Visibility forecasting mode evaluation method and system

Abstract

The invention belongs to the technical field of weather forecast, and discloses a visibility forecast mode evaluation method and system. The method comprises the steps of sample screening and key meteorological element selection, delinearization trend preprocessing, calculation of the relation between visibility and key meteorological elements, observation sample significance testing and visibility forecast mode simulation effect evaluation. According to the visibility forecasting mode evaluation method based on analysis of the relationship between the visibility and the meteorological elements, the simulation effect of the relationship between the visibility and the meteorological elements in an existing visibility forecasting mode can be evaluated, and an important reference is provided for pertinently improving the visibility forecasting effect of the visibility forecasting mode; according to the method, the relationship between the visibility and other related meteorological elements in the actual observation and visibility forecasting mode is calculated and analyzed, the difference between the two is compared, and the relationship between the visibility and the meteorological elements in the visibility forecasting mode is evaluated, so that a basis is provided for the targeted improvement of the visibility forecasting mode.

Description

Technical Field

[0001] The invention belongs to the technical field of weather forecasting, and in particular relates to a visibility forecasting model evaluation method and system. Background Art

[0002] Visibility is correlated with meteorological factors such as air temperature and humidity. At present, many formulas for the relationship between meteorological factors such as air humidity and visibility have been given at home and abroad. Most visibility forecast models use the forecast results of various meteorological variables other than visibility to calculate visibility based on the relationship formula between them and visibility. However, for different regions, the relationship between visibility and other meteorological factors is more complicated, and it is impossible to use a single calculation formula to give a result consistent with the actual visibility observation. It needs to be adjusted in conjunction with other influencing factors, and this complex influencing relationship has not yet been determined.

[0003] For the existing visibility forecast model evaluation methods, most of them evaluate the visibility forecast effect of the visibility forecast model by verifying the accuracy of the visibility forecast by the model, that is, checking the consistency between the visibility forecast result and the visibility observation result. However, the defects and shortcomings of the existing methods are that they cannot find the source of the error of the visibility forecast model, that is, they cannot determine which meteorological factors have what impact on the visibility forecast error, and cannot put forward targeted improvement suggestions for the forecast method of the visibility forecast model.

[0004] Through the above analysis, the problems and defects of the existing technology are as follows: the existing visibility forecast model evaluation method is too simple and single, cannot find out the source of error of the visibility forecast model in a specific area, cannot determine the impact of other meteorological factors on the visibility forecast error, and cannot put forward targeted improvement suggestions for the forecasting method of the visibility forecast model. Summary of the invention

[0005] In order to overcome the problems existing in the related art, the disclosed embodiment of the present invention provides a visibility forecast mode evaluation method and system, and the technical solution is as follows:

[0006] The present invention is achieved by providing a visibility forecast model evaluation method, comprising:

[0007] S1, sample screening and key meteorological element selection: obtain forecast data and observation site data of the area to be evaluated, match the two, and construct visibility sample sets and meteorological element sample sets through feature condition screening;

[0008] S2, de-linear trend preprocessing: Before data calculation, all sample set data are preprocessed to de-linear trend, and the meteorological element sample x value and visibility sample vis value are obtained;

[0009] S3, calculates the relationship between visibility and key meteorological elements;

[0010] S4, conduct observation sample significance test;

[0011] S5, for each visibility level, the model's simulation effect on the relationship between visibility and key meteorological elements is evaluated by judging whether each relationship feature calculated from the simulation results is consistent with the observation.

[0012] In step S1, sample screening and key meteorological element selection include:

[0013] Select the land and sea areas to be forecasted, obtain the meteorological observation station data with atmospheric horizontal visibility observations and visibility forecast model forecast data within the geographical range to be forecasted, and match the observation station data according to the spatial resolution d and grid point position of the forecast data;

[0014] For the selected grid points and all stations, select the same time resolution as the forecast data, select the grid points and stations with daily average atmospheric horizontal visibility VIS≤10km in the local foggy season, and the number of days satisfying the daily average VIS≤10km in the selected time length exceeds 60% of the selected time length, and only use the data of VIS≤10km for calculation, with visibility <0.5km, 0.5-1km, 1-2km, 2-4km, >4km as the standard, divided into 5 visibility levels as the visibility sample set;

[0015] Multiple relevant ground meteorological elements at the same time are extracted from the screened station and grid visibility samples, including wind speed and direction 10 m above the ground, temperature 2 m above the ground, dew point temperature 2 m above the ground, temperature-dew point difference 2 m above the ground, relative humidity 2 m above the ground and atmospheric horizontal visibility; meteorological elements missing in the forecast data or observation data are eliminated, and the meteorological elements to be evaluated are determined to form a sample set of meteorological elements at the same time as the visibility sample.

[0016] Furthermore, the observation site data are matched according to the spatial resolution d and grid position of the forecast data, including:

[0017] For station data within 0.1° of the grid point distance, the nearest station is selected for direct time resolution matching; for station data within 0.1° of the grid point distance, the nearest station is selected for direct time resolution matching. For the station data, the two stations with the closest distance are selected and the weighted average of the distance is taken for matching; for the grid point distance If there is only one site in the grid, directly select this site for matching; Grid points without sites are removed, and sites without grid point matches are also removed.

[0018] In step S2, de-linear trend preprocessing is performed, including:

[0019] At the selected grid points and stations where the sample X is located, all data moments of the month in which the sample moment is located are numbered sequentially to obtain the time series t and the length n of the time series of the month t , calculate the sample average of the month and the time series mean

[0020] Calculate the slope γ, intercept β, and linear trend Subtract the linear trend value of the month from the value X at each time The abnormal data of each variable is obtained as the processed sample x value;

[0021] The visibility sample VIS is preprocessed to remove the linear trend and obtain the visibility sample vis value.

[0022] The calculation formula for the slope γ is:

[0023]

[0024] Among them, t is the time series of the current month, n t is the length of the time series for the current month, is the summation symbol, indicating X i and t i Continue to take values ​​from X1 and t1 to X i is the value of the sample data of sort i, t i is the value of the time series of sort i;

[0025] The calculation formula for intercept β is:

[0026]

[0027] Linear Trend The calculation formula is:

[0028]

[0029] In step S3, the relationship between visibility and key meteorological elements is calculated, including:

[0030] For each visibility level meteorological element sample set, all the screened sample x values ​​are sorted in order of value size, and the sample x values ​​are separated based on the median value to obtain two x subsets greater than the median value and less than the median value;

[0031] Calculate the average of the two x subsets and the average of the vis values ​​corresponding to the two x subsets;

[0032] The ratio of the average values ​​of the two x subsets and the average difference of the vis values ​​corresponding to the two x subsets is recorded as the relationship between the visibility level and the meteorological element x.

[0033] Further, calculate the average of the two x subsets using the formula:

[0034]

[0035] in, is the summation symbol, representing x i Keep taking values ​​from x1 to x n , x i is the x value of sequence number i, and n is the length of the subset;

[0036] The ratio of the average value of two x subsets to the average difference of the vis values ​​corresponding to the two x subsets is:

[0037] D(vis) / D(x)

[0038] Where D(vis) is the average value of the two x subsets, and D(x) is the average difference of the vis values ​​corresponding to the two x subsets.

[0039] In step S4, a significance test of the observed samples is performed, including: a bilateral test is performed on the difference between the sample mean value at each station selected for each single meteorological element and the mean value of all data within the selected time length at the station;

[0040] Calculate the sample test statistic a;

[0041]

[0042] Among them, for each single meteorological element, is the average value of all samples X at the selected site, is the average value of all data within the selected time length at the site, n is the sample size of sample X at the site, n0 is the size of the selected time length at the site, and b is the pooled variance;

[0043] The calculation formula of b is as follows:

[0044]

[0045] Among them, S is the sample variance of sample X at the site, and S0 is the sample variance of all data within the selected time length at the site;

[0046] The p-value associated with the test statistic a and the degrees of freedom n+n0-2 is queried, the test standard is a=0.05, and the range of uncertainty of the climate state mean corresponding to the variable characteristics obtained in the sample X is defined within a 95% confidence interval.

[0047] In step S5, the visibility forecast model simulation effect evaluation includes:

[0048] Determine whether the relationship between the model simulation results and the station observations at each visibility level is statistically significant or insignificant. The judgment method is to compare the orders of magnitude between the two. The larger order of magnitude is statistically significant, and the smaller order of magnitude is statistically insignificant.

[0049] For statistically significant relationship characteristics, determine whether the signs of the relationship characteristics in the simulation results of the model are consistent with those of the relationship characteristics obtained by site observation. If both are positive or negative, the signs are consistent; if the positive and negative values ​​are opposite, the signs are inconsistent.

[0050] If the relationship characteristics in the model simulation results are opposite in sign to the relationship characteristics obtained from site observations; or when the relationship characteristics obtained from observations are statistically small to insignificant, while the simulation results obtain statistically significant relationship characteristics; or when the relationship characteristics of the observations are statistically significant, while the relationship characteristics obtained from the model simulation results are statistically insignificant, then the model simulation results are considered to be inconsistent with the relationship characteristics obtained from observations; if none of the conditions occur, then the relationship characteristics simulated by the model are considered to be consistent with the observations.

[0051] Another object of the present invention is to provide a visibility forecast mode evaluation system, which is used to control the visibility forecast mode evaluation method, and the system comprises:

[0052] The sample screening and key meteorological element selection module is used to obtain the forecast data and observation site data of the area to be evaluated, match the two, and construct the visibility sample set and meteorological element sample set through feature condition screening;

[0053] The de-linear trend preprocessing module is used to pre-process all sample set data for de-linear trend before data calculation to obtain meteorological element sample x values ​​and visibility sample vis values;

[0054] Element relationship calculation module, used to calculate the relationship between visibility and key meteorological elements;

[0055] The observation sample significance test module performs a bilateral test on the difference between the sample mean value at each selected station and the mean value of all data within the selected time length at the station for the meteorological element X;

[0056] The visibility forecast model simulation effect evaluation module is used to evaluate the simulation effect of the model on the relationship between visibility and key meteorological elements by judging whether each relationship feature calculated from the simulation results is consistent with the observation for each visibility level.

[0057] Combining all the above technical solutions, the advantages and positive effects of the present invention are as follows:

[0058] The visibility forecast model evaluation method based on the relationship analysis between visibility and meteorological elements of the present invention can evaluate the simulation effect of the relationship between visibility and meteorological elements in the existing visibility forecast model, and provide an important reference for targetedly improving the visibility forecast effect of the visibility forecast model. The present invention evaluates the simulation effect of the relationship between visibility and meteorological elements in the specified area by calculating and analyzing the relationship between visibility and other relevant meteorological elements in the actual observation and visibility forecast model in the specified area, comparing the difference between the two, and evaluating the simulation effect of the visibility forecast model on the relationship between visibility and meteorological elements in the specified area. It can determine the influence of other meteorological elements in the evaluated model on the visibility forecast error in the specified area, such as underestimating the influence of relative humidity on visibility, overestimating the influence of temperature change on visibility, etc., guide the visibility model to adjust the calculation relationship between other meteorological elements and visibility, or add other meteorological elements with significant relationship in visibility simulation, so as to provide a basis for targeted improvement of the evaluated visibility forecast model.

[0059] The visibility forecast model evaluation method of the present invention can provide the key points of the error sources of the visibility forecast of the evaluated model, that is, the calculation relationship between the visibility and the meteorological variables that are greatly different from the actual situation. The present invention evaluates the visibility forecast model, provides targeted suggestions for improving the forecast effect of the model, improves the simulation effect of the evaluated model, and improves the quality and value of the forecast product of the evaluated model. The existing visibility forecast model evaluation method is too simple and single, and cannot find the error source of the visibility forecast model in a specific area, cannot determine the influence of other meteorological elements on the visibility forecast error, and cannot make targeted improvement suggestions for the forecast method of the visibility forecast model. The present invention can determine the influence of other meteorological elements on the visibility forecast error of the evaluated model in the specified area, guide the visibility model to adjust the visibility calculation method, or add other meteorological elements with significant relationship in the visibility simulation, thereby providing a basis for targeted improvement of the evaluated visibility forecast model. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] The accompanying drawings herein are incorporated in and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description, serve to explain the principles of the present disclosure;

[0061] Figure 1 It is a flow chart of the visibility forecast model evaluation method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0062] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention, so the present invention is not limited by the specific implementation disclosed below.

[0063] The innovation of the present invention is that the evaluation method of the present invention is different from the evaluation method of directly testing the accuracy of the visibility forecast model for the visibility forecast result. Instead, it calculates and evaluates the accuracy of the relationship between other variables in the model and visibility, and then finds out the key elements and optimization directions that may increase the error in the process of the model calculating visibility through other variables. The present invention protects the overall method steps and the sample screening and key meteorological element selection modules, and the other four modules are all existing individual method steps, which are reorganized and applied in this method.

[0064] Embodiment 1, as Figure 1 As shown, the visibility forecast mode evaluation method provided by the embodiment of the present invention comprises the following steps:

[0065] S1, sample screening and key meteorological element selection: obtain forecast data and observation site data of the area to be evaluated, match the two, and construct visibility sample sets and meteorological element sample sets through feature condition screening;

[0066] S2, de-linear trend preprocessing: Before data calculation, all sample set data are preprocessed to de-linear trend, and the meteorological element sample x value and visibility sample vis value are obtained;

[0067] S3, calculates the relationship between visibility and key meteorological elements;

[0068] S4, conduct observation sample significance test;

[0069] S5, for each visibility level, the model's simulation effect on the relationship between visibility and key meteorological elements is evaluated by judging whether each relationship feature calculated from the simulation results is consistent with the observation.

[0070] The sample screening and key meteorological element selection in step S1 provided in the embodiment of the present invention include:

[0071] Step 1.1: Select the land and sea areas that need to be forecasted, obtain the meteorological observation station data and visibility forecast model forecast data with atmospheric horizontal visibility observations within this geographical range, and match the observation station data according to the spatial resolution d and grid position of the forecast data: for station data within 0.1° of the grid point distance, select the nearest station for direct time resolution matching; .... For the station data, the two stations with the closest distance are selected and the weighted average of the distance is taken for matching; for the grid point distance If there is only one site in the grid, directly select this site for matching; Grid points without sites are removed, and sites without grid point matches are also removed.

[0072] Step 1.2: In order to screen out samples with long-term low visibility characteristics, for the grid points and all stations screened out in the previous step, select the same time resolution as the forecast data, screen out the grid points and stations with daily average atmospheric horizontal visibility VIS≤10km in the local foggy season, and the number of days satisfying the daily average VIS≤10km in the selected time length exceeds 60% of the selected time length, and only use the data of VIS≤10km for calculation, with visibility <0.5km, 0.5-1km, 1-2km, 2-4km, >4km as the standard, divided into 5 visibility levels as the visibility sample set;

[0073] Step 1.3: Extract multiple relevant ground meteorological elements at the same time for the above-screened site and grid visibility samples, including wind speed (m / s) and wind direction (°) at 10 m above the surface, temperature (K) at 2 m above the surface, dew point temperature (K) at 2 m above the surface, temperature-dew point difference (K) at 2 m above the surface, relative humidity (%) at 2 m above the surface, and atmospheric horizontal visibility (km). Eliminate the meteorological elements missing in the forecast data or observation data, determine the meteorological elements to be evaluated, and form a sample set of meteorological elements at the same time as the visibility sample.

[0074] The de-linear trend preprocessing in step S2 provided in the embodiment of the present invention is to preprocess all data to de-linear trend before performing data calculation, which specifically includes the following steps:

[0075] Step 2.1: At the selected grid points and stations where the sample X is located, sequentially number all the data moments of the month in which the sample moment is located, and obtain the time series t and the length of the time series n for the month t , calculate the sample average of the month and the time series mean

[0076] Step 2.2: Calculate the slope γ:

[0077]

[0078] Among them, t is the time series of the current month, n t is the length of the time series for the current month, is the summation symbol, indicating X i and t i Continue to take values ​​from X1 and t1 to X i is the value of the sample data of sort i, t i is the value of the time series of sort i;

[0079] Step 2.3: Calculate the intercept β:

[0080]

[0081] Step 2.4: Calculate the linear trend

[0082]

[0083] Subtract the linear trend value of the month from the value X at each time The abnormal data of each variable is obtained as the processed sample x value;

[0084] Step 2.5: The visibility sample VIS undergoes the same de-linear trend preprocessing steps to obtain the visibility sample vis value.

[0085] The calculation of the relationship between visibility and key meteorological elements in step S3 provided in the embodiment of the present invention includes:

[0086] Step 3.1: Sort all the sample x values ​​selected from the meteorological element sample set of each visibility level in order of value size, separate the sample x values ​​based on their median values, and obtain two x subsets greater than the median value and less than the median value;

[0087] Step 3.2: Calculate the mean of the two x subsets, that is And the average value of the vis values ​​corresponding to the two x subsets;

[0088] Step 3.3: Record the ratio of the difference between the two groups of mean values, i.e. D(vis) / D(x), as the relationship between the visibility level and the meteorological element x.

[0089] In step S4 provided by the embodiment of the present invention, a bilateral test is performed on the difference between the sample average value at each site selected by each single meteorological element and the average value of all data within the selected time length at the site;

[0090] The specific steps include:

[0091] Step 4.1: Calculate the sample test statistic a;

[0092]

[0093] Among them, for each single meteorological element, is the average value of all samples X at the selected site, is the average value of all data within the selected time length at the site, n is the sample size of sample X at the site, n0 is the size of the selected time length at the site, and b is the pooled variance;

[0094] The calculation formula of b is as follows:

[0095]

[0096] Among them, S is the sample variance of sample X at the site, and S0 is the sample variance of all data within the selected time length at the site;

[0097] Step 4.2: Query the p-value associated with the test statistic a and the degrees of freedom n+n0-2. The test standard is a=0.05. The range of uncertainty of the climate state mean corresponding to the variable characteristics obtained in sample X is defined within a 95% confidence interval.

[0098] The visibility forecast model simulation effect evaluation in step S5 provided in the embodiment of the present invention is to evaluate the simulation effect of the model on the relationship between visibility and key meteorological elements by judging whether each relationship feature calculated by the simulation result is consistent with the observation for each visibility level, and specifically includes the following steps:

[0099] Step 5.1: Determine whether the relationship between the model simulation results and the site observations at each visibility level is statistically significant or insignificant. The judgment method is to compare the orders of magnitude between the two. The one with a larger order of magnitude is statistically significant, and the one with a smaller order of magnitude can be considered statistically insignificant.

[0100] Step 5.2: For statistically significant relationship characteristics, determine whether the signs of the relationship characteristics in the model simulation results are consistent with those of the relationship characteristics obtained from site observations. If both are positive or negative, the signs are consistent; if the signs are opposite, the signs are inconsistent.

[0101] Step 5.3: If the relationship characteristics in the model simulation results are opposite in sign to the relationship characteristics obtained from the site observations; when the relationship characteristics obtained from the observations are statistically small to insignificant, while the simulation results obtain statistically significant relationship characteristics; when the relationship characteristics of the observations are statistically significant, while the relationship characteristics obtained from the model simulation results are statistically insignificant, then the model simulation results are considered to be inconsistent with the relationship characteristics obtained from the observations. If none of these conditions occur, the relationship characteristics simulated by the model are considered to be consistent with the observations.

[0102] Embodiment 2: The visibility forecast mode evaluation system provided by the embodiment of the present invention comprises:

[0103] The sample screening and key meteorological element selection module is used to obtain the forecast data and observation site data of the area to be evaluated, match the two, and construct the visibility sample set and meteorological element sample set through feature condition screening;

[0104] The de-linear trend preprocessing module is used to pre-process all sample set data for de-linear trend before data calculation to obtain meteorological element sample x values ​​and visibility sample vis values;

[0105] Element relationship calculation module, used to calculate the relationship between visibility and key meteorological elements;

[0106] The observation sample significance test module performs a bilateral test on the difference between the sample mean value at each selected station and the mean value of all data within the selected time length at the station for the meteorological element X;

[0107] The visibility forecast model simulation effect evaluation module is used to evaluate the simulation effect of the model on the relationship between visibility and key meteorological elements by judging whether each relationship feature calculated from the simulation results is consistent with the observation for each visibility level.

[0108] To further demonstrate the positive effects of the above embodiments, the present invention conducts the following experiments based on the above technical solutions. The data used in this study include observation data of sites where visibility observation instruments are deployed in Qingzhou in 2023, visibility forecast data of the EC-THIN forecast model, and forecast data of other meteorological elements.

[0109] 1.1 Assuming that the time resolution of the model's visibility forecast is 1 hour, the assessment area is determined to be a certain area in Qing, and a matching screening is performed on the spatial positions of all visibility stations in Qing and the grid points of the forecast model.

[0110] 1.2 Further screen the selected stations and grid points in terms of time length, select the fog season from April to July in a certain area of ​​Qing, determine the evaluation years as 2020-2023, count the daily average visibility from April to July in 2020-2023, count the number of days with daily average visibility ≤10km, and judge whether the statistical number of days at each station and grid point exceeds 60% of all days from April to July in 2020-2023. Grid points and stations that do not reach 60% are further eliminated, and only the hourly observation data and model prediction data of the day with daily average visibility ≤10km are used as samples for the next step of calculation. At the same time, the samples with hourly visibility ≤10km are divided into 5 levels.

[0111] 1.3 Assume that the meteorological factors in the forecast model are x1, x2…x10, eliminate the meteorological factors that are not in the observation data, determine x1, x2, x3 as the meteorological elements for evaluation, select the meteorological elements on the hourly station observations and model grid points with daily average visibility ≤10km to form a data set, and also form a sample set of 5 levels according to the visibility classification.

[0112] 2. De-linear trend processing is performed on each sample. For example, for the visibility VIS at 01:00 on April 1, 2020, the value of the linear change trend of visibility in April 2020 at 01:00 on April 1, 2020 needs to be subtracted. The visibility difference vis is obtained as the processed sample value, and the meteorological element X at this moment is processed in the same way, and the linear trend value of the meteorological element is subtracted to obtain the meteorological element difference x as the processed sample value.

[0113] 3. Sort the meteorological element sample values ​​x at each station and grid point by size x1, x2…x n , with the median x i Divide x into two subsets x1, x2…x i and x i+1 ,x i+2 …x n , calculate the average values ​​p1 and p2 of the two subsets respectively, and then calculate the average values ​​vis1 and vis2 of the vis values ​​corresponding to the x values ​​in the two subsets. The relationship between meteorological elements and visibility is D = (p1-p2) / (vis1-vis2). Through calculation, D1 = 3, D2 = 4, D3 = 0.1 in site observations, and D10 = -3, D20 = 3.2, D30 = 10 in the forecast model.

[0114] 4. Perform a significance test on the meteorological element X and visibility VIS in the observations of the sites screened in step 1 to determine whether the sample is representative.

[0115] 5. The forecast model was evaluated and the relationship between meteorological elements and visibility in site observations and forecasts was compared. It was found that D1 and D10 had different signs, D3 was relatively significant and D30 was small to insignificant, and D2 and D20 had the same signs and were both significant. Therefore, the relationship between the meteorological element x2 and visibility in the forecast model was consistent with the observations, while the relationship between x1, x3 and visibility was inconsistent with the observations, and targeted adjustments and improvements were needed in the model.

[0116] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principles of the present invention should be covered within the protection scope of the present invention.

Claims

1. A visibility forecast model evaluation method, characterized in that: The method includes: S1, sample screening and key meteorological element selection: obtain forecast data and observation site data of the area to be evaluated, match the two, and construct visibility sample sets and meteorological element sample sets through feature condition screening; S2, de-linear trend preprocessing: Before data calculation, all sample set data are preprocessed to de-linear trend, and the meteorological element sample x value and visibility sample vis value are obtained; S3, calculates the relationship between visibility and key meteorological elements; S4, conduct observation sample significance test; S5, for each visibility level, the model's simulation effect on the relationship between visibility and key meteorological elements is evaluated by judging whether each relationship feature calculated from the simulation results is consistent with the observation.

2. The visibility forecast model evaluation method according to claim 1, characterized in that: In step S1, sample screening and key meteorological element selection include: Select the land and sea areas to be forecasted, obtain the meteorological observation station data with atmospheric horizontal visibility observations and visibility forecast model forecast data within the geographical range to be forecasted, and match the observation station data according to the spatial resolution d and grid point position of the forecast data; For the selected grid points and all stations, select the same time resolution as the forecast data, select the grid points and stations with daily average atmospheric horizontal visibility VIS≤10km in the local foggy season, and the number of days satisfying the daily average VIS≤10km in the selected time length exceeds 60% of the selected time length, and only use the data of VIS≤10km for calculation, with visibility <0.5km, 0.5-1km, 1-2km, 2-4km, >4km as the standard, divided into 5 visibility levels as the visibility sample set; Multiple relevant ground meteorological elements at the same time are extracted from the screened station and grid visibility samples, including wind speed and direction 10 m above the ground, temperature 2 m above the ground, dew point temperature 2 m above the ground, temperature-dew point difference 2 m above the ground, relative humidity 2 m above the ground and atmospheric horizontal visibility; meteorological elements missing in the forecast data or observation data are eliminated, and the meteorological elements to be evaluated are determined to form a sample set of meteorological elements at the same time as the visibility sample.

3. The visibility forecast model evaluation method according to claim 2, characterized in that: According to the spatial resolution d and grid position of the forecast data, the observation site data are matched, including: For station data within 0.1° of the grid point distance, the nearest station is selected for direct time resolution matching; for station data within 0.1° of the grid point distance, the nearest station is selected for direct time resolution matching. For the station data, the two stations with the closest distance are selected and the weighted average of the distance is taken for matching; for the grid point distance If there is only one site in the grid, directly select this site for matching; Grid points without sites are removed, and sites without grid point matches are also removed.

4. The visibility forecast model evaluation method according to claim 1, characterized in that: In step S2, de-linear trend preprocessing is performed, including: At the selected grid points and stations where the sample X is located, all data moments of the month in which the sample moment is located are numbered sequentially to obtain the time series t and the length n of the time series of the month t , calculate the sample average of the month and the time series mean Calculate the slope γ, intercept β, and linear trend Subtract the linear trend value of the month from the value X at each time The abnormal data of each variable is obtained as the processed sample x value; The visibility sample VIS is preprocessed with delinear trending to obtain the visibility sample vis value.

5. The visibility forecast model evaluation method according to claim 4, characterized in that: The calculation formula for the slope γ is: Among them, t is the time series of the current month, n t is the length of the time series for the current month, is the summation symbol, indicating X i and t i Continue to take values ​​from X1 and t1 to and X i is the value of the sample data of sort i, t i is the value of the time series of sort i; The calculation formula for intercept β is: Linear Trend The calculation formula is:

6. The visibility forecast model evaluation method according to claim 1, characterized in that: In step S3, the relationship between visibility and key meteorological elements is calculated, including: For each visibility level meteorological element sample set, all the screened sample x values ​​are sorted in order of value size, and the sample x values ​​are separated based on the median value to obtain two x subsets greater than the median value and less than the median value; Calculate the average of the two x subsets and the average of the vis values ​​corresponding to the two x subsets; The ratio of the average values ​​of the two x subsets to the average difference of the vis values ​​corresponding to the two x subsets is recorded as the relationship between the visibility level and the meteorological element x.

7. The visibility forecast model evaluation method according to claim 6, characterized in that: Calculate the average of two subsets of x using the formula: in, is the summation symbol, representing x i Keep taking values ​​from x1 to x n , x i is the x value of sequence number i, and n is the length of the subset; The ratio of the average value of two x subsets to the average difference of the vis values ​​corresponding to the two x subsets is: D(vis) / D(x) Where D(vis) is the average value of the two x subsets, and D(x) is the average difference of the vis values ​​corresponding to the two x subsets.

8. The visibility forecast model evaluation method according to claim 1, characterized in that: In step S4, a significance test of the observed samples is performed, including: a bilateral test is performed on the difference between the sample mean value at each station selected for each single meteorological element and the mean value of all data within the selected time length at the station; Calculate the sample test statistic a; Among them, for each single meteorological element, is the average value of all samples X at the selected site, is the average value of all data within the selected time length at the site, n is the sample size of sample X at the site, n0 is the size of the selected time length at the site, and b is the pooled variance; The calculation formula of b is as follows: Among them, S is the sample variance of sample X at the site, and S0 is the sample variance of all data within the selected time length at the site; The p-value associated with the test statistic a and the degrees of freedom n+n0-2 is queried, the test standard is a=0.05, and the range of uncertainty of the climate state mean corresponding to the variable characteristics obtained in the sample X is defined within a 95% confidence interval.

9. The visibility forecast model evaluation method according to claim 1, characterized in that: In step S5, the visibility forecast model simulation effect evaluation includes: Determine whether the relationship between the model simulation results and the station observations at each visibility level is statistically significant or insignificant. The judgment method is to compare the orders of magnitude between the two. The larger order of magnitude is statistically significant, and the smaller order of magnitude is statistically insignificant. For statistically significant relationship characteristics, determine whether the signs of the relationship characteristics in the simulation results of the model are consistent with those of the relationship characteristics obtained by site observation. If both are positive or negative, the signs are consistent; if the positive and negative values ​​are opposite, the signs are inconsistent. If the relationship characteristics in the model simulation results are opposite in sign to the relationship characteristics obtained from site observations; or when the relationship characteristics obtained from observations are statistically small to insignificant, while the simulation results obtain statistically significant relationship characteristics; or when the relationship characteristics of the observations are statistically significant, while the relationship characteristics obtained from the model simulation results are statistically insignificant, then the model simulation results are considered to be inconsistent with the relationship characteristics obtained from observations; if none of the conditions occur, then the relationship characteristics simulated by the model are considered to be consistent with the observations.

10. A visibility forecast model evaluation system, characterized in that: The system is used to control the visibility forecast model evaluation method described in any one of claims 1 to 9, and the system includes: The sample screening and key meteorological element selection module is used to obtain the forecast data and observation site data of the area to be evaluated, match the two, and construct the visibility sample set and meteorological element sample set through feature condition screening; The de-linear trend preprocessing module is used to pre-process all sample set data for de-linear trend before data calculation to obtain meteorological element sample x values ​​and visibility sample vis values; Element relationship calculation module, used to calculate the relationship between visibility and key meteorological elements; The observation sample significance test module performs a bilateral test on the difference between the sample mean value at each selected station and the mean value of all data within the selected time length at the station for the meteorological element X; The visibility forecast model simulation effect evaluation module is used to evaluate the simulation effect of the model on the relationship between visibility and key meteorological elements by judging whether each relationship feature calculated from the simulation results is consistent with the observation for each visibility level.

Images