A Multi-Factor Quantitative Attribution Method for Weather Events Based on Linear Mixture Model Decomposition

By employing a linear mixture model decomposition method and clustering algorithm, the problem of quantitative separation of influencing factors of short-term heavy precipitation was solved, enabling quantitative analysis of weather patterns and topographic effects and providing stable attribution analysis results.

CN123116649APending Publication Date: 2026-11-17ZHEJIANG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611620445.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-10-20
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing technologies are insufficient to quantitatively separate the effects of weather patterns, topography, and the combined effects of multiple factors on the occurrence of short-term heavy precipitation from observational data. Traditional methods are also insufficient to effectively capture local high-impact events and lack regional adaptability.

Method used

A linear mixture model-based decomposition method was adopted. By acquiring meteorological station data, atmospheric reanalysis data, and digital elevation topographic data, a linear mixture attribution decomposition model was constructed to decompose the frequency of short-term heavy precipitation into the independent and composite contributions of each influencing factor. A clustering algorithm was used to classify the weather patterns, and a self-organizing map algorithm was used for objective classification.

Benefits of technology

It achieves quantitative separation of weather patterns, topography, and the combined effects of multiple factors, overcoming the uncertainty and insufficient regional adaptability of traditional methods, and providing stable quantitative attribution analysis results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123116649A_ABST
    Figure CN123116649A_ABST
Patent Text Reader

Abstract

This invention provides a multi-factor quantitative attribution method for weather events based on linear mixture model decomposition, belonging to the field of meteorological disaster attribution analysis technology. The method includes: acquiring hourly precipitation data, reanalysis data, and topographic data of the area to be analyzed; identifying short-duration heavy precipitation events based on precipitation data; extracting multi-layer meteorological elements at the corresponding time of the event to construct a circulation feature vector, and using a self-organizing map algorithm to cluster various weather pattern types; classifying the underlying surface topography based on topographic data; constructing a linear mixture attribution decomposition model with specified weather patterns, topography, and time as the baseline state, representing the frequency of short-duration heavy precipitation as a linear combination of the independent and composite contributions of each factor; and calculating the quantitative contribution of each factor relative to the baseline state to the frequency. This invention can quantitatively separate the contributions of weather patterns, topography, and the composite effects of various factors from observational data, providing an objective basis for weather event attribution.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of meteorological disaster attribution analysis technology, and in particular to a multi-factor quantitative attribution method for weather events based on linear mixture model decomposition. Background Technology

[0002] Short-duration heavy rainfall is a typical type of severe convective weather. The areas where it forms are usually characterized by complex terrain and are affected by the alternation or combined influence of multiple weather systems.

[0003] In identifying short-duration heavy rainfall, fixed threshold methods or percentage threshold methods are commonly used. Fixed threshold methods define short-duration heavy rainfall events using a uniform rainfall threshold, while percentage threshold methods use a certain percentile of the historical rainfall sequence at a station as the threshold. However, these traditional methods are difficult to effectively capture localized high-impact short-duration heavy rainfall events, and the regional adaptability of the identification results is insufficient.

[0004] To clarify the dominant mechanisms of short-duration heavy precipitation, weather pattern classification is a classic and effective method. However, existing weather pattern studies typically focus only on the weather pattern itself, neglecting the combined effects of underlying topography and weather patterns, and making it even more difficult to reveal the quantitative differences in the effects of topography under different weather patterns. In terms of quantitative analysis of topographic effects, high-resolution numerical simulation is the mainstream method, but it is affected by physical parameterization schemes, contains uncertainties, and is difficult to quantitatively separate the contributions of different factors directly from observational data. Currently, there is a lack of an attribution analysis method that can quantitatively analyze the impact of weather patterns, topography, and the combined effects of multiple factors on the occurrence of short-duration heavy precipitation based on observational data. Summary of the Invention

[0005] The purpose of this invention is to provide a quantitative attribution method for multiple factors of weather events based on linear mixture model decomposition, so as to solve the technical problem in the prior art that it is impossible to quantitatively separate the influence of weather patterns, topography and multiple factors on the occurrence of short-term heavy precipitation from observation data.

[0006] To achieve the above objectives, this invention provides a multi-factor quantitative attribution method for weather events based on linear mixture model decomposition, comprising the following steps: Step S1: Obtain hourly precipitation observation data, atmospheric reanalysis data, and digital elevation topographic data of meteorological stations in the area to be analyzed; Step S2: Based on hourly precipitation observation data, calculate the short-term heavy precipitation intensity index of each meteorological station according to the hourly precipitation and its historical precipitation threshold, and identify the hourly precipitation that meets the preset intensity level as a short-term heavy precipitation event. Step S3: Based on atmospheric reanalysis data, extract multiple meteorological elements at multiple altitude layers at the time corresponding to short-term heavy precipitation events, construct circulation feature vectors for each event, and use a clustering algorithm to cluster the circulation feature vectors of all events to obtain various weather pattern types. Step S4: Based on digital elevation topographic data, the underlying surface of the area to be analyzed is divided into different topographic categories; Step S5: Using the specified weather pattern type, specified terrain category, and atmospheric state at the specified time as the baseline state, construct a linear hybrid attribution decomposition model to represent the frequency of short-term heavy precipitation as a linear combination of the independent contribution terms and composite contribution terms of each influencing factor. Step S6: Based on the linear mixed attribution decomposition model, using the actual frequency of short-term heavy precipitation under different weather conditions, different terrain types, and different atmospheric conditions at different times, calculate the quantitative contribution of each influencing factor to the frequency of short-term heavy precipitation relative to the baseline state. The relative magnitude of this contribution does not depend on the selected baseline state, that is, the quantitative attribution of the occurrence of short-term heavy precipitation does not change with the change of the baseline state.

[0007] Preferably, in step S2, based on hourly precipitation observation data, the short-term heavy precipitation intensity index of each meteorological station is calculated according to the hourly precipitation and its historical precipitation threshold, including: The 95th percentile of the historical hourly precipitation at each meteorological station is taken as the precipitation threshold for that meteorological station. Calculate the standard deviation of precipitation exceeding the precipitation threshold in the historical hourly precipitation data for each meteorological station; Calculate the short-duration heavy precipitation intensity index: ; in, This is the short-duration heavy rainfall intensity index. This refers to the hourly precipitation at the meteorological station. This is the 95th percentile of the station's historical precipitation. greater than The standard deviation of precipitation at each station.

[0008] Preferably, in step S3, constructing the circulation feature vector for each event includes: For each regional short-duration heavy precipitation event, atmospheric reanalysis data were extracted from three key time points: one hour before the event, the peak time with the most short-duration heavy precipitation stations, and one hour after the event ended. At each key time point, the meridional and zonal winds at 200hPa, 500hPa, and 850hPa, the geopotential height at 500hPa, and the precipitable water data were extracted. The 24 meteorological elements from the three key time periods were flattened and spliced ​​together according to the same rule to form a one-dimensional feature vector, which served as the circulation feature vector of this regional short-term heavy precipitation event.

[0009] Preferably, the clustering algorithm in step S3 is the self-organizing map algorithm.

[0010] Preferably, the linear mixed attribution decomposition model in step S5 is: ; in, The frequency of short-duration heavy rainfall. An independent contribution to weather patterns. An independent contribution to the terrain. An independent contribution to atmospheric conditions. The combined contribution of weather patterns and topography, It contributes to the combined effects of weather patterns and atmospheric conditions. It contributes to the combined effects of topography and atmospheric conditions. It is a composite contribution of three factors: weather conditions, topography, and atmospheric state. This is the error term.

[0011] Preferably, in step S6, the quantitative contribution value includes the independent contribution and composite contribution of each influencing factor. The composite contribution includes pairwise composite contribution and three-factor composite contribution. Both the independent contribution and the composite contribution are determined based on the frequency change, which is the difference between the actual occurrence frequency of short-term heavy precipitation and the actual occurrence frequency under the baseline state.

[0012] Preferably, the independent contributions of each influencing factor include: No. Average independent contribution of weather conditions at any given time: ; No. Total average contribution of terrain at time: ; No. The purely independent contribution of atmospheric state at any given time: ; in, For the first The total average contribution of weather-like phenomena on a diurnal scale. This represents the total average contribution of non-baseline topography on a diurnal variation scale. For the first The purely independent contribution of atmospheric state at any given moment. These are the sequence numbers for weather situation types other than the baseline weather situation type. For time sequence number, , As the baseline terrain, This is a non-baseline terrain. For the first The individual contribution of weather-like patterns, The separate contribution of non-baseline terrain. For the first Weather patterns and the first The combined contribution of atmospheric conditions at any given time. For non-baseline terrain and the first The combined contribution of atmospheric conditions at any given time. For the first Weather patterns, baseline topography, and the first The frequency change of a given time relative to the reference state. For baseline weather situation type, baseline topography, and the first The frequency change of a given time relative to the reference state. For baseline weather pattern type, non-baseline terrain, the first The frequency change of a given time relative to the reference state. For baseline weather situation type, baseline topography, and the first The actual frequency of short-duration heavy rainfall at a given time. This includes the baseline weather pattern type, baseline topography, and the actual frequency of short-duration heavy precipitation at 00:00.

[0013] Preferably, pairwise composite contributions include: The combined contribution of weather conditions and topography: ; Weather conditions and the The total composite contribution of atmospheric state at time: ; Topography and the The total composite contribution of atmospheric state at time: ; in, For the first The combined contribution of weather-like conditions and non-baseline terrain. For the first Weather patterns and the first The total composite contribution of atmospheric state at any given time. For non-baseline terrain and the first The total composite contribution of atmospheric state at any given moment, For the first The second-order composite contribution of weather-like conditions and non-baseline terrain. For the first Weather patterns, non-baseline terrain and the first The combined contribution of three factors related to atmospheric conditions at any given time. For the first Weather patterns, non-baseline terrain, and the first The frequency change of a given time relative to the reference state. For the first Weather patterns, baseline topography, and the first The frequency change of a given time relative to the reference state. For baseline weather pattern type, non-baseline terrain, the first The frequency change of a given time relative to the reference state. For baseline weather situation type, baseline topography, and the first The frequency change of a given time relative to the reference state. For the first Weather-like conditions, baseline terrain, and frequency changes at time 00 relative to the baseline. This includes the baseline weather pattern type, the non-baseline terrain, and the frequency change of time 00 relative to the baseline. For the first Weather-like conditions, non-baseline terrain, and frequency changes at time 00 relative to the baseline state.

[0014] Preferred, the composite contribution of three factors for: ; in, For the first Weather patterns, baseline topography, and frequency changes at 00:00.

[0015] Preferably, the method further includes a verification step: changing the setting of the baseline state, repeatedly calculating the quantitative contribution value of each influencing factor, and if the relative order of the contribution values ​​of each factor remains unchanged, then the linear mixed attribution decomposition model is determined to be stable.

[0016] Therefore, the present invention employs the above-mentioned multi-factor quantitative attribution method for weather events based on linear mixture model decomposition, and the beneficial technical effects are as follows: (1) This invention, for the first time, starts from observation data and realizes the quantitative separation and contribution calculation of the influence of weather conditions, topography and multiple factors on the occurrence of short-term heavy precipitation, overcoming the defects of traditional numerical simulation methods that are affected by physical parameterization schemes and have uncertainties; (2) By constructing a linear hybrid attribution decomposition model, this invention decomposes the frequency of short-term heavy precipitation into the independent and composite contributions of each factor, which can quantitatively reveal the differential effects of topography under different weather conditions and overcome the shortcomings of traditional methods in capturing local high-impact events and insufficient regional adaptability. (3) This invention uses a self-organizing mapping algorithm to objectively classify regional short-term heavy precipitation events, avoiding the uncertainty caused by subjective classification. The classification results have good objectivity and continuity. (4) This invention does not rely on a specific reference state selection. By changing the reference state, the stability and universality of its results can be verified, and it is applicable to the attribution analysis of short-term heavy precipitation in different regions. Attached Figure Description

[0017] Figure 1 This is a diurnal variation diagram of the frequency of short-duration heavy precipitation at hilly and plain stations under different circulation patterns in this embodiment of the invention; wherein, Figure 1 (a) in the table represents the sum of all types. Figure 1 (b) in the text represents a coastal cyclone type. Figure 1 (c) in the diagram represents a subtropical high-pressure system. Figure 1 (d) in the diagram represents the summer front type. Figure 1 (e) in the text refers to the spring jet stream type; Figure 2 This is a flowchart of the multi-factor quantitative attribution method for weather events based on linear mixture model decomposition according to the present invention. Detailed Implementation

[0018] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0019] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0020] Example 1 This embodiment takes the middle and lower reaches of the Yangtze River as the area to be analyzed, and combines actual observation data from 2010 to 2022 for a total of 13 years to provide a detailed description of the technical solution of the present invention.

[0021] In this embodiment, the four weather patterns affecting short-term heavy rainfall in the middle and lower reaches of the Yangtze River are: Weather Pattern Type 1 (hereinafter referred to as Type 1): Coastal cyclone type; Weather Pattern Type 2 (hereinafter referred to as Type 2): Subtropical high-pressure system type; Weather Pattern Type 3 (hereinafter referred to as Type 3): Summer front type; Weather Pattern Type 4 (hereinafter referred to as Type 4): Spring jet stream type. Baseline topography. Take plain terrain, non-baseline terrain The terrain is hilly. The afternoon atmospheric conditions are represented by 5 PM.

[0022] Reference Figure 2 A multi-factor quantitative attribution method for weather events based on linear mixture model decomposition includes the following steps: Step S1: Obtain hourly precipitation observation data, atmospheric reanalysis data, and digital elevation topographic data of meteorological stations in the area to be analyzed.

[0023] Hourly precipitation data from 2010 to 2022 were obtained from 614 meteorological stations in the middle and lower reaches of the Yangtze River from the National Meteorological Information Center, excluding stations with incomplete records. The fifth-generation global atmospheric reanalysis dataset (ERA5) for the same period was acquired, with a temporal resolution of 1 hour and a spatial resolution of 0.25°, extracting variables such as temperature, geopotential height, wind field, specific humidity, and precipitable water. Digital elevation topographic data at a 2-meter resolution was then acquired and resampled to a 9-km resolution to match the distribution density of meteorological stations.

[0024] The region's topography is characterized by a high elevation in the south and a low elevation in the north, with the eastern part bordering the sea. 614 meteorological stations are distributed throughout the region.

[0025] Step S2: Based on hourly precipitation observation data, calculate the short-term heavy precipitation intensity index for each meteorological station according to the hourly precipitation and its historical precipitation threshold, and identify the hourly precipitation that meets the preset intensity level as a short-term heavy precipitation event.

[0026] For each meteorological station, calculate the 95th quantile of its historical hourly precipitation series. And calculate all values ​​greater than Standard deviation of precipitation Calculate the short-duration heavy precipitation intensity index for each meteorological station at each time point. : ; in, This represents the hourly precipitation at the meteorological station.

[0027] When a certain weather station at a certain hour When the value is greater than or equal to 0.78, the hourly precipitation is recorded as a short-duration heavy precipitation event.

[0028] A regional short-duration heavy precipitation event is further defined as follows: if at least five meteorological stations simultaneously experience short-duration heavy precipitation at a certain moment, that moment is taken as the start moment; the event continues to be tracked until the last moment when at least five meteorological stations simultaneously experience short-duration heavy precipitation, and no further short-duration heavy precipitation occurs at any station for the next two hours, then that moment is taken as the end moment. The period between the start and end moments is defined as a regional short-duration heavy precipitation event. A total of 1855 regional short-duration heavy precipitation events were identified.

[0029] Step S3: Based on atmospheric reanalysis data, extract multiple meteorological elements at multiple altitude layers at the corresponding time of short-term heavy precipitation events, construct circulation feature vectors for each event, and use a clustering algorithm to cluster the circulation feature vectors of all events to obtain various weather pattern types.

[0030] For each regional short-duration heavy precipitation event, three key time points were selected: one hour before the start of the event, the peak time with the highest number of stations experiencing short-duration heavy precipitation during the event, and one hour after the end of the event. At each key time point, meridional and zonal wind data at 200 hPa, 500 hPa, and 850 hPa, 500 hPa geopotential height, and precipitable water were extracted, totaling eight physical quantities. Therefore, for each event, a total of... Each meteorological element field is flattened into a one-dimensional vector according to the spatial grid. Then, the 24 one-dimensional vectors are concatenated end to end to form a circulation feature vector for the event.

[0031] The circulation feature vectors of all 1855 events were used as input, and clustering was performed using the Self-Organizing Map (SOM) algorithm. The SOM network was set as follows: The topological structure, after iterative training, yielded four cluster nodes, each corresponding to a weather pattern type. Through synthetic analysis, the four types were named: coastal cyclone type (Type 1), subtropical high-controlled type (Type 2), summer front type (Type 3), and spring jet stream type (Type 4). These four types accounted for 18.17%, 33.69%, 30.30%, and 17.84% of the total events, respectively, and mainly occurred in August, July, June, and April.

[0032] Step S4: Based on digital elevation topographic data, the underlying surface of the area to be analyzed is divided into different topographic categories.

[0033] Based on their altitude, the 614 stations were divided into two categories: those with an altitude ≤ 100m were plain stations, and those with an altitude > 100m were hilly stations. For example... Figure 1 As shown, looking at the sum of all weather types ( Figure 1In (a) of the data, the frequency of short-duration heavy rainfall in hilly and plain areas shows the same trend over the next day, exhibiting a bimodal pattern of morning and afternoon peaks. However, when different weather types are separated, the four types show significant differences: Type 1 ( Figure 1 (b) and type 2 ( Figure 1 In (c) of the data, the frequency of short-duration heavy rainfall in plains is consistently greater than that in hilly areas; while in type 3 ( Figure 1 (d) and type 4 ( Figure 1 In (e) of the data, the frequency of short-term heavy rainfall is greater in hilly areas than in plains, which directly shows from the observation that topography has different effects on short-term heavy rainfall under different weather conditions.

[0034] Step S5: Using the specified weather pattern type, specified terrain category, and specified atmospheric state at a specified time as the baseline state, construct a linear hybrid attribution decomposition model to represent the frequency of short-term heavy precipitation as a linear combination of the independent and composite contributions of each influencing factor.

[0035] The baseline state is defined as: Type 1 (coastal cyclone), plain terrain, and atmospheric conditions at 00:00 (Beijing time). The frequency of short-duration heavy precipitation under this baseline state is used as the basis for the analysis. For reference only.

[0036] Based on the linear mixed attribution decomposition model: ; in, The frequency of short-duration heavy rainfall. An independent contribution to weather patterns. An independent contribution to the terrain. An independent contribution to atmospheric conditions. The combined contribution of weather patterns and topography, It contributes to the combined effects of weather patterns and atmospheric conditions. It contributes to the combined effects of topography and atmospheric conditions. It is a composite contribution of three factors: weather conditions, topography, and atmospheric state. This is the error term.

[0037] For any state (the... Weather patterns, the first Type of terrain, first (Time), the frequency of its short-duration heavy precipitation is recorded as Then, the frequency change relative to the reference state for: ; in, For weather situation type 1, plain terrain, the actual frequency of short-duration heavy precipitation at 00:00; To facilitate the solution of each contribution term from the observation data, the above equation is transformed into a difference space relative to the reference state, yielding the attribution decomposition equation in difference form: .

[0038] Step S6: Based on the linear mixed attribution decomposition model, using the actual frequency of short-term heavy precipitation under different weather conditions, different terrain types, and different atmospheric conditions at different times, calculate the quantitative contribution of each influencing factor to the frequency of short-term heavy precipitation relative to the baseline state. The relative magnitude of this contribution does not depend on the selected baseline state, that is, the quantitative attribution of the occurrence of short-term heavy precipitation does not change with the change of the baseline state.

[0039] This step uses the successive difference method to solve for each contribution in the difference-form attribution decomposition equation. The core idea is to fix other factors at the baseline state, successively change the target factor, and directly separate the contribution of the target factor by the difference in the observation frequency.

[0040] Specifically: Calculate the individual contribution of hilly terrain. : ; in, For weather pattern type 1, hilly terrain, and frequency variation at 00:00, For weather pattern type 1, hilly terrain, the actual frequency of short-duration heavy precipitation at 00:00. For weather situation type 1, plain terrain, the actual frequency of short-term heavy precipitation at 00:00.

[0041] In this embodiment, the calculated individual contribution of hilly terrain is -47.8%, which means that, compared to plains, hilly terrain alone reduces the frequency of short-duration heavy rainfall by 47.8%.

[0042] Similarly, calculate the individual contributions of atmospheric state at different times. : ; in, Weather situation type 1, plain terrain, the first The frequency change of a given time relative to the reference state. Weather situation type 1, plain terrain, the first The actual frequency of short-duration heavy rainfall at a given time. For weather situation type 1, plain terrain, the actual frequency of short-term heavy precipitation at 00:00.

[0043] Individual contributions of various weather conditions : ; in, For the first Weather patterns, plain terrain, and frequency changes at 00:00. For the first Weather patterns, plain terrain, and the actual frequency of short-duration heavy rainfall at 00:00. This refers to the sequence number of weather situation types other than the baseline weather situation type (Type 1).

[0044] The combined contribution of topography and atmospheric state : ; in, Weather situation type 1, hilly terrain, the first The frequency change of a given time relative to the reference state.

[0045] The combined contribution of weather conditions and topography : ; in, For the first Weather patterns, hilly terrain, and frequency changes at 00:00.

[0046] The combined contribution of weather conditions and atmospheric conditions : ; in, For the first Weather patterns, plains topography, and the first The frequency change of a given time relative to the reference state.

[0047] Three-factor composite contribution : ; in, For the first Weather patterns, hilly terrain, and the first The frequency change of a given time relative to the reference state.

[0048] To obtain the average contribution of each factor, the average is calculated over all times (00:00–23:00). For example, the... The average independent contribution of weather-like conditions for: ; Similarly, the total average contribution of terrain for: ; The pure independent contribution of afternoon atmospheric conditions for: ; Then calculate the pairwise composite contribution: The combined contribution of weather conditions and topography : ; The combined contribution of weather conditions and atmospheric state : ; The combined contribution of topography and afternoon atmospheric conditions : ; In this embodiment, the quantitative contributions of each factor to the frequency of short-duration heavy precipitation are calculated and shown in Table 1 (based on type 1, plain, and 00:00): Table 1. Quantitative Contribution Table of Each Factor with Type 1 as the Baseline State

[0049] In Table 1, each contribution value is a relative contribution to the reference state. A value of 0 indicates no contribution relative to the reference state, a value greater than 0 indicates a positive contribution (more favorable than the reference state), and a value less than 0 indicates a negative contribution (more unfavorable than the reference state).

[0050] The results show that in hilly areas, the combination of type 3 (summer front type) and topography is most conducive to the occurrence of short-term heavy precipitation (the total contribution of weather conditions and hills is +185.2%). At 17:00 in the afternoon, the total contribution of type 3 in the afternoon is +235.0%, which exceeds the +222.5% of type 2 and ranks first, indicating that type 3 has the strongest topographic thermal forcing effect in the afternoon.

[0051] Step S7: Model stability verification.

[0052] To demonstrate that the above contribution calculation results do not depend on the choice of the baseline state, all contribution terms were recalculated when the baseline state was changed to type 4, plain, or 00. The verification results are shown in Table 2. Table 2. Quantitative Contribution Table of Each Factor with Type 4 as the Baseline State

[0053] The results showed that the relative order of the contributions of each factor was completely consistent with the previous findings: the individual contribution of weather conditions was consistently Type 3 > Type 2 > Type 1 > Type 4; the total contribution of hilly conditions was consistently Type 3 > Type 4 > Type 1 > Type 2; and the total contribution of afternoon conditions was consistently Type 3 > Type 2 > Type 1 > Type 4. This demonstrates the robustness of the linear mixed attribution decomposition model and the universality of its quantitative conclusions.

[0054] Step S8: Application of Results.

[0055] Based on the above attribution analysis results, the dominant factors influencing short-duration heavy precipitation under different weather conditions can be identified. Under the subtropical high-controlled type (Type 2), short-duration heavy precipitation in plains areas is mainly driven by afternoon thermal instability; under the summer frontal type (Type 3), the dynamic uplift of hilly terrain and afternoon thermal forcing work together to result in the highest frequency of short-duration heavy precipitation. These quantitative conclusions can provide a scientific basis for the forecasting and early warning of short-duration heavy precipitation in this region, and can also be extended to attribution studies of short-duration heavy precipitation in other complex terrain areas.

[0056] It is worth noting that all contents not described in detail in this invention are existing technologies and are well known to those skilled in the art.

[0057] Therefore, the present invention adopts the above-mentioned multi-factor quantitative attribution method for weather events based on linear mixture model decomposition, which can quantitatively separate the contributions of weather conditions, topography, atmospheric state and the combined effects of various factors to the frequency of short-term heavy precipitation from observation data, and provide a scientific basis for the attribution analysis and forecasting and early warning of short-term heavy precipitation.

[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A multi-factor quantitative attribution method for weather events based on linear mixture model decomposition, characterized in that, Includes the following steps: Step S1: Obtain hourly precipitation observation data, atmospheric reanalysis data, and digital elevation topographic data of meteorological stations in the area to be analyzed; Step S2: Based on hourly precipitation observation data, calculate the short-term heavy precipitation intensity index of each meteorological station according to the hourly precipitation and its historical precipitation threshold, and identify the hourly precipitation that meets the preset intensity level as a short-term heavy precipitation event. Step S3: Based on atmospheric reanalysis data, extract multiple meteorological elements at multiple altitude layers at the time corresponding to short-term heavy precipitation events, construct circulation feature vectors for each event, and use a clustering algorithm to cluster the circulation feature vectors of all events to obtain various weather pattern types. Step S4: Based on digital elevation topographic data, the underlying surface of the area to be analyzed is divided into different topographic categories; Step S5: Using the specified weather pattern type, specified terrain category, and atmospheric state at the specified time as the baseline state, construct a linear hybrid attribution decomposition model to represent the frequency of short-term heavy precipitation as a linear combination of the independent contribution terms and composite contribution terms of each influencing factor. Step S6: Based on the linear mixed attribution decomposition model, calculate the quantitative contribution of each influencing factor to the frequency of short-term heavy precipitation by using the actual occurrence frequency of short-term heavy precipitation under different weather conditions, different terrain categories, and different atmospheric conditions at different times.

2. The method for quantitative attribution of multiple factors in weather events based on linear mixture model decomposition according to claim 1, characterized in that, In step S2, based on hourly precipitation observation data, and according to the hourly precipitation at each meteorological station and its historical precipitation thresholds, the short-term heavy precipitation intensity index is calculated for each meteorological station, including: The 95th percentile of the historical hourly precipitation at each meteorological station is taken as the precipitation threshold for that meteorological station. Calculate the standard deviation of precipitation exceeding the precipitation threshold in the historical hourly precipitation data for each meteorological station; Calculate the short-duration heavy precipitation intensity index: ; in, This is the short-duration heavy rainfall intensity index. This refers to the hourly precipitation at the meteorological station. This is the 95th percentile of the station's historical precipitation. greater than The standard deviation of precipitation at each station.

3. The method for quantitative attribution of multiple factors in weather events based on linear mixture model decomposition according to claim 1, characterized in that, Step S3 involves constructing the circulation feature vectors for each event, including: For each regional short-duration heavy precipitation event, atmospheric reanalysis data were extracted from three key time points: one hour before the event, the peak time with the most short-duration heavy precipitation stations, and one hour after the event ended. At each key time point, the meridional and zonal winds at 200hPa, 500hPa, and 850hPa, the geopotential height at 500hPa, and the precipitable water data were extracted. The 24 meteorological elements from the three key time periods were flattened and spliced ​​together according to the same rule to form a one-dimensional feature vector, which served as the circulation feature vector of this regional short-term heavy precipitation event.

4. The method for quantitative attribution of multiple factors in weather events based on linear mixture model decomposition according to claim 1, characterized in that, The clustering algorithm in step S3 is the self-organizing map algorithm.

5. The method for quantitative attribution of multiple factors in weather events based on linear mixture model decomposition according to claim 1, characterized in that, The linear mixed attribution decomposition model in step S5 is as follows: ; in, The frequency of short-duration heavy rainfall, An independent contribution to weather patterns. An independent contribution to the terrain. An independent contribution to atmospheric conditions. The combined contribution of weather patterns and topography, It contributes to the combined effects of weather patterns and atmospheric conditions. It contributes to the combined effects of topography and atmospheric conditions. It contributes to the combined effects of weather conditions, topography, and atmospheric state. This is the error term.

6. The method for quantitative attribution of multiple factors in weather events based on linear mixture model decomposition according to claim 5, characterized in that, In step S6, the quantitative contribution value includes the independent contribution and composite contribution of each influencing factor. The composite contribution includes pairwise composite contribution and three-factor composite contribution. Both the independent contribution and the composite contribution are determined based on the frequency change, which is the difference between the actual occurrence frequency of short-term heavy precipitation and the actual occurrence frequency under the baseline state.

7. The method for quantitative attribution of multiple factors in weather events based on linear mixture model decomposition according to claim 6, characterized in that, The independent contributions of each influencing factor include: No. Average independent contribution of weather conditions at any given time: ; No. Total average contribution of terrain at time: ; No. The purely independent contribution of the atmospheric state at any given time: ; in, For the first The total average contribution of weather-like phenomena on a diurnal scale. This represents the total average contribution of non-baseline topography on a diurnal variation scale. For the first The purely independent contribution of atmospheric state at any given moment. These are the sequence numbers for weather situation types other than the baseline weather situation type. For time sequence number, , As the baseline terrain, This is a non-baseline terrain. For the first The individual contribution of weather-like patterns, The separate contribution of non-baseline terrain. For the first Weather patterns and the first The combined contribution of atmospheric conditions at any given time. For non-baseline terrain and the first The combined contribution of atmospheric conditions at any given time. For the first Weather patterns, baseline topography, and the first The frequency change of a given time relative to the reference state. For baseline weather situation type, baseline topography, and the first The frequency change of a given time relative to the reference state. For baseline weather pattern type, non-baseline terrain, the first The frequency change of a given time relative to the reference state. For baseline weather situation type, baseline topography, and the first The actual frequency of short-duration heavy rainfall at a given time. This includes the baseline weather pattern type, baseline topography, and the actual frequency of short-duration heavy precipitation at 00:

00.

8. The method for quantitative attribution of multiple factors in weather events based on linear mixture model decomposition according to claim 7, characterized in that, The pairwise composite contributions include: The combined contribution of weather conditions and topography: ; Weather conditions and the The total composite contribution of atmospheric state at time: ; Topography and the The total composite contribution of atmospheric state at time: ; in, For the first The combined contribution of weather-like conditions and non-baseline terrain. For the first Weather patterns and the first The total composite contribution of atmospheric state at any given moment, For non-baseline terrain and the first The total composite contribution of atmospheric state at any given moment, For the first The second-order composite contribution of weather-like conditions and non-baseline terrain. For the first Weather patterns, non-baseline terrain and the first The combined contribution of three factors related to atmospheric conditions at any given time. For the first Weather patterns, non-baseline terrain, and the first The frequency change of a given time relative to the reference state. For the first Weather patterns, baseline topography, and the first The frequency change of a given time relative to the reference state. For baseline weather pattern type, non-baseline terrain, the first The frequency change of a given time relative to the reference state. For baseline weather situation type, baseline topography, and the first The frequency change of a given time relative to the reference state. For the first Weather-like conditions, baseline terrain, and frequency changes at time 00 relative to the baseline. This includes the baseline weather pattern type, the non-baseline terrain, and the frequency change of time 00 relative to the baseline. For the first Weather-like conditions, non-baseline terrain, and frequency changes at time 00 relative to the baseline state.

9. The method for quantitative attribution of multiple factors in weather events based on linear mixture model decomposition according to claim 8, characterized in that, Three-factor composite contribution for: ; in, For the first Weather patterns, baseline topography, and frequency changes at 00:

00.

10. The method for quantitative attribution of multiple factors in weather events based on linear mixture model decomposition according to claim 6, characterized in that, The method also includes a verification step: changing the baseline setting and repeatedly calculating the quantitative contribution value of each influencing factor. If the relative order of the contribution values ​​of each factor remains unchanged, the linear mixed attribution decomposition model is determined to be stable.