtain and process global ground observation data and global model forecast data to obtain near-surface element observation data and near-surface element forecast data, including: Acquire global ground observation data, and perform quality control on the global ground observation data based on the observation data quality control plan of the numerical forecast center to obtain near-surface element observation data; Global model forecast data are obtained, and format conversion, decompression and near-surface element extraction are performed on the global model forecast data to obtain near-surface element forecast data.

4. The global model near-surface element verification method according to claim 3, characterized in that: The observation data quality control program of the Numerical Prediction Center includes climate extreme value check, inter-element consistency check, station height and station pressure consistency check, time consistency check and continuity check; the inspection contents of the climate extreme value check include whether the wind speed, temperature, altitude, ground pressure, sea level pressure and humidity are within the corresponding value range; the inspection contents of the inter-element consistency check include wind direction, wind speed and temperature; the inspection contents of the station height and station pressure consistency check include whether the station's altitude matches the ground pressure; the inspection contents of the time consistency check include whether the rate of change of the observation element over time is within the corresponding value range; the continuity check is implemented by the standard deviation method or the adjacent observation time data difference method. The inspection contents of the standard deviation method are whether the standard deviation of the observation data within a given time range is greater than a given criterion, and the inspection contents of the adjacent observation time data difference method are whether the maximum value of all adjacent observation time data differences within a given time range is greater than a given criterion.

5. The global model near-surface element verification method according to claim 1, characterized in that: According to the near-surface element observation data and the near-surface element forecast data, the inspection product parameters of each near-surface element are calculated respectively, and the inspection product file is generated, including: Selecting observation sites and actual observation conditions according to a set inspection observation station table, and extracting observation data from the near-surface element observation data; Extracting corresponding forecast data from the near-surface element forecast data according to the actual verification time, and interpolating the forecast data to the observation site using a bilinear interpolation method; the corresponding forecast data includes grid forecast data of different forecast time effectiveness with different starting times; Based on different observation sites, different observation conditions and different near-surface elements, the observation data and the forecast data are compared and the inspection product parameters are calculated to obtain relevant inspection results; The relevant inspection results are written into a file named by the inspection actual time and forecast timeliness, which contains the forecast, actual situation and deviation of each observation site, to obtain an inspection product file.

6. The global model near-surface element verification method according to claim 1, characterized in that: The inspection product parameters include general statistical inspection results of elements and special statistical inspection results of precipitation; the general statistical inspection results of elements include root mean square error, average error, average absolute error, correlation coefficient, accuracy under different error thresholds, total number of effective inspection stations, number of accurate forecast stations under different error thresholds, total number of stations, number of stations with errors less than the first threshold, number of stations with errors less than the second threshold, and number of stations with errors less than the third threshold; the special statistical inspection results of precipitation include two-category inspection results of five different precipitation levels under two classification standards, number of stations with correct reports, number of stations with missed reports, and number of stations with empty reports and the number of stations that correctly predicted no rain; the two-category test results include forecast test score, fair forecast score, true skill score, handicap ratio, forecast frequency deviation, false alarm rate, missed alarm rate, empty alarm rate, forecast accuracy rate, hit rate and fairness score; wherein, for temperature, the first threshold is 1°C, the second threshold is 2°C, and the third threshold is 3°C; for wind speed, the first threshold is 1m / s, the second threshold is 2m / s, and the third threshold is 3m / s; for precipitation, the first threshold is 1mm, the second threshold is 3mm, and the third threshold is 5mm.

7. A global model near-surface element verification system, applied to the global model near-surface element verification method according to claim 1, characterized in that: include: The data pre-processing module is used to obtain and process global ground observation data and global model forecast data to obtain near-ground element observation data and near-ground element forecast data; A forecast verification module, used to calculate the verification product parameters of each near-surface element according to the near-surface element observation data and the near-surface element forecast data, and generate a verification product file; The near-surface elements include temperature, precipitation and wind speed, wherein the wind speed is composed of a U wind component and a V wind component; The inspection map making module is used to draw the inspection result map according to the inspection product file; the inspection result map includes the error distribution map of each near-ground element and the binary classification inspection map of precipitation.

8. The global model near-surface element verification system according to claim 7, characterized in that: Also includes: The inspection product online module is used to summarize the inspection result graphs and upload them to the inspection product display website for display.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the global model near-surface element verification method according to any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the global model near-surface element verification method described in any one of claims 1 to 6 is implemented.

Images