Progressive forecasting method for regional persistent rainstorm based on key influence factors

Abstract

The invention discloses a progressive forecasting method for regional persistent rainstorm based on key impact factors, which comprises the following steps of: firstly, comparing and analyzing difference characteristics of a regional persistent rainstorm process and a general rainstorm process in a traditional physical quantity factor, a physical quantity anomaly factor and a comprehensive dynamic factor, and screening a key impact factor set; then, extracting a circulation field anomaly factor before rainstorm occurrence to construct a quantitative index system, and establishing a three-level linkage criterion in combination with a mid-high weft circulation form, a mid-low weft power lifting threshold value and water vapor conveying strength; and finally, determining an optimal factor combination and weight distribution scheme by taking the TS score as a judgment criterion, constructing a probability forecasting model by fusing a kernel density estimation technology, and outputting a rainstorm falling area spatial distribution probability. According to the method, a traditional single forecasting mode is broken through, a progressive forecasting technology system from rainstorm process forecasting to falling area probability forecasting is innovatively established, and technical support is provided for early warning decision making and disaster prevention and reduction deployment of extreme weather events.

Description

Technical Field

[0001] This invention belongs to the field of weather forecasting technology, and in particular relates to a progressive forecasting method for regional persistent heavy rainfall based on key influencing factors. Background Technology

[0002] Unlike typical precipitation events, persistent heavy rainfall is characterized by its high intensity, wide range, long duration, and stable location. This stability is attributed to factors such as the maintenance of anomalous atmospheric circulation, the continuous convergence of cold and warm air masses, strong upward motion, and persistent unstable conditions. Currently, heavy rainfall forecasting primarily utilizes numerical weather prediction and ensemble forecast products, and has developed techniques based on meteorological diagnostic analysis, objective correction ensembles, probabilistic forecasting, and artificial intelligence. However, research specifically on methods for forecasting persistent heavy rainfall is still relatively limited.

[0003] Although significant progress has been made in the research of theories and methods for heavy rainfall forecasting, some problems still exist. For example, the quality control of observational data and the comprehensive application of multi-source data need to be strengthened; the accurate description of physical processes in models and the coordination performance of dynamic frameworks are still insufficient to reflect the dynamic and thermodynamic processes of actual heavy rainfall system development, making it difficult to achieve refined, precise, and quantitative heavy rainfall forecasts; and there are still many difficulties in understanding the fine structure and development mechanism of mesoscale systems in heavy rainfall. These are all bottlenecks that hinder the rapid improvement of heavy rainfall forecasting capabilities.

[0004] Based on the above situation, it is necessary to develop forecasting methods for persistent heavy rainfall, building upon existing heavy rainfall forecasting technologies. Summary of the Invention

[0005] Purpose of the Invention: The purpose of this invention is to provide a progressive forecasting method for regional persistent heavy rainfall based on key influencing factors. By further forecasting the location of persistent heavy rainfall after determining whether a regional persistent heavy rainfall event has occurred, the method improves the forecasting capability for persistent heavy rainfall and provides technical support for early warning decision-making in extreme weather events.

[0006] Technical solution: The present invention provides a progressive forecasting method for regional persistent heavy rainfall based on key influencing factors, comprising the following steps:

[0007] Step 1: Compare and analyze the differences between continuous rainstorm processes and general rainstorm processes in traditional physical quantity factors, physical quantity anomaly factors, and comprehensive dynamic factors, and screen the set of key influencing factors with significant discriminative ability.

[0008] Step 2: Extract the anomaly factor of the circulation field before the rainstorm and construct a quantitative index system. Combine the mid-to-high latitude circulation pattern, the mid-to-low latitude dynamic lifting threshold and the water vapor transport intensity to establish a three-level linkage criterion to identify regional continuous rainstorm processes.

[0009] Step 3: Using the TS score as the evaluation criterion, based on the set of key influencing factors, determine the optimal factor combination and weight allocation scheme, integrate kernel density estimation technology to construct a probability forecast model, and output the spatial distribution probability of the rainstorm area.

[0010] Furthermore, step 1 specifically includes the following steps:

[0011] Step 1.1: Statistically analyze the historical continuous and general rainstorm events and their occurrence times within the study area. Based on the ECMWF ERA-Interim reanalysis data with a spatial resolution of 0.75°×0.75°, combined with literature data and forecasting experience, calculate the values ​​of traditional physical quantity factors, physical quantity anomaly factors, and comprehensive dynamic factors closely related to rainstorms.

[0012] Step 1.2: Compare the differences in 24 traditional physical quantity factors between persistent and general rainstorm processes using kernel density estimation, box plots, and analysis of variance. Calculate the probability distribution of these factors using kernel density estimation, analyze the probability distribution maps, and identify factors with different distribution ranges. Combine the results of box plots and analysis of variance to select a set of key influencing factors with significant discriminative power, including 13 factors: 500 hPa, 850 hPa specific humidity, 850 hPa and 925 hPa water vapor flux divergence, atmospheric precipitable water, 850 hPa and 925 hPa vorticity, 200 hPa, 850 hPa and 925 hPa divergence, 850 hPa and 925 hPa vertical velocity, and 925 hPa temperature advection.

[0013] Step 1.3: By comparing the differences in anomaly factors of 24 physical quantities between persistent rainstorm processes and general rainstorm processes using kernel density estimation, box plots, and analysis of variance, a set of key influencing factors with significant discriminative power was selected, including 14 factors: 500 hPa, 700 hPa, 850 hPa specific humidity, 850 hPa water vapor flux divergence, atmospheric precipitable water, 700 hPa, 850 hPa vorticity, 200 hPa, 850 hPa divergence, 850 hPa, 925 hPa vertical velocity, 850 hPa, 925 hPa temperature advection, and 850 hPa pseudo-equivalent potential temperature.

[0014] Step 1.4: The differences between five comprehensive dynamic factors of persistent rainstorm processes and general rainstorm processes are compared and analyzed by kernel density estimation, box plots and variance analysis. All five factors are included in the set of key influencing factors.

[0015] Furthermore, in step 1.1, there are a total of 24 traditional physical quantity factors, including specific humidity at 500 hPa, 700 hPa, 850 hPa, and 925 hPa, water vapor flux divergence and atmospheric precipitable water at 700 hPa, 850 hPa, and 925 hPa, characterizing water vapor conditions; vorticity at 700 hPa, 850 hPa, and 925 hPa, divergence and vertical velocity at 700 hPa, 850 hPa, and 925 hPa, characterizing dynamic lifting conditions; and pseudo-equivalent potential temperature at 700 hPa, 850 hPa, and 925 hPa, characterizing thermodynamic and unstable conditions. hPa temperature advection; the comprehensive dynamic factors consist of 5 components, including hydrothermal advection parameters, thermal helicity, vertical divergence flux, water vapor divergence flux, and thermal wave intensity; the physical quantity anomaly factor is calculated from 24 traditional physical quantity factors according to the standardized anomaly formula, as follows:

[0016]

[0017] In the formula, SA is the standardized anomaly, which is a function of spatial point x and time t, OBS(x,t) is the actual situation, MEAN_clim(x,t) is the climate mean, and SD_clim(x,t) is the climate standard deviation.

[0018] Furthermore, the kernel density estimation is as follows:

[0019] set up Given n independently distributed sample points, let their probability density function be f. What is the probability density function of a point x? for:

[0020]

[0021] In the formula, n is the number of samples, and h is the window width. x is the kernel density estimation function; i For the i-th point, using the standard Gaussian function as the kernel function, we take:

[0022]

[0023] Window width is an important parameter in nonparametric kernel estimation, and the optimal window width value is calculated as follows:

[0024] h d = σ d [ 4 n(p+2) ] 1 p+4

[0025] In the formula, To achieve the optimal window width; denoted as d, where is the standard deviation of the d-dimensional variable distribution; p represents the dimension of the variable, the fitted object is a single variable, and the kernel density estimate is p=1.

[0026] Furthermore, step 2 specifically includes the following steps:

[0027] Step 2.1: Using NCEP / NCAR reanalysis data with a spatial resolution of 2.5°×2.5°, subjectively classify the mid-to-high latitude circulation pattern of historical regional persistent heavy rainfall events at 500 hPa based on forecasting experience;

[0028] Step 2.2: Using NCEP / NCAR reanalysis data with a spatial resolution of 2.5°×2.5°, calculate the 500 hPa geopotential height anomaly field of historical regional persistent rainstorm events according to the standardized anomaly formula. Perform a composite analysis of the geopotential height anomaly field of regional persistent rainstorm events by month and by circulation pattern, and perform t-test.

[0029] Step 2.3: Monthly statistical analysis of the key influence areas of the 500 hPa upper-level trough that match the rainfall areas of historical persistent heavy rainfall events. Synthetic analysis of the 500 hPa geopotential height anomaly field of regional persistent heavy rainfall events, and t-test. In the key influence area of ​​the upper-level trough, a rectangular box is used to delineate the negative anomaly areas that pass the significance test of α=0.05 as the key areas of mid- and low-latitude dynamic lifting anomalies. Whether the average geopotential height anomaly of the key areas is negative is determined as the criterion for mid- and low-latitude dynamic lifting during the occurrence of regional persistent heavy rainfall.

[0030] Step 2.4: Perform a composite analysis of the 850 hPa and 925 hPa water vapor flux divergence fields of historical persistent heavy rainfall events in the region on a monthly basis. Identify obvious water vapor convergence and rising areas that match the heavy rainfall areas, and delineate these areas as key water vapor transport zones using rectangular frames. Calculate the average value of the 850 hPa and 925 hPa water vapor flux divergence of each historical persistent heavy rainfall event in the key zones, and statistically analyze the average and minimum values ​​of all events. Determine whether the average value of the 850 hPa or 925 hPa water vapor flux divergence in the key zones is negative and less than the minimum value as the criterion for water vapor transport during persistent heavy rainfall in the region.

[0031] Step 2.5: Based on the numerical forecast products of the European Centre for Medium-Range Weather Forecasts (ECMWF) global forecast model, and combined with the three-level criteria of mid-to-high latitude circulation, mid-to-low latitude dynamic lifting, and water vapor transport, the identification of regional persistent heavy rainfall processes is achieved.

[0032] Furthermore, in step 2.1, the subjective classification includes five types: two-trough-one-ridge type, two-trough-two-ridge type, Ou type, Baiu type, and typhoon type.

[0033] Further, step 2.2 specifically involves: on the composite map of geopotential height anomalies in the two-trough-one-ridge type region, the geopotential height anomalies of the two-trough-one-ridge are obvious. Using rectangles, areas that pass the significance test of α=0.05 are delineated at the locations of the two-trough-one-ridge, serving as key areas of mid-to-high latitude circulation anomalies. The average geopotential height anomaly of the positive anomaly key areas is subtracted from the average geopotential height anomaly of the negative anomaly key areas to determine the quantitative index as the criterion for mid-to-high latitude circulation patterns in the occurrence of persistent heavy rainfall in the circulation type region.

[0034] Furthermore, step 3 specifically includes the following steps:

[0035] Step 3.1: Construction of the forecast factor matrix; Using the kernel density estimation method, the probability of 13 key traditional physical quantity factors is estimated to obtain the cumulative probability distribution curves corresponding to the 13 traditional physical quantities. Using the probability values ​​corresponding to the 13 traditional physical quantities as forecast factors, 13 factor matrices are constructed. These 13 matrices contain 1 to 13 factors respectively. Taking a matrix containing 3 factors as an example, each row of the matrix represents a combination of 3 factors extracted from the 13 factors. According to the principle of permutation and combination, there are 364 possible combinations of 3 factors. The size of the matrix containing 3 factors is 364×3, as shown below:

[0036]

[0037] Step 3.2: Construction of the forecast factor weight coefficient matrix; Matching 13 forecast factor matrices, 13 weight coefficient matrices are also constructed. Taking the 3-factor weight coefficient matrix corresponding to the 3-factor matrix as an example, each row of the matrix contains the 3 weight coefficients corresponding to the 3 factors. To avoid excessive computation, the coefficient values ​​range from 0.1 to 0.9. According to the principle of permutation and combination, and simultaneously satisfying the condition that the sum of the 3 weight coefficients is 1, there are 36 possible combinations of 3-factor weight coefficients. The constructed 3-factor weight coefficient matrix is ​​36×3 in size, as shown below:

[0038]

[0039] Step 3.3: Determining the probability forecast threshold for persistent heavy rainfall; The probability of persistent heavy rainfall at the meteorological station is calculated using the forecast factor matrix and the forecast factor weight coefficient matrix. Forecast thresholds of 30%, 35%, 40%, 45%, 50%, 55%, 60%, 65%, 70%, 75%, and 80% are selected sequentially. That is, if the probability of persistent heavy rainfall is greater than the forecast threshold, persistent heavy rainfall is predicted. The TS score is used as the evaluation standard. The highest TS score corresponds to the optimal factor combination, weight allocation scheme, and persistent heavy rainfall probability forecast threshold. The calculated formula for the persistent heavy rainfall probability forecast model corresponding to the highest TS score among various combinations of traditional physical quantity forecast factors and forecast factor weight coefficients is as follows:

[0040] PRO1=0.1×X1+0.4×X2+0.3×X3+0.2×X4

[0041] In the formula, PRO1 is the probability of continuous heavy rain, X1 is the divergence at 925 hPa, X2 is the specific humidity at 850 hPa, X3 is the vorticity at 925 hPa, and X4 is the vertical velocity at 850 hPa. The forecast threshold for the probability of continuous heavy rain is 50%, that is, PRO1≥40% predicts the occurrence of continuous heavy rain.

[0042] The TS scoring formula is as follows:

[0043]

[0044] In the formula, TP stands for True Positive, which means that the predicted event actually occurred, i.e., a correct forecast; FP stands for False Positive, which means that the predicted event occurred but did not actually occur, i.e., a false alarm; FN stands for False Negative, which means that the predicted event did not occur but actually occurred, i.e., a missed forecast; the closer the TS score is to 1, the better the forecast performance.

[0045] Step 3.4: Following steps 3.1-3.3, the formula for the probability forecast model of persistent heavy rainfall based on the physical quantity anomaly factor is as follows:

[0046] PRO2=0.3×X5+0.4×X6+0.2×X7+0.1×X8

[0047] In the formula, PRO2 is the probability of continuous heavy rain, X5 is the specific humidity at 850 hPa, X6 is the temperature advection at 925 hPa, X7 is the atmospheric precipitable water, and X8 is the divergence at 200 hPa. The forecast threshold for the probability of continuous heavy rain is 35%, that is, PRO2≥35% predicts the occurrence of continuous heavy rain.

[0048] Step 3.5: Following steps 3.1-3.3, the formula for the probability forecast model of persistent heavy rainfall based on comprehensive dynamic factors is as follows:

[0049] PRO3 = 0.5 × X9 + 0.3 × X 10 +0.1×X 11 +0.1×X 12

[0050] In the formula, PRO3 represents the probability of continuous heavy rainfall, X9 represents the hydrothermal advection parameter, and X... 10 X is the thermal helicity. 11 Let X be the water vapor divergence flux. 12 The thermal wave intensity is the density of the event; the probability forecast threshold for persistent heavy rainfall is 60%, that is, PRO3≥60% predicts persistent heavy rainfall.

[0051] Step 3.6: Based on the formulas of the three probability forecast models for persistent heavy rainfall, and combined with the numerical forecast products of the European Centre for Medium-Rainfall Global Forecasting Model, calculate PRO1, PRO2, and PRO3, and select the forecast result with the highest TS score of the previous day among the three models as the final forecast opinion.

[0052] The present invention also discloses a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method of the present invention.

[0053] The present invention also discloses a computer-readable storage medium having a computer program / instructions stored thereon, which, when executed by a processor, implements the steps of the method of the present invention.

[0054] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0055] (1) This invention considers the maintenance mechanism of large-scale circulation on persistent rainstorms, extracts the influencing factors that are significantly different from those of general rainstorms, breaks through the traditional single forecasting model, and establishes a progressive forecasting technology system from rainstorm process forecasting to landing area probability forecasting.

[0056] (2) The mid-to-high latitude circulation pattern, mid-to-low latitude dynamic lifting threshold and water vapor transport intensity quantitative indicators were constructed in sequence to form a three-level linkage criterion for regional continuous rainstorm process, thus realizing the objective forecast of regional continuous rainstorm process.

[0057] (3) Using the highest TS score as the evaluation criterion, the optimal combination of forecast factors and forecast factor weight coefficients was selected, and the constructed regional persistent heavy rainfall area forecast model showed good forecast performance and had a good indicative role in the forecast of regional persistent heavy rainfall. When applied to the forecast of regional persistent heavy rainfall in Hunan in June 2023, the model achieved an average TS score of 20.1, which was better than the average TS score of 14.7 of the European Centre for Medium-Range Weather Forecasting (ECMWF) global forecast model for heavy rainfall. Attached Figure Description

[0058] Figure 1 This is a flowchart illustrating the present invention;

[0059] Figure 2 This is the forecast result of the continuous heavy rain process in Hunan Province in the next 24 hours as of 08:00 on July 4, 2022; among them, (a) is the 500 hPa geopotential height (isolines) and anomaly (color spots); (b) is the 925 hPa water vapor flux divergence field (color spots) and wind field (wind vane).

[0060] Figure 3The forecast results for the continuous heavy rain process in Hunan Province in the next 24 hours are as of 08:00 on July 4, 2022. Among them, (a) is the actual precipitation at 08:00 on July 5; (b) is the probability forecast of the continuous heavy rain at 08:00 on July 5. Detailed Implementation

[0061] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0062] This invention provides a method for forecasting regional persistent heavy rainfall, specifically taking the forecasting of regional persistent heavy rainfall in Hunan Province in July as an example. Figure 1 As shown, it includes the following steps:

[0063] Step 1: Selection and calculation of key influencing factors of the persistent regional rainstorm in Hunan in July.

[0064] Step 1 includes the following steps:

[0065] Step 1.1: Statistically analyze the occurrence times of persistent and general heavy rainfall events in Hunan Province during July from 1979 to 2020. Using ECMWF ERA-Interim reanalysis data, calculate the specific humidity (500 hPa, 700 hPa, 850 hPa, 925 hPa), water vapor flux divergence (700 hPa, 850 hPa, 925 hPa), and atmospheric precipitable water at the corresponding times, representing water vapor conditions. Also calculate the vorticity (700 hPa, 850 hPa, 925 hPa), divergence (200 hPa, 700 hPa, 850 hPa, 925 hPa), and vertical velocity (700 hPa, 850 hPa, 925 hPa), representing dynamic lifting conditions. Finally, calculate the pseudo-equivalent potential temperature (700 hPa, 850 hPa, 925 hPa), and temperature advection (700 hPa, 850 hPa, 925 hPa), representing thermal and instability conditions. A total of 24 traditional physical quantities (hPa) were calculated. Simultaneously, the SA values ​​of these 24 traditional physical quantities were calculated, along with the values ​​of five comprehensive dynamic factors: hydrothermal advection parameters, thermal helicity, vertical divergence flux, water vapor divergence flux, and thermal wave intensity.

[0066] Step 1.2: By comparing the differences in 24 traditional physical quantity factors between persistent rainstorm processes and general rainstorm processes using methods such as probability distribution, box plots, and analysis of variance, 13 key influencing factors were selected, as shown in Table 1.

[0067] Table 1. Factors of 13 key traditional physical quantities

[0068]

[0069] Step 1.3: By comparing the differences in the anomaly factors of 24 physical quantities between persistent rainstorm processes and general rainstorm processes using methods such as probability distribution, box plots, and analysis of variance, 14 key influencing factors were selected, as shown in Table 2.

[0070] Table 2 Anomaly Factors for 14 Key Physical Quantities

[0071]

[0072] Step 1.4: By comparing and analyzing the differences in five comprehensive dynamic factors between persistent rainstorm processes and general rainstorm processes using kernel density estimation and box plot methods, all five factors are included in the set of key influencing factors, as shown in Table 3.

[0073] Table 3. Five Comprehensive Dynamic Factors

[0074]

[0075] Step 2: Forecast of the continuous heavy rain process in Hunan in July.

[0076] Step 2 includes the following steps:

[0077] Step 2.1: Using NCEP / NCAR reanalysis data, the subjective classification of the mid-to-high latitude circulation pattern of the regional persistent rainstorm events in Hunan in July from 1961 to 2020 in the Eurasian region (40°–180°E) at 500 hPa can be divided into five types: two troughs and one ridge, Ou-U type, Bai-U type, two troughs and two ridges type, and typhoon type.

[0078] Step 2.2: Using NCEP / NCAR reanalysis data, calculate the SA field of 500 hPa geopotential height for regional persistent heavy rainfall events in Eurasia from 1961 to 2020. Perform synthetic analysis on the anomaly field of geopotential height to identify key areas and criteria for mid-to-high latitude circulation anomalies, as shown in Table 4. Due to the limited number of typhoon cases in July, the statistical results are not representative, and the criteria for typhoon type are not provided.

[0079] Table 4. Key areas and criteria for determining persistent heavy rainfall in various circulation patterns in Hunan Province in July using the 500 hPa mid-to-high latitude circulation.

[0080]

[0081] Step 2.3: Statistically analyze the key areas of low-latitude dynamic lifting anomalies in the composite field of geopotential height anomalies during the July regional persistent rainstorm event in Hunan Province. Determine the criteria for low-latitude dynamic lifting anomalies during the occurrence of regional persistent rainstorms as follows: When the average value of geopotential height anomalies in Hunan Province (107°-115° E, 25°-32° N) is negative, or the maximum value of geopotential height in the key area affected by the subtropical high (110°-120° E, 20°-23° N) is ≥588 dagpm and lasts for 2 days or more, the dynamic conditions for persistent rainstorms in Hunan Province are considered to be met.

[0082] Step 2.4: Perform composite analysis on the 850 hPa and 925 hPa water vapor flux divergences of the persistent regional rainstorm event in Hunan in July, delineate the key area for water vapor transport, and determine the water vapor transport criterion for the occurrence of persistent regional rainstorms as follows: when the average value of the 850 hPa or 925 hPa water vapor flux divergence in the Hunan region (110°-115° E, 25°-30° N) is negative and the minimum value is ≤ -0.3×10-5 g·cm³ -2 ·hPa -1 ·s -1 If the water vapor conditions for continuous heavy rainfall in Hunan Province are met, and the rainfall lasts for 2 days or more, then the conditions are considered to be met.

[0083] Step 2.5: Based on forecast products from the European Centre for Medium-Range Global Forecasting Model, combined with mid-to-high latitude circulation and mid-to-low latitude dynamic lifting...

[0084] Based on the three-level criteria for water vapor transport, a persistent heavy rainfall event in Hunan Province from July 5th to 6th, 2022, was identified. During this event, the high-latitude geopotential height anomaly field was of the Ordos-Ulanhot type, with the three key areas exhibiting a "+-+" distribution, meeting the first-level identification criteria. The average geopotential height anomaly in Hunan Province was negative, meeting the second-level identification criteria. The water vapor flux divergence magnitude in the 925 hPa key area met the third-level identification criteria, indicating a forecast of persistent heavy rainfall in the region. See details... Figure 2 .

[0085] Step 3: Probability forecast of continuous heavy rain in Hunan in July.

[0086] Step 3 includes the following steps:

[0087] Step 3.1: Using the TS score as the evaluation criterion, the following formula is obtained for the probability forecast model of persistent heavy rainfall in Hunan in July based on traditional physical quantity factors:

[0088] PRO1=0.1×X1+0.4×X2+0.3×X3+0.2×X4

[0089] In the formula, PRO1 represents the probability of persistent heavy rainfall, X1 is the divergence at 925 hPa, X2 is the specific humidity at 850 hPa, X3 is the vorticity at 925 hPa, and X4 is the vertical velocity at 850 hPa. The forecast threshold for the probability of persistent heavy rainfall is 50%, that is, PRO1 ≥ 40% predicts the occurrence of persistent heavy rainfall.

[0090] Step 3.2: Using the TS score as the evaluation criterion, the following formula is obtained for the probability forecast model of persistent heavy rainfall in Hunan in July based on the physical quantity anomaly factor:

[0091] PRO2=0.3×X5+0.4×X6+0.2×X7+0.1×X8

[0092] In the formula, PRO2 is the probability of continuous heavy rain, X5 is the specific humidity at 850 hPa, X6 is the temperature advection at 925 hPa, X7 is the atmospheric precipitable water, and X8 is the divergence at 200 hPa. The forecast threshold for the probability of continuous heavy rain is 35%, that is, PRO2≥35% predicts the occurrence of continuous heavy rain.

[0093] Step 3.3: Using the TS score as the evaluation criterion, the following formula is obtained for the probability forecast model of persistent heavy rainfall in Hunan in July based on comprehensive dynamic factors:

[0094] PRO3 = 0.5 × X9 + 0.3 × X 10 +0.1×X 11 +0.1×X 12

[0095] In the formula, PRO3 represents the probability of continuous heavy rainfall, X9 represents the hydrothermal advection parameter, and X... 10 X is the thermal helicity. 11 Let X be the water vapor divergence flux. 12 The thermal wave intensity is the density of the event; the probability forecast threshold for persistent heavy rainfall is 60%, that is, PRO3≥60% predicts persistent heavy rainfall.

[0096] Step 3.4: Based on the identification of the persistent heavy rainfall event in Hunan Province from July 5th to 6th, 2022, the values ​​of each forecast factor are calculated according to the European Centre for Medium-Range Weather Forecasts (ECMWF) global forecast model. Combined with the Hunan July persistent heavy rainfall probability forecast model, the rainfall area for this event is forecasted. PRO1, PRO2, and PRO3 are calculated according to the formulas of three persistent heavy rainfall probability forecast models. Among the three models, the one with the highest TS score the previous day is the traditional physical quantity forecast model. The forecast result PRO1 from this model is used as the final forecast opinion. Verification shows that the 24-hour probability forecast area of ​​this model is in good agreement with the actual situation, with an average TS score of 19.4, indicating a good forecast effect. See details... Figure 3 .

Claims

1. A progressive forecasting method for regional persistent heavy rainfall based on key influencing factors, characterized in that, Includes the following steps: Step 1: Compare and analyze the differences between continuous rainstorm processes and general rainstorm processes in traditional physical quantity factors, physical quantity anomaly factors, and comprehensive dynamic factors, and screen the set of key influencing factors with significant discriminative ability. Step 2: Extract the anomaly factor of the circulation field before the rainstorm and construct a quantitative index system. Combine the mid-to-high latitude circulation pattern, the mid-to-low latitude dynamic lifting threshold and the water vapor transport intensity to establish a three-level linkage criterion to identify regional continuous rainstorm processes. Step 3: Using the TS score as the evaluation criterion, based on the set of key influencing factors, determine the optimal factor combination and weight allocation scheme, integrate kernel density estimation technology to construct a probability forecast model, and output the spatial distribution probability of the rainstorm area.

2. The progressive forecasting method for regional persistent heavy rainfall based on key influencing factors according to claim 1, characterized in that, Step 1 specifically includes the following steps: Step 1.1: Statistically analyze the historical continuous and general rainstorm events and their occurrence times within the study area. Based on the ECMWF ERA-Interim reanalysis data with a spatial resolution of 0.75°×0.75°, combined with literature data and forecasting experience, calculate the values ​​of traditional physical quantity factors, physical quantity anomaly factors, and comprehensive dynamic factors closely related to rainstorms. Step 1.2: Compare the differences in 24 traditional physical quantity factors between persistent and general rainstorm processes using kernel density estimation, box plots, and analysis of variance. Calculate the probability distribution of these factors using kernel density estimation, analyze the probability distribution maps, and identify factors with different distribution ranges. Combine the results of box plots and analysis of variance to select a set of key influencing factors with significant discriminative power, including 13 factors: 500 hPa, 850 hPa specific humidity, 850 hPa and 925 hPa water vapor flux divergence, atmospheric precipitable water, 850 hPa and 925 hPa vorticity, 200 hPa, 850 hPa and 925 hPa divergence, 850 hPa and 925 hPa vertical velocity, and 925 hPa temperature advection. Step 1.3: By comparing the differences in anomaly factors of 24 physical quantities between persistent rainstorm processes and general rainstorm processes using kernel density estimation, box plots, and analysis of variance, a set of key influencing factors with significant discriminative power was selected, including 14 factors: 500 hPa, 700 hPa, 850 hPa specific humidity, 850 hPa water vapor flux divergence, atmospheric precipitable water, 700 hPa, 850 hPa vorticity, 200 hPa, 850 hPa divergence, 850 hPa, 925 hPa vertical velocity, 850 hPa, 925 hPa temperature advection, and 850 hPa pseudo-equivalent potential temperature. Step 1.4: The differences between five comprehensive dynamic factors of persistent rainstorm processes and general rainstorm processes are compared and analyzed by kernel density estimation, box plots and variance analysis. All five factors are included in the set of key influencing factors.

3. The progressive forecasting method for regional persistent heavy rainfall based on key influencing factors according to claim 2, characterized in that, In step 1.1, there are a total of 24 traditional physical quantity factors, including specific humidity at 500 hPa, 700 hPa, 850 hPa, and 925 hPa, water vapor flux divergence and atmospheric precipitable water at 700 hPa, 850 hPa, and 925 hPa, characterizing water vapor conditions; vorticity at 700 hPa, 850 hPa, and 925 hPa, divergence and vertical velocity at 700 hPa, 850 hPa, and 925 hPa, characterizing dynamic lifting conditions; and pseudo-equivalent potential temperature at 700 hPa, 850 hPa, and 925 hPa, characterizing thermodynamic and unstable conditions. hPa temperature advection; the comprehensive dynamic factors consist of 5 components, including hydrothermal advection parameters, thermal helicity, vertical divergence flux, water vapor divergence flux, and thermal wave intensity; the physical quantity anomaly factor is calculated from 24 traditional physical quantity factors according to the standardized anomaly formula, as follows: ； In the formula, SA is the standardized anomaly, which is a function of spatial point x and time t, OBS(x,t) is the actual situation, MEAN_clim(x,t) is the climate mean, and SD_clim(x,t) is the climate standard deviation.

4. The progressive forecasting method for regional persistent heavy rainfall based on key influencing factors according to claim 2, characterized in that, The kernel density estimation is as follows: set up Given n independently distributed sample points, let their probability density function be f. Let the probability density function of a point x be f. for: In the formula, n is the number of samples, and h is the window width. x is the kernel density estimation function; i For the i-th point, using the standard Gaussian function as the kernel function, we take: Window width is a parameter in nonparametric kernel estimation, and the optimal window width value is calculated as follows: In the formula, To achieve the optimal window width; denoted as d, where is the standard deviation of the d-dimensional variable distribution; p represents the dimension of the variable, the fitted object is a single variable, and the kernel density estimate is p=1.

5. The progressive forecasting method for regional persistent heavy rainfall based on key influencing factors according to claim 1, characterized in that, Step 2 specifically includes the following steps: Step 2.1: Using NCEP / NCAR reanalysis data with a spatial resolution of 2.5°×2.5°, subjectively classify the 500 hPa mid-to-high latitude circulation pattern of historical regional persistent heavy rainfall events based on forecasting experience; Step 2.2: Using NCEP / NCAR reanalysis data with a spatial resolution of 2.5°×2.5°, calculate the 500 hPa geopotential height anomaly field of historical regional persistent rainstorm events according to the standardized anomaly formula. Perform a composite analysis of the geopotential height anomaly field of regional persistent rainstorm events by month and by circulation pattern, and perform t-test. Step 2.3: Monthly statistical analysis of the key influence areas of the 500 hPa upper-level trough that match the rainfall areas of historical persistent heavy rainfall events. Synthetic analysis of the 500 hPa geopotential height anomaly field of regional persistent heavy rainfall events, and t-test. In the key influence area of ​​the upper-level trough, a rectangular box is used to delineate the negative anomaly areas that pass the significance test of α=0.05 as the key areas of mid- and low-latitude dynamic lifting anomalies. Whether the average geopotential height anomaly of the key areas is negative is determined as the criterion for mid- and low-latitude dynamic lifting during the occurrence of regional persistent heavy rainfall. Step 2.4: Perform a composite analysis of the 850 hPa and 925 hPa water vapor flux divergence fields of historical persistent heavy rainfall events in the region on a monthly basis. Identify obvious water vapor convergence and rising areas that match the heavy rainfall areas, and delineate these areas as key water vapor transport zones using rectangular frames. Calculate the average value of the 850 hPa and 925 hPa water vapor flux divergence of each historical persistent heavy rainfall event in the key zones, and statistically analyze the average and minimum values ​​of all events. Determine whether the average value of the 850 hPa or 925 hPa water vapor flux divergence in the key zones is negative and less than the minimum value as the criterion for determining water vapor transport during persistent heavy rainfall in the region. Step 2.5: Based on the numerical forecast products of the European Centre for Medium-Range Weather Forecasts (ECMWF) global forecast model, and combined with the three-level criteria of mid-to-high latitude circulation, mid-to-low latitude dynamic lifting, and water vapor transport, the identification of regional persistent heavy rainfall processes is achieved.

6. The progressive forecasting method for regional persistent heavy rainfall based on key influencing factors according to claim 5, characterized in that, In step 2.1, the subjective classification includes five types: two-trough-one-ridge type, two-trough-two-ridge type, Ou type, Baiu type, and typhoon type.

7. The progressive forecasting method for regional persistent heavy rainfall based on key influencing factors according to claim 6, characterized in that, Step 2.2 specifically involves: On the composite map of geopotential height anomalies in the two-trough-one-ridge type region, the geopotential height anomalies of the two-trough-one-ridge are obvious. Using rectangles, areas that pass the significance test of α=0.05 are delineated at the locations of the two-trough-one-ridge, which are taken as key areas of mid-to-high latitude circulation anomalies. The average geopotential height anomaly of the positive anomaly key areas is subtracted from the average geopotential height anomaly of the negative anomaly key areas to determine the quantitative index as the criterion for mid-to-high latitude circulation patterns in the occurrence of persistent rainstorms in the circulation type region.

8. The progressive forecasting method for regional persistent heavy rainfall based on key influencing factors according to claim 2, characterized in that, Step 3 specifically includes the following steps: Step 3.1: Construction of forecast factor matrix; Using the kernel density estimation method, the probability of 13 key traditional physical quantity factors is estimated to obtain the cumulative probability distribution curves corresponding to the 13 traditional physical quantities. The probability values ​​corresponding to the 13 traditional physical quantities are used as forecast factors to construct 13 factor matrices. Each of the 13 matrices contains 1 to 13 factors. Taking a matrix with 3 factors as an example, each row of the matrix represents a combination of 3 factors drawn from 13 factors. According to the principle of permutations and combinations, there are 364 possible combinations of 3 factors. The size of the matrix with 3 factors is 364×3. The 3-factor matrix is ​​shown below: ； Step 3.2: Construction of the forecast factor weight coefficient matrix; Match the 13 forecast factor matrices, and similarly construct 13 weight coefficient matrices. Taking the 3-factor weight coefficient matrix corresponding to the 3-factor matrix as an example, each row of the matrix contains the 3 weight coefficients corresponding to the 3 factors. The coefficient values ​​range from 0.1 to 0.

9. According to the principle of permutation and combination, and simultaneously satisfying the condition that the sum of the 3 weight coefficients is 1, there are 36 possible combinations of 3-factor weight coefficients. The constructed 3-factor weight coefficient matrix is ​​36×3 in size, as shown below: ； Step 3.3: Determining the probability forecast threshold for persistent heavy rainfall; The probability of persistent heavy rainfall at the meteorological station is calculated using the forecast factor matrix and the forecast factor weight coefficient matrix. Forecast thresholds of 30%, 35%, 40%, 45%, 50%, 55%, 60%, 65%, 70%, 75%, and 80% are selected sequentially. That is, if the probability of persistent heavy rainfall is greater than the forecast threshold, persistent heavy rainfall is predicted. The TS score is used as the evaluation standard. The highest TS score corresponds to the optimal factor combination, weight allocation scheme, and persistent heavy rainfall probability forecast threshold. The calculated formula for the persistent heavy rainfall probability forecast model corresponding to the highest TS score among various combinations of traditional physical quantity forecast factors and forecast factor weight coefficients is as follows: PRO1=0.1×X1+0.4×X2+0.3×X3+0.2×X4 In the formula, PRO1 is the probability of continuous heavy rain, X1 is the divergence at 925 hPa, X2 is the specific humidity at 850 hPa, X3 is the vorticity at 925 hPa, and X4 is the vertical velocity at 850 hPa. The forecast threshold for the probability of continuous heavy rain is 50%, that is, PRO1≥40% predicts the occurrence of continuous heavy rain. The TS scoring formula is as follows: ； In the formula, TP stands for True Positive, which means that the predicted event actually occurred, i.e., a correct forecast; FP stands for False Positive, which means that the predicted event occurred but did not actually occur, i.e., a false alarm; FN stands for False Negative, which means that the predicted event did not occur but actually occurred, i.e., a missed forecast; the closer the TS score is to 1, the better the forecast performance. Step 3.4: Following steps 3.1-3.3, the formula for the probability forecast model of persistent heavy rainfall based on the physical quantity anomaly factor is as follows: PRO2=0.3×X5+0.4×X6+0.2×X7+0.1×X8 In the formula, PRO2 is the probability of continuous heavy rain, X5 is the specific humidity at 850 hPa, X6 is the temperature advection at 925 hPa, X7 is the atmospheric precipitable water, and X8 is the divergence at 200 hPa. The forecast threshold for the probability of continuous heavy rain is 35%, that is, PRO2≥35% predicts the occurrence of continuous heavy rain. Step 3.5: Following steps 3.1-3.3, the formula for the probability forecast model of persistent heavy rainfall based on comprehensive dynamic factors is as follows: PRO3=0.5×X9+0.3×X 10 +0.1×X 11 +0.1×X 12 In the formula, PRO3 represents the probability of continuous heavy rainfall, X9 represents the hydrothermal advection parameter, and X... 10 X is the thermal helicity. 11 Let X be the water vapor divergence flux. 12 The thermal wave intensity is the density of the event; the probability forecast threshold for persistent heavy rainfall is 60%, that is, PRO3≥60% predicts persistent heavy rainfall. Step 3.6: Based on the formulas of the three probability forecast models for persistent heavy rainfall, and combined with the numerical forecast products of the European Centre for Medium-Rainfall Global Forecasting Model, calculate PRO1, PRO2, and PRO3, and select the forecast result with the highest TS score of the previous day among the three models as the final forecast opinion.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method of claim 1.

10. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method of claim 1.

Images