Attribution analysis method for cold season composite low temperature and heavy snow events based on circulation classification

Abstract

The present invention provides an attribution analysis method for cold season composite low temperature and heavy snow events based on circulation classification, including: based on the collected measured meteorological and hydrological data, global climate model data and atmospheric circulation data of the study area, bias correction calculation of the climate model and snowfall calculation are performed; self-organizing map neural network is used to perform circulation classification on atmospheric circulation variables, identify composite extreme low temperature and heavy snow events under different circulation types and perform event feature analysis; based on the frequency change of circulation type, regularized optimal fingerprint method is used to perform external forced detection attribution analysis and constraint prediction analysis of global climate models in the future period; climate change separation method is used to separate thermal effects, dynamic effects and thermal-dynamic interactions through the frequency change trend of composite extreme events. The present invention provides a new perspective for the attribution of climate change and human activity impacts on composite extreme climate events.

Description

Technical Field

[0001] The present invention relates to the technical field of climate data processing, and in particular to an attribution analysis method for cold season composite low-temperature heavy snow events based on circulation classification. Background Art

[0002] As global warming intensifies, extreme weather and climate events occur more frequently, more intensely and over a wider range. These changes have brought unprecedented challenges to human production and life as well as natural ecosystems. In recent years, many scholars have conducted extensive and in-depth characteristic analysis and risk assessment of high-impact climate extreme events caused by warm seasons (such as extreme droughts, heat waves, heavy rains and floods). Nevertheless, in the context of global warming, cold season extreme climate events have received relatively less attention than their warm season counterparts.

[0003] Extreme low temperatures and rain, snow and freezing are the two main natural disasters that affect human production and life in the cold season. The two sometimes occur at the same time. The impact of compound low temperature and heavy snow extreme events on ecosystems and human society is often more serious. In addition, at present, the attribution analysis of extreme low temperature and heavy snow events is mostly conducted from the perspective of climate variables such as temperature and precipitation, and the influence of atmospheric circulation patterns is often ignored. Therefore, based on the atmospheric circulation classification, analyzing the thermodynamic and kinetic effects that affect cold season low temperature and heavy snow events, and accurately assessing the risks of different types of cold season low temperature and heavy snow events can improve people's understanding of cold season low temperature and heavy snow events, thereby strengthening the risk prevention and control of extreme low temperature and heavy snow events, which is of great significance. Summary of the invention

[0004] The present invention provides an attribution analysis method for cold season composite low temperature and heavy snow events based on circulation typing, so as to solve the defect of ignoring the influence of atmospheric circulation typing in the attribution analysis of low temperature and heavy snow extreme events in the prior art.

[0005] In a first aspect, the present invention provides an attribution analysis method for cold season composite low temperature and heavy snow events based on circulation classification, comprising:

[0006] Collect measured meteorological and hydrological data, global climate model data and atmospheric circulation data in the study area, and conduct climate model evaluation, bias correction calculation and snowfall calculation of global climate model data;

[0007] The self-organizing map network method is used to classify the atmospheric circulation variables, identify the extreme low temperature and heavy snow events under different circulation types, and analyze the event characteristics of the extreme low temperature and heavy snow events;

[0008] Based on the frequency changes of circulation patterns, the regularized optimal fingerprint method is used to perform attribution analysis of external forcing detection. Based on the attribution scale factors of historical periods, constrained predictions of global climate patterns for future periods are made. The thermal effects, dynamic effects, and thermal-dynamic interactions are separated through the frequency change trends of complex extreme events.

[0009] According to an attribution analysis method for cold season composite low temperature and heavy snow events based on circulation classification provided by the present invention, measured meteorological and hydrological data, global climate model data and atmospheric circulation data of the study area are collected, and climate model evaluation, bias correction calculation and snowfall calculation of global climate model data are performed, including:

[0010] Collect CN05.1 grid data set, CMIP6 global climate model data, and ERA5, JRA-55, MERRA2 and NCEP atmospheric reanalysis data sets, and use bilinear interpolation to interpolate the CN05.1 grid data set, the CMIP6 global climate model data, and the ERA5, JRA-55, MERRA2 and NCEP atmospheric reanalysis data sets to a preset resolution grid;

[0011] Six different CMIP6 model data were used to evaluate the climate model based on the measured temperature and precipitation data of the CN05.1 grid data set as the benchmark data set. Cold season data were extracted, and four extreme temperature indices and four extreme precipitation indices recommended by the climate change index ETCCDI were selected. The Taylor comprehensive index was used to comprehensively evaluate the average climate state and temporal variation of temperature and precipitation simulated by the cold season climate model in the historical period.

[0012] Based on PQM and QDM, PQDM is obtained, and PQDM is used to correct the bias of global climate model data;

[0013] Using a dynamic threshold parameterization scheme and combining the melting process of snowfall to the ground, the high and low temperature thresholds are updated according to the altitude and meteorological conditions of each grid point to divide the total precipitation into snowfall, rainfall, and the intermediate amount of snowfall and rainfall.

[0014] According to an attribution analysis method for cold season composite low temperature and heavy snow events based on circulation classification provided by the present invention, PQDM is obtained based on PQM and QDM, and PQDM is used to perform bias correction on global climate model data, including:

[0015] Based on the POT and the 95th percentile threshold, the meteorological time series corresponding to the global climate model data is divided into extreme values ​​and non-extreme values;

[0016] performing bias correction on the non-extreme value using an empirical cumulative distribution function, wherein the empirical cumulative distribution function is determined by the QDM;

[0017] Using GP as a transfer function to fit the extreme value, and using the maximum likelihood method to estimate the parameters of the extreme value;

[0018] The QDM is used to make incremental adjustments and PQDM corrections are performed on the hist and hist-nat simulations of the CN05.1 grid data set and the CMIP6 global climate model data.

[0019] According to the attribution analysis method of cold season composite low temperature heavy snow events based on circulation classification provided by the present invention, a dynamic threshold parameterization scheme is used, combined with the melting process of snowfall falling to the ground, and the high temperature threshold and low temperature threshold are updated according to the altitude and meteorological conditions of each grid point, and the snowfall, rainfall and the intermediate amount of snowfall and rainfall in the total precipitation are divided, including:

[0020] Based on the near-surface relative humidity, the dry-bulb temperature and the altitude, the high temperature threshold and the low temperature threshold are obtained;

[0021] The wet-bulb temperature is obtained by an empirical formula, and the snowfall, the rainfall, and the intermediate amount of snowfall and rainfall are obtained by using the wet-bulb temperature, the high temperature threshold, and the low temperature threshold.

[0022] According to an attribution analysis method for cold season composite low temperature and heavy snow events based on circulation typing provided by the present invention, a self-organizing mapping network method is used to perform circulation typing on atmospheric circulation variables, identify extreme low temperature and heavy snow events under different circulation types, and perform event feature analysis on the extreme low temperature and heavy snow events, including:

[0023] Determine the consistency and reliability of atmospheric reanalysis products ERA5, JRA-55, NCEP and MERRA2, and compare and analyze winter sea level pressure, 500hPa geopotential height and 300hPa wind speed;

[0024] Self-organizing map neural network was used to conduct sensitivity test of circulation typing;

[0025] Based on the sensitivity test of circulation typing of self-organizing map neural network, the combination of circulation variables and the hyperparameters of circulation typing of self-organizing map neural network are determined to obtain different circulation types after cluster typing, and the characteristics of the different circulation types after cluster typing are analyzed to obtain the circulation analysis results;

[0026] According to the circulation analysis results, based on the CN05.1 measured temperature and precipitation data, the climate variables corresponding to each circulation type of the self-organizing mapping neural network are extracted to identify and extract the extreme low temperature and heavy snow events.

[0027] According to an attribution analysis method for cold season composite low temperature and heavy snow events based on circulation typing provided by the present invention, a circulation typing sensitivity test is performed using a self-organizing map neural network, comprising:

[0028] Step 1: Initialize the self-organizing map neural network and normalize each neuron in the network output layer using the initialization weight vector;

[0029] Step 2: Calculate the distance between the input vector and each weight vector to get the winning neuron;

[0030] Step 3: Determine a neighborhood function according to the network topology, and use the neighborhood function to update the weights of the winning neuron and the neurons in the neighborhood of the winning neuron;

[0031] Step 4: Repeat steps 2 to 4 until the network learning rate is determined to be less than the learning rate threshold and the network reaches convergence, map each input vector to the winning neuron, and obtain the input data clustering result.

[0032] According to the attribution analysis method of cold season composite low temperature and heavy snow events based on circulation classification provided by the present invention, the regularized optimal fingerprint method is used to perform attribution analysis of external forced detection based on the frequency change of circulation type, and the global climate model of the future period is constrained and estimated based on the attribution scale factor of the historical period. The thermal effect, dynamic effect and thermal-dynamic interaction are separated through the frequency change trend of composite extreme events, including:

[0033] The regularized optimal fingerprint method and total least squares method are used to detect and attribute the changes in circulation pattern frequency under climate change;

[0034] Based on the optimal estimation results of the attribution scale factors in the historical period, an attribution constraint estimation analysis is conducted on the frequency changes of circulation patterns in the future period;

[0035] Based on the self-organizing map neural network clustering algorithm, the main atmospheric circulation patterns are identified and the contribution of thermodynamic and dynamic changes to the changes in extreme events is determined.

[0036] According to the attribution analysis method of cold season composite low temperature and heavy snow events based on circulation classification provided by the present invention, based on the self-organizing map neural network clustering algorithm, the main atmospheric circulation patterns are identified, and the contribution of thermodynamic and dynamic changes to extreme event changes is determined, including:

[0037] Thermodynamic-dynamic separation method is used to separate and quantify the relative contributions of thermal and dynamic effects to the changing trend of the frequency of cold season composite low-temperature heavy snow events.

[0038] Based on the time mean of the frequency of compound events and the frequency of any circulation pattern, as well as the deviation of the frequency of compound events and the time mean of the frequency of any circulation pattern, the time division is carried out;

[0039] By taking the derivative with respect to time, the relative contributions of the thermal effect, dynamic effect and thermal-dynamic interaction related to any circulation pattern can be obtained from the trend of the frequency of composite low-temperature heavy snow events.

[0040] In a second aspect, the present invention further provides an attribution analysis system for cold season composite low temperature and heavy snow events based on circulation classification, comprising:

[0041] The data processing module is used to collect measured meteorological and hydrological data, global climate model data and atmospheric circulation data in the study area, and to perform climate model evaluation, bias correction calculation and snowfall calculation on global climate model data;

[0042] A circulation classification module is used to classify atmospheric circulation variables using a self-organizing mapping network method, identify extreme low temperature and heavy snow events under different circulation types, and analyze event characteristics of the extreme low temperature and heavy snow events;

[0043] The attribution analysis module is used to perform attribution analysis of external forcing detection based on the frequency changes of circulation patterns using the regularized optimal fingerprint method. It constrains the prediction of global climate patterns in the future based on the attribution scale factors of historical periods, and separates thermal effects, dynamic effects, and thermal-dynamic interactions through the frequency change trends of composite extreme events.

[0044] In a third aspect, the present invention also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, an attribution analysis method for cold season composite low temperature and heavy snow events based on circulation typing as described in any one of the above-mentioned methods is implemented.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] 1. Based on global climate model data and atmospheric circulation data, the present invention improves the deviation correction method of global climate model data, so that the evaluation and optimization of global climate model data are more in line with the needs of practical applications.

[0047] 2. Based on the dual-variable dynamic threshold parameterization scheme, the present invention calculates the rainfall and snowfall in the cold season, and then identifies the cold season compound extreme low temperature and heavy snow events, on this basis, enhancing the understanding of the characteristics of compound extreme low temperature and heavy snow events;

[0048] 3. This invention is the first to propose an attribution analysis method and system for cold season compound low temperature and heavy snow events based on the self-organizing map neural network circulation typing method, combined with the regularized optimal fingerprint method and the thermal-dynamic action separation method, providing a new perspective on the attribution of climate change and the impact of human activities on compound low temperature and heavy snow events. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0050] Figure 1 This is one of the flow diagrams of the attribution analysis of cold season compound low temperature and heavy snow events based on circulation classification provided by the present invention;

[0051] Figure 2 This is the second flow diagram of the attribution analysis of cold season compound low temperature and heavy snow events based on circulation classification provided by the present invention;

[0052] Figure 3 It is a simulation evaluation diagram of the CMIP6 climate model provided by the present invention;

[0053] Figure 4 It is a schematic diagram of the climate model simulation effect before and after the bias correction using the PQDM method provided by the present invention;

[0054] Figure 5 It is a result diagram of the SOM circulation typing sensitivity test provided by the present invention;

[0055] Figure 6 It is a SOM circulation characteristic analysis result diagram provided by the present invention;

[0056] Figure 7 It is a characteristic analysis result diagram of low temperature and heavy snow events under various circulation patterns provided by the present invention;

[0057] Figure 8 It is the attribution result diagram of the optimal fingerprint method of circulation type frequency in historical periods provided by the present invention;

[0058] Fig. 9 The present invention provides a spatial distribution map of the frequency change level of low temperature and heavy snow event trends in the historical period and future circulation patterns before and after attribution constraints;

[0059] Fig.10 This is a graph showing the decomposition results of the thermal and dynamic effects of a composite extreme low temperature and heavy snow event provided by the present invention;

[0060] Fig.11 It is a structural schematic diagram of the attribution analysis system for cold season composite low temperature and heavy snow events based on circulation classification provided by the present invention;

[0061] Fig.12 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] Figure 1 This is one of the flow diagrams of attribution analysis of cold season composite low temperature and heavy snow events based on circulation classification provided by an embodiment of the present invention, such as Figure 1 As shown, including:

[0064] Step 100: Collect measured meteorological and hydrological data, global climate model data and atmospheric circulation data of the study area, and perform climate model evaluation, bias correction calculation and snowfall calculation of global climate model data;

[0065] Step 200: using a self-organizing mapping network method to classify atmospheric circulation variables, identifying extreme low temperature and heavy snow events under different circulation types, and performing event feature analysis on the extreme low temperature and heavy snow events;

[0066] Step 300: Based on the frequency changes of circulation patterns, the regularized optimal fingerprint method is used to perform attribution analysis of external forcing detection. Based on the attribution scale factors of the historical period, the global climate model for the future period is constrained and estimated. The thermal effect, dynamic effect and thermal-dynamic interaction are separated through the frequency change trend of composite extreme events.

[0067] Specifically, the event identification and attribution analysis method of cold season composite low temperature and heavy snow events based on circulation classification is proposed in the embodiment of the present invention, such as Figure 2 As shown, the following steps are included:

[0068] Step 1: Based on the measured meteorological and hydrological data, global climate model data and atmospheric circulation data of the study area, bias correction calculations of global climate model data and snowfall calculations are performed.

[0069] Step 1 further includes the following sub-steps:

[0070] Step 1.1: Collection of basic meteorological and hydrological data. The measured meteorological data uses the CN05.1 grid data set, the global climate model data uses the CMIP6 data, and the four atmospheric reanalysis data sets of ERA5, JRA-55, MERRA2, and NCEP are used as atmospheric reanalysis data sets. The bilinear interpolation method is used to interpolate the measured meteorological data, climate model data, and atmospheric reanalysis data to the preset resolution grid.

[0071] The measured meteorological data CN05.1 is a grid data set obtained by interpolating observation data from more than 2,400 stations in China using the anomaly approximation method, with a spatial resolution of 0.25°×0.25°; the DEM elevation data is selected from the 90m elevation data of SRTMDM (Shuttle Radar Topography Mission); the surface pressure data is selected from the surface pressure data of ERA5 (ECMWF Reanalysis v5); the global climate models (GCMs) are from CMIP6, using the experimental data of "historical" for full forced simulation in historical periods, "hist-nat" for natural forced simulation only, and "ssp245" and "ssp245-nat" for future periods. The r1i1p1f1 experimental data represent the real-world simulation and the counterfactual simulation under natural forced only; through CDO (Climate Data The bilinear interpolation technique in the CMIP6 Scenario Interpolation Operators is used to interpolate the data described in the present invention to the same resolution grid of 0.25°×0.25°. The time length of 1961 / 1 / 1-2020 / 12 / 31 is uniformly used in the historical period, and the time length of 2021 / 1 / 1-2100 / 12 / 31 is used in the future period. The "ssp245" of the future medium development path SSP245 scenario of the "historical" scenario of CMIP6 supplements the data from 2015 to 2020, which is called "hist" in the present invention, and the time resolution is daily.

[0072] Step 1.2: Evaluation of global climate model data. Six different CMIP6 model data were used to evaluate the climate model based on the CN05.1 measured temperature and precipitation data as the benchmark data set, and cold season data were extracted, where the cold season is defined as five months from November to December of each year to January to March of the following year. Four extreme temperature indices and four extreme precipitation indices recommended by ETCCDI (Expert Team on Climate Change Detection and Indices) were selected to comprehensively evaluate the ability of the cold season climate model to simulate the average climate state and temporal changes of temperature and precipitation in the historical period.

[0073] Table 1 shows the selected global climate model data and experimental scenario information

[0074]

[0075] Table 2 shows the extreme temperature index and extreme precipitation index information recommended by ETCCDI

[0076]

[0077] In order to compare the differences between the simulations of cold season temperature and precipitation by different global climate models and the measured reference data, the Taylor comprehensive index of climate mean state and time variation is selected for comprehensive evaluation. The Taylor index is derived from the Taylor diagram, which combines the correlation coefficient, standard deviation and central root mean square error. The calculation formula of the Taylor comprehensive index (TS) is as follows:

[0078]

[0079] Where R is the correlation coefficient, and Rmax is the maximum correlation coefficient. The closer TS is to 1, the better the simulation performance.

[0080] Correlation coefficient (R):

[0081]

[0082] Central Root Mean Square Error (C-RMSE):

[0083]

[0084] Standard Deviation:

[0085]

[0086]

[0087] Where x is the observed field and y is the simulated field; when evaluating the climate mean state, i represents different spatial grid points and N is the total number of regional grid points; when evaluating temporal changes, i represents different years and N is the total number of years;

[0088] China is divided into 9 sub-regions according to the characteristics of its eco-geographical system, and the comprehensive Taylor score of each sub-region is evaluated separately. The detailed information of each sub-region is shown in the table below.

[0089] Table 3 shows the specific information of each sub-region in China.

[0090]

[0091] The Taylor Composite Index was used to evaluate the ability of the six CMIP6 climate models to simulate the spatial climate mean and time series of the cold season temperature and precipitation in each sub-region of China. Due to the different units of the extreme climate indices, the values ​​of the six models were normalized, and the means of the four extreme temperature indices and the four extreme precipitation indices were calculated respectively. The evaluation results are shown in Figure 3 , where the closer the TS value is to 1, the better the mode performance is.

[0092] Step 1.3: Bias correction of global climate model data. In addition to random errors, there are certain systematic deviations between climate model data and the actual climate state. Moreover, the resolution of global climate model data is generally coarse, and it is difficult to directly apply it to regional-scale climate change and its impact research. Based on the PQM (Piecewise-Quantile Mapping method) method proposed by Zhang et al. in 2022, the present invention proposes a piecewise quantile delta mapping method (PQDM, Piecewise-Quantile Delta Mapping method) for bias correction of climate model data, which comprehensively considers the advantages of the QDM method and the PQM method in sequence inconsistency and extreme value correction, respectively.

[0093] The QDM method attempts to create a transfer function that makes the simulated cumulative distribution function (CDF) better match the observed CDF. The QDM method preserves the relative changes in the model across quantiles by adding increments (for air temperature correction) or multiplying increments (for precipitation correction). The transfer function and increments are defined as follows:

[0094]

[0095] Where x bc (t) is the correction value, x mf (t) is the original model output; F m,f and denote the CDF and inverse function of the model output for the future and historical periods, respectively, Represents the inverse function of the observed values ​​of climate variables in the historical period, Δ m (t) represents the future increment of the model relative to the historical period.

[0096] First, based on the peak over threshold (POT) method and the 95th percentile threshold, the meteorological time series is divided into extreme value and non-extreme value parts. For the non-extreme value part, the empirical cumulative distribution function is used for bias correction. For the extreme value part, the generalized Pareto distribution (GP) is used as the transfer function to fit the extreme value, and the maximum likelihood method is used for parameter estimation. The POT / GP model has been widely used in extreme flood and precipitation frequency analysis. The cumulative probability density distribution function of GP is as follows:

[0097]

[0098] where x represents the climate variable, k represents the shape parameter, u represents the location parameter, and σ represents the scale parameter.

[0099] Subsequently, the QDM method is used for incremental adjustments. PQDM corrections are performed on the CN05.1 meteorological observations and the hist and hist-nat simulations of CMIP6 to make the CDFs of the hist and hist-nat scenarios consistent with the observations while retaining the differences between the hist and hist-nat scenarios. The PQDM method can also be used to perform corrections using future counterfactual scenarios (ssp245-nat) based on bc_ssp245 (PQDM bias-corrected model data) and ssp245 simulations.

[0100] like Figure 4 As shown, after bias correction, the uncertainties of the six GCMs are reduced and the distribution of daily extreme climate variables in the historical simulations is captured, the differences between the factual and counterfactual scenarios in the historical period, and the trends of the future scenarios are also preserved.

[0101] Step 1.4: Use the dynamic threshold parameterization scheme to perform snowfall analysis and calculation. The two-variable dynamic threshold scheme (Dynamic Threshold Parameterization Scheme, DTPS) of precipitation phase state published by Ding et al. in Journal of Hydrology in 2014 is used to consider the melting process of snowfall falling to the ground, and update the high and low threshold temperatures according to the altitude and meteorological conditions of each grid point to divide the snowfall and rainfall in the total precipitation. The calculation formula is as follows:

[0102]

[0103] Where T w Wet bulb temperature, calculated using empirical formula; T max and T min are two threshold temperatures, and the calculation formulas are as follows:

[0104]

[0105] ΔS=2.374-1.634×RH

[0106] ΔT=0.215-0.099×RH+1.018×RH 2

[0107] T 0 =-5.87-0.1042×Z+0.0885×Z 2 +16.06×RH-9.614×RH 2

[0108] Where T is the dry bulb temperature (°C), RH is the near-surface relative humidity (%), and Z is the altitude (km).

[0109] The present invention uses the wet_bulb_temperature function of the metpy library in Python to calculate the wet bulb temperature, and uses the Normand's rule method to calculate the wet bulb temperature according to the surface air pressure, the dry bulb temperature, and the dew point temperature; wherein the dew point temperature Td is calculated from the dry bulb temperature and the relative humidity according to the dewpoint_from_relative_humidity function. The present invention evenly distributes sleet into rain and snow when using DTPS.

[0110] Step 2: Based on the four sets of atmospheric reanalysis data sets, the spatiotemporal distribution of atmospheric circulation variables of the four sets of atmospheric reanalysis data sets are compared, and then the appropriate atmospheric reanalysis data set is selected for SOM (self-organizing map network) circulation typing, extreme low temperature and heavy snow events under different circulation types are identified, and event characteristics are analyzed.

[0111] Step 2 further includes the following sub-steps:

[0112] Step 2.1: Comparison of four atmospheric reanalysis datasets. Four atmospheric reanalysis products were selected: ERA5, JRA-55, NCEP, and MERRA2. ERA5 is the fifth generation of atmospheric reanalysis datasets of the global climate from January 1950 to the present by the European Centre for Medium-Range Weather Forecasts (ECMWF); JRA-55 is the reanalysis data released by the Japan Meteorological Agency (JMA) in 2013; the NCEP / NCAR reanalysis dataset is jointly produced by the National Center for Meteorological and Environmental Prediction (NCEP) and the National Center for Atmospheric Research (NCAR); MERRA-2 is an atmospheric reanalysis data product released by the Global Modeling and Assimilation Office (GMAO) of the Goddard Space Flight Center (GSFC) of the National Aeronautics and Space Administration (NASA). The detailed information of the four atmospheric reanalysis product datasets is shown in Table 4.

[0113] This invention first tests the consistency and reliability of winter sea level pressure (SLP), 500hPa geopotential height (Z500), and 300hPa wind speed (W300) of four sets of reanalysis data. SLP, Z500, and W300 capture the information of near-surface, middle-troposphere, and upper-troposphere circulations, respectively. This paper selects the Chinese region as the research object, and the atmospheric circulation research area is selected from 0 to 70°N, 40°E to 160°E. The selection of the atmospheric circulation research area takes into account the activity range of various circulation systems, such as the Ural Mountain blocking high pressure, the Okhotsk blocking high pressure, the western Pacific subtropical high pressure, and the tropical circulation in the Northern Hemisphere. The sea level pressure field (hPa), 500hPa geopotential height field (dagpm), and 300hPa wind field (m 3 / s) in the selected area during the cold season (November, December, January, February, and March) from 1961 to 2020.

[0114] Table 4 shows the information of four atmospheric reanalysis datasets.

[0115]

[0116]

[0117] Step 2.2: Use the self-organizing map (SOM) neural network to conduct a circulation typing sensitivity test.

[0118] SOM is an unsupervised artificial neural network clustering method that can map high-dimensional input data to a low-dimensional output space, thereby preserving and visualizing the topological structure of the input vector. The specific steps are as follows:

[0119] (1) Initialize the network

[0120] For each neuron in the output layer, initialize a weight vector and normalize it:

[0121] W = {ω ij :j=1,…N;i=1,…D}

[0122] Where N is the number of nodes, D is the latitude, W is the weight of each neuron, ω ij Represents the weight of each node.

[0123] (2) Calculate the winning neuron.

[0124] Calculate the input vector The distance d between each weight vector j , find the neuron with the closest distance, which is the winning neuron j * :

[0125]

[0126] d j (x) represents the distance between the input vector and each weight vector.

[0127] (3) Update the weights of the winning neuron and its neighboring neurons

[0128] Define a neighborhood function based on the topological structure of the network Update the weights of the winning neuron and the neurons in its neighborhood:

[0129]

[0130] Where: ij (t) and ω ij (t+1) is the weight of the jth neuron in the ith node at training time t and t+1, respectively, η(t) is the learning rate at training time t, is the pth input vector value in the i-th node.

[0131] (4) Iteration and convergence judgment

[0132] Repeat steps (2) to (4) until the learning rate η(t) is less than the threshold η min When , the network reaches convergence. By mapping each input vector to the winning neuron, the clustering result of the input data can be obtained.

[0133] The present invention performs SOM cluster analysis based on the MiniSOM library in Python. First, the atmospheric circulation variables are normalized by subtracting the multi-year average of each grid and dividing by the standard deviation, and then multiplied by the square root of the latitude cosine for area weighting, and then the normalized and area-weighted SLP, Z500 and W300 data are used as the input data of SOM.

[0134] Firstly, the sensitivity test of circulation variables was carried out, and the circulation variable combinations were made according to SLP, Z500 and W300, including ①SLP, ②Z500, ③W300, ④SLP+Z500, ⑤SLP+W300, ⑥Z500+W300, and ⑦SLP+Z500+W300.

[0135] In addition, it is also necessary to optimize the hyperparameters of the SOM network, mainly including: som_shape, sigma, learning_rate, epochs, initiate_method, train_method, etc., and it is necessary to test different SOM hyperparameters and classification node sizes: 1×3, 1×4, 2×2, 1×5, 2×3, 2×4, 3×3, 2×5, 3×4, 3×5, 4×4, 3×6, 4×5, etc. The present invention uses the Hyperopt library in python to perform Bayesian optimization of hyperparameters. The objective function combines the topological error (TE) and quantization error (QE) of SOM network training, and the calculation formula is as follows:

[0136] loss=0.01×QE(data)+TE(data)

[0137] Where QE is the average Euclidean distance of the SOM node relative to the input data, indicating the size of the intra-cluster error of each neuron; TE represents the percentage of non-adjacent neurons of the first matching neuron and the second matching neuron of the input data, reflecting the topological quality of the SOM network. The smaller the QE and TE values, the higher the quality of the SOM circulation typing. The results of the SOM circulation typing sensitivity test are shown in Figure 5 .

[0138] Step 2.3: Based on the SOM circulation typing sensitivity test, the SLP+Z500 circulation variable combination was selected, and the 2×2 typing node size was selected, where "neighborhood_function" was selected as "gaussian", "topology" adopted the "hexagonal" hexagonal topology structure, "initiate_method" adopted "random", and "train_method" adopted "batch". The iterative calculation was performed 5000 times to obtain 4 different circulation types after cluster typing. Then the characteristic analysis of the optimal circulation type was carried out, mainly analyzing the frequency of the circulation type (the number of times each type of circulation occurs each year), the average duration (the average duration of each type of circulation each year, days), the maximum duration (the maximum duration of each type of circulation each year, days), and the time distribution of each type of circulation type within the season.

[0139] The results of SOM circulation classification and characteristic analysis are shown in Figure 6 The results of circulation classification show that the correlation coefficient between the daily atmospheric circulation field and the corresponding SOM circulation type is relatively large, while the correlation coefficient between different circulation types is relatively small, indicating that the number of SOM classification nodes is more appropriate. Figure 6, CP1 and CP3 circulation types account for 42.34% and 43.36% of the circulation types in China's cold season, respectively, which is a very large proportion. CP1 is the dominant circulation type in January and December of the cold season, and CP3 is the dominant circulation type in March. When the CP1 circulation type occurs, the Chinese region is mainly positive sea level pressure anomaly, while the entire Asian circulation area is occupied by negative geopotential height anomaly. When the CP3 circulation type occurs, the Chinese region is mainly negative sea level pressure anomaly, while the Asian circulation area is mainly positive geopotential height anomaly. The spatial distribution of the CP2 and CP3 circulation types presents an opposite state, and the occurrence time of this circulation type is basically complementary to that of CP3, mainly occurring in December and January of the cold season. The CP4 circulation type has the lowest occurrence frequency, only 2.74%, and its spatial distribution mainly shows an obvious latitudinal distribution. The north is occupied by positive sea level pressure anomaly and positive geopotential height anomaly, and the south is occupied by negative sea level pressure anomaly and negative geopotential height anomaly.

[0140] Step 2.4: According to the results of circulation classification, based on the measured temperature and precipitation data of CN05.1, the daily precipitation, temperature and other climate variables corresponding to each SOM circulation type are extracted to further identify and extract composite extreme low temperature and heavy snow events. The composite low temperature and heavy snow event is defined as: on the same day, the temperature is lower than the 90th percentile of the cold season historical period, and the daily snowfall is higher than the 90th percentile of the cold season historical period. Identify and extract the low temperature and heavy snow events under each circulation type, and perform feature analysis on the composite low temperature and heavy snow events. The main feature definitions are as follows:

[0141]

[0142] Where Frequency is the frequency, which is defined as the number of extreme days in a year; Intensity is the intensity of extreme days, which is defined as the product of the normalized values ​​of the daily minimum temperature and snowfall intensity relative to the 25th and 75th percentiles.

[0143] The characteristic analysis of the composite low temperature and heavy snow events under various circulation patterns is shown in Figure 7 The results show that the characteristics of the composite extreme low temperature and heavy snow events under various circulation patterns are shown in Figure 7. The results show that under the CP2 and CP4 circulation patterns, the frequency of compound extreme low temperature and heavy snow events in the cold season shows a significant decreasing trend in the entire Chinese region, among which the decreasing trend is the strongest in the middle and lower reaches of the Yangtze River Basin, and the frequency of compound extreme low temperature and heavy snow events is less than 2 times per year. Under the CP1 circulation pattern, the frequency of compound extreme low temperature and heavy snow events shows an increasing trend in western and northern China, and a decreasing trend in southeastern China; from the perspective of time series, the multi-year average frequency of compound events in CP1 is the highest compared with the other three circulation patterns, with an average of 6 days. Under the CP3 circulation pattern, the frequency of compound extreme low temperature and heavy snow events in the cold season shows a significant increasing trend in the entire Chinese region, among which the increasing trend is the strongest in Northeast and Northwest China, and the average multi-year average frequency of compound events reaches 4 days.

[0144] Step 3: Human activities and natural variability can affect the spatial pattern of cold season temperature, precipitation, and snowfall in China by affecting large-scale circulation. Since the global climate model simulates atmospheric circulation better than precipitation, the present invention uses the optimal fingerprint method to perform external forced detection attribution analysis based on the circulation classification results and the frequency change of the circulation type; based on the attribution scale factor of the historical period, the global climate model of the future period is constrained and estimated, and the optimal fingerprint method detection attribution analysis is further performed; based on the circulation type decomposition and climate change separation method, the thermal effect, dynamic effect, and thermal-dynamic interaction are separated through the frequency change trend of compound extreme events.

[0145] Step 3.1: Detect and attribute the anthropogenic forcing signal based on the frequency changes of the circulation pattern in the historical period. The regularized optimal fingerprint method (ROF) proposed by Ribes et al. in 2013 is used to detect and attribute the frequency changes of the circulation pattern under climate change. The ROF method is based on the generalized linear regression model and assumes that the external forcing response is additive. The calculation formula is as follows:

[0146]

[0147] Among them, y is the observed value, x i is the response of the climate system to the ith external forcing, β i represents the scaling factor that matches the simulated change of external forcing with the observed change, and ε represents the internal variability of the climate.

[0148] Considering that the simple ensemble average of the global climate model itself cannot completely remove the impact of internal climate variability, the present invention adopts the overall least squares method, and the regression model is:

[0149]

[0150] When the scaling factor β of the external forcing is within the 90% confidence interval and significantly greater than zero, it means that the model simulation response of the external forcing is identifiable in the observational data; if this condition is not met, the signal simulated by the climate model is considered to be unidentifiable in the observation. In addition, if the value of the scaling factor β is significantly greater than zero and its 90% confidence interval includes 1, it can be further judged that the changes in the observations are mainly caused by the specified external forcing, while excluding the influence of other external forcings or factors. If the optimal estimate of the scaling factor β exceeds 1 (or is less than 1), it indicates that the model's predicted response to the external forcing is underestimated (or overestimated).

[0151] The present invention uses the sea level pressure field and 500hPa potential height field data of 6 different CMIP6 climate models. First, it is necessary to evaluate the simulation ability of the climate model for the circulation type to ensure that the climate model can reproduce the spatial characteristics and frequency characteristics of the circulation type. The SOM method is used to classify the sea level pressure of each model member in the historical period of 1961-2020, and 6 sets of 2×2 circulation types are obtained. For each circulation type, the spatial correlation coefficient between a single circulation type and the reanalysis sea level pressure field is obtained. If the spatial correlation coefficients are all high, it means that the full forced simulation of these models in the historical period can better capture the circulation mode characteristics.

[0152] Furthermore, the optimal fingerprint method is used to combine the observed data with the climate model data of full forcing, natural forcing only, and pre-industrial revolution simulation control experiment (piControl), and the attribution analysis is based on the frequency change of circulation patterns. piControl is used to estimate the internal natural variability. Table 5 lists the 16 CMIP6 models used to estimate the internal natural variability in this example. With a research period of 60 years, the piControl data is divided into 223 sets of data, half of which are used for optimal estimation and the other half for residual verification.

[0153] Table 5 is a list of CMIP6 models used to estimate internal natural variability

[0154] sequence model Number of years Number of sets 1 ACCESS-CM2 500 8 2 CanESM5 1052 17 3 IPSL-CM6A-LR 2000 33 4 MIROC6 800 13 5 NorESM2-LM 500 8 6 ACCESS-ESM1-5 1000 16 7 CMCC-ESM2 500 8 8 CNRM-CM6-1 500 8 9 CESM2 1200 20 10 CESM2-WACCM-FV2 500 8 11 GFDL-CM4 500 8 12 GFDL-ESM4 500 8 13 HadGEM3-GC31-LL 2000 33 14 MRI-ESM2-0 701 11 15 MPI-ESM1-2-LR 1000 16 16 NorESM2-MM 500 8 SUM 223

[0155] In order to improve the signal-to-noise ratio of long-term changes, the circulation pattern frequency needs to be processed first. The average climate state from 1980 to 2010 is used as the reference period, and then the frequency anomaly sequence is processed for non-overlapping continuous 5 years to eliminate the impact of interannual fluctuations. The frequency series of each circulation pattern are combined to obtain the circulation pattern frequency distribution change series. The circulation pattern frequency distribution change series under the reanalysis data and external forcing scenario is applied to the optimal fingerprint attribution analysis.

[0156] The anomaly time series of historical circulation pattern frequencies simulated by climate models with all forcing (ALL) and natural forcing only (NAT) relative to 1981-2010 and the single-signal and dual-signal optimal fingerprint attribution results are shown in Figure 8 From the single signal detection attribution of ALL, it can be seen that the scale factors and their 95% confidence intervals corresponding to CP1 and CP2 include 0, which means that the ALL forced signal is not significantly detected. The scale factors and confidence intervals corresponding to CP3 and CP4 are all greater than 0 and include 1, indicating that the forced signal can be significantly detected; from the dual signal detection attribution of ANT and NAT, it can be seen that the scale factors and confidence intervals corresponding to ANT and NAT forced signals of CP1 and CP2 include 0, indicating that the forced signal cannot be detected; the scale factors and their 95% confidence intervals corresponding to ANT signals of CP3 and CP4 are all greater than 0 and include 1, indicating that the forced signal can be significantly detected; the scale factors and their 95% confidence intervals corresponding to NAT signals of CP3 and CP4 include 0, indicating that the forced signal cannot be detected.

[0157] Step 3.2: According to the optimal estimation result of the scale factor provided by the detection attribution in the historical period, the frequency change of the circulation type in the future period is estimated and analyzed with attribution constraints, which helps to more accurately reveal the systematic deviation of the simulation of the circulation type frequency under the influence of external forcing by the climate model. Using this correction capability, the present invention further constrains the estimation of the circulation type frequency in the future, that is, corrects the original estimation result of the model by multiplying the corresponding optimal scaling factor. The attribution constraint idea is to assume that the simulation deviation of the climate model in the historical period will continue into the future. However, the attribution constraint method is currently less used at the regional scale, because the attribution constraint requires that it can be significantly attributed in the historical stage, but the current climate model has a large error in simulating precipitation at the regional scale, especially in China, which is significantly affected by the East Asian monsoon and complex terrain. The present invention adopts the attribution constraint idea to correct the deviation of the model in simulating the external forced circulation frequency, and estimates the future circulation frequency changes, aiming to explore the future change pattern of extreme events of low temperature and heavy snow in China's cold season.

[0158] First of all, it is necessary to consider whether the atmospheric circulation type simulated by the climate model in the future period is consistent with the circulation of the reanalysis data in the historical period after SOM classification in the trough and ridge position and the anomaly field, whether the distribution of occurrence time is similar, and whether the quality of SOM classification is stable. Based on the attribution results of the circulation type frequency in the historical period, it is found that the CP3 and CP4 circulation types in the historical period can significantly detect external forcing signals, indicating that the climate model underestimates the changes in the circulation type frequency under the action of external forcing. Therefore, the frequency deviations of the CP3 and CP4 circulation types simulated by the six climate models of SSP245 and SSP245-NAT in the future period are corrected by attribution constraints by multiplying them by the scale factor. Comparing the frequency changes of the circulation types before and after the attribution constraints in the historical period (1961-2020) and the future period (2021-2100), the constraint estimation results are shown in Figure 2. Fig. 9 After attribution constraints, under ALL forcing, the frequency of CP3 circulation type in the future period shows an increasing trend, while the frequency of CP4 circulation type changes relatively steadily, but both are higher than in the historical period; under NAT forcing, after attribution constraints, the frequency of CP3 circulation type in the future period is higher than in the historical period, while the frequency of CP4 circulation type is lower than in the historical period.

[0159] Step 3.3: Separate the thermal and dynamic effects based on the circulation type decomposition method. The thermodynamic effects of extreme low temperature and heavy snow events are mainly related to surface radiation, water vapor flux or land cover changes, while the dynamic effects are mainly related to changes in atmospheric circulation. Based on the SOM clustering algorithm, the main atmospheric circulation patterns are identified to further determine the contribution of thermodynamic and dynamic changes to changes in extreme events. The present invention adopts the thermodynamic-dynamic effect separation method proposed by Horton et al. in 2015 to separate and quantify the relative contributions of thermal and dynamic effects to the changing trend of the frequency of cold season composite low temperature and heavy snow events. The calculation formula is as follows:

[0160]

[0161] In the formula, E represents the frequency of occurrence of the compound event, f i represents the frequency of the ith circulation pattern, E i represents the frequency of compound events when the i-th circulation pattern appears, and K represents the total number of circulation patterns.

[0162] Further, considering the division of time, the calculation formula is as follows:

[0163]

[0164] In the formula and They represent the frequency of occurrence of composite events and the time mean of the frequency of the i-th circulation pattern, E i ′ and f i′ represents the deviation value of the corresponding time mean.

[0165] Further derivative with respect to time yields:

[0166]

[0167] The left side of the equation shows the changing trend of the frequency of composite low-temperature heavy snow events, and the right side of the equation shows the relative contributions of thermal action, dynamic action and thermal-dynamic interaction related to the i-th circulation pattern from left to right.

[0168] The results of the decomposition of thermal and dynamic effects are shown in Fig.10 The results show that: for the percentage of the change trend of the frequency of composite events under each circulation pattern, CP1, CP2, and CP4 mainly make negative contributions, while CP3 makes a positive contribution. For dynamic effects, CP1-CP4 all make positive contributions, and CP4 has the largest positive contribution, reaching 90%. For thermal effects, CP1-CP4 all make positive contributions, and CP3 has the largest positive contribution, reaching 51%. For thermal and dynamic interactions, the contributions of CP1, CP2, and CP4 are all negative, while CP3 makes a positive contribution.

[0169] The attribution analysis system for cold season compound low temperature and heavy snow events based on circulation typing provided by the present invention is described below. The attribution analysis system for cold season compound low temperature and heavy snow events based on circulation typing described below and the attribution analysis method for cold season compound low temperature and heavy snow events based on circulation typing described above can be referenced to each other.

[0170] Fig.11 is a schematic diagram of the structure of an attribution analysis system for cold season composite low temperature and heavy snow events based on circulation classification provided by an embodiment of the present invention, such as Fig.11 As shown, it includes: a data processing module 1101, a circulation classification module 1102 and an attribution analysis module 1103, wherein:

[0171] The data processing module 1101 is used to collect measured meteorological and hydrological data, global climate model data and atmospheric circulation data in the study area, and to perform climate model assessment, bias correction calculation and snowfall calculation on the global climate model data; the circulation classification module 1102 is used to perform circulation classification on atmospheric circulation variables using a self-organizing mapping network method, identify extreme low temperature and heavy snow events under different circulation types, and perform event characteristic analysis on the extreme low temperature and heavy snow events; the attribution analysis module 1103 is used to perform external forced detection attribution analysis based on the frequency changes of circulation types using the regularized optimal fingerprint method, constrain the global climate model for the future period based on the attribution scale factors of the historical period, and separate the thermal effects, dynamic effects and thermal-dynamic interactions through the frequency change trends of compound extreme events.

[0172] Fig.12 An example of a physical structure diagram of an electronic device is shown in FIG. Fig.12 As shown, the electronic device may include: a processor (processor) 1210, a communication interface (Communications Interface) 1220, a memory (memory) 1230 and a communication bus 1240, wherein the processor 1210, the communication interface 1220, and the memory 1230 communicate with each other through the communication bus 1240. The processor 1210 can call the logic instructions in the memory 1230 to execute the attribution analysis method of cold season composite low temperature and heavy snow events based on circulation classification, which method includes: collecting measured meteorological and hydrological data, global climate model data and atmospheric circulation data in the study area, and performing climate model evaluation, bias correction calculation and snowfall calculation on the global climate model data; using the self-organizing mapping network method to perform circulation classification on atmospheric circulation variables, identify extreme low temperature and heavy snow events under different circulation types, and perform event feature analysis on the extreme low temperature and heavy snow events; based on the frequency change of the circulation type, using the regularized optimal fingerprint method to perform external forced detection attribution analysis, based on the attribution scale factor of the historical period, constraining the global climate model for the future period, and separating the thermal effect, dynamic effect and thermal-dynamic interaction through the frequency change trend of composite extreme events.

[0173] In addition, the logic instructions in the above-mentioned memory 1230 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0174] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0175] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0176] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An attribution analysis method for cold season composite low temperature and heavy snow events based on circulation classification, characterized in that: include: Collect measured meteorological and hydrological data, global climate model data and atmospheric circulation data in the study area, and conduct climate model evaluation, bias correction calculation and snowfall calculation of global climate model data; The self-organizing mapping network method is used to classify the atmospheric circulation data, identify the extreme low temperature and heavy snow events under different circulation types, and analyze the event characteristics of the extreme low temperature and heavy snow events; Based on the frequency changes of circulation patterns, the regularized optimal fingerprint method is used to perform attribution analysis of external forcing detection. Based on the attribution scale factors of the historical period, the global climate model of the future period is constrained and estimated. The thermal effect, dynamic effect and thermal-dynamic interaction are separated through the frequency change trend of composite extreme events. Among them, the regularized optimal fingerprint method is used to perform attribution analysis of external forced detection based on the frequency change of circulation patterns. Based on the attribution scale factors of the historical period, the global climate model of the future period is constrained and estimated. The thermal effect, dynamic effect and thermal-dynamic interaction are separated through the frequency change trend of composite extreme events, including: The regularized optimal fingerprint method and total least squares method are used to detect and attribute the changes in circulation pattern frequency under climate change; Based on the optimal estimation results of the attribution scale factors in the historical period, an attribution constraint estimation analysis is conducted on the frequency changes of circulation patterns in the future period; Based on the self-organizing map neural network clustering algorithm, the main atmospheric circulation patterns are identified and the contribution of thermodynamic and dynamic changes to the changes in extreme events is determined; Based on the self-organizing map neural network clustering algorithm, the main atmospheric circulation patterns are identified and the contribution of thermodynamic and dynamic changes to extreme event changes is determined, including: Thermodynamic-dynamic separation method is used to separate and quantify the relative contributions of thermal and dynamic effects to the changing trend of the frequency of extreme low temperature and heavy snow events. Based on the time mean of the frequency of compound events and the frequency of any circulation pattern, as well as the deviation of the frequency of compound events and the time mean of the frequency of any circulation pattern, the time division is carried out; By taking the derivative with respect to time, the relative contributions of the thermal effect, dynamic effect and thermal-dynamic interaction related to any circulation pattern can be obtained from the trend of the frequency of composite low-temperature heavy snow events.

2. The attribution analysis method for cold season composite low temperature and heavy snow events based on circulation classification according to claim 1 is characterized in that: Collect measured meteorological and hydrological data, global climate model data and atmospheric circulation data in the study area, and conduct climate model evaluation, bias correction calculation and snowfall calculation of global climate model data, including: Collect CN05.1 grid data set, CMIP6 global climate model data, and ERA5, JRA-55, MERRA2 and NCEP atmospheric reanalysis data sets, and use bilinear interpolation to interpolate the CN05.1 grid data set, the CMIP6 global climate model data, and the ERA5, JRA-55, MERRA2 and NCEP atmospheric reanalysis data sets to a preset resolution grid; Six different CMIP6 model data were used to evaluate the climate model based on the measured temperature and precipitation data of the CN05.1 grid data set as the benchmark data set. Cold season data were extracted, and four extreme temperature indices and four extreme precipitation indices recommended by the climate change index ETCCDI were selected. The Taylor comprehensive index was used to comprehensively evaluate the average climate state and temporal variation of temperature and precipitation simulated by the cold season climate model in the historical period. Based on the piecewise quantile mapping method PQM and the quantile incremental mapping method QDM, the piecewise quantile incremental mapping method PQDM is obtained, and the PQDM is used to correct the bias of the global climate model data. Using a dynamic threshold parameterization scheme and combining the melting process of snowfall to the ground, the high and low temperature thresholds are updated according to the altitude and meteorological conditions of each grid point to divide the total precipitation into snowfall, rainfall, and the intermediate amount of snowfall and rainfall.

3. The attribution analysis method for cold season composite low temperature and heavy snow events based on circulation classification according to claim 2 is characterized in that: Based on PQM and QDM, PQDM is obtained, and PQDM is used to correct the bias of global climate model data, including: Based on the peak over threshold method POT and the 95th percentile threshold, the meteorological time series corresponding to the global climate model data is divided into extreme values ​​and non-extreme values; performing bias correction on the non-extreme value using an empirical cumulative distribution function, wherein the empirical cumulative distribution function is determined by the QDM; The generalized Pareto distribution GP is used as a transfer function to fit the extreme value, and the maximum likelihood method is used to estimate the parameters of the extreme value; The QDM is used to make incremental adjustments and PQDM corrections are performed on the hist and hist-nat simulations of the CN05.1 grid data set and the CMIP6 global climate model data.

4. The attribution analysis method for cold season composite low temperature and heavy snow events based on circulation classification according to claim 2 is characterized in that: Using the dynamic threshold parameterization scheme, combined with the melting process of snow falling to the ground, the high temperature threshold and low temperature threshold are updated according to the altitude and meteorological conditions of each grid point, and the snowfall, rainfall and the intermediate amount of snowfall and rainfall in the total precipitation are divided, including: Based on the near-surface relative humidity, the dry-bulb temperature and the altitude, the high temperature threshold and the low temperature threshold are obtained; The wet-bulb temperature is obtained by an empirical formula, and the snowfall, the rainfall, and the intermediate amount of snowfall and rainfall are obtained by using the wet-bulb temperature, the high temperature threshold, and the low temperature threshold.

5. The attribution analysis method for cold season composite low temperature and heavy snow events based on circulation classification according to claim 1 is characterized in that: The self-organizing mapping network method is used to classify the atmospheric circulation data, identify the extreme low temperature and heavy snow events under different circulation types, and analyze the event characteristics of the extreme low temperature and heavy snow events, including: Determine the consistency and reliability of atmospheric reanalysis products ERA5, JRA-55, NCEP and MERRA2, and compare and analyze winter sea level pressure, 500hPa geopotential height and 300hPa wind speed; Self-organizing map neural network was used to conduct sensitivity test of circulation typing; Based on the sensitivity test of circulation typing of self-organizing map neural network, the combination of circulation variables and the hyperparameters of circulation typing of self-organizing map neural network are determined to obtain different circulation types after cluster typing, and the characteristics of the different circulation types after cluster typing are analyzed to obtain the circulation analysis results; According to the circulation analysis results, based on the CN05.1 measured temperature and precipitation data, the climate variables corresponding to each circulation type of the self-organizing mapping neural network are extracted to identify and extract the extreme low temperature and heavy snow events.

6. The attribution analysis method for cold season composite low temperature and heavy snow events based on circulation classification according to claim 5 is characterized in that: The self-organizing map neural network was used to conduct sensitivity tests on circulation typing, including: Step 1: Initialize the self-organizing map neural network and normalize each neuron in the network output layer using the initialization weight vector; Step 2: Calculate the distance between the input vector and each weight vector to get the winning neuron; Step 3: Determine a neighborhood function according to the network topology, and use the neighborhood function to update the weights of the winning neuron and the neurons in the neighborhood of the winning neuron; Step 4: Repeat steps 2 to 4 until the network learning rate is determined to be less than the learning rate threshold and the network reaches convergence, map each input vector to the winning neuron, and obtain the input data clustering result.

7. An attribution analysis system for cold season composite low temperature and heavy snow events based on circulation classification, characterized in that: include: The data processing module is used to collect measured meteorological and hydrological data, global climate model data and atmospheric circulation data in the study area, and to perform climate model evaluation, bias correction calculation and snowfall calculation on global climate model data; A circulation classification module is used to classify atmospheric circulation data using a self-organizing mapping network method, identify extreme low temperature and heavy snow events under different circulation types, and analyze event characteristics of the extreme low temperature and heavy snow events; The attribution analysis module is used to perform attribution analysis of external forcing detection based on the frequency changes of circulation patterns using the regularized optimal fingerprint method. Based on the attribution scale factors of the historical period, it constrains the global climate model for the future period and separates the thermal effect, dynamic effect and thermal-dynamic interaction through the frequency change trend of compound extreme events. The attribution analysis module is specifically used for: The regularized optimal fingerprint method and total least squares method are used to detect and attribute the changes in circulation pattern frequency under climate change; Based on the optimal estimation results of the attribution scale factors in the historical period, an attribution constraint estimation analysis is conducted on the frequency changes of circulation patterns in the future period; Based on the self-organizing map neural network clustering algorithm, the main atmospheric circulation patterns are identified and the contribution of thermodynamic and dynamic changes to the changes in extreme events is determined; Based on the self-organizing map neural network clustering algorithm, the main atmospheric circulation patterns are identified and the contribution of thermodynamic and dynamic changes to extreme event changes is determined, including: Thermodynamic-dynamic separation method is used to separate and quantify the relative contributions of thermal and dynamic effects to the changing trend of the frequency of extreme low temperature and heavy snow events. Based on the time mean of the frequency of compound events and the frequency of any circulation pattern, as well as the deviation of the frequency of compound events and the time mean of the frequency of any circulation pattern, the time division is carried out; By taking the derivative with respect to time, the relative contributions of the thermal effect, dynamic effect and thermal-dynamic interaction related to any circulation pattern can be obtained from the trend of the frequency of composite low-temperature heavy snow events.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, it implements the attribution analysis method for cold season composite low temperature and heavy snow events based on circulation classification as described in any one of claims 1 to 6.

Images