An improved method for forecasting convective winds by assimilating densely packed ground observation data.

Abstract

This invention discloses a method for improving the forecasting of convective strong winds by assimilating dense ground-based observation data. The method includes collecting observation data from dense automatic weather stations in a preset area; setting up a convection forecasting model and conducting experiments; extracting precipitation characteristics and analyzing convection evolution based on the observation data to obtain evolution data; obtaining atmospheric circulation patterns based on the evolution data; performing ground-grid nudging assimilation and simulation evaluation on the ground meteorological element field simulation; using ground-grid nudging technology to identify key ground meteorological elements affecting the triggering and development of convection in warm-sector heavy precipitation, and obtaining the influencing mechanisms; the key meteorological elements are water vapor, temperature, and wind field; the influencing mechanisms include the influence of ground water vapor, surface temperature, and surface wind field; determining the ground-grid nudging duration for numerical forecasting of warm-sector heavy precipitation through experiments; optimizing the bias of the convection forecasting model based on the analysis of the influencing mechanisms; and outputting the forecast results.

Description

Technical Field

[0001] This invention relates to the field of atmospheric coupling, and more particularly to an improved forecasting method for convective winds by assimilating dense ground observation data. Background Technology

[0002] Existing numerical models exhibit significant uncertainties in forecasting convective systems. Previous research indicates that the formation of convective winds is influenced not only by convective systems but also by a variety of complex factors, including surface cold pools, the South China Sea monsoon, sea-land breezes, and topography. These multi-scale atmospheric processes are difficult to accurately represent in models, resulting in relatively poor forecasting skills for convective winds. While nudging four-dimensional data assimilation techniques can improve the simulation performance of numerical models, current research on the impact of surface-grid nudging techniques on the numerical forecasting (simulation) of convective winds is insufficient.

[0003] The main drawback of existing technologies lies in the low accuracy of numerical models in predicting the triggering and evolution of convection, making it impossible to reliably predict the triggering time and location of such convection. Specifically, operational numerical forecasts often suffer from missed or false alarms, and the predicted convection location and intensity deviate significantly from the actual situation.

[0004] Therefore, there is an urgent need to invent a new method for assimilating ground-based dense observation data to improve the forecasting of convective winds. By utilizing ground-based gridded nudging technology, the ability of numerical models to simulate and forecast water in convective winds can be improved, making the simulation results closer to the actual occurrence and evolution of convection. This will provide a more accurate reference for the forecasting and early warning of convective winds, and provide stronger support for power grid and other industries to resist the impact of strong convective winds. Summary of the Invention

[0005] The purpose of this invention is to provide an improved forecasting method for convective winds by assimilating dense ground observation data.

[0006] To achieve the above objectives, the present invention is implemented according to the following technical solution:

[0007] This invention includes the following steps:

[0008] Collect observation data from pre-defined, densely packed automatic weather stations, and set up models and experiments for the convection forecasting model;

[0009] Based on observational data, radar echo characteristics (reflecting convection intensity and structure) are extracted and convection evolution analysis is performed to obtain evolution data, and atmospheric circulation patterns are obtained based on the evolution data;

[0010] The surface meteorological element field simulation is performed using surface grid relaxation approximation assimilation and simulation evaluation. The key surface meteorological elements affecting the triggering and development of convection are identified using surface grid relaxation approximation technology, and the influencing mechanisms are obtained. The key meteorological elements are water vapor, temperature, and wind field. The influencing mechanisms include the influence of surface water vapor, surface temperature, and surface wind field.

[0011] The relaxation approximation time of the ground grid in the numerical forecast was determined through experiments. Based on the influence mechanism analysis, the deviation of the convection forecast model was optimized, and the forecast results were output.

[0012] Furthermore, the method for setting up the model and experiment of the convection forecasting model includes:

[0013] A mesoscale numerical model was adopted, with a high-resolution nested mesh set in three layers. Initialization and boundary conditions were set, and the model was configured in conjunction with physical processes. The physical processes included the Thompson cloud microphysical parameterization scheme, the multi-scale Cairns cumulus convection parameterization scheme, the RRTM longwave radiation scheme, the Duhay shortwave radiation scheme, the turbulent kinetic boundary layer parameterization scheme, the Moning-Olhoff near-surface scheme, and the Noah-MP land surface model scheme.

[0014] Multiple sets of comparative experiments were designed to observe that relaxation approximation and high-grid relaxation approximation had a weak impact on the results, while ground grid relaxation approximation was the key to determining convection triggering and precipitation simulation.

[0015] Based on the ground grid relaxation approximation, one set of control experiments and seven sets of sensitivity experiments were set up. The control experiments assimilated all ground elements, while the sensitivity experiments turned off the assimilation of some elements. Through comparative studies, the independent and combined effects of water vapor, temperature, and wind field were clarified. Based on the evaluation of precipitation and convection simulation analysis, the influence mechanism of each element was analyzed, the assimilation time was optimized, and the improved convection forecast model was output.

[0016] Furthermore, the method for obtaining evolution data by extracting radar echo characteristics (reflecting convection intensity and structure) and analyzing convection evolution based on observational data includes:

[0017] The process involves acquiring observational data and reanalyzing information, calculating the cumulative precipitation at each station over 18 hours, interpolating to generate a spatial distribution map to identify the location of the rain belt, obtaining the maximum cumulative rainfall and the coverage area of ​​the rain belt, and obtaining the spatial distribution of cumulative observed precipitation based on the maximum cumulative rainfall and the coverage area of ​​the rain belt. The observational data includes automatic weather station data, radar data, and radiosonde data.

[0018] Extract the hourly precipitation rate from the station with the highest precipitation, plot the time series, obtain the period of heavy precipitation and the peak time, and obtain the temporal evolution of precipitation rate based on the period of heavy precipitation and the peak time; where heavy precipitation is defined as precipitation of ≥20 mm / hour and lasting for 5 hours.

[0019] When the radar reflectivity first exceeds 35 dBZ, it is determined to be convection triggering. The radar combined reflectivity, equivalent ground potential temperature, and wind field are tracked hourly. Combined with wind field analysis, the spatiotemporal evolution of the convection system is carried out. The key processes of the spatiotemporal evolution of the convection system include the triggering stage, the organizing stage, and the dissipation stage. The triggering stage is characterized by the thermal gradient in the wet tongue region promoting uplift. The organizing stage is characterized by the formation of loose, zonal convection along the coastline.

[0020] Environmental form diagnosis is conducted based on large-scale background and local influences. Convection evolution analysis is performed through convection triggering determination, spatiotemporal evolution of the convection system, and environmental diagnosis. Circulation diagnosis is conducted based on large-scale background and local influences. Thermodynamic conditions require high convection effective energy potential and low convection inhibition energy. Dynamic conditions require convection to be idiosyncratic rather than organized.

[0021] Furthermore, the method for obtaining atmospheric circulation patterns based on the evolution data includes:

[0022] The large-scale circulation background was analyzed using a global analysis dataset to obtain geopotential height, wind field, equivalent potential temperature and precipitable water. Isobaric surface superposition analysis was carried out by combining radiosonde data and auxiliary observation data, including isobaric surfaces of 500 hPa, 850 hPa and 950 hPa.

[0023] The form analysis is performed at 500 hPa based on the location of the trough and ridge and the subtropical high. The low-level jet stream and water vapor transport are obtained at 850 hPa based on the wind field and water vapor flux. The boundary layer characteristics are obtained at 950 hPa based on the warm and humid air flow and the land-sea temperature difference. The dynamic-thermal coupling zone is identified by superimposing the 500 hPa trough line, the 850 hPa jet stream and the 950 hPa equivalent potential temperature field.

[0024] Based on unstable energy, uplift conditions, and wind shear, sounding stratification analysis was conducted. By analyzing the configuration of the trough, jet stream, and warm and humid tongue, the synergistic effects of dynamic uplift, water vapor transport, and energy supply were obtained, and water vapor parameters and energy parameters were calculated. The water vapor parameters include precipitable water, specific humidity, and water vapor flux, while the energy parameters include convective effective potential energy, convective inhibition energy, uplift condensation height, and free convection height.

[0025] The correlation of convection triggering conditions outputs the atmospheric circulation pattern. Convection triggering conditions include thermodynamic, dynamic, and water vapor conditions. Thermodynamic conditions are high convective effective potential energy, low convective suppression potential energy, low lifting condensation height, or low free convection height. Dynamic conditions are that the low-level jet provides lifting, and the weak trough has no significant suppression. Water vapor conditions are that warm and humid transport passes through the advection and turbulent mixing of high equivalent potential temperature air masses.

[0026] Furthermore, methods for ground-grid relaxation approximation assimilation and simulation evaluation of ground meteorological element fields include:

[0027] Acquire and reanalyze observation data from ground-based encrypted automatic observation stations;

[0028] Data from ground-based automatic weather stations and reanalysis data were collected. Control and sensitivity experiments were designed, and relaxation approximation parameters were set to perform ground-based grid-point relaxation approximation assimilation. The relaxation approximation parameters included an assimilation layer that only applied to the ground, and exponentially decaying time and spatial weights.

[0029] Surface meteorological elements are assessed based on water vapor mixing ratio, temperature, and wind speed; precipitation simulation assessment is obtained by comparing the spatial distribution of 18-hour cumulative precipitation and statistically analyzing the proportion of stations with precipitation at different thresholds; and convective systems are assessed based on triggering time, spatial structure, and organization morphology.

[0030] The interaction mechanisms of water vapor, temperature, and wind field are analyzed, and the assimilation parameters are optimized based on these mechanisms.

[0031] Furthermore, the method for influencing surface water vapor includes:

[0032] Five comparative experiments were designed, using the relative humidity, water vapor mixing ratio, convective available potential energy, and convective suppression energy output by the model as data sources.

[0033] Rapid humidification of the boundary layer, given the budget equation, is expressed as follows:

[0034]

[0035] Where q v This represents the average water vapor mixing ratio. This represents the local variation of the average water vapor mixing ratio. For horizontal and vertical transport items, The horizontal and vertical transport terms are represented by QBL, the effect of boundary layer turbulent mixing, QCM, the condensation and evaporation effects from cumulus convection and microphysical processes, QNU, the relaxation approximation tendency of the average water vapor mixing ratio, and R, the residual term. The calculation formula is as follows: QBL, QCM, and QNU use mode output values;

[0036] The local variation of the mean water vapor mixing ratio was calculated using the central difference method:

[0037]

[0038] Where Δt is 6 min, and the average water vapor mixing ratio at time t+Δt is q. v │ t+Δt The average water vapor mixing ratio at time t-Δt is q v │ t-ΔtThe local variation term of the average water vapor mixing ratio at time t is:

[0039] The boundary layer humidification process was verified based on the evolution of relative humidity and the increase of water vapor mixing ratio. The water vapor budget was quantified, the convection inhibition energy was increased and decreased, the condensation height was reduced, the vertical velocity disturbance was enhanced, the convection was triggered and organized, the effects of precipitation and convection were assessed, and the results of the surface water vapor impact were obtained.

[0040] Furthermore, the method for influencing ground temperature includes:

[0041] Four sets of comparative experiments were set up. Temperature relaxation approximation was used to slow down the cooling based on the cooling rate and the holding temperature to maintain the potential temperature perturbation. Vertical profile and time series were used to diagnose the potential temperature perturbation.

[0042] Calculate thermal buoyancy:

[0043]

[0044] Where g is the acceleration due to gravity, in milliseconds (ms). -2 , Potential temperature perturbation, in K. The region's average potential temperature, in Kelvin (q). v The water vapor mixing ratio, q represents the average water vapor mixing ratio. c q represents the cloud water content. r Rainwater content;

[0045] To maximize thermal buoyancy, enhance thermal buoyancy and increase vertical velocity, morphological comparisons are conducted after early convection triggering, and an organized evaluation is performed. The cross-effect verification is carried out based on the synergistic contribution of thermal buoyancy growth in the thermal buoyancy components and the contribution of single elements. Temperature and humidity assimilation synergy is achieved, and the temperature and water vapor synergistic effect is obtained based on the maximization of thermal buoyancy contribution.

[0046] Furthermore, the method for influencing the surface wind field includes:

[0047] Four sets of comparative experiments were designed. The model wind speed was reduced by relaxation approximation. The corrected wind speed was obtained by data verification based on the assimilation wind speed error. The wind direction convergence was adjusted for the comparative experiments. The four sets of comparative experiments included one reference experiment and three comparative experiments.

[0048] To obtain the convective monomers and organizational morphology of the comparative experiment, the low-level warm and humid airflow convergence was reduced by relaxation approximation, thus limiting energy supply;

[0049] The necessity of triggering convection was verified by comparing wind field assimilation alone. The synergistic effect of wind field assimilation and temperature and humidity assimilation was obtained by combining wind field assimilation with temperature and humidity assimilation in the reference experiment.

[0050] Furthermore, the method for determining the surface grid relaxation approximation time of numerical forecasts for heavy precipitation in warm regions through experiments includes:

[0051] Design a benchmark experiment with fixed assimilation parameters and only adjust the assimilation duration;

[0052] The hit rate exceeding the rainfall threshold is used as the TS score. The precipitation score is calculated based on the root mean square error of the spatial distribution of cumulative rainfall and the TS score. The convection trigger time deviation is obtained by simulating the first trigger time of convection and the time difference between observation. The boundary layer variable improvement rate is obtained by comparing the observed temperature, humidity, wind speed and the simulation correlation coefficient. The precipitation score, convection trigger time deviation, and boundary layer variable improvement rate are used as evaluation indicators.

[0053] Calculate the time decay rate of the relaxation approximation tendency term in the benchmark test and determine the critical time point at which the relaxation approximation influence will reach 10%.

[0054] By comparing the response speed of the physical processes of the model after assimilation, the autonomous recovery of the convective forecasting model is assessed based on whether the evolution of convective available potential energy and vertical velocity transitions naturally.

[0055] The duration that simultaneously satisfies the following conditions is the high TS score with the lowest root mean square error, the smallest convection trigger time deviation, and the smooth transition of the boundary layer, and the output is the relaxation approximation duration of the ground grid.

[0056] Furthermore, the method for bias optimization of the convection forecast model based on the aforementioned influence mechanism analysis includes:

[0057] Quality control is performed on ground-based densified observations, data larger than 3 standard deviations are removed, and when interpolating the model grid, the observation error covariance matrix is ​​retained. The model short-time forecast is used as the background field, and the background field error covariance is calculated.

[0058] Given an objective function, the expression is:

[0059]

[0060] Where J is the objective function, x is the model variable, and x is the model variable. obs Let Z be the observed value, Q be the observation error covariance matrix, and Z be the observed value. bg The background field is L1, and the dynamic constraint operator is L1. For spatial gradient operators, Here, B is the model's physical process tendency term, D is the background error covariance, and W is the model tendency error covariance. λ2 is the observation fitting term, λ3 is the smoothing strength parameter, λ1 is the tendency constraint weight, and λ1 is the regularization parameter, which controls the dynamic coordination weight.

[0061] Computation of dynamic constraint operators:

[0062]

[0063] Where ζ is vorticity, and θ e For equivalent potential temperature, This represents the vertical gradient of vorticity. For wind field divergence, The sensitivity of temperature relative to humidity is equivalent. This represents the sensitivity of potential temperature to temperature.

[0064] We introduce water vapor assimilation optimization parameters to constrain abrupt changes in the humidity field and maintain the Reifelli balance between the humidity and temperature fields; we introduce temperature assimilation optimization parameters to adjust the temperature field and maintain boundary layer stability; we introduce wind field assimilation optimization parameters to limit the wind field adjustment range and maintain the natural evolution of low-level convergence.

[0065] When the decrease in the continuous objective function value is less than 0.01, the iteration stops and the optimized convection forecast model is output.

[0066] The beneficial effects of this invention are:

[0067] This invention is an improved forecasting method for convective winds that assimilates densely packed ground observation data. Compared with existing technologies, this invention has the following technical advantages:

[0068] This invention improves the accuracy of numerical models in simulating pre-flood warm-sector heavy precipitation in region A by employing preprocessing, obtaining satisfactory deviations, obtaining fuzzy intentions, obtaining correction intentions, obtaining target intentions, and model building steps. It can more accurately reproduce convection and heavy precipitation processes. It clarifies the influence mechanism of different nudging surface meteorological elements on the triggering and development of warm-sector heavy precipitation convection, providing theoretical support for subsequent research. It determines an appropriate nudging duration, providing a reference for operational assimilation applications, which is beneficial for improving disaster early warning capabilities and reducing losses. Attached Figure Description

[0069] Figure 1 This is a flowchart illustrating the steps of an improved convective wind forecasting method that assimilates densely packed ground observation data according to the present invention.

[0070] Figure 2 This is a time evolution diagram of the average MCAPE and MCIN in the convection triggering region in a specific embodiment of the present invention;

[0071] Figure 3 This is a time evolution diagram of the convection triggering regions of EXP1 and EXP5 in a specific embodiment of the present invention;

[0072] Figure 4 This is a simulation result diagram in a specific embodiment of the present invention;

[0073] Figure 5 This is a graph showing the evolution of the average thermal buoyancy and its influencing factors in the convection-triggered region over time in a specific embodiment of the present invention.

[0074] Figure 6 This is a time evolution diagram of water vapor mixing ratio, temperature, and wind speed in a specific embodiment of the present invention. Detailed Implementation

[0075] The present invention will be further described below through specific embodiments. The illustrative embodiments and descriptions herein are used to explain the present invention, but are not intended to limit the present invention.

[0076] The present invention provides a method for improving the forecasting of convective winds by assimilating densely observed ground data, comprising the following steps:

[0077] like Figure 1-6 As shown, this embodiment includes the following steps:

[0078] Collect observation data from pre-defined, densely packed automatic weather stations, and set up models and experiments for the convection forecasting model;

[0079] In the actual assessment, the observational data consisted of observational and reanalysis data. Hourly observations from a regional densified automatic weather station provided by a certain unit were used for weather condition analysis, grid nudging during model integration, and model result verification. Surface observation elements included hourly precipitation, 2m air temperature, 2m relative humidity, and 10m wind field. A three-dimensional network product was established using observations from four radars in regions A1, A2, A3, and A4 as a reference for convective evolution analysis and model assessment. The atmospheric circulation pattern before precipitation was analyzed using NCEP-FNL global analysis data at 20:00 on May 29, 2020. At the same time, the atmospheric stratification structure before the precipitation event was analyzed using radiosonde observations from a certain station B.

[0080] Based on observational data, precipitation characteristics are extracted and convection evolution analysis is performed to obtain evolution data, and atmospheric circulation patterns are obtained based on the evolution data.

[0081] The surface meteorological element field simulation is performed using surface grid relaxation approximation assimilation and simulation evaluation. The key surface meteorological elements affecting the triggering and development of convection in warm-sector heavy precipitation are identified using surface grid relaxation approximation technology, and the influencing mechanisms are obtained. The key meteorological elements are water vapor, temperature, and wind field. The influencing mechanisms include the influence of surface water vapor, surface temperature, and surface wind field.

[0082] The relaxation approximation time of the surface grid in the numerical forecast of heavy precipitation in the warm sector was determined by experiments. Based on the influence mechanism analysis, the bias of the convection forecast model was optimized, and the forecast results were output.

[0083] In this embodiment, the method for setting up the model and experiment of the convection forecasting model includes:

[0084] A mesoscale numerical model was adopted, with a high-resolution nested mesh set in three layers. Initialization and boundary conditions were set, and the model was configured in conjunction with physical processes. The physical processes included the Thompson cloud microphysical parameterization scheme, the multi-scale Cairns cumulus convection parameterization scheme, the RRTM longwave radiation scheme, the Duhay shortwave radiation scheme, the turbulent kinetic boundary layer parameterization scheme, the Moning-Olhoff near-surface scheme, and the Noah-MP land surface model scheme.

[0085] Multiple sets of comparative experiments were designed to observe that relaxation approximation and high-grid relaxation approximation had a weak impact on the results, while ground grid relaxation approximation was the key to determining convection triggering and precipitation simulation.

[0086] Based on the ground grid relaxation approximation, one set of control experiments and seven sets of sensitivity experiments were set up. The control experiments assimilated all ground elements, while the sensitivity experiments turned off the assimilation of some elements. Through comparative studies, the independent and combined effects of water vapor, temperature, and wind field were clarified. Based on the evaluation of precipitation and convection simulation analysis, the influence mechanism of each element was analyzed, the assimilation time was optimized, and the improved convection forecast model was output.

[0087] In the actual evaluation, the WRF-ARWv4.1.1 model was used, with a three-layer nesting structure and horizontal resolutions of 12, 4, and 1.333 km. The number of grid points from the outside to the inside were 313×202, 571×334, and 703×448, respectively. The outermost center of the model was at 22.5°N and 112°E. The vertical grid of the model contained 50 η layers, with the top of each layer at 50 hPa. The initial boundary values ​​of the model were the NCEP-FNL analysis field with a 6-hour interval and a horizontal resolution of 0.25°. The integration time was from 20:00 on May 29, 2020 to 14:00 on May 30, 2020.

[0088] Previous comparative experiments showed that turning on and off the observation relaxation approximation had little impact on the results and failed to reasonably reproduce convection and precipitation. Similarly, the impact of turning on and off only the high-grid relaxation approximation was small and could be ignored. The ground grid relaxation approximation directly determined the development of convection and the formation of subsequent precipitation. To assess the effectiveness of the ground grid relaxation approximation in simulating warm-sector heavy rainfall in South China, a total of 8 sets of experiments were designed. The relaxation approximation of all ground meteorological elements served as the control experiment (EXP1), while EXP2-EXP8 were sensitivity experiments of different ground elements. Based on the 10m wind, 2m temperature, and humidity of the relaxation approximation ground elements, the parameter configurations of the 8 sets of sensitivity experiments are shown in Table 1.

[0089] Table 1. Ground grid point nudging test setup

[0090]

[0091] The results of this simulation with and without observational relaxation approximation were very similar. All experiments were conducted with observational relaxation approximation disabled, focusing only on the sensitivity of ground features in the ground grid relaxation approximation. The ground grid analysis field used for the grid relaxation approximation was output by the Obsgrid program using model features and observational data. Given that the quality of the ground grid analysis field directly affects the relaxation approximation results, the ground grid analysis field was validated. Based on the comparison between the initial model observations and the ground grid analysis field, the ground grid analysis field is consistent with the observations, and its performance is equivalent to regional high-resolution reanalysis data.

[0092] In this embodiment, the method for obtaining evolution data by extracting precipitation characteristics and analyzing convection evolution based on observational data includes:

[0093] The process involves acquiring observational data and reanalyzing information, calculating the cumulative precipitation at each station over 18 hours, interpolating to generate a spatial distribution map to identify the location of the rain belt, obtaining the maximum cumulative rainfall and the coverage area of ​​the rain belt, and obtaining the spatial distribution of cumulative observed precipitation based on the maximum cumulative rainfall and the coverage area of ​​the rain belt. The observational data includes automatic weather station data, radar data, and radiosonde data.

[0094] Extract the hourly precipitation rate from the station with the highest precipitation, plot the time series, obtain the period of heavy precipitation and the peak time, and obtain the temporal evolution of precipitation rate based on the period of heavy precipitation and the peak time; where heavy precipitation is defined as precipitation of ≥20 mm / hour and lasting for 5 hours.

[0095] When the radar reflectivity first exceeds 35 dBZ, it is determined to be convection triggering. The radar combined reflectivity, equivalent ground potential temperature, and wind field are tracked hourly, and the spatiotemporal evolution of the convective system is analyzed in conjunction with the wind field to determine the convective movement path. The key processes of the spatiotemporal evolution of the convective system include the triggering stage, the organization stage, and the dissipation stage. The triggering stage is characterized by the thermal gradient in the wet tongue region promoting uplift. The organization stage is characterized by the formation of loose, banded convection along the coastline. The dissipation stage is characterized by the system moving eastward into the sea and the reflectivity decreasing.

[0096] Environmental pattern diagnosis is conducted based on large-scale background and local influences. Convection evolution analysis is performed through convection triggering determination, spatiotemporal evolution of convection systems, and environmental diagnosis. Circulation diagnosis is based on large-scale background and local influences. Thermodynamic conditions require high convective effective energy potential and low convection inhibition energy. Dynamic conditions require convection to be in a state of flux rather than organized.

[0097] In the actual assessment, the spatial distribution of cumulative precipitation observed over 18 hours from 20:00 on May 29, 2020 to 14:00 on May 30, 2020 was obtained. The precipitation belt was mainly concentrated in the coastal areas of AI Province, with a southwest-northeast distribution. The maximum cumulative hourly precipitation observed by the automatic observation station in K1 District of K City was 264.9 mm and 74.4 mm, respectively. Although the warm-sector precipitation was not strong, the operational numerical forecast completely missed the precipitation. The temporal evolution of precipitation rate at the stations with the maximum cumulative precipitation and the maximum hourly precipitation was observed. The precipitation increased rapidly from 04:00, reached its peak at 07:00, lasted for 1 hour, and then rapidly weakened. After 2 hours, the precipitation rate dropped to below 10 mmh-1. The hourly precipitation exceeding 20 mm lasted for 5 hours during the entire process, from 04:00 to 09:00, accounting for about 83% of the 18-hour cumulative precipitation.

[0098] The spatiotemporal evolution of the convective system from 21:00 on May 29th to 10:00 on May 30th was obtained. Based on the definition that a radar reflectivity greater than 35 dBZ is the trigger for convection, at 02:00 on May 30th, multiple scattered convective cells formed along the western coast of M1 province, with a maximum radar echo value exceeding 45 dBZ. The occurrence and development of convection correspond to the surface wet tongue; the warm and moist airflow from the southern ocean has a potential temperature 3 K higher than the atmosphere over land, and the temperature gradient between the two is conducive to the rise of the warm air mass, thus triggering convection. The convective system moved generally northeastward, forming a banded distribution along the coastline at 07:00, but the organization of the convective system was relatively loose. At 10:00, the convective system gradually weakened and moved eastward into the ocean. Corresponding to the development of the convective system, as the linear convective system gradually formed, precipitation gradually increased, reaching its peak at 07:00, during which time the southerly winds were relatively weak.

[0099] In this embodiment, the method for obtaining atmospheric circulation patterns based on the evolution data includes:

[0100] The large-scale circulation background was analyzed using a global analysis dataset to obtain geopotential height, wind field, equivalent potential temperature and precipitable water. Isobaric surface superposition analysis was carried out by combining radiosonde data and auxiliary observation data, including isobaric surfaces of 500 hPa, 850 hPa and 950 hPa.

[0101] The form analysis is performed at 500 hPa based on the location of the trough and ridge and the subtropical high. The low-level jet stream and water vapor transport are obtained at 850 hPa based on the wind field and water vapor flux. The boundary layer characteristics are obtained at 950 hPa based on the warm and humid air flow and the land-sea temperature difference. The dynamic-thermal coupling zone is identified by superimposing the 500 hPa trough line, the 850 hPa jet stream and the 950 hPa equivalent potential temperature field.

[0102] Based on unstable energy, uplift conditions, and wind shear, sounding stratification analysis was conducted. By analyzing the configuration of the trough, jet stream, and warm and humid tongue, the synergistic effects of dynamic uplift, water vapor transport, and energy supply were obtained, and water vapor parameters and energy parameters were calculated. The water vapor parameters include precipitable water, specific humidity, and water vapor flux, while the energy parameters include convective effective potential energy, convective inhibition energy, uplift condensation height, and free convection height.

[0103] The system correlates convection triggering conditions and outputs atmospheric circulation patterns. Convection triggering conditions include thermodynamic, dynamic, and water vapor conditions. Thermodynamic conditions include high convective effective potential energy, low convective suppression potential energy, low lifting condensation height, or low free convection height. Dynamic conditions include lifting provided by low-level jet streams and no significant suppression by weak troughs. Water vapor conditions include warm and humid transport through advection and turbulent mixing of high equivalent potential temperature air masses.

[0104] In the actual assessment, the atmospheric circulation pattern at 20:00 on May 29, 2020, indicated that the direct impact of 500 hPa on the coastal areas of M1 was relatively small; a large amount of water vapor transport from a certain sea caused the precipitable amount in the convection triggering area to exceed 62 mm; at 850 hPa, the wind speed north of the convection triggering area exceeded 10 m / s. -1 At 950 hPa, the southern part of the convection triggering area is characterized by warm and humid southerly airflow with a wind speed of approximately 8 m / s. -1 A radiosonde reading from Station B at 20:00 on May 29, 2020, indicated that the atmospheric convective potential energy (CAPE) reached 1306 J / kg. -1 The convection suppression energy (CIN) is only 2 J / kg. -1 The condensation level (LCL) is 932 hPa, the free convection level (LFC) is 923 hPa, and the atmospheric precipitation is about 61 mm. Low CIN, LCL and LFC, and high CAPE are conducive to the formation of convection. The horizontal wind changes from a low-level southwesterly wind to a westerly wind as the altitude increases.

[0105] In this embodiment, the method for performing ground grid relaxation approximation assimilation and simulation evaluation of ground meteorological element field simulation includes:

[0106] Acquire and reanalyze observation data from ground-based encrypted automatic observation stations;

[0107] Data from ground-based automatic weather stations and reanalysis data were collected. Control and sensitivity experiments were designed, and relaxation approximation parameters were set to perform ground-based grid-point relaxation approximation assimilation. The relaxation approximation parameters included an assimilation layer that only applied to the ground, and exponentially decaying time and spatial weights.

[0108] Surface meteorological elements are assessed based on water vapor mixing ratio, temperature, and wind speed; precipitation simulation assessment is obtained by comparing the spatial distribution of 18-hour cumulative precipitation and statistically analyzing the proportion of stations with precipitation at different thresholds; and convective systems are assessed based on triggering time, spatial structure, and organization morphology.

[0109] The interaction mechanisms of water vapor, temperature, and wind field are analyzed, and the assimilation parameters are optimized based on these mechanisms.

[0110] In practical assessments, the temporal evolution of regional average 2m water vapor mixing ratio, 2m temperature, and 10m wind speed was obtained from observations, EXP1, and the closed relaxation approximation experiment. Regarding the 2m water vapor mixing ratio, the simulated early-stage value of the closed Nudging experiment was lower than the observed value, while the EXP1 data was closer to the observation. At night, the 2m temperature in the closed relaxation approximation experiment was lower and cooled significantly faster than the observation; in contrast, EXP1 showed a more consistent temperature change with the observation. The closed relaxation approximation experiment overestimated the 10m wind speed, while EXP1 was closer to the observation. The closed relaxation approximation experiment showed a significant deviation from the observation, while the ground grid relaxation approximation method significantly improved the accuracy of the ground meteorological field simulation.

[0111] The spatial distribution of 18-hour accumulated precipitation was obtained from eight sets of experiments. EXP2, EXP5, EXP6, and EXP7 failed to simulate the coastal rainband, with maximum 18-hour accumulated precipitation along the coast of 61.5, 37.2, 84.9, and 60.0 mm, respectively. EXP4 and EXP8 successfully simulated the coastal rainband, but the precipitation intensity and location deviated significantly from the observations. Comparing the experiments, EXP1 and EXP3 reproduced the spatial distribution of heavy precipitation better, with maximum values ​​of 233.6 mm and 281.1 mm, respectively. Close to the maximum observed value of 264.9 mm, during the simulation period, the percentages of stations with 18-hour cumulative precipitation exceeding 10 mm, 50 mm, and 100 mm were 45.98%, 11.36%, and 4.16%, respectively. The corresponding grid percentages in EXP1 were 48.14%, 9.32%, and 2.90%, while in EXP3, these percentages were 53.11%, 18.54%, and 8.28%. In summary, EXP1 most closely approximates the observed intensity and spatial distribution of the simulated heavy precipitation belt along the coast of M1 province.

[0112] Compared to radar observations, EXP1 reproduces the overall characteristics of convection triggering and development better. The location of convection triggering is close to the western coast of M1 province, but delayed by about 2 hours; the southwest-northeast trending coastal linear convection is also close to radar observations; the time delay of convection triggering in EXP1 may be related to the model's start-up time; initial conditions and physical schemes both affect the triggering time of the simulated convection system. Although there are temporal differences, EXP1 is closest to radar observations in terms of the spatial location, horizontal scale, and structure of convection triggering and development.

[0113] Compared with radar observations, EXP1 reproduces the overall characteristics of convection triggering and development. The location of convection triggering is close to the western coast of M1 province, but delayed by about 2 hours. The southwest-northeast trending coastal linear convection is also close to radar observations. The reason for the delay in convection triggering time in EXP1 may be related to the start-up time of the model. Initial conditions and physical schemes will affect the triggering time of the simulated convection system. Although there are time differences, EXP1 is closest to radar observations in terms of spatial location, horizontal scale and structure of convection triggering and development.

[0114] In this embodiment, the method for influencing surface water vapor includes:

[0115] Five comparative experiments were designed, using the relative humidity, water vapor mixing ratio, convective available potential energy, and convective suppression energy output by the model as data sources.

[0116] Rapid humidification of the boundary layer, given the budget equation, is expressed as follows:

[0117]

[0118] Where q v This represents the average water vapor mixing ratio. This represents the local variation of the average water vapor mixing ratio. For horizontal transport items, The vertical transport term is represented by QBL, which represents the effect of boundary layer turbulent mixing. QCM represents the condensation and evaporation effects generated by cumulus convection and microphysical processes. QNU represents the relaxation approximation tendency of the average water vapor mixing ratio. R is the residual term, and the calculation formula is as follows: QBL, QCM, and QNU use mode output values;

[0119] The local variation of the mean water vapor mixing ratio was calculated using the central difference method:

[0120]

[0121] Where Δt is 6 min, and the average water vapor mixing ratio at time t+Δt is q. v │ t+Δt The average water vapor mixing ratio at time t-Δt is q v │ t-Δt The local variation term of the average water vapor mixing ratio at time t is:

[0122] The boundary layer humidification process was verified based on the evolution of relative humidity and the increase of water vapor mixing ratio. The water vapor budget was quantified, the convection inhibition energy was increased and decreased, the condensation height was reduced, the vertical velocity disturbance was enhanced, the convection was triggered and organized, the effects of precipitation and convection were assessed, and the results of the surface water vapor impact were obtained.

[0123] In the actual evaluation, the combined radar reflectivity, surface relative humidity, and spatiotemporal evolution of wind field were obtained for EXP1, EXP3, EXP4, EXP5, and EXP8. EXP1, EXP3, EXP4, and EXP8 all relaxed the approximation of surface water vapor. The relative humidity of all four sets of experiments increased significantly, triggering convection in the coastal areas of M1 province. The convective systems of all four sets of experiments developed northeastward along the coast. Due to differences in other relaxation approximation settings, there were significant differences in the intensity, spatial location, and organization of the convective systems. EXP5 closed the surface water vapor relaxation approximation, and the humidification process in the coastal areas was slow, failing to simulate the occurrence and development of convection. Therefore, relaxing the approximation of surface water vapor can effectively improve water vapor conditions and directly determine the triggering of convection.

[0124] The temporal evolution of the mean relative humidity profiles in the convection-triggered regions (21.6°N–22.2°N, 111.5°E–112.5°E) of EXP1 and EXP5 was obtained. Initially, both sets of experiments had the same relative humidity profiles. At 21:30, significant differences in relative humidity began to appear below 850 hPa, with EXP1 reaching as high as 96% at 975 hPa, while EXP5 was only 81%. At convection triggering, EXP1 had a relative humidity exceeding 98% at 950 hPa, while EXP5 had only about 80% at that altitude. The boundary layer water vapor content of EXP1 rapidly increased to near saturation within three hours.

[0125] The temporal evolution of the mean water vapor mixing ratio in the convection-triggered regions of EXP1 and EXP5 at an altitude of 950 hPa shows that the mean water vapor mixing ratio in EXP1 increased rapidly in the early stage, followed by a slowing trend, with the mean water vapor mixing ratio decreasing from 16.62 g kg. -1 Increased to 21.52 g kg at convection triggering. -1 The average water vapor mixing ratio in EXP5 increases slowly, eventually reaching 17.58 g / kg. -1 To further quantify and diagnose the boundary layer humidification process in the convection-triggered region, the single-layer water vapor budget at 950 hPa was calculated. The average water vapor mixing ratio budget equation is as follows:

[0126]

[0127] Where q v This represents the average water vapor mixing ratio. This represents the local variation of the average water vapor mixing ratio. For horizontal and vertical transport items, The horizontal and vertical transport terms are represented by QBL, the effect of boundary layer turbulent mixing, QCM, the condensation and evaporation effects from cumulus convection and microphysical processes, QNU, the relaxation approximation tendency of the average water vapor mixing ratio, and R, the residual term. The calculation formula is as follows: The local variation of the mean water vapor mixing ratio was calculated using the central difference method:

[0128]

[0129] Δt is 6 min; QBL, QCM, and QNU use model output values. The horizontal transport term is obtained by subtracting the vertical transport term from the total water vapor transport term output by the model; the mean time evolution of all terms at 950 hPa in the convection triggering region for both sets of experiments; in the early stage of EXP1 startup, the relaxation approximation tendency of the mean water vapor mixing ratio leads to a sharp increase in boundary layer water vapor, and the effect of relaxation approximation rapidly decays after the model value approaches the observed value; in contrast, the horizontal transport and boundary layer turbulent mixing terms have a smaller impact, while cumulus convection and microphysical processes enhance condensation before convection triggering; from 21:06 to 23:00, the vertical transport term becomes the main source of boundary layer humidification; comparative analysis shows that due to the closure of the relaxation approximation water vapor in EXP5, vertical transport becomes the main source of boundary layer humidification, and its humidification effect is largely offset by the negative horizontal transport; the 950 hPa water vapor budget diagnosis in the northwest region of a river estuary shows that, in the absence of vertical water vapor transport, the boundary layer humidification process dominated by the relaxation approximation tendency cannot trigger convection;

[0130] By relaxing the adjustment of water vapor approaching the surface, the boundary layer rapidly humidifies, thereby enhancing lower atmospheric instability within the convection-triggered region; from 20:00 to 23:00 on May 29, the mean maximum convective available potential energy (MCAPE) in the convection-triggered region increased from 1130.1 J kg. -1 Increased to 4227.1 J kg -1 In EXP5, however, MCAPE only increased to 1487.0 J kg. -1 Furthermore, the maximum convection suppression energy (MCIN) of EXP1 is 17.9 J kg. -1 Reduced to 0.8J kg -1 This is close to the actual CIN calculated by the B-sounding at 20:00; the Lifted Condensation Level (LCL) and Free Convection Level (LFC) decreased from 706m and 1457m to 387m and 453m respectively during this period, which is consistent with the observations; vertical profiles of the over-convection-triggered areas of EXP1 and EXP5 were obtained, with a larger CAPE below 1km in EXP1, and the high CAPE extending from the ocean to the inland and continuously increasing from 21:00 to 23:00; at 21:00, the CIN at a height of 1km at 21.75°N exceeded 60J kg. -1Vertical motion is relatively weak; the CIN weakens significantly from 22:00 to 23:00, and a strong vertical velocity disturbance below 2 km overcomes the CIN. A strong CAPE helps the air mass to further lift, accelerating convection triggering. In contrast, EXP5 exhibits a large CIN (greater than 100 J kg) at altitudes of 0.3–1.2 km. -1 The lower-level vertical motion is weak, which is not conducive to the generation of convection. In addition, below 1 km, the CAPE of EXP5 is much smaller than that of EXP1, indicating a weaker potential for convection development. Therefore, the improvement of water vapor conditions and changes in CIN and lifting conditions are conducive to triggering convection, while the increase in CAPE provides the energy conditions for the rapid occurrence of subsequent convection.

[0131] In this embodiment, the method for influencing ground temperature includes:

[0132] Four sets of comparative experiments were set up. Temperature relaxation approximation was used to slow down the cooling based on the cooling rate and the holding temperature to maintain the potential temperature perturbation. Vertical profile and time series were used to diagnose the potential temperature perturbation.

[0133] Calculate thermal buoyancy:

[0134]

[0135] Where g is the acceleration due to gravity, in milliseconds (ms). -2 , Potential temperature perturbation, in K. The region's average potential temperature, in Kelvin (q). v The water vapor mixing ratio, q represents the average water vapor mixing ratio. c q represents the cloud water content. r Rainwater content;

[0136] To maximize thermal buoyancy, enhance thermal buoyancy and increase vertical velocity, morphological comparisons are conducted after early convection triggering, and an organized evaluation is performed. The cross-effect verification is carried out based on the synergistic contribution of thermal buoyancy growth in the thermal buoyancy components and the contribution of single elements. Temperature and humidity assimilation synergy is achieved, and the temperature and water vapor synergistic effect is obtained based on the maximization of thermal buoyancy contribution.

[0137] In the actual assessment, based on the combined reflectivity of radars EXP1, EXP2, EXP4, and EXP7, the equivalent surface potential temperature, and the spatiotemporal evolution of the wind field, EXP2 showed almost no radar echoes at night, with only a few weak radar echoes appearing on the ocean surface southeast of a river estuary at 08:00 on the 30th; EXP7 only activated the surface temperature relaxation approximation, and only sporadic weak radar echoes appeared near the coast at 08:00 on the 30th; it can be seen that under unfavorable water vapor conditions, the surface temperature relaxation approximation alone is insufficient to trigger convection, which is different from the isolated case triggered by the urban heat island on May 7, 2017; compared with EXP2, the relaxation approximation setting of EXP7 did not improve the simulation of nighttime coastal convection; at 04:00 on the 30th, EXP4 triggered convection in the coastal area of ​​M1 province, compared with E XP1 was delayed by 5 hours. A convective cell with a diameter greater than 10 km formed near (21.8°N, 112.1°E), while three adjacent convective cells with diameters less than 10 km formed near (21.9°N, 113.1°E), more than 130 km apart. Compared to the control experiment (EXP1), EXP4 shut down surface temperature nudging. After convection triggering, unlike the 300 km long linear convection formed in EXP1, the convective cell in EXP5 ultimately formed two convective systems only 100 km long. Therefore, both experiments show that nudging surface temperature can accelerate nighttime convection triggering and better promote the organization and development of convective systems, thus producing stronger precipitation.

[0138] The spatiotemporal evolution of surface temperature and wind field in EXP1 and EXP4 was obtained. At the initial moment of the model, the initial temperature fields of EXP1 and EXP4 were the same. After the model started, the coastal surface temperature of EXP4 decreased rapidly, and after 3 hours of model integration, the surface temperature on the land side dropped below 27°C. In contrast, the surface temperature in EXP1 was well maintained, and the cooling rate was slow. In summary, relaxing the near-surface temperature can slow down the near-surface cooling at night, which is conducive to the occurrence and development of convection.

[0139] The thermal buoyancy B before convection is triggered is calculated using the following formula:

[0140]

[0141] Where g is the acceleration due to gravity, in milliseconds (ms). -2 , Potential temperature perturbation, in K. The region's average potential temperature, in Kelvin (q). v The water vapor mixing ratio, q represents the average water vapor mixing ratio. c q represents the cloud water content. r Rainwater content;

[0142] Vertical profiles of thermal buoyancy were obtained for EXP1 and EXP4. After one hour of mode operation, the thermal buoyancy of EXP1 decreased from the ocean towards the inland areas, with the maximum thermal buoyancy occurring at an altitude of approximately 1–2 km above the ocean, reaching 0.06 ms. -2 As time progresses, the thermal buoyancy of EXP1 gradually increases and extends inland, reaching a maximum of 0.08 ms when convection is triggered at 23:00. -2 Furthermore, at 21:00, EXP1 exhibited a significant potential temperature disturbance with an intensity of 0.1 ms over a location between 21.51°N and 21.67°N. -1 Multiple vertical velocity disturbance centers were observed; at 22:00, the vertical velocity disturbances intensified and expanded in area; at 23:00, the updraft moved northward, with the latter's velocity exceeding 0.6 ms. -1 In contrast, EXP4 exhibits relatively small changes in potential temperature perturbation, corresponding to smaller thermal buoyancy and upward motion.

[0143] The evolution of average parameters over time in the convection-triggered region of the boundary layer (below 1 km) in EXP1, EXP4, and EXP5 was compared. The thermal buoyancy in EXP1 was 0.0087 ms at 20:00. -2 Increased to 0.036ms by 23:00 -2 The main contributor is potential temperature perturbation; in addition, the water vapor component increases slowly, and the cloud and rain component has little impact. In contrast, the thermal buoyancy effect at 23:00 in EXP4 and EXP5 is relatively small, increasing to 0.021 and 0.025 ms respectively. -2 Although both EXP1 and EXP5 relax to approach the ground temperature, before 23:00, the potential temperature perturbation term of EXP5 was lower than that of EXP1, increasing only to 0.021ms. -2 Comparing EXP1 and EXP4 shows that when only the water vapor term is relaxed, the water vapor effect is weaker than when both the temperature and water vapor terms are relaxed simultaneously. In summary, relaxing only the temperature term leads to a rapid increase in potential temperature perturbation, and relaxing only the water vapor term leads to a rapid increase in the water vapor effect. Both contribute less than their respective contributions when both temperature and water vapor are relaxed simultaneously, which means that the interaction between temperature and water vapor promotes their contributions to some extent.

[0144] In this embodiment, the method for influencing the surface wind field includes:

[0145] Four sets of comparative experiments were designed. The model wind speed was reduced by relaxation approximation. The corrected wind speed was obtained by data verification based on the assimilation wind speed error. The wind direction convergence was adjusted for the comparative experiments. The four sets of comparative experiments included one reference experiment and three comparative experiments.

[0146] To obtain the convective monomers and organizational morphology of the comparative experiment, the low-level warm and humid airflow convergence was reduced by relaxation approximation, thus limiting energy supply;

[0147] The necessity of triggering convection was verified by comparing wind field assimilation alone. The synergistic effect of wind field assimilation and temperature and humidity assimilation was obtained by combining wind field assimilation with temperature and humidity assimilation in the reference experiment.

[0148] In actual assessments, a comparison between EXP1 and EXP3 shows that relaxing the near-surface wind field can weaken the intensity of the convective system. During convection triggering, convective cells in the triggering region exhibit a wavy arrangement, with greater quantity and intensity than in EXP1. Wind speeds over the ocean are stronger, and the convergence of southwest and southeast winds is more pronounced. In the subsequent development of convection, due to the adjustment of the relaxed near-surface wind field, the southern ocean surface of the convective system in EXP1 is dominated by southwest winds with lower speeds and weaker wind convergence. In EXP4, the near-surface winds on the southwest and southeast sides of the convective system... The stratospheric winds consist of southwesterly winds and stronger southeasterly winds, with higher wind speeds and stronger wind direction convergence. In addition, the comparison between EXP2 and EXP6 shows that EXP2 overestimates the surface wind field, while EXP6 reduces surface wind speeds by numerating the surface wind field, but fails to trigger convection at night due to insufficient moisture. Overall, numerating the surface wind field, by adjusting the wind speed and direction of the airflow on the south side of the convective system, makes the simulation results closer to the observations, reduces the convergence of warm and humid airflows to some extent, and thus suppresses the problem of excessively strong convective systems.

[0149] In this embodiment, the method for determining the surface grid relaxation approximation time of numerical forecasts for heavy precipitation in warm regions through experiments includes:

[0150] Design a benchmark experiment with fixed assimilation parameters and only adjust the assimilation duration;

[0151] The hit rate exceeding the rainfall threshold is used as the TS score. The precipitation score is calculated based on the root mean square error of the spatial distribution of cumulative rainfall and the TS score. The convection trigger time deviation is obtained by simulating the first trigger time of convection and the time difference between observation. The boundary layer variable improvement rate is obtained by comparing the observed temperature, humidity, wind speed and the simulation correlation coefficient. The precipitation score, convection trigger time deviation, and boundary layer variable improvement rate are used as evaluation indicators.

[0152] Calculate the time decay rate of the relaxation approximation tendency term in the benchmark test and determine the critical time point at which the relaxation approximation influence will reach 10%.

[0153] By comparing the response speed of the physical processes of the model after assimilation, the autonomous recovery of the convective forecasting model is assessed based on whether the evolution of convective available potential energy and vertical velocity transitions naturally.

[0154] The duration that simultaneously satisfies the following conditions is the high TS score with the lowest root mean square error, the smallest convection trigger time deviation, and the smooth transition of the boundary layer, and the output is the relaxation approximation duration of the ground grid.

[0155] In this embodiment, the method for bias optimization of the convection forecast model based on the influence mechanism analysis includes:

[0156] Quality control is performed on ground-based densified observations, data larger than 3 standard deviations are removed, and when interpolating the model grid, the observation error covariance matrix is ​​retained. The model short-time forecast is used as the background field, and the background field error covariance is calculated.

[0157] Given an objective function, the expression is:

[0158]

[0159] Where J is the objective function, x is the model variable, and x is the model variable. obs Let x be the observed value, Q be the observation error covariance matrix, and x be the observed value. bg The background field is L1, and the dynamic constraint operator is L1. For spatial gradient operators, Here, B is the model's physical process tendency term, D is the background error covariance, and W is the model tendency error covariance. λ2 is the observation fitting term, λ3 is the smoothing strength parameter, λ1 is the tendency constraint weight, and λ1 is the regularization parameter, which controls the dynamic coordination weight.

[0160] Computation of dynamic constraint operators:

[0161]

[0162] Where ζ is vorticity, and θ e For equivalent potential temperature, This represents the vertical gradient of vorticity. For wind field divergence, The sensitivity of temperature relative to humidity is equivalent. This represents the sensitivity of potential temperature to temperature.

[0163] We introduce water vapor assimilation optimization parameters to constrain abrupt changes in the humidity field and maintain the Reifelli balance between the humidity and temperature fields; we introduce temperature assimilation optimization parameters to adjust the temperature field and maintain boundary layer stability; we introduce wind field assimilation optimization parameters to limit the wind field adjustment range and maintain the natural evolution of low-level convergence.

[0164] When the decrease in the continuous objective function value is less than 0.01, the iteration stops and the optimized convection forecast model is output.

[0165] Figure 2The temporal evolution of the average MCAPE and MCIN of the convection triggering regions of EXP1 and EXP5 from 20:00 to 23:00 on May 29, 2020;

[0166] Figure 3 The average convection triggering regions of EXP1 and EXP5 from 20:00 to 23:00 on May 29, 2020 are represented by a, where a represents the temporal evolution of MCAPE and MCIN, and b represents the temporal evolution of LCL and LFC.

[0167] Figure 4 Figure a shows the height-latitude profiles of thermal buoyancy (filled in), equivalent potential temperature (black contour lines), vertical velocity (black curves are contour lines), and composite wind field (arrows, vertical velocity magnified 10 times) along the line from 21:00 to 23:00 on May 29, 2020; (a–c) and (d–f) are the simulation results of EXP1 and EXP4, respectively.

[0168] Figure 5 The evolution of the average thermal buoyancy and its influence terms in the convection-triggered area below 1 km from 20:00 to 23:00 on May 29, 2020: a is EXP1, b is EXP4, c is EXP5; the circle represents thermal buoyancy, the triangle represents potential temperature disturbance, the square represents water vapor effect, and the pentagram represents cloud and rainwater effect.

[0169] Figure 6 The data represents the regional average (21.5°N~23.5°N, 111.0°E~115.5°E) from 20:00 on May 29 to 14:00 on May 30, 2020. a represents the 2m water vapor mixing ratio, b represents the 2m temperature, and c represents the 10m wind speed over time. The pentagrams represent observations, the squares represent pure WRF simulations, and the triangles represent full-field nudging (EXP1).

[0170] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for improving the forecasting of convective winds by assimilating densely observed ground data, characterized in that, Includes the following steps: Collect observation data from pre-defined, densely packed automatic weather stations, and set up models and experiments for the convection forecasting model; Based on observational data, precipitation characteristics are extracted and convection evolution analysis is performed to obtain evolution data, and atmospheric circulation patterns are obtained based on the evolution data. The surface meteorological element field simulation is performed using surface grid relaxation approximation assimilation and simulation evaluation. The key surface meteorological elements affecting the triggering and development of convection in warm-sector heavy precipitation are identified using surface grid relaxation approximation technology, and the influencing mechanisms are obtained. The key meteorological elements are water vapor, temperature, and wind field. The influencing mechanisms include the influence of surface water vapor, surface temperature, and surface wind field. The relaxation approximation time of the surface grid in the numerical forecast of heavy precipitation in the warm sector was determined by experiments. Based on the influence mechanism analysis, the bias of the convection forecast model was optimized, and the forecast results were output. The method for setting up the model and experiment of the convection forecasting model includes: A mesoscale numerical model was adopted, with a high-resolution nested mesh set in three layers. Initialization and boundary conditions were set, and the model was configured in conjunction with physical processes. The physical processes included the Thompson cloud microphysical parameterization scheme, the multi-scale Cairns cumulus convection parameterization scheme, the RRTM longwave radiation scheme, the Duhay shortwave radiation scheme, the turbulent kinetic boundary layer parameterization scheme, the Moning-Olhoff near-surface scheme, and the Noah-MP land surface model scheme. Multiple sets of comparative experiments were designed to observe that relaxation approximation and high-grid relaxation approximation had a weak impact on the results, while ground grid relaxation approximation was the key to determining convection triggering and precipitation simulation. Based on the ground grid relaxation approximation, one set of control experiments and seven sets of sensitivity experiments were set up. The control experiments assimilated all ground elements, while the sensitivity experiments turned off the assimilation of some elements. Through comparative studies, the independent and combined effects of water vapor, temperature, and wind field were clarified. Based on the evaluation of precipitation and convection simulation analysis, the influence mechanism of each element was analyzed, the assimilation time was optimized, and the improved convection forecast model was output. Methods for ground-grid relaxation approximation assimilation and simulation evaluation of ground meteorological element fields include: Acquire and reanalyze observation data from ground-based encrypted automatic observation stations; Data from ground-based automatic weather stations and reanalysis data were collected. Control and sensitivity experiments were designed, and relaxation approximation parameters were set to perform ground-based grid-point relaxation approximation assimilation. The relaxation approximation parameters included an assimilation layer that only applied to the ground, and exponentially decaying time and spatial weights. Surface meteorological elements are assessed based on water vapor mixing ratio, temperature, and wind speed; precipitation simulation assessment is obtained by comparing the spatial distribution of 18-hour cumulative precipitation and statistically analyzing the proportion of stations with precipitation at different thresholds; and convective systems are assessed based on triggering time, spatial structure, and organization morphology. The interaction mechanisms of water vapor, temperature, and wind field are analyzed, and the assimilation parameters are optimized based on these mechanisms.

2. The method for improving convective wind forecasting by assimilating densely observed ground data according to claim 1, characterized in that, The method for obtaining evolution data by extracting precipitation characteristics and analyzing convection evolution based on observational data includes: The process involves acquiring observational data and reanalyzing information, calculating the cumulative precipitation at each station over 18 hours, interpolating to generate a spatial distribution map to identify the location of the rain belt, obtaining the maximum cumulative rainfall and the coverage area of ​​the rain belt, and obtaining the spatial distribution of cumulative observed precipitation based on the maximum cumulative rainfall and the coverage area of ​​the rain belt. The observational data includes automatic weather station data, radar data, and radiosonde data. Extract the hourly precipitation rate from the station with the highest precipitation, plot the time series, obtain the period of heavy precipitation and the peak time, and obtain the temporal evolution of precipitation rate based on the period of heavy precipitation and the peak time; where heavy precipitation is defined as precipitation of ≥20 mm / hour and lasting for 5 hours. When the radar reflectivity first exceeds 35 dBZ, it is determined to be convection triggering. The radar combined reflectivity, equivalent ground potential temperature, and wind field are tracked hourly, and the spatiotemporal evolution of the convective system is analyzed in conjunction with the wind field to determine the convective movement path. The key processes of the spatiotemporal evolution of the convective system include the triggering stage, the organization stage, and the dissipation stage. The triggering stage is characterized by the thermal gradient in the wet tongue region promoting uplift. The organization stage is characterized by the formation of loose, banded convection along the coastline. The dissipation stage is characterized by the system moving eastward into the sea and the reflectivity decreasing. Environmental form diagnosis is conducted based on large-scale background and local influences. Convection evolution analysis is performed through convection triggering determination, spatiotemporal evolution of the convection system, and environmental diagnosis. Circulation diagnosis is conducted based on large-scale background and local influences. Thermodynamic conditions require high convection effective energy potential and low convection inhibition energy. Dynamic conditions require convection to be idiosyncratic rather than organized.

3. The method for improving convective wind forecasting by assimilating densely observed ground data according to claim 1, characterized in that, The method for obtaining atmospheric circulation patterns based on the evolution data includes: The large-scale circulation background was analyzed using a global analysis dataset to obtain geopotential height, wind field, equivalent potential temperature and precipitable water. Isobaric surface superposition analysis was carried out by combining radiosonde data and auxiliary observation data, including isobaric surfaces of 500 hPa, 850 hPa and 950 hPa. The form analysis is performed at 500 hPa based on the location of the trough and ridge and the subtropical high. The low-level jet stream and water vapor transport are obtained at 850 hPa based on the wind field and water vapor flux. The boundary layer characteristics are obtained at 950 hPa based on the warm and humid air flow and the land-sea temperature difference. The dynamic-thermal coupling zone is identified by superimposing the 500 hPa trough line, the 850 hPa jet stream and the 950 hPa equivalent potential temperature field. Based on unstable energy, uplift conditions, and wind shear, sounding stratification analysis was conducted. By analyzing the configuration of the trough, jet stream, and warm and humid tongue, the synergistic effects of dynamic uplift, water vapor transport, and energy supply were obtained, and water vapor parameters and energy parameters were calculated. The water vapor parameters include precipitable water, specific humidity, and water vapor flux, while the energy parameters include convective effective potential energy, convective inhibition energy, uplift condensation height, and free convection height. The correlation of convection triggering conditions outputs the atmospheric circulation pattern. Convection triggering conditions include thermodynamic, dynamic, and water vapor conditions. Thermodynamic conditions are high convective effective potential energy, low convective suppression potential energy, low lifting condensation height, or low free convection height. Dynamic conditions are that the low-level jet provides lifting, and the weak trough has no significant suppression. Water vapor conditions are that warm and humid transport passes through the advection and turbulent mixing of high equivalent potential temperature air masses.

4. The method for improving convective wind forecasting by assimilating densely observed ground data according to claim 1, characterized in that, The method for influencing surface water vapor includes: Five sets of comparative experiments were designed, using the relative humidity, water vapor mixing ratio, convective available potential energy, and convective suppression energy output by the model as data sources. Rapid humidification of the boundary layer, given the budget equation, is expressed as follows: in This represents the average water vapor mixing ratio. ∂qv∂t This represents the local variation of the average water vapor mixing ratio. For horizontal and vertical transport items, For horizontal and vertical transport items, The influence of boundary layer turbulent mixing, The condensation and evaporation effects are caused by cumulus convection and microphysical processes. The relaxation approximation tendency of the average water vapor mixing ratio is given by R, which is the residual term, and is calculated using the following formula: QBL, QCM, and QNU use mode output values; The local variation of the mean water vapor mixing ratio was calculated using the central difference method: in It is 6 minutes, time The average water vapor mixing ratio is ,time The average water vapor mixing ratio is The local variation term of the average water vapor mixing ratio at time t is: ; The boundary layer humidification process was verified based on the evolution of relative humidity and the increase of water vapor mixing ratio. The water vapor budget was quantified, the convection inhibition energy was increased and decreased, the condensation height was reduced, the vertical velocity disturbance was enhanced, the convection was triggered and organized, the effects of precipitation and convection were assessed, and the results of the surface water vapor impact were obtained.

5. The method for improving convective wind forecasting by assimilating densely observed ground data according to claim 1, characterized in that, The method for influencing ground temperature includes: Four sets of comparative experiments were set up. Temperature relaxation approximation was used to slow down the cooling based on the cooling rate and the holding temperature to maintain the potential temperature perturbation. Vertical profile and time series were used to diagnose the potential temperature perturbation. Calculate thermal buoyancy: Where g is the acceleration due to gravity, in milliseconds (ms). −2 , Potential temperature perturbation, in K. The average potential temperature of the region, in K. The water vapor mixing ratio, This represents the average water vapor mixing ratio. Cloud water content, Rainwater content; To maximize thermal buoyancy, enhance thermal buoyancy and increase vertical velocity, morphological comparisons are conducted after early convection triggering, and an organized evaluation is performed. The cross-effect verification is carried out based on the synergistic contribution of thermal buoyancy growth in the thermal buoyancy components and the contribution of single elements. Temperature and humidity assimilation synergy is achieved, and the temperature and water vapor synergistic effect is obtained based on the maximization of thermal buoyancy contribution.

6. The method for improving convective wind forecasting by assimilating densely observed ground data according to claim 1, characterized in that, The method for influencing the surface wind field includes: Four sets of comparative experiments were designed. The model wind speed was reduced by relaxation approximation. The corrected wind speed was obtained by data verification based on the assimilation wind speed error. The wind direction convergence was adjusted for the comparative experiments. The four sets of comparative experiments included one reference experiment and three comparative experiments. To obtain the convective monomers and organizational morphology of the comparative experiment, the low-level warm and humid airflow convergence was reduced by relaxation approximation, thus limiting energy supply; The necessity of triggering convection was verified by comparing wind field assimilation alone. The synergistic effect of wind field assimilation and temperature and humidity assimilation was obtained by combining wind field assimilation with temperature and humidity assimilation in the reference experiment.

7. The method for improving convective wind forecasting by assimilating densely observed ground data according to claim 1, characterized in that, The method for determining the surface grid relaxation approximation time of numerical forecasts for heavy precipitation in warm regions through experiments includes: Design a benchmark experiment with fixed assimilation parameters and only adjust the assimilation duration; The hit rate exceeding the rainfall threshold is used as the TS score. The precipitation score is calculated based on the root mean square error of the spatial distribution of cumulative rainfall and the TS score. The convection trigger time deviation is obtained by simulating the first trigger time of convection and the time difference between observation. The boundary layer variable improvement rate is obtained by comparing the observed temperature, humidity, wind speed and the simulation correlation coefficient. The precipitation score, convection trigger time deviation, and boundary layer variable improvement rate are used as evaluation indicators. Calculate the time decay rate of the relaxation approximation tendency term in the benchmark test and determine the critical time point at which the relaxation approximation influence will reach 10%. By comparing the response speed of the physical processes of the model after assimilation, the autonomous recovery of the convective forecasting model is assessed based on whether the evolution of convective available potential energy and vertical velocity transitions naturally. The duration that simultaneously satisfies the following conditions is the high TS score with the lowest root mean square error, the smallest convection trigger time deviation, and the smooth transition of the boundary layer, and the output is the relaxation approximation duration of the ground grid.

8. The method for improving convective wind forecasting by assimilating densely observed ground data according to claim 1, characterized in that, The method for bias optimization of the convection forecast model based on the aforementioned influence mechanism analysis includes: Quality control is performed on ground-based densified observations, data larger than 3 standard deviations are removed, and when interpolating the model grid, the observation error covariance matrix is ​​retained. The model short-time forecast is used as the background field, and the background field error covariance is calculated. Given an objective function, the expression is: in Let be the objective function. For pattern variables, Let Q be the observed value, and let Q be the observation error covariance matrix. As background scene, For dynamic constraint operators, For spatial gradient operators, B represents the physical process tendency term of the model, B represents the background error covariance, and D represents the weight matrix, which is related to the terrain. For the model bias error covariance, For the observation fitting term, For smoothing intensity parameters, To favor the constraint weights, This is a regularization parameter that controls the dynamic coordination weights. Computation of dynamic constraint operators: in vorticity, For equivalent potential temperature, This represents the vertical gradient of vorticity. For wind field divergence, The sensitivity of relative potential temperature to relative humidity, This represents the sensitivity of potential temperature to temperature. We introduce water vapor assimilation optimization parameters to constrain abrupt changes in the humidity field and maintain the Reifelli balance between the humidity and temperature fields; we introduce temperature assimilation optimization parameters to adjust the temperature field and maintain boundary layer stability; we introduce wind field assimilation optimization parameters to limit the wind field adjustment range and maintain the natural evolution of low-level convergence. When the decrease in the continuous objective function value is less than 0.01, the iteration stops and the optimized convection forecast model is output.

Images