CN113468482B

CN113468482B - Storm weather simulation forecasting method based on WRF mode

Info

Publication number: CN113468482B
Application number: CN202110761768.0A
Authority: CN
Inventors: 陈东辉; 刘长空; 王亮; 郭刚; 董欢欢; 徐涛; 任林涛
Original assignee: 31440 Unit Of Chinese Pla
Current assignee: 31440 Unit Of Chinese Pla
Priority date: 2021-07-06
Filing date: 2021-07-06
Publication date: 2024-05-03
Anticipated expiration: 2041-07-06
Also published as: CN113468482A

Abstract

The invention provides a rainstorm weather simulation forecasting method based on a WRF mode, which comprises the following steps of: (1) Carrying out statistical analysis on the historical stormwater weather and carrying out weather classification; (2) Selecting a plurality of cases from each type of storm, and performing sensitivity test on different storm simulation parameterization schemes in each type of storm by adopting a WRF mode to obtain optimal parameterization schemes of various types of storm simulation; (3) And carrying out test forecast evaluation, and carrying out comparison analysis on test forecast results and existing numerical forecast products to obtain a storm weather simulation parameterization scheme. The WRF mode is adopted to conduct parameterization scheme sensitivity test research on different types of storm simulation in northeast areas, and forecast, evaluation and inspection are conducted to give an optimal parameterization scheme for numerical simulation of various types of storm in northeast areas.

Description

Storm weather simulation forecasting method based on WRF mode

Technical Field

The invention relates to the technical field of weather forecast, in particular to a storm weather simulation forecast method based on a WRF mode.

Background

In recent years, there have been many studies on the climate characteristics of abnormal rainfall in summer in northeast, and it is believed that regional and extensive heavy rain in northeast is mainly the result of western wind zone, subtropical zone and tropical loop system interactions.

For storm prediction, two forms of numerical simulation research and aggregate prediction exist, and the numerical prediction achieves quite accurate degree no matter for situation prediction or for precipitation prediction; the method for forecasting the aggregate dynamic factor storm in the aggregate forecast uses the dynamic forecast result to release heat on the basis of the mode forecast, can correct the rainfall forecast of the mode correspondingly, and has obvious advantages in the aspect of forecasting the storm drop zone.

However, the current numerical forecasting mode has weaker capability of forecasting sudden storm weather, the accurate description of the physical process of the mode and the coordination performance of the power frame are insufficient to reflect the dynamic and thermal process of the development of an actual storm system, and the requirements on the refinement, fixed point, quantification, no gap and the like of the storm forecasting are difficult to realize.

The number of members in the set forecast is limited, and it is impossible to describe the probability density function of the current atmospheric state completely and exactly, but only the samples sampled from the current atmospheric probability density function can be represented, so how to sample the probability density function capable of reflecting the current atmospheric state as much as possible is critical.

The forecasting forms can not distinguish the storm weather system, and the forecasting accuracy can not be ensured.

Disclosure of Invention

Aiming at the technical problems, the invention aims to provide a rainstorm weather simulation forecasting method based on a WRF mode, which solves the problems that a rainstorm weather system cannot be distinguished, the forecasting accuracy is low and the like and provides a basis for the refined forecasting of the rainstorm in northeast China.

The invention provides a rainstorm weather simulation forecasting method based on a WRF mode, which comprises the following steps:

(1) Carrying out statistical analysis on the historical stormwater weather and carrying out weather classification;

(2) Selecting a plurality of cases from each type of storm, and performing sensitivity test on different storm simulation parameterization schemes in each type of storm by adopting a WRF mode to obtain optimal parameterization schemes of various types of storm simulation;

(3) And carrying out test forecast evaluation, and carrying out comparison analysis on test forecast results and existing numerical forecast products to obtain a storm weather simulation parameterization scheme.

Further, in the step (1), the historical heavy rain weather includes 199 regional heavy rains and a large scale heavy rain occurring in the northeast region of 1981-2017.

Further, in the step (1), the type of storm obtained after the weather typing includes: shear type storm, cyclone type storm, cold vortex type storm and typhoon type storm.

Further, in the step (2), a plurality of cases are selected in each type of heavy rain by analyzing the weather situation in the 500hPa weather map.

Further, in the step (2), the global forecast data for driving the WRF mode is forecast data issued by the united states atmospheric environment forecast center.

Further, in the step (2), the WRF mode adopts a double nesting scheme, the center positions of the two layers of nesting areas are located at (46.7 DEG N,125.0 DEG E), the outer areas are 110 DEG-140 DEG E,30 DEG-60 DEG N, and the resolution is 27km; the inner region is: 115 DEG to 135 DEG E,34 DEG to 54 DEG N, the resolution is 9km, the vertical direction is divided into 28 layers, and the top of the mode layer is 50hPa.

Further, in the step (2), the parameterization scheme includes a microphysics scheme, a short-and-short wave radiation scheme, a land process, a boundary layer scheme and a cloud accumulation parameterization scheme, and two schemes are selected in each sensitivity test respectively to form 32 scheme combinations.

Further, in the step (3), the existing numerical forecasting product includes a T639 global set forecasting system and a middle-term weather forecasting center mode in europe.

Further, in the step (3), the inspection data includes 0.1 degree x 0.1 degree precipitation data, 0.1 degree gpm precipitation data and site precipitation data of the automatic station and CMORPH precipitation product fusion of China provided by the China weather science data sharing service network.

Further, in the step (3), the WRF forecast precipitation field is interpolated to a grid matched with the live data for scoring inspection.

According to the WRF mode-based storm weather simulation forecasting method, the WRF mode is adopted to conduct parameterization scheme sensitivity test research on different types of storm simulation in northeast areas, and forecasting evaluation inspection is conducted to give the optimal parameterization scheme of the northeast areas and various types of storm numerical simulation thereof. The method can effectively solve the problem that the storm weather system cannot be distinguished in the existing weather forecast process, can greatly improve the forecast accuracy, and realizes the fine forecast of the storm in northeast areas. The method lays a solid foundation for early release of storm disaster early warning information and research of storm characteristic rules in northeast areas. The method has the advantages that the precise forecasting guarantee is carried out on the intensity, the falling area and the occurrence time of the storm in northeast areas, and the method is ready for releasing the early warning information of the storm disaster in advance and reducing the influence caused by the storm disaster to the maximum extent.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic flow chart of a method for simulating and forecasting heavy rain weather in the invention;

Fig. 2 is a graph showing the predictive effect score of sh_1 in example 1;

FIG. 3 is a graph of the forecast bias for SH_1 in example 1;

fig. 4 is a graph showing the prediction effect score of sh_2 in example 1;

FIG. 5 is a graph of the forecast bias for SH_2 in example 1;

Fig. 6 is a graph showing the prediction effect score of sh_3 in example 1;

FIG. 7 is a graph of the forecast bias for SH_3 in example 1;

FIG. 8 is a graph showing the evaluation of the effect of 50mm of each case of heavy rain on the precipitation threshold.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.

Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

Some embodiments of the present invention are described in detail below with reference to the accompanying drawings. The following embodiments and features of the embodiments may be combined with each other without conflict.

The method for simulating and forecasting the heavy rain weather based on the WRF mode is mainly used for carrying out statistical analysis on heavy rain disastrous weather in the northeast area from 1981 to 2017, summarizing the types and the characteristics of the heavy rain disastrous weather in the area, further carrying out simulation research on various heavy rain weather historical samples by applying the mesoscale WRF mode, and calculating various localized physical parameterization schemes with the best heavy rain forecasting effect in the northeast area through design of historical sample simulation tests.

The technical route of the storm weather simulation forecasting method is shown in figure 1.

Mainly comprises the following steps:

(3) And carrying out test forecast evaluation, comparing and analyzing test forecast results with the existing numerical forecast products, and respectively carrying out quantitative inspection on the mode calculation efficiency, the running time and the forecast effect to obtain a storm weather simulation parameterization scheme.

The WRF mode adopts a double nesting scheme, the central positions of the two layers of nesting areas are located at (46.7 DEG N,125.0 DEG E), the outer areas are 110 DEG-140 DEG E,30 DEG-60 DEG N, and the resolution is 27km; the inner region is: 115 DEG to 135 DEG E,34 DEG to 54 DEG N, the resolution is 9km, the vertical direction is divided into 28 layers, and the top of the mode layer is 50hPa.

Through sensitivity test research, an optimal parameterization scheme for simulating different types of heavy rain is formed.

The storm weather simulation and prediction method based on the WRF mode mainly comprises the following steps:

Statistical analysis is carried out on the heavy rain disastrous weather in the northeast area of China in 1981, the types and the characteristics of the heavy rain weather are summarized, and 199-time regional heavy rains and large-scale heavy rains in the northeast area in 1981-2017 are subjected to weather classification. The stormwater weather situation in northeast areas can be divided into 4 major categories, namely shear stormwater, cyclone stormwater, cold vortex stormwater and typhoon type stormwater;

2-3 typical cases are selected for each type of storm, sensitivity test research is carried out on storm simulation parameterization schemes in each case by adopting a WRF mode, 2 schemes are respectively selected from 5 main parameterization schemes, namely a microphysics scheme, a short-and-short wave radiation scheme, a land process, a boundary layer scheme and a cloud accumulation parameterization scheme, 32 scheme combinations are combined, sensitivity test is carried out, and finally, the optimal parameterization scheme of various types of storm simulation is given;

And (3) carrying out test forecast evaluation test, comparing and analyzing a test forecast result TS score with a T639 global set forecast system and a European middle weather forecast center mode, and respectively carrying out quantitative test on the mode calculation efficiency, the running time and the forecast effect.

The invention selects the forecast data of the global spectrum mode (Global Spectral Model, GSM) for the mode driving field, and specifically comprises the following steps:

The method comprises the steps of selecting forecast data issued by a Global Forecast System (GFS) of an atmospheric environment forecast center for mode driving, wherein the triangular cutoff wave number of a global spectrum mode is 254, the global Gaussian grid is 768 multiplied by 384, the global Gaussian grid is approximately equivalent to 0.5 degree multiplied by 0.5 degree, the vertical layering is 64 sigma layers, and the time interval is 6 hours from the ground to about 2.7 hPa;

the precipitation data of a conventional observation station and an automatic weather station in 1981 of northeast China are selected as historical reference data;

the automatic station and CMORPH precipitation product fusion hour-by-hour 0.1 degree x 0.1 degree precipitation data, 0.1 degree gpm precipitation data, site precipitation data and the like provided by the China meteorological science data sharing service network are used as detection data.

The following describes the detection data of the selected typical examples:

The simulated inspection data of the heavy rain examples of 1 month in 2008, 20 months in 2010, 16 months in 2013, 7 months and 9 months in 2010 are hour-by-hour 0.1 degree x 0.1 degree precipitation data of the integration of automatic stations of China and CMORPH precipitation products provided by the China meteorological science data sharing service network;

The selected test data of the case of the heavy rain of the 8 th 2017 and the 3 rd is 0.1 DEG gpm precipitation data;

The test data selected from the group consisting of 7.8.21 days in 1997, 6.21 days in 1981, 6.16 days in 1984 and 16 days in 1994 are site precipitation data.

According to the invention, 2 schemes are respectively selected from 5 main parameterization schemes, namely a microphysics scheme, a long-short wave radiation scheme, a land process, a boundary layer scheme and a cloud accumulation parameterization scheme, and the sensitivity test is carried out by combining 32 schemes to the 5 th power of 2.

In the storm types obtained from the early statistical classification, typical representative examples of various types of storm weather are screened from each type of storm by analyzing weather conditions in a 500hPa weather diagram, 2-3 typical examples are selected from each type of storm, sensitivity research of different parameterization schemes is carried out on the typical examples, and an optimal parameterization scheme capable of simulating the type of storm is finally calculated through repeated debugging.

The screening process and standard of the typical examples in each type of storm include:

In a typical example selection of cut-type stormwater weather, a low groove appears in a weather diagram of 500hPa (35-55 DEG N, 110-140 DEG E), a latitude area in the whole east Asia continent presents two grooves and one ridge, two grooves are respectively formed in the northeast areas of Siberia and China, a ridge is formed near the Bagai lake, the lowest potential height of the groove area in the northeast area is less than 5800gpm, the groove line is positioned at Mo River, the red peak and the Beijing first line, and the groove line is continuously developed and moves towards the northeast area within 12 hours, and at the moment, the groove line is positioned at Huma, the Changchun and the Dandong first line, and the cut-type stormwater weather is selected as a typical example of cut-type stormwater weather when the minimum potential height of the groove area is less than 5800 gpm.

In a typical example selection of cyclone type heavy rain weather, in a weather diagram of 500hPa (50-60 DEG N, 105-125 DEG E), namely closed low value centers appear in the east to northeast regions of inner Mongolia in Begar lake, the central air pressure value is lower than 550hPa, and groove lines extending from the low value centers are positioned in full, red peaks and first-line in the sun; the low-pressure center continues to move to northeast areas within 12 hours in the future, the groove line moves to the first line of the Hulleber, the Changchun and the Dandong, and the central air pressure value continues to be kept below 550hPa, so that the typical example of cyclone type heavy rain weather is selected.

In the typical selection of the cold vortex type storm weather, a closed contour line appears in a weather diagram of 500hPa (45-55 DEG N, 105-125 DEG E), the central air pressure value is below 560hPa, and the cold center or obvious cold groove is matched, the life history is maintained for at least 3 days, the low-pressure center continues to move to the northeast area within 12 hours in the future, and the intensity of the low-pressure center is maintained unchanged, so that the cold vortex type storm weather is defined as the typical example of the cold vortex type storm weather.

In the typical selection of typhoon type storm weather, the typhoon center is positioned in a yellow sea area in a 500hPa weather diagram, typhoons continue to land on the north near a Dalian area within 12 hours, at the moment, the typhoon system is caught by a middle latitude Western wind system after landing is weakened within the range of 35-55 DEG N and 100-140 DEG E, and the typhoon system continues to move to the northeast area under the action of a Western wind belt, so that the typhoon type storm weather is selected as the typical example.

In the invention, the WRF forecast precipitation field is interpolated on a grid matched with the live data for scoring and checking, the d02 area is checked, and the marine precipitation is not checked and analyzed.

The microphysics scheme of the WRF mode in the invention adopts Thompson graupel scheme and WSM 6-class graupel scheme, which are respectively represented by numerals 8 and 6 in the mode; the long-short wave radiation scheme adopts an RRTM long-wave radiation scheme and a Goddard short-wave radiation scheme, and the long-short wave radiation scheme and the Goddard short-wave radiation scheme are respectively represented by numbers 4 and 5 in modes; the land procedure adopts unified Noah scheme and RUC scheme, which are respectively denoted by the numerals 2 and 3 in the mode; the boundary layer scheme adopts a YSU scheme and a Mellor-Yamada-Janjic (Eta) TKE scheme, and the modes are respectively represented by numerals 1 and 2; meanwhile, when the boundary layer scheme is a YSU scheme, the near-ground scheme is a Monin-Obukhov scheme, and when the boundary layer scheme selects a Mellor-Yamada-Janjic (Eta) TKE scheme, the near-ground layer scheme is a MYJ Monin-Obukhov scheme correspondingly; the cloud parameterization scheme adopts a Betts-Miller-Janjic scheme and a Tiedtke scheme, which are respectively represented by numerals 1 and 6 in the mode; the different parameterization schemes described above are explained below.

Wherein Thompson graupel scheme improves the earlier Reisner scheme, and adopts cooper formula to replace ice crystal nucleation process of Fletehe curve; walkoetal is utilized in the automatic conversion process instead of the original Kessler process; the index distribution process is replaced by the generalized gamma distribution; the drag coefficient of the aragonite is determined by the mixing ratio of the aragonite rather than by a constant; according to improvement 3, the freezing growth of snow exceeds the de-sublimation growth of snow before the snow is converted into aragonite; utilizing a drag coefficient of a temperature-dependent dimensional distribution of snow; the distribution of the dimensional drag coefficient of rain is subject to the magnitude of the rain mix ratio and can therefore be used to calculate the drop course of raindrops and small raindrops.

The WSM 6-class graupel scheme adds the forecast variable of the aragonite and some processes related to the aragonite, so that the forecast amount of water-phase substances reaches 6, and the micro-physical process is more complex. WSM 6-class graupel scheme saturation adjustment the ice and water saturation process was handled separately as Dudhia and Hong et al, optimizing the calculated magnitude, reducing the sensitivity of the scheme to the mode time step, and being suitable for high resolution simulations.

The radiation delivery of the RRTM long-wave radiation scheme uses a method of correlation with K (correlation K method) to calculate the flux and cooling rate of the atmospheric long-wave spectral domain (10-3000 cm-1). Molecular species contemplated by the model include water vapor, ozone, carbon dioxide, methane, nitrogen dioxide, and halocarbons. The K distribution is obtained directly from LBRTM line-by-line modes, which provide the absorption coefficient required for RRTM, and the look-up table is preset to accurately represent the long wave process due to absorption of the above molecular species.

The Goddard short wave radiation scheme includes an atmospheric flow mode, a mesoscale mode, and a cloud mode. It calculates the solar radiation flux due to the absorption of moisture, ozone, carbon dioxide, oxygen, clouds and aerosols, and due to the scattering of clouds, aerosols and various gases.

The unified Noah scheme can operate alone as a one-dimensional single point mode or can be coupled with an atmospheric mode. This model has 4 layers of soil (0.1, 0.4, 1.0, 2.0 m), the calculation of soil temperature adopts the soil heat conduction equation, the soil humidity adopts the Richard equation, and the runoff calculation adopts a simple water balance method. It uses finite difference space division method and Crank-Nicholson time integration scheme when integrating the control equation. The unifiedNoah scheme can forecast the influence of icing and snow on soil, improves the capability of treating urban ground, and considers the property of a ground emitter.

The RUC scheme contains 6 soil layers and 2 snow layers, and the influences of frozen soil and snow cover in the energy and water conveying process, vegetation effect and canopy water are carefully considered in the soil icing process, uneven snow lands and temperature and density differences of snow.

The YSU scheme uses the inverse gradient term to represent the flux caused by non-localization and explicitly processes the rolling layer on top of the planetary boundary layer. From the results of the large scale vortex pattern study, the tendrils were expressed as amounts proportional to the surface buoyancy flux.

The Mellor-Yamada-Janjic (Eta) TKE scheme uses a boundary layer and turbulence parameterization process in the free atmosphere instead of the 2.5-order turbulence closed model of Mellor-Yamada, which predicts turbulent kinetic energy and mixes locally vertically. The scheme calls SLAB (thin layer) mode to calculate the temperature of the ground; the exchange coefficients were calculated with similar theory before SLAB, and the vertical fluxes were calculated with an implicit diffusion scheme after SLAB. When the boundary layer regime is this regime, the near-ground layer regime generally corresponds to the MYJ Monin-Obukhov regime.

The Betts-Miller-Janjic scheme makes relaxation adjustments to the thermal profile over a given period of time, during which time the convective mass flux can consume some effective buoyancy. The scheme determines buffering time and a convection profile according to cloud efficiency representing the characteristics of the river basin, and modifies a trigger mechanism so as to adapt to higher horizontal resolution.

The Tiedtke scheme is a mass flux type scheme, and under the condition of removing a time scale, the deep convection, a wind accumulation cloud area and a subtropical organized convection phenomenon can be effectively described by considering shallow convection effective potential energy and momentum transmission.

In the WRF mode, FNL data is used as the driving field, and when the selected case is too early, the driving field data is ERA-Interm data. The double nesting scheme is adopted, the central positions of the two layers of nesting areas are located at (46.7 DEG N,125.0 DEG E), the resolution of a coarse grid is 27km, the resolution of a fine grid is 3km, the map projection adopts Langmuir projection, the vertical direction is 27 non-equidistant sigma layers, the top of a mode layer is 50hPa, the integral step length is 120s, and the front 12h of mode integration is taken as spin-up time.

The settings of the specific experimental parameterization scheme in the invention are shown in the following table:

in four different types of stormwater weather, 2-3 typical examples were selected for each type of stormwater, and a total of 9 stormwater courses were selected, as shown in the following table:

In the trial forecast evaluation, the forecast effect is evaluated mainly for the forecast field and the observation field, and the following statistics are used to evaluate the forecast effect. The following are the calculation formulas of hit rate forecast score (Threat score, TS), miss rate (MISSING RATE, MR), miss rate (FALSEALARM RATIO, FAR) and forecast bias (Frequency bias index, FBI), respectively:

the definitions of a, b, c, d in formulas (1) - (4) are shown in the following table:

In the formula (1), TS is used for evaluating the forecasting effect of a precipitation event meeting a certain precipitation threshold, the value range is 0-1, and when the precipitation event is accurately forecasted (b=0, c=0), the value is equal to 1, and the forecasting effect is the best; when the rainfall event prediction is inaccurate (a=0), its value is 0, and there is no prediction skill.

In the formula (2), MR represents the specific gravity of the area which is not reported in the actual precipitation area and occupies the whole actual precipitation area, and the value range is 0-1, and the smaller the value, the better.

In the formula (3), the area where no precipitation actually exists in the forecast precipitation area accounts for the specific gravity of the total forecast precipitation area, and the value range is 0-1, and the smaller the value, the better.

In the formula (4), the FBI is mainly used for measuring the forecast deviation of a model on precipitation of a certain magnitude, and the score is equal in value to the ratio of the total grid number meeting a certain precipitation threshold value to the total grid number of the corresponding live precipitation in the forecast area, so as to reflect the overall forecast effect of the precipitation. When no empty report and no missing report (b=0, c=0), the value is equal to 1, and the forecast is completely accurate; when the missing report is greater than the empty report times (c > b), the value is less than 1, and the forecast is insufficient; and when the empty report is larger than the number of missed reports (b > c), the value is larger than 1, and the prediction is excessive.

The test data of SH_1, SH_2, CY_2 and VO_2 storm are automatic station and CMORPH hour precipitation fusion products with resolution of 0.1 degree multiplied by 0.1 degree, the test data of SH_3 storm is 0.1 degree gpm precipitation data, and the site precipitation data of CY_1, VO_1, TY_1 and TY_2 storm are selected. The WRF forecast precipitation field is interpolated to a grid matched with the live data, and then comparison analysis is carried out, only the d02 area is inspected at this time, and the marine precipitation is not inspected and analyzed.

Example 1

According to the WRF mode-based stormwater weather simulation forecasting method, sensitivity test research is conducted on a simulation parameterization scheme of cut-off stormwater, and development and explanation are conducted.

In this example, 3 cases were selected from the group consisting of 1/8/2008, 20/7/2010, and 3/8/2017.

In 2008, an example of precipitation observation data is SH_1, namely an automatic station and CMORPH hour precipitation fusion product, and FNL data is used as a WRF mode driving field. The simulation time is from 31 days 0000UTC of 7 months in 2008 to 1200UTC of 1 days in 8 months, and the first 12 hours is Spin-up time.

In order to compare the simulation effect of each parameterization scheme on the example, relevant statistics are calculated according to formulas (1) - (3), and the selected precipitation thresholds respectively take 10mm, 25mm and 50mm and respectively correspond to medium rain, heavy rain and heavy rain.

Referring to fig. 2, the 32 schemes are plotted on the abscissa, the first to third columns of pictures represent TS, MR and FAR scores, respectively, and the first to third rows represent precipitation thresholds taking the corresponding precipitation forecast scores of 10mm, 25mm and 50mm, respectively. Different horizontal lines represent different average forecast scores. The box represents the optimal parameterization scheme for this case.

It can be seen that the stormwater TS scores for the 32 protocols simulated this time are all lower than 0.26, less than T639, the forecaster and the forecast scores for the EC mode, with the scores for protocol 1, protocol 7, protocol 17, protocol 18, and protocol 21 being higher. The scores for scheme 7, scheme 8 and scheme 11 were lower in the MR scores. 32 of the FAR scores were overall higher than the forecaster, T639, and EC modes of forecasting scores, with lower scores for scheme 1, scheme 7, scheme 18, and scheme 21. In view of the above, although the stormwater TS score for scenario 18 is not the highest, the empty report rate and the miss report rate for scenario 18 are the lowest, so the simulation results for scenario 18 in SH_1 are optimal.

In order to intuitively analyze the overall rainfall simulation effect of meeting a certain rainfall threshold in 32 scheme prediction areas in SH_1, the FBI value is calculated according to a formula (4), and the selected rainfall thresholds are respectively 10mm, 25mm and 50mm and respectively correspond to medium rain, heavy rain and heavy rain.

Referring to fig. 3, (a) a prediction bias corresponding to 10mm is taken for the precipitation threshold, (b) a prediction bias corresponding to 25mm is taken for the precipitation threshold, and (c) a prediction bias corresponding to 50mm is taken for the precipitation threshold, with the middle horizontal line representing an FBI value of 1.

It can be seen that, compared with actual precipitation, in the medium rain forecast, there are 10 kinds of scheme simulation forecast results with underestimation phenomenon, the FBI values of scheme 10 and scheme 28 are close to 1, and there are 20 kinds of scheme simulation forecast results with overestimation phenomenon, wherein scheme 18 has underestimation phenomenon; in heavy rain and heavy rain forecasting, overestimation exists in the simulation forecasting results of 32 schemes.

The shear type northeast storm process in sh_1 is mainly focused on 1200UTC at 31, 7, 2008 to 1200UTC at 1,8 months.

As can be seen, the live precipitation drop zone is in a lump shape, and the storm center in the 24-hour accumulated precipitation is mainly distributed in Liaoning and Jilin, and the maximum precipitation can reach more than 60 mm. In the simulation result of the scheme 18, the storm center of the 24-hour accumulated precipitation extends to the south-to-east area of the Heilongjiang river, and the maximum precipitation can reach more than 80 mm. In contrast to the live, the simulation results of scheme 18 generally agree with the observations, but the local area is too heavy in rain, the center of the storm extends in the northeast-southwest direction, and the storm intensity is higher than the live. Scheme 18 provides a greater difference from the live precipitation simulation results for the north regions of the black longjiang.

In the case of SH_2 for 7 th 2010 and 20 th, the precipitation observation data is a fusion product of automatic standing and CMORPH hours precipitation, and FNL data is used as a WRF mode driving field. The simulation time is 1200UTC from 7 months of 2010 to 21 days of 7 months and 0000UTC, and the first 12 hours are Spin-up time.

In order to intuitively analyze the simulation effect of each parameterization scheme on the example, relevant statistics are calculated according to formulas (1) - (3), and the selected precipitation thresholds respectively take 10mm, 25mm and 50mm and respectively correspond to medium rain, heavy rain and heavy rain.

As can be seen in fig. 4, there are 11 schemes with a stormwater TS score greater than 0.32, higher than the forecaster, T639 and EC modes of forecasting. The 32 protocols had a lower MR, FAR score for storms as a whole than the forecaster, T639 and EC modes of forecasting, with the MR, FAR score for protocol 14 being lowest. In view of the above, although the stormwater TS score for scenario 14 is not the highest, the empty report rate and the miss report rate for scenario 14 are the lowest, so the simulation results for scenario 14 in SH_2 are optimal.

In order to intuitively analyze the overall rainfall simulation effect of meeting a certain rainfall threshold in 32 scheme prediction areas in SH_2, the FBI value is calculated according to a formula (4), and the selected rainfall thresholds are respectively 10mm, 25mm and 50mm and respectively correspond to medium rain, heavy rain and heavy rain.

Referring to fig. 5, (a) a prediction bias corresponding to 10mm is taken for the precipitation threshold, (b) a prediction bias corresponding to 25mm is taken for the precipitation threshold, and (c) a prediction bias corresponding to 50mm is taken for the precipitation threshold, with the middle horizontal line representing an FBI value of 1.

Compared with actual precipitation, the simulation and forecast results of 32 schemes have overestimation in the forecast of medium and heavy rain; in the storm forecast, there are 6 kinds of scheme simulation forecast results with underestimation phenomenon, and 26 kinds of scheme simulation forecast results with overestimation phenomenon, wherein scheme 14 with overestimation phenomenon.

The shear type northeast storm process in sh_2 is mainly focused on 7 months 20 days 0000UTC to 7 months 21 days 0000UTC in 2010.

As can be seen, the live precipitation area is in southwest-northeast trend, and the storm is mainly distributed in the middle of Liaoning, jilin and the south of Heilongjiang, and the maximum precipitation amount can reach 60mm. Compared with the live condition, the scheme 14 can basically simulate the main rain band of the precipitation and the trend thereof, and the center of the heavy rain is closer to the live condition but the intensity and the range of the heavy rain are larger. Scheme 14 has poor effect on simulating the precipitation conditions in the north of the Heilongjiang river.

In 2017, 3 days 8 are abbreviated as SH_2, and the precipitation observation data is GPM observation data, and FNL data is WRF mode driving field. GPM observations are generally more intense than automatic standing and CMORPH hour precipitation fusion products, so this case has a higher rate of missing reports. The simulation time is 2017, 8 months, 2 days 0000UTC to 8 months, 3 days, 1200UTC, and the first 12 hours is Spin-up time.

As can be seen in fig. 6, the stormwater TS score for scheme 17 is about 0.21, higher than the other 31 schemes, but lower than the forecaster, T639 and EC modes. The stormwater MR score for scheme 17 was lowest, approximately 0.74, higher than the forecaster, EC mode, and lower than the forecast score of T639. The stormwater FAR score of scheme 17 is lower than the forecaster, T639 and EC mode forecasting scores. In view of the above, in sh_3, although scheme 17 has a higher rate of false alarm than other schemes, the result of scheme 17 is optimal as a whole.

In order to intuitively analyze the overall rainfall simulation effect of meeting a certain rainfall threshold in 32 scheme prediction areas in SH_3, the FBI value is calculated according to a formula (4), and the selected rainfall thresholds are respectively 10mm, 25mm and 50mm and respectively correspond to medium rain, heavy rain and heavy rain.

Referring to fig. 7, (a) a prediction bias corresponding to 10mm is taken for the precipitation threshold, (b) a prediction bias corresponding to 25mm is taken for the precipitation threshold, and (c) a prediction bias corresponding to 50mm is taken for the precipitation threshold, with the middle horizontal line representing an FBI value of 1.

It can be seen that, compared with actual precipitation, in the medium rain forecast, there are 5 kinds of scheme simulation forecast results with underestimation phenomenon, the FBI value of scheme 4 is close to 1, there are 26 kinds of scheme simulation forecast results with overestimation phenomenon, and the scheme 17 has overestimation phenomenon; in heavy rain and heavy rain forecast, the simulation forecast results of 32 schemes have underestimation phenomenon, and accord with the characteristic of strong precipitation in GPM observation data, wherein the FBI value of scheme 17 is closer to 1 than other schemes.

The shear type northeast storm process in sh_3 is mainly focused on 1200UTC at 8 months 2 to 1200UTC at 3 months 8.

It can be seen that the storm center of live precipitation is in northeast-southwest trend, and is mainly located in Liaoning, jilin, southeast of inner Mongolia and southeast of Heilongjiang, and the accumulated maximum precipitation amount for 24 hours can reach more than 90 mm. Compared with the live condition, the main position of the storm center of the simulation result of the scheme 17 is consistent with the observation, but the simulated storm center is more divergent and has smaller intensity, and weaker storm centers exist in the middle part and the north part of the Heilongjiang.

In the embodiment, 2 precipitation observation data are selected from 3 cases of shear type heavy rain, namely GPM observation data and automatic standing and CMORPH hour precipitation fusion products. The GPM observation data selected by SH_3 is generally stronger in precipitation compared with the automatic station selected by SH_1 and SH_2 and the CMORPH hour precipitation fusion product, so that the missing report rate of SH_3 is higher.

Through respectively carrying out sensitivity test researches on 3 northeast region cut-off type storm examples, the cut-off type storm simulation result is found to be ideal, the test is accurate for simulating a precipitation fall region and a storm center, and the simulation forecast results of the cut-off type storm in the northeast region are good through the scheme 14, the scheme 17 and the scheme 18. The parameterization settings for the 3 schemes are shown in the table below.

The microphysics, long and short waves, land and boundary layer parameterization schemes selected for scheme 17 and scheme 18 are consistent, and scheme 14 differs significantly from scheme 17 and scheme 18. By analyzing fig. 4, it can be seen that scheme 18 also responds well to the case of a 7 th 2010 20 th storm. The method is characterized in that for numerical simulation of tangential storm in northeast, WSM 6-class graupel scheme is selected for parameterization of the microphysics process, RRTM scheme is selected for parameterization of short-and-long-wave radiation, unified Noah scheme is selected for land process, YSU scheme is selected for parameterization of boundary layer, and cloud accumulation parameterization scheme is selected according to specific conditions, so that ideal simulation results can be obtained.

Example 2

According to the WRF mode-based stormwater weather simulation forecasting method, sensitivity test research is conducted on a cyclone type stormwater simulation parameterization scheme, and the research process is the same as that of the method in the embodiment 1.

In this example, 2 cases were selected from the group consisting of 21/8/1997 and 16/7/2013.

Through respectively carrying out sensitivity test researches on 2 cyclone type storm examples in northeast areas, the simulation of a storm falling area is found to be ideal, wherein the result of the cyclone type storm simulation in the northeast areas is good in scheme 8. The parameterization settings for the 2 schemes are shown in the table below.

This shows that for numerical simulation of cyclone type storm in northeast, the micro-physical process parameterization selects Thompson graupel scheme, the short-wave radiation parameterization selects RRTM scheme, the land-based process selects RUC scheme, the boundary layer parameterization selects Mellor-Yamada-Janjic (Eta) TKE scheme, and the cloud accumulation parameterization selects Tiedtke scheme, a better simulation effect is obtained.

Example 3

According to the WRF mode-based stormwater weather simulation forecasting method, sensitivity test research is conducted on a simulated parameterization scheme of cold vortex type stormwater, and the research process is the same as that of the method in the embodiment 1.

In this example, 2 cases were selected from the group consisting of 21/6/8/9/1981.

Through the sensitivity test research on 2 cold vortex storm examples in northeast areas, the falling area for simulating the cold vortex storm is found to be ideal, wherein the result of simulating the cold vortex storm in northeast areas is good in scheme 10 and scheme 11. The parameterization settings for the 2 schemes are shown in the table below.

The optimal schemes corresponding to the 2 cases are identical except for different boundary layer parameterization schemes. This shows that for numerical simulation of cold vortex storm in northeast, the micro-physical process parameterization selects Thompson graupel scheme, the short-wave radiation parameterization selects Goddard scheme, the land process selects unified Noah scheme, the cloud accumulation parameterization selects Tiedtke scheme, and the boundary layer parameterization scheme can be selected according to specific conditions, so that a better simulation effect can be obtained.

Example 4

According to the WRF mode-based stormwater weather simulation forecasting method, sensitivity test research is conducted on a typhoon type stormwater simulation parameterization scheme, and the research process is the same as that of the method in the embodiment 1.

In this example, 2 cases were selected from 16 days of 6 months of 1984 and 16 days of 8 months of 1994.

Through the sensitivity test study on 2 typhoon type storm examples in northeast areas, the simulation intensity of typhoon type storm is ideal, wherein the simulation results of typhoon type storm in northeast areas are good in scheme 15 and scheme 7. The parameterization settings for the 2 schemes are shown in the table below.

The 2 schemes are identical except that the short-and-short wave radiation parameterization schemes are selected differently. This shows that for numerical simulation of typhoons and storms in northeast areas, the micro-physical process parameterization selects Thompson graupel scheme, the land process selects RUC scheme, the boundary layer parameterization selects Mellor-Yamada-Janjic (Eta) TKE scheme, the cloud accumulation parameterization selects Betts-Miller-Janjic scheme, and the short-wave radiation parameterization scheme is judged according to specific conditions, so that a better simulation effect is obtained.

By performing parameterization scheme sensitivity tests on 9 cases corresponding to 4 types of northeast storm in northeast regions, namely tangential storm, cyclone storm, cold vortex storm and typhoon storm respectively, it can be found that the optimal parameterization schemes corresponding to the various types of storm have large difference. In the service operation process, it is often not known which type of storm is specific, and in order to select a parameterization scheme for numerical simulation as soon as possible, a general parameterization scheme needs to be provided for storm simulation in northeast areas.

Fig. 8 shows that the precipitation threshold was scored in TS of 50mm each case, and the abscissa shows each case, (a) shows the case of 1 st of 2008, b) shows the case of 20 th of 2010, c) shows the case of 3 rd of 2017, d) shows the case of 21 st of 1997, e) shows the case of 16 th of 2013, f) shows the case of 21 st of 1981, g) shows the case of 9 th of 2010, h) shows the case of 16 th of 1984, and i) shows the case of 16 th of 1994. The middle horizontal line represents the forecaster average forecast score, the lower horizontal line represents the T639 mode average forecast score, and the upper horizontal line represents the EC mode average forecast score. The box represents the selection of the optimal scheme for the numerical simulation of the heavy rain in northeast areas.

When each case was selected for regimen 7, the TS scores were at a higher level than the other regimens in each case, except for the 2013, 7, 16, and 21, 6, and 1981, heavy rains. The overall TS score of each of the 2013 7 and 16 heavy rains is far lower than that of the other cases. The occurrence time of the 21 st storm in 6 th 1981 is earliest, and the quality of the data for simulation and evaluation is poor. The 7.16-2013 and 21-1981 cases are poorly represented.

In view of the above, the recommended scheme 7 is a general parameterized scheme for numerical simulation of heavy rain in northeast areas. When the numerical simulation is needed for the storm in northeast area and the type of the storm in northeast area is unknown, the micro-physical process parameterization selects Thompson graupel scheme, the short-wave radiation parameterization selects RRTM scheme, the land process selects RUC scheme, the boundary layer parameterization selects Mellor-Yamada-Janjic (Eta) TKE scheme, and the cloud accumulation parameterization selects Betts-Miller-Janjic scheme, so that a more proper simulation effect can be obtained.

By adopting the WRF mode-based storm weather simulation forecasting method, the precise forecasting of the storm in northeast area can be realized, and a unified storm weather simulation parameterization scheme suitable for the northeast area can be obtained after simulation analysis and research on different storm types, so that the forecasting accuracy is effectively improved.

The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. The stormwater weather simulation forecasting method based on the WRF mode is characterized by comprising the following steps of:

(2) Selecting a plurality of cases from each type of storm, and performing sensitivity test on different storm simulation parameterization schemes in each type of storm by adopting a WRF mode to obtain optimal parameters of various types of storm simulation;

(3) Performing test forecast evaluation, and comparing and analyzing test forecast results with the existing numerical forecast products to obtain storm weather simulation parameters;

In the step (1), statistical analysis is performed on storm disasters in the northeast area of China in 1981, and the analysis process is to classify 199-time regional storms and large-scale storms occurring in the northeast area in 1981-2017 into 4 major categories, namely shear type storms, cyclone type storms, cold vortex type storms and typhoon type storms, by summarizing the types and the characteristics of the storms;

In the step (2), 2-3 typical cases are selected for each type of storm, sensitivity test research is carried out on storm simulation parameterization schemes in each case by adopting a WRF mode, 2 schemes are respectively selected from 5 parameterization schemes, namely a microphysics scheme, a short-and-short wave radiation scheme, a land process, a boundary layer scheme and a cloud accumulation parameterization scheme, 32 scheme combinations are used for carrying out sensitivity test, and finally, the optimal parameterization scheme of each type of storm simulation is provided;

The microphysics scheme adopts Thompson graupel scheme and WSM 6-classgraupel scheme, which are respectively indicated by numerals 8 and 6 in the mode; the long-short wave radiation scheme adopts an RRTM long-wave radiation scheme and a Goddard short-wave radiation scheme, and the long-short wave radiation scheme and the Goddard short-wave radiation scheme are respectively represented by numbers 4 and 5 in modes; the land procedure adopts unified Noah scheme and RUC scheme, which are respectively denoted by the numerals 2 and 3 in the mode; the boundary layer scheme adopts a YSU scheme and a Mellor-Yamada-Janjic (Eta) TKE scheme, and the modes are respectively represented by numerals 1 and 2; meanwhile, when the boundary layer scheme is a YSU scheme, the near-ground scheme is a Monin-Obukhov scheme, and when the boundary layer scheme selects a Mellor-Yamada-Janjic (Eta) TKE scheme, the near-ground layer scheme is a MYJ Monin-Obukhov scheme correspondingly; the cloud parameterization scheme adopts a Betts-Miller-Janjic scheme and a Tiedtke scheme, which are respectively represented by numerals 1 and 6 in the mode;

In the step (3), the method for performing contrast analysis includes the following steps: in four different types of heavy rain weather, 2-3 typical cases are selected for each type of heavy rain, and a total of 9 heavy rain processes are selected for specific comparison analysis;

In the trial prediction evaluation, the following statistics are used to evaluate the quality of the prediction effect for the prediction field and the observation field, and the following calculation formulas are respectively shown as hit rate prediction score Ts, miss rate M _R, empty rate F _AR and prediction deviation F _BI:

Wherein a represents the correct number of forecast, i.e. forecast with precipitation and actually observed precipitation, b represents the number of empty report, i.e. forecast with precipitation but not observed precipitation in real time, c represents the number of missed report, i.e. forecast without precipitation but actually observed precipitation;

In the formula (1), hit rate forecast score Ts is used for evaluating the forecast effect of a precipitation event meeting a certain precipitation threshold, the value range is 0-1, and when the precipitation event forecast is accurate, b=0, c=0, the value is equal to 1, and the forecast effect is the best; a=0 when the rainfall event forecast is inaccurate, its value is 0;

in the formula (2), the missing report rate M _R represents the specific gravity of the missing report area in the actual precipitation area to occupy all the actual precipitation areas, the value range is 0-1, and the smaller the value is, the better the value is;

In the formula (3), the blank report rate F _AR is the specific gravity of the area without precipitation in the forecast precipitation area to the total forecast precipitation area, the value range is 0-1, and the smaller the value is, the better;

In the formula (4), the forecast deviation F _BI is used for measuring the forecast deviation of a model to a certain level of precipitation, the score is equal to the ratio of the total grid number meeting a certain precipitation threshold value to the total grid number of the corresponding live precipitation in a forecast area in value, the total forecast effect of the precipitation is reflected, and when no empty report and no missing report are b=0 and c=0, the value is equal to 1, and the forecast is completely accurate; when the missing report is greater than the empty report times c > b, the value is less than 1, and the forecast is insufficient; and when the number of times b > c of the empty report is larger than 1, the value is larger than 1, the forecast is excessive, the relevant statistics are calculated according to the formulas (1) - (3), the F _BI value is calculated according to the formula (4), and the selected precipitation thresholds are respectively 10mm, 25mm and 50mm and respectively correspond to medium rain, heavy rain and heavy rain.