Rainfall mechanism analysis method based on regional climate data

Abstract

The invention relates to the field of climate change and meteorological data processing, and discloses a rainfall mechanism analysis method based on regional climate data, and the method comprises the following steps: selecting a target research region, and collecting rainfall indexes and climate driving factor data; based on an ensemble empirical mode decomposition method, decomposing the time sequence into components of different time scales; performing power spectrum analysis on the decomposed data, and identifying periodic characteristics; continuous wavelet transform, cross wavelet transform and wavelet coherence analysis are utilized to research the relevance and phase relation of the rainfall index and the climate driving factor in a time-frequency space; quantifying contributions of the driving factors to rainfall indexes, and determining primary and secondary driving factors; and a main climate driving mechanism of rainfall change is clarified in combination with a multi-time-scale analysis result. According to the method, the complex driving process of regional rainfall change can be comprehensively analyzed, and scientific basis and technical support are provided for climate change research, water resource management and adaptive strategies.

Description

Technical Field

[0001] The present invention relates to the technical field of climate change and meteorological data processing, and in particular to a precipitation mechanism analysis method based on regional climate data. Background Art

[0002] As global climate change intensifies, the temporal and spatial distribution characteristics of regional precipitation and its changing patterns have become an important part of climate research. Precipitation changes are directly related to ecosystem stability, agricultural productivity, and water resource security. Therefore, in-depth research on the driving mechanism of precipitation changes is of great significance for predicting future climate change trends and formulating response strategies. However, the precipitation process is jointly affected by a variety of natural climate factors and human activities. The effects of these factors have complex nonlinear and multi-time scale characteristics, which brings great challenges to the study of precipitation mechanisms.

[0003] In the existing technology, traditional precipitation mechanism analysis methods are often based on simple correlation or linear regression models, and usually only focus on the relationship between driving factors and precipitation changes on a single time scale. The limitations of this method are mainly reflected in two aspects: first, it cannot effectively reveal the nonlinear correlation characteristics between precipitation and climate driving factors on different time scales, resulting in insufficient understanding of the driving forces of long-term trends and short-term fluctuations; second, it fails to accurately quantify the relative contributions of multiple driving factors, especially in the complex background of multiple driving factors acting together, lacking quantitative attribution analysis of the dominant factors. In addition, existing studies have rarely studied the interaction between natural climate factors (such as El Niño-Southern Oscillation, Pacific Decadal Oscillation, Atlantic Multidecadal Oscillation) and anthropogenic forcing factors (such as greenhouse gases, aerosols) and their periodic impact on regional precipitation, which limits the comprehensiveness and accuracy of precipitation mechanisms.

[0004] In response to the above problems, there is an urgent need for a method that can fully consider the multi-time scale characteristics of precipitation changes, quantitatively analyze the contribution rate of climate driving factors, and combine physical mechanisms to deeply study the dominant factors of precipitation changes, so as to break through the limitations of existing technologies and provide scientific tools and theoretical basis for the study of regional precipitation change mechanisms. Summary of the invention

[0005] In view of the shortcomings of the existing technology, the present invention provides a precipitation mechanism analysis method based on regional climate data, which solves the technical problems in the existing technology that it is impossible to effectively analyze the multi-time scale characteristics of precipitation changes, quantify the contribution of driving factors, and fully reveal the climate driving mechanism.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A precipitation mechanism analysis method based on regional climate data, comprising the following steps:

[0007] Select the target study area and collect climate data, including precipitation index and climate driving factor data;

[0008] Based on the ensemble empirical mode decomposition method, the precipitation index and driving factor data are decomposed into components of different time scales;

[0009] Perform power spectrum analysis on the decomposed data to identify significant periodic features at each time scale;

[0010] The correlation and phase relationship between precipitation index and climate driving factors at different time scales are analyzed using continuous wavelet transform, cross wavelet transform and wavelet coherence analysis.

[0011] Through sliding correlation analysis and statistical methods, the impact of driving factors on precipitation index is quantified, and the main and secondary driving factors are identified;

[0012] Based on the relationship between precipitation index and driving factors on different time scales, the main climate driving mechanism of regional precipitation changes is analyzed.

[0013] Preferably, the climate data include tree ring width data, drought and flood index data, instrument observation data and climate reanalysis data, and the precipitation index is obtained by multi-source data fusion, including comprehensive analysis results of tree ring width, historical climate records and modern meteorological observation data.

[0014] Preferably, the climate driving factor data include El Nino-Southern Oscillation, Pacific Decadal Oscillation, Atlantic Multi-Decadal Oscillation, total solar radiation and anthropogenic forcing factor data.

[0015] Preferably, the ensemble empirical mode decomposition method comprises the following steps:

[0016] Add low-intensity white noise to the raw climate data;

[0017] The empirical mode decomposition method is used to decompose the data after adding white noise into multiple intrinsic mode functions;

[0018] Repeat the above steps, using different white noise sequences, to decompose the data n times;

[0019] The intrinsic mode functions obtained by n decompositions are averaged to obtain the final components.

[0020] Preferably, the power spectrum analysis uses a frequency analysis method based on Fourier transform to decompose the total energy of the climate data into different frequency components, and determines the dominant period by calculating the variance contribution of each frequency component.

[0021] Preferably, the significant periodic features identified by the power spectrum analysis include interannual cycles, decadal cycles, multi-decadal cycles and century cycles.

[0022] Preferably, in the step of analyzing the correlation and phase relationship between the precipitation index and the climate driving factors on different time scales:

[0023] Continuous wavelet transform is used to identify non-stationary periodic signals of precipitation index and driving factor data in the time-frequency space;

[0024] Cross-wavelet transform is used to analyze the common periodic characteristics and relative phases of precipitation index and driving factors in time-frequency space;

[0025] Wavelet coherence analysis was used to identify areas of significant correlation between precipitation indices and driving factors.

[0026] Preferably, quantifying the impact of driving factors on precipitation index through the sliding correlation analysis includes:

[0027] Determine the time series of driving factors and precipitation indices;

[0028] Select the sliding window length and adjust the window range according to the analysis requirements;

[0029] The correlation coefficient between the driving factor and the precipitation index is calculated in each sliding window to obtain the dynamic correlation changing over time;

[0030] Analyze the changing characteristics of correlations over time and identify significant correlations in key time periods.

[0031] Preferably, the step of determining the primary and secondary driving factors comprises:

[0032] The R language relaimpo package was used to conduct quantitative attribution analysis on driving factors;

[0033] By calculating the relative importance weight of each driving factor, the contribution rate of different driving factors to the precipitation index is quantified;

[0034] According to the contribution rate, the main driving factors and secondary driving factors are determined.

[0035] Preferably, the step of analyzing the main climate driving mechanism of regional precipitation change based on the relationship between precipitation index and driving factors on different time scales includes:

[0036] Based on the correlation analysis results between precipitation index and driving factors at interannual, interdecadal, multidecadal and century scales, the main driving factors at different time scales are determined;

[0037] The specific impact mechanisms of driving factors on precipitation changes are analyzed for different time scales, including:

[0038] The impact mechanism of El Niño-Southern Oscillation on interannual precipitation variation;

[0039] The mechanism of the impact of the Pacific Decadal Oscillation on multi-decadal precipitation changes;

[0040] The impact mechanism of Atlantic multidecadal oscillation and total solar radiation on century-scale precipitation changes;

[0041] Compare the relative effects of natural driving factors and anthropogenic forcing factors on precipitation changes before and after the Industrial Revolution, and determine the differences in the impact of climate driving factors on precipitation in different periods.

[0042] The present invention provides a precipitation mechanism analysis method based on regional climate data. It has the following beneficial effects:

[0043] 1. The present invention decomposes the complex precipitation index and climate driving factor data into components at the interannual, interdecadal, multidecadal and century scales through the ensemble empirical mode decomposition (EEMD) method, which can accurately capture the characteristics of precipitation changes on different time scales and provide a scientific basis for subsequent periodic analysis and mechanism research. This multi-scale decomposition method effectively solves the problem of mode aliasing in traditional empirical mode decomposition and ensures the stability and reliability of the decomposition results.

[0044] 2. The present invention can accurately identify the significant periodic characteristics of precipitation index and climate driving factors on different time scales through power spectrum analysis and red noise background comparison. This method ensures the scientificity and rigor of the periodic analysis results, and provides a key basis for further analysis of the temporal characteristics of precipitation changes and the role of driving factors.

[0045] 3. The present invention uses continuous wavelet transform, cross wavelet transform and wavelet coherence analysis to reveal the correlation and phase relationship between precipitation index and climate driving factors at different time scales, and can intuitively display the time-frequency domain correlation characteristics between the two. This method overcomes the deficiency of traditional correlation analysis being limited to a single time scale, and enables the complex interactions in the climate system to be fully revealed.

[0046] 4. This invention uses sliding correlation analysis and statistical attribution methods to accurately quantify the contribution of different climate driving factors to the precipitation index and identify the main and secondary driving factors. This quantitative analysis improves the accuracy of the research and provides a scientific basis for understanding the dominant role of driving factors in different periods and time scales.

[0047] 5. The present invention is not only suitable for analyzing precipitation mechanisms under historical climate change, but can also be applied to current and future climate change research, providing scientific support and decision-making basis for regional climate forecasting, water resources management and climate change adaptation strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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.

[0050] Please refer to the attached Figure 1 The present invention provides a precipitation mechanism analysis method based on regional climate data, which comprehensively reveals the spatiotemporal variation characteristics of precipitation in the target area and its main driving factors through multi-time scale decomposition, correlation analysis and climate driving mechanism research.

[0051] like Figure 1 As shown, the precipitation mechanism analysis method based on regional climate data may include the following steps:

[0052] S1. Select the target study area and collect climate data;

[0053] S2, multi-time scale component decomposition based on ensemble empirical mode decomposition method;

[0054] S3, identifying periodic features through power spectrum analysis;

[0055] S4, analysis of correlation and phase relationship based on wavelet transform;

[0056] S5. Quantify the impact of driving factors through sliding correlation analysis;

[0057] S6. Comprehensively analyze the climate-driven mechanisms of regional precipitation.

[0058] Each step of the method of the present invention is described in detail below.

[0059] For step S1, in this embodiment, the precipitation mechanism analysis method for regional climate data first needs to select a target research area and collect data related to the regional climate, including precipitation index and climate driving factor data, to provide a basis for subsequent precipitation mechanism analysis.

[0060] The target research area is determined according to the research needs. In this example, the northern and central regions of China are selected as the research object. The precipitation changes in this region are significant and have important research value. At the same time, it is susceptible to the combined influence of multiple climate driving factors (such as El Nino-Southern Oscillation, Pacific Decadal Oscillation, Atlantic Multidecadal Oscillation, and total solar radiation) and human activities (such as greenhouse gas emissions and aerosol emissions).

[0061] After the target area is determined, the regional climate data is comprehensively collected and sorted in this embodiment, including precipitation index and driving factor data. The precipitation index data comes from the integration of multiple climate reconstruction and observation methods, including tree ring width, historical drought and flood records, and modern meteorological observation data. Through the comprehensive use of these multi-source data, a precipitation index series with high temporal and spatial resolution that reflects long-term climate change can be obtained.

[0062] For driving factor data, the following types of data are mainly considered in this embodiment:

[0063] El Nino-Southern Oscillation (ENSO) index data, which can be used to reflect the abnormal variation characteristics of sea surface temperature in the tropical Pacific region, usually expressed as the Southern Oscillation Index (SOI) or Sea Surface Temperature Anomalies Index (SST Anomalies);

[0064] The Pacific Decadal Oscillation (PDO) index data describes the variation characteristics of sea surface temperature in the North Pacific region on an interdecadal scale, which can be obtained through the principal component analysis of temperature anomalies;

[0065] The Atlantic Multidecadal Oscillation (AMV) index data is used to characterize the variation characteristics of sea surface temperature in the North Atlantic region on a multidecadal scale, usually calculated by the regional average of temperature anomalies in the Atlantic region;

[0066] Total Solar Irradiance (TSI) data, which reflects the direct impact of solar activity on the climate system. The data can be derived from long-term solar activity observations and reconstructions;

[0067] Human activity forcing data, including data on greenhouse gas emission concentrations and aerosol emissions, are usually derived from climate model simulations or long-term observations.

[0068] In this embodiment, all data need to be preprocessed after collection to ensure the consistency and accuracy of subsequent analysis. The preprocessing methods include but are not limited to:

[0069] Perform linear trend removal on time series data to eliminate possible non-periodic trend interference;

[0070] Data standardization, which converts time series into a form with zero mean and unit standard deviation, to facilitate comparison and analysis between different data types;

[0071] Data interpolation processing is to perform reasonable interpolation on the time points of missing data to ensure the integrity of the time series.

[0072] For the time range of precipitation index and climate driving factors, the time interval from 1470 to modern times (such as 2000) is selected in this example, and the specific time range depends on the availability of data. This time period can cover natural climate fluctuations (such as the Little Ice Age, the modern warming period) and the stage of enhanced human activities, laying the foundation for analyzing precipitation changes and driving factors in different historical periods.

[0073] Through the collection and processing of the above data, this embodiment provides comprehensive and high-quality basic data support for subsequent multi-time scale decomposition, periodic feature identification and climate driving mechanism analysis.

[0074] For step S2, in this embodiment, in order to perform multi-time scale decomposition analysis on the precipitation index and climate driving factor data, the ensemble empirical mode decomposition (EEMD) method is used. This method can effectively decompose nonlinear and non-stationary time series data into multiple intrinsic mode function (IMF) components and a residual component (trend term), which respectively reflect the characteristic changes on different time scales.

[0075] The application of the EEMD method in this embodiment is based on the following technical principles and specific operation steps:

[0076] This method first adds low-intensity white noise to the original time series data, and uses the uniform distribution characteristics of white noise in the time and frequency domain to enhance the local characteristics of the signal, thereby solving the problem of modal aliasing that is prone to occur in the traditional empirical mode decomposition (EMD) method. In each decomposition process, the randomness of white noise can make the decomposition results of the original data more stable and accurate.

[0077] The decomposition of the original climate data time series X(t) is achieved through the following process:

[0078] 1. Add low-intensity white noise N with an amplitude equal to the standard deviation σ of the original data to the time series data X(t) i (t), generate a new time series X i (t) = X (t) + N i (t).

[0079] 2. Add white noise to the time series X i (t) Perform empirical mode decomposition (EMD) to decompose it into multiple intrinsic mode functions (IMFs) i,j(t) and a residual trend component R i (t). The decomposition result can be expressed as:

[0080]

[0081] Where m is the number of IMFs obtained through decomposition.

[0082] 3. Repeat steps 1 and 2 N times (as in this embodiment, set N = 1000), adding a different white noise sequence N each time i (t), the intrinsic mode function obtained from each decomposition is averaged to obtain the final stable IMF component:

[0083]

[0084] The final residual trend component is obtained in a similar way:

[0085]

[0086] In this embodiment, the precipitation index and driving factor data are decomposed into multiple intrinsic mode functions (IMFs) by the above EEMD method, and each IMF component corresponds to a specific time scale. For example:

[0087] The interannual scale (1-10 years) mainly corresponds to the first and second IMF components;

[0088] The decadal scale (10-30 years) mainly corresponds to the 3rd and 4th IMF components;

[0089] Multidecadal scale (30-100 years), mainly corresponding to the 5th and 6th IMF components;

[0090] Century scale (>100 years), mainly corresponding to the 7th and 8th IMF components;

[0091] The residual trend component reflects the long-term background variation trend.

[0092] In the EEMD process, the amplitude of the added white noise is set to 20% of the standard deviation of the original time series, and the specific value can be adjusted according to the stability and characteristics of the data. In this embodiment, the number of EEMD decomposition repetitions is set to 1000 times to ensure the stability and reliability of the decomposition results.

[0093] To further verify the validity of the decomposition results, this embodiment uses wavelet bandpass filtering analysis to cross-validate the IMF components after EEMD decomposition. Wavelet bandpass filtering selects signals within a specific frequency band to confirm whether each IMF component accurately reflects the characteristic changes of the corresponding time scale.

[0094] Through the EEMD decomposition of this embodiment, complex nonlinear and non-stationary climate data can be converted into components on different time scales, thereby laying a solid foundation for subsequent periodic feature identification, correlation analysis and climate driving mechanism research.

[0095] For step S3, in this embodiment, in order to further reveal the periodic characteristics of the precipitation index and its driving factors on different time scales, the intrinsic mode function (IMF) components obtained based on the ensemble empirical mode decomposition method (EEMD) are subjected to power spectrum analysis. The main purpose of power spectrum analysis is to identify significant cycles on each time scale and provide a reliable periodic basis for subsequent correlation analysis.

[0096] In this embodiment, power spectrum analysis is implemented by a frequency domain method based on Fourier transform. Fourier transform can convert time series signals from time domain to frequency domain to obtain power spectrum density distribution. Power spectrum density is a way to describe the energy distribution of a signal at different frequencies, reflecting the intensity of periodic fluctuations in a time series.

[0097] The specific implementation steps are as follows:

[0098] In this embodiment, each IMF component IMF j (t) performs discrete Fourier transform (DFT), the formula is:

[0099]

[0100] Among them, F(f) is the complex frequency domain signal corresponding to frequency f, T is the length of the time series, and i is the imaginary unit. According to the discrete Fourier transform results, the power spectral density (PSD) is calculated, and the formula is:

[0101]

[0102] Where |F(f)| 2 It represents the square of the amplitude of the frequency domain signal, which represents the energy of the signal at frequency f.

[0103] In order to determine which frequencies in the power spectrum density correspond to significant periods, this embodiment introduces a red noise spectrum for comparison. Red noise is a common background noise in climate time series. Its power spectrum density increases as the frequency decreases. It is usually obtained by fitting a first-order autoregressive model (AR(1) model). The calculation formula for the red noise spectrum is:

[0104]

[0105] Among them, σ 2 is the variance of the time series, ρ is the autoregressive coefficient, and f is the frequency.

[0106] The power spectrum of the IMF component is compared with the red noise spectrum. The frequency component whose power spectrum density is significantly higher than the 95% confidence level of the red noise spectrum is determined to be a significant cycle. The calculation steps of the significant cycle are as follows:

[0107] According to the red noise model, generate the corresponding significance threshold curve (95% confidence level);

[0108] Compare the power spectrum density curve with the significance threshold curve, and extract the frequency f corresponding to the significant period;

[0109] According to the relationship between period and frequency formula Calculate significant cycles.

[0110] In this example, after power spectrum analysis of the IMF components of the precipitation index and its driving factors, the following significant cycles are obtained:

[0111] Interannual scale: The main period is 3-8 years, corresponding to the interannual oscillation in the climate system;

[0112] Decadal scale: The main period is 10-30 years, which is related to the decadal changes in the climate system;

[0113] Multidecadal scale: The main period is 30-100 years, corresponding to multidecadal oscillations of climate drivers;

[0114] Century scale: The main cycle is more than 100 years, reflecting long-term climate background changes.

[0115] Through power spectrum analysis, this example effectively identified the dominant period and significant characteristics of each IMF component, laying a solid data foundation for the correlation analysis and mechanism research between precipitation index and driving factors in the subsequent steps. The results of power spectrum analysis further prove the effectiveness and accuracy of EEMD decomposition, which helps to accurately understand the multi-time scale characteristics of precipitation changes.

[0116] For step S4, in this embodiment, in order to fully reveal the correlation and phase relationship between precipitation index and climate driving factors at different time scales, continuous wavelet transform, cross wavelet transform and wavelet coherence analysis are used to perform time-frequency domain comprehensive analysis on the data. These methods can simultaneously study the relationship between non-stationary time series from two dimensions of time and frequency, and are particularly suitable for multi-scale correlation analysis of climate data.

[0117] In this embodiment, the mathematical basis of the wavelet transform is the continuous wavelet transform. For a given time series X(t), the definition of the continuous wavelet transform is:

[0118]

[0119] Among them, ψ is the wavelet mother function, a is the scale parameter (inversely proportional to the frequency), b is the translation parameter (corresponding to time), and the symbol * represents the complex conjugate.

[0120] In this embodiment, the Morlet wavelet function is selected as the wavelet mother function. The Morlet wavelet has good time-frequency localization characteristics, and its expression is:

[0121]

[0122] Here, ω0 is the center frequency, and ω0=6 is usually taken to ensure balanced time-frequency resolution.

[0123] The local power spectrum of the precipitation index or driving factor time series in time and frequency (or scale) can be obtained through continuous wavelet transform. The significance of the power spectrum is verified by comparing it with the 95% confidence level of the red noise background power spectrum. The results of continuous wavelet transform are used to preliminarily identify the significant cycles in the time series and their changes over time.

[0124] This embodiment further studies the common periodicity characteristics between precipitation index and climate driving factors through cross wavelet transform. The definition of cross wavelet transform is:

[0125]

[0126] Among them, W X and W0 are the wavelet transform results of two time series X(t) and Y(t), respectively. W Y The power spectrum of the cross wavelet transform can be used to analyze the common period and relative intensity of two time series in the time-frequency domain.

[0127] Based on the cross wavelet transform, this embodiment uses wavelet coherence analysis to further quantify the correlation strength between two time series. The definition of wavelet coherence analysis is:

[0128]

[0129] Among them, S is the smoothing operator, R 2 (a,b) represents the coherence strength at time scale a and time b, and its value range is [0,1]. The high coherence area indicates that the two time series have significant correlation at this time scale.

[0130] In this embodiment, the wavelet coherence analysis also uses arrows to indicate the phase relationship. Arrows pointing to the right indicate that the two sequences change in phase, arrows pointing to the left indicate that the two sequences change in opposite phases, and arrows pointing up or down indicate that the two sequences have a lagging or leading relationship.

[0131] Through the wavelet analysis method of this embodiment, the following results are obtained:

[0132] On the interannual scale (1-10 years), the precipitation index shows a significant inverse relationship with the El Niño-Southern Oscillation index, indicating that El Niño events (warm phase) usually lead to reduced precipitation in the study area.

[0133] On the interdecadal scale (10-30 years), the precipitation index and the Pacific Decadal Oscillation index show a significant in-phase relationship, indicating that the warm phase of the Pacific Decadal Oscillation is closely related to increased precipitation.

[0134] On the multi-decadal scale (30-100 years), the precipitation index has a high correlation with the Atlantic Multidecadal Oscillation index, and the warm phase of the Atlantic Multidecadal Oscillation usually corresponds to an increase in precipitation.

[0135] On a century scale (>100 years), the precipitation index shows a strong in-phase relationship with total solar radiation, indicating that precipitation tends to increase when solar activity increases.

[0136] Through the above method, this embodiment effectively analyzes the correlation and phase relationship between the precipitation index and climate driving factors on different time scales, providing an important basis for subsequent quantitative analysis of driving factors and research on precipitation mechanisms.

[0137] For step S5, in this embodiment, in order to quantitatively evaluate the impact of climate driving factors on the precipitation index, sliding correlation analysis and statistical attribution methods are used to quantify the role of each driving factor from two dimensions: time and contribution rate, and further determine the main and secondary driving factors on different time scales.

[0138] In the sliding correlation analysis, this embodiment uses a sliding correlation method based on a time window to analyze the dynamic changes in the correlation between the precipitation index and the climate driving factors over time. The basic principle of sliding correlation analysis is to define a sliding window of fixed length on the time series, calculate the correlation coefficient within the window, and gradually shift the window to obtain the dynamic correlation characteristics that change over time.

[0139] Assuming that the precipitation index time series is X(t), the climate driving factor time series is Y(t), and the window length is L, the calculation formula of the sliding correlation coefficient r(t) is:

[0140]

[0141] in, and Represent the means of X(t) and Y(t) within the window respectively.

[0142] In this embodiment, the window length L is selected according to the analysis target. For example, in the interannual scale analysis, L is set to 10 years; in the multi-decadal scale analysis, L can be set to 30 years or longer. The sliding correlation result is plotted into a curve with time as the horizontal axis and the correlation coefficient as the vertical axis, which can intuitively reflect the changes in the strength of the correlation in different time periods.

[0143] In the statistical attribution analysis, this embodiment adopts the relative importance weight method to quantify the relative impact of each driving factor by calculating the variance contribution of each driving factor to the precipitation index. The specific method is as follows:

[0144] 1. Use the stepwise regression method to establish the precipitation index X(t) and various driving factors Y1(t), Y2(t),...,Y n (t) is a multiple linear regression model, the model form is:

[0145] X(t)=β0+β1Y1(t)+β2Y2(t)+...+β n Y n (t)+∈

[0146] Among them, β0 is a constant term, β i is the regression coefficient, ∈ is the random error.

[0147] 2. Use the R language "relaimpo" package to perform relative importance analysis on the regression model and calculate the contribution rate of each driving factor to the total variance. The calculation of relative importance weight is based on the contribution of the driving factor in the decomposition model to the explanatory power of the target variable. The specific formula is:

[0148]

[0149] in, Indicates that only the driving factor Y is considered i is the predicted variance of the target variable X(t).

[0150] 3. According to the calculation results, the driving factors are sorted according to their contribution rate. The factor with the highest contribution rate is the main driving factor, followed by the secondary driving factor.

[0151] The analysis results of this example show that the main effects of driving factors are different at different time scales.

[0152] For example:

[0153] On the interannual scale, El Niño-Southern Oscillation (ENSO) is the main driving factor, contributing up to 62.5% to the precipitation index;

[0154] On the multidecadal scale, the Pacific Decadal Oscillation (PDO) and the Atlantic Multidecadal Oscillation (AMV) have higher contribution rates;

[0155] On a century scale, total solar radiation (TSI) is the main driving factor, contributing 62.7% to the precipitation index.

[0156] In addition, this example further analyzes the difference in the impact of driving factors on precipitation changes before and after the Industrial Revolution. Before the Industrial Revolution, the precipitation index was mainly affected by natural driving factors (such as total solar radiation and Atlantic multidecadal oscillation); after the Industrial Revolution, the contribution of anthropogenic forcing factors (such as greenhouse gas and aerosol emissions) to the precipitation index gradually increased, showing a significant change in time correlation.

[0157] Through sliding correlation analysis and statistical attribution analysis, this embodiment can accurately quantify the impact of climate driving factors on the precipitation index, and effectively distinguish the main and secondary driving factors in different periods and time scales, providing an important quantitative basis for subsequent precipitation mechanism analysis.

[0158] For step S6, in this embodiment, in order to fully reveal the main climate driving mechanism of regional precipitation changes, combined with the analysis results of the correlation and contribution rate of the precipitation index and climate driving factors in the previous steps, from the interannual, interdecadal, multi-decadal and century scales, the main driving factors and their action mechanisms on different time scales are deeply explored.

[0159] The core of this embodiment is to comprehensively summarize the analysis results of multiple time scales and systematically explain how driving factors affect regional precipitation changes through the climate system from the perspective of physical mechanisms.

[0160] On an interannual scale, this example shows that the El Nino-Southern Oscillation (ENSO) is the dominant factor. The warm phase of ENSO (El Nino) directly affects the atmospheric circulation pattern in the Pacific-Indian Ocean region by causing abnormal sea surface temperatures in the tropical Pacific, including the eastward shift and weakening effect of the Walker circulation. Specifically, it is manifested as follows:

[0161] ΔP∝-ΔSST

[0162] Among them, ΔP represents the change in precipitation, and ΔSST represents the anomaly of sea surface temperature. The East Asian monsoon circulation weakens due to the increase in Pacific SST, resulting in a decrease in water vapor transport to northern and central China, which in turn causes a decrease in precipitation. Conversely, during the cold phase (La Niña) of ENSO, the sea surface temperature anomaly is negative, the Walker circulation strengthens, water vapor transport increases, and precipitation increases.

[0163] On the decadal scale, the Pacific Decadal Oscillation (PDO) is found to be the main driving factor in this example. The warm phase of the PDO affects the intensity and position of the western Pacific subtropical high pressure by regulating the sea surface temperature anomaly in the North Pacific region. The study shows that:

[0164] ΔH ∝ ΔPDO

[0165] Among them, ΔH represents the position change of the subtropical high pressure, and ΔPDO represents the PDO index. The warm phase of PDO causes the subtropical high pressure to move southward, weakening the ability of water vapor transport to the study area, thereby reducing precipitation; while in the cold phase of PDO, the subtropical high pressure moves northward, water vapor transport is enhanced, and precipitation increases.

[0166] On a multidecadal scale, the Atlantic Multidecadal Oscillation (AMV) is found to be the main driving factor in this example. AMV triggers a remote response effect of atmospheric circulation by changing the North Atlantic sea surface temperature anomaly, especially by heating the middle and upper troposphere of Eurasia, regulating the monsoon circulation. According to the Thermal Wind Effect, the warm phase of AMV enhances the thermal gradient between the Indian Ocean and Eurasia:

[0167] ΔT∝ΔAMV

[0168] Among them, ΔT represents the change in temperature gradient, and ΔAMV represents the AMV index. This process ultimately strengthens the Indian summer monsoon and the East Asian summer monsoon, increases the water vapor transport in northern and central China, and leads to increased precipitation; while in the cold phase of the AMV, the opposite effect is manifested.

[0169] On a century scale, this example found that total solar radiation (TSI) is the main driving factor. TSI affects the intensity of the monsoon system and the water vapor transport capacity by changing the surface radiation balance and the land-sea thermal contrast. When solar activity increases, surface radiation increases, the land warms faster than the ocean, the land-sea thermal difference increases, and the monsoon intensity increases, resulting in increased precipitation. The basic process can be expressed as:

[0170] ΔR∝ΔTSI

[0171] Among them, ΔR represents the change in water vapor transport, and ΔTSI represents the change in total solar radiation. Conversely, when solar activity weakens, precipitation decreases significantly.

[0172] In addition, this example also analyzes the phased changes in the effects of driving factors before and after the Industrial Revolution. Before the Industrial Revolution, regional precipitation changes were mainly controlled by natural driving factors (such as TSI and AMV); after the Industrial Revolution, the influence of anthropogenic forcing factors (such as greenhouse gas and aerosol emissions) increased significantly. For example, aerosol emissions change the thermal contrast of the region by weakening the surface shortwave radiation, which in turn has an important impact on the monsoon circulation and precipitation distribution. This effect of anthropogenic forcing can be expressed as:

[0173] ΔQ∝-ΔA

[0174] Among them, ΔQ represents the change in precipitation, and ΔA represents the change in aerosol concentration.

[0175] This example provides a theoretical basis for understanding the complexity of regional precipitation changes through a systematic analysis of the physical mechanisms of the main driving factors at different time scales. This multi-time scale driving factor analysis method can not only explain the historical changes in precipitation, but also provide scientific support for future climate change adaptation strategies.

[0176] In general, the present invention uses ensemble empirical mode decomposition, multi-time scale decomposition, power spectrum analysis, wavelet analysis, sliding correlation and statistical attribution techniques to comprehensively reveal the correlation characteristics and phase relationship between precipitation index and climate driving factors at interannual, interdecadal, multi-decadal and century scales. By quantitatively analyzing the contribution rates of major and minor driving factors at different time scales, and combining the physical mechanisms of the climate system, the dominant factors and their mechanisms of action of regional precipitation changes are deeply explained. This method is highly scientific and practical, and provides a theoretical basis and technical support for understanding the climate-driven process of precipitation changes and responding to climate change.

[0177] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A precipitation mechanism analysis method based on regional climate data, characterized in that: The following steps are involved: Select the target study area and collect climate data, including precipitation index and climate driving factor data; Based on the ensemble empirical mode decomposition method, the precipitation index and driving factor data are decomposed into components of different time scales; Perform power spectrum analysis on the decomposed data to identify significant periodic features at each time scale; The correlation and phase relationship between precipitation index and climate driving factors at different time scales are analyzed using continuous wavelet transform, cross wavelet transform and wavelet coherence analysis. Through sliding correlation analysis and statistical methods, the impact of driving factors on precipitation index is quantified, and the main and secondary driving factors are identified; Based on the relationship between precipitation index and driving factors on different time scales, the main climate driving mechanism of regional precipitation changes is analyzed.

2. The precipitation mechanism analysis method based on regional climate data according to claim 1 is characterized in that: The climate data include tree ring width data, drought and flood index data, instrument observation data and climate reanalysis data. The precipitation index is obtained by multi-source data fusion, including the comprehensive analysis results of tree ring width, historical climate records and modern meteorological observation data.

3. The precipitation mechanism analysis method based on regional climate data according to claim 1 is characterized in that: The climate driving factor data include El Niño-Southern Oscillation, Pacific Decadal Oscillation, Atlantic Multi-Decadal Oscillation, total solar radiation and anthropogenic forcing factor data.

4. The precipitation mechanism analysis method based on regional climate data according to claim 1 is characterized in that: The ensemble empirical mode decomposition method comprises the following steps: Add low-intensity white noise to the raw climate data; The empirical mode decomposition method is used to decompose the data after adding white noise into multiple intrinsic mode functions; Repeat the above steps, using different white noise sequences, to decompose the data n times; The intrinsic mode functions obtained by n decompositions are averaged to obtain the final components.

5. The precipitation mechanism analysis method based on regional climate data according to claim 1 is characterized in that: The power spectrum analysis adopts a frequency analysis method based on Fourier transform to decompose the total energy of climate data into different frequency components, and determines the dominant period by calculating the variance contribution of each frequency component.

6. The precipitation mechanism analysis method based on regional climate data according to claim 1 is characterized in that: The significant periodic features identified by the power spectrum analysis include interannual cycles, decadal cycles, multi-decadal cycles and century cycles.

7. The precipitation mechanism analysis method based on regional climate data according to claim 1 is characterized in that: In the step of analyzing the correlation and phase relationship between precipitation index and climate driving factors at different time scales: Continuous wavelet transform is used to identify non-stationary periodic signals of precipitation index and driving factor data in the time-frequency space; Cross-wavelet transform is used to analyze the common periodic characteristics and relative phases of precipitation index and driving factors in time-frequency space; Wavelet coherence analysis was used to identify areas of significant correlation between precipitation indices and driving factors.

8. The precipitation mechanism analysis method based on regional climate data according to claim 1 is characterized in that: The impact of driving factors on the precipitation index quantified by the sliding correlation analysis includes: Determine the time series of driving factors and precipitation indices; Select the sliding window length and adjust the window range according to the analysis requirements; The correlation coefficient between the driving factor and the precipitation index is calculated in each sliding window to obtain the dynamic correlation changing over time; Analyze the changing characteristics of correlations over time and identify significant correlations in key time periods.

9. The precipitation mechanism analysis method based on regional climate data according to claim 1, characterized in that: The steps of determining primary and secondary driving factors include: The R language relaimpo package was used to conduct quantitative attribution analysis on driving factors; By calculating the relative importance weight of each driving factor, the contribution rate of different driving factors to the precipitation index is quantified; According to the contribution rate, the main driving factors and secondary driving factors are determined.

10. The precipitation mechanism analysis method based on regional climate data according to claim 1, characterized in that: The steps of analyzing the main climate driving mechanism of regional precipitation change based on the relationship between precipitation index and driving factors on different time scales include: Based on the correlation analysis results between precipitation index and driving factors at interannual, interdecadal, multidecadal and century scales, the main driving factors at different time scales are determined; The specific impact mechanisms of driving factors on precipitation changes are analyzed for different time scales, including: The impact mechanism of El Niño-Southern Oscillation on interannual precipitation variation; The mechanism of the impact of the Pacific Decadal Oscillation on multi-decadal precipitation changes; The impact mechanism of Atlantic multidecadal oscillation and total solar radiation on century-scale precipitation changes; Compare the relative effects of natural driving factors and anthropogenic forcing factors on precipitation changes before and after the Industrial Revolution, and determine the differences in the impact of climate driving factors on precipitation in different periods.

Images