CN114330850A - Abnormal relative tendency generation method and system for climate prediction - Google Patents
Abnormal relative tendency generation method and system for climate prediction Download PDFInfo
- Publication number
- CN114330850A CN114330850A CN202111570454.9A CN202111570454A CN114330850A CN 114330850 A CN114330850 A CN 114330850A CN 202111570454 A CN202111570454 A CN 202111570454A CN 114330850 A CN114330850 A CN 114330850A
- Authority
- CN
- China
- Prior art keywords
- abnormal
- prediction
- climate
- relative tendency
- recent
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 152
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000009499 grossing Methods 0.000 claims abstract description 16
- 238000004364 calculation method Methods 0.000 claims description 27
- 230000004044 response Effects 0.000 claims description 17
- 238000004590 computer program Methods 0.000 claims description 6
- 230000001932 seasonal effect Effects 0.000 claims description 5
- 230000000737 periodic effect Effects 0.000 claims description 4
- 238000012935 Averaging Methods 0.000 claims description 3
- 230000002708 enhancing effect Effects 0.000 claims description 3
- 230000005856 abnormality Effects 0.000 abstract description 9
- 238000001914 filtration Methods 0.000 description 27
- 238000004422 calculation algorithm Methods 0.000 description 13
- 230000001364 causal effect Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 8
- 230000008859 change Effects 0.000 description 7
- 238000011160 research Methods 0.000 description 7
- 230000007774 longterm Effects 0.000 description 6
- 238000012545 processing Methods 0.000 description 6
- 238000012417 linear regression Methods 0.000 description 5
- 230000004048 modification Effects 0.000 description 4
- 238000012986 modification Methods 0.000 description 4
- 238000001228 spectrum Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000002547 anomalous effect Effects 0.000 description 2
- 230000002238 attenuated effect Effects 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000002787 reinforcement Effects 0.000 description 2
- 238000005316 response function Methods 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses an abnormal relative tendency generation method and system for climate prediction. Carrying out time smoothing on the climate variable original data, and removing high-frequency variability; subtracting the climate state of the smooth data to obtain the pitch; on the basis, according to the target frequency band, selecting the length of the early-stage average time period at each moment, and defining a corresponding recent abnormal background; subtracting a recent abnormal background, and removing low-frequency variability to obtain an abnormal relative tendency; modeling and predicting the relative tendency of the anomaly, substituting and adding low-frequency variability serving as a known near-term anomaly background, and realizing the prediction of the pitch and the original field. The method solves the problems of time boundary and multi-scale of climate prediction, only historical and current data are needed to extract target frequency range information at the end of a time sequence, the relative tendency of high-frequency abnormality on the background of the known recent abnormality can be highlighted, only the relative tendency of abnormality needs to be predicted, errors caused by the introduction of predicted low-frequency variability are avoided, and the accuracy and the stability of climate prediction are effectively improved.
Description
Technical Field
The invention belongs to the field of digital signal processing and climate prediction, and particularly relates to a target frequency range information extraction method only using historical and current data and a climate prediction method based on the method, which can be used for climate prediction of seasons, sub-seasons and other time scales.
Background
The current common filtering method for climate research. Climate research usually focuses on changes of climate systems in certain specific frequency bands, and filters (filters) need to be applied to original data in order to extract information of the concerned frequency bands and weaken interference of other frequency bands on the research. The filters commonly used in climate research are currently mainly of the FFT filter (steiner, 2021), Lanczos filter (Lanczos, 1956; Duchon,1979) and Butterworth filter (Butterworth, 1930; Murakami,1979) three types. The FFT filter belongs to a frequency domain filter and is based on a theoretical basis through discrete Fourier transform; lanczos and Butterworth filters belong to time-domain filters and are based on discrete-time fourier transforms.
The existing common filtering method is not applicable to climate prediction. Climate prediction usually uses a climate variability strong signal on a certain time scale as a prediction factor to predict the climate state of the time scale in the future, which requires that information of climate variables on the corresponding time scale is extracted in real time, and filtering is realized at the end of an observation time sequence. The equivalent time domain calculation of the frequency domain FFT filter relates to the period extension of an input time sequence, the end filtering result is influenced by the initial end data, and the change of the length of the input sequence can cause the change of the filtering result at a fixed moment, so the method is not suitable for the real-time filtering of continuously adding the observation data and is only suitable for the situation of fixed data length. The calculation of the Lanczos filter relates to data at a future moment, the calculation cannot be performed at the end of a sequence, the filtered data at the end is lost, the Lanczos filter needs a higher order, namely more filter coefficients, to obtain a better frequency response characteristic, and the more the data at the end is lost. The Butterworth filter has a nonlinear phase response characteristic, which may cause waveform distortion after filtering, and the calculation of the Butterworth filter involves recursion, and values outside the boundary are assumed to start the recursion at the end points of the sequence, for example, zero padding continuation outside the boundary; in practical application, in order to counteract nonlinear phase response caused by the Butterworth filter, forward-backward bidirectional filtering is performed, and an intermediate filtering result at a future moment is assumed during backward filtering. The three common filters for climate research are non-causal systems, that is, historical, current and future data are used for calculation in the time domain or equivalent time domain calculation in the frequency domain, so that the filters cannot be operated at the tail end of a real-time observation record, cannot provide target frequency band component information required by climate prediction, and can only be applied to retrospective research. From the perspective of continuation, the three common filtering methods respectively perform cycle, null and zero padding continuation outside the end boundary of the time sequence, and cannot obtain an end filtering result or obtain an unreal end filtering result.
Existing solutions to the climate forecast end problem. Aiming at the problem that the traditional filter can not realize filtering at the end of a time sequence, two types of solutions of boundary continuation (Mann, 2004; Mann, 2008; Frankcombe et al, 2015) and algorithm modification (Wheeler and Hendon, 2004; Arguez, 2008; Kikuchi et al, 2012; Lee et al, 2013; Kiladis et al, 2014; Hsu et al, 2015; Qian et al, 2019) exist. The boundary continuation scheme keeps the full-sequence filtering algorithm unchanged, future time data are set artificially in the forms of zero padding, boundary symmetry and the like, and the non-causal filtering algorithm can run in a tail end interval. The algorithm modification scheme does not relate to data extension outside the boundary, but modifies a filtering algorithm in the terminal interval into a causal algorithm, only uses historical and current data for calculation, and performs special treatment on the terminal interval.
The problem of the existing climate forecast end problem solution exists. The boundary continuation scheme does not modify a filtering algorithm, but introduces false future data for terminal filtering calculation to obtain unreal terminal filtering results, and introduces a serious initial value error for climate prediction. The algorithm modification scheme adopts a special algorithm in a terminal interval, the magnitude and phase of the processed data may be inconsistent with the time sequence obtained by traditional filtering, and the processing mode of the front and back inconsistency can obviously reduce the return skill of the real-time prediction skill (Wang et al, 2020); in addition, the formulation of the end-specific algorithm depends on specific historical data of specific problems, and the maximum correlation coefficient is obtained by solving the traditional filtering result to determine relevant parameters of the specific algorithm, so that the statistical and empirical limitations exist, and the general property and the general applicability are lacked.
The invention aims to solve the technical problem. The present invention is directed to solving the problems presented by the prior art described above. (1) Causality problem: the abnormal relative tendency generation method provided by the invention belongs to a causal system, and only history and current data are used for calculation, so that the difficulty that the traditional filter as a non-causal system cannot filter at the end of a time sequence is overcome; (2) initial value error problem: the abnormal relative tendency provided by the invention does not relate to any extended future data in calculation, and an initial value error cannot be introduced into climate prediction due to the processing process of the invention; (3) handling the consistency problem: the method provided by the invention has the function of a non-traditional filter, can be used for processing all data including terminal points consistently, ensures that the obtained time sequences are consistent in magnitude and continuous in phase, and is favorable for establishing a prediction model with stable performance; (4) the universality problem is as follows: the method provided by the invention is used as a non-traditional filter, various general properties of the time domain and the frequency domain are completely determined by filter coefficients of the non-traditional filter, and the non-traditional filter has general applicability in climate prediction of multiple scales such as seasons, sub-seasons and the like.
Disclosure of Invention
The purpose of the invention is as follows: aiming at the problems that the conventional common filtering method cannot be applied at the end of a time sequence in climate research and the problems that the conventional solution of the end problem can introduce an initial value error to climate prediction, the magnitude and phase front and back are inconsistent, and the universal property and universality are lacked, the invention provides the abnormal relative tendency generation method and the system for the climate prediction, which are used as a causal system.
The technical scheme is as follows: in order to achieve the purpose, the invention adopts the technical scheme that:
an abnormal relative tendency generation method for climate prediction comprises the following steps:
(1) carrying out time smoothing on the original data of the climate variables, and removing high frequency variability, for example, calculating a ten-day average value from day-by-day data, removing a weather scale variability, or calculating a seasonal average value from month-by-month data, removing a sub-seasonal variability;
(2) subtracting the climate state of the smooth data to obtain the pitch, for example, calculating a ten-day climate state annual cycle from the long-term ten-day smooth data, and subtracting the climate state to obtain the ten-day pitch;
(3) selecting the length of the early-stage average period of each moment according to the target frequency band, and defining a corresponding Recent Abnormal Background (RAB), wherein the step is essentially low-pass filtering to the distance and extracting low-frequency variability;
(4) the recent abnormal background is subtracted from the pitch, the low frequency variability is removed, and an Abnormal Relative Trend (ART) is obtained.
When the climate variable original data acquired in the step (1) is daily average, averaging the daily data by ten days (10 days) or 5 days), and removing the weather scale variability; and when the acquired climate variable original data is a monthly average value, carrying out seasonal average on the monthly data, and removing the sub-seasonal variability.
The concrete calculation of the step (3) is as follows: and (3) representing the data obtained in the step (2) from any n time in the horizontal time series by x [ n ]. Defining the recent abnormal background of x [ n ] as the average value of the previous B continuous time data, and expressing the recent abnormal background obtained by the step (3) by r [ n ]:
wherein w is a weight coefficient satisfyingThe specific value of the parameter B can be selected from 1,2,3 and other values according to the amplitude response curve characteristic of the abnormal relative tendency filter.
The concrete calculation of the step (4) is as follows: and (5) representing the data of the abnormal relative trend time sequence obtained in the step (4) at any time n by y [ n ], wherein y [ n ] is equal to the difference between the distance x [ n ] and the recent abnormal background r [ n ]:
the method for predicting the climate by using the abnormal relative tendency generation method comprises the following steps:
obtaining the abnormal relative tendency of a prediction factor and a prediction object by adopting the abnormal relative tendency generation method for climate prediction; a climate prediction model is established by utilizing the prediction factors and the abnormal relative tendency of the prediction object, the future abnormal relative tendency of the prediction object is predicted, the known recent abnormal backgrounds are substituted and added to realize the prediction of the distance, and the climate state is further substituted and added to realize the prediction of the original field.
Based on the same inventive concept, the invention provides an abnormal relative tendency generation system for climate prediction, which comprises:
the time smoothing module is used for performing time smoothing on the climate variable original data and removing high frequency variability;
the pitch calculation module is used for subtracting the climate state of the smooth data to obtain the pitch;
the recent abnormal background calculation module is used for selecting the length of the early-stage average time period at each moment according to the target frequency band and defining a corresponding recent abnormal background;
and the abnormal relative tendency calculation module is used for subtracting the recent abnormal background from the pitch, removing the low-frequency variability and obtaining the abnormal relative tendency.
Based on the same inventive concept, the invention provides a climate prediction system based on abnormal relative tendency, which comprises:
the time smoothing module is used for performing time smoothing on the climate variable original data and removing high frequency variability;
the pitch calculation module is used for subtracting the climate state of the smooth data to obtain the pitch;
the recent abnormal background calculation module is used for selecting the length of the early-stage average time period at each moment according to the target frequency band and defining a corresponding recent abnormal background;
the abnormal relative tendency calculation module is used for subtracting the recent abnormal background from the distance to the horizontal plane and removing the low-frequency variability to obtain the abnormal relative tendency;
and the modeling prediction module is used for establishing a climate prediction model by utilizing the prediction factor and the abnormal relative tendency of the prediction object, predicting the future abnormal relative tendency of the prediction object, substituting and adding the known recent abnormal backgrounds to realize the prediction of the pitch, and further substituting and adding the climate states to realize the prediction of the original field.
Based on the same inventive concept, the invention provides a computer system, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the computer program is loaded to the processor to realize the abnormal relative tendency generation method for climate prediction or the climate prediction method based on the abnormal relative tendency.
Has the advantages that: compared with the prior art, the invention has the following advantages:
(1) the method solves the problem of time boundary of climate prediction, and does not introduce initial value errors. The abnormal relative tendency generation method provided by the invention belongs to a causal system, only historical and current data are needed to extract target frequency band information at the end of a time sequence, future data continuation is not involved, and an initial value error is not introduced into climate prediction due to the processing process of the method.
(2) The invention has the advantages of processing consistency and easy reducibility. The abnormal relative tendency generation method provided by the invention can be applied to all data including terminal points, and can ensure the consistency of time sequences on magnitude and phase in predictive modeling; the method provided by the invention decomposes a pitch time sequence into two parts of abnormal relative tendency and recent abnormal background, predicts the abnormal relative tendency, substitutes and adds the recent abnormal background, can restore the pitch, substitutes and adds the climate state, and can restore the original field of the variable.
(3) The invention has universal properties and universal applicability. The method for generating the abnormal relative tendency is essentially a non-traditional filter, the property of the method is completely determined by the filter coefficient in the time domain, and the method does not depend on the comparison between specific data with specific problems and the traditional filtering method; by selecting a proper recent abnormal background algorithm, the method is generally applicable to climate prediction of multiple scales such as seasons, sub-seasons and the like.
(4) The invention solves the multi-scale problem of climate prediction. The prominent abnormal relative tendency of the invention is a higher frequency component on the known recent abnormal background, when the climate prediction is carried out, only the abnormal relative tendency needs to be predicted, the low frequency variability is taken as the known recent abnormal background and substituted and added, the prediction of the pitch and the original field is realized, the error caused by the introduced prediction of the low frequency variability is avoided, and the accuracy and the stability of the climate prediction are effectively improved.
Drawings
FIG. 1 is a block diagram of an abnormal relative tendency generation system for climate prediction and a climate prediction system based on abnormal relative tendency disclosed herein;
FIG. 2 is a diagram of an amplitude response function of an abnormal relative trend filter according to an embodiment of the present invention, where a parameter B takes values of 1,2, and 3, respectively, and determines a weight coefficient of a recent abnormal background based on a binomial coefficient;
FIG. 3 is a time series diagram of Nino3.4 pitch and relative tendency of anomaly in the winter of 1950-2020 in the embodiment of the present invention, where the value of parameter B is 2, and the weight coefficient of recent anomaly background is taken
FIG. 4 is a power spectrum density chart of Nino3.4 pitch and relative tendency of anomaly in the winter of 1950-2020 in the embodiment of the present invention, wherein the value of parameter B is 2, and the weight coefficient of recent anomaly background is taken
Fig. 5 is a distance-level correlation coefficient diagram of the actual and return of domestic rainfall distance in summer in each year of the cross-checking and independent return period in the embodiment of the invention, and specifically, a unitary linear regression climate prediction model is established by respectively taking the winter nino3.4 distance and the abnormal relative tendency as prediction factors and taking the next summer europe rainfall distance or the abnormal relative tendency as prediction objects, so as to predict the next summer europe rainfall distance. Carrying out the same data preprocessing on the Nino3.4 index in winter and the European-Asia rainfall field in the next summer, expressing the distance between the two indexes as 0, taking the parameter B as 2 for the abnormal relative tendency, and taking the weight coefficient of the recent abnormal backgroundThe modeling and cross checking period is 30 years in 1981-2010, the independent return period is 10 years in 2011-2020, and lattice points in a Chinese range are selected to calculate a distance correlation coefficient;
FIG. 6 is a graph showing the return results obtained by using European-Asia rainfall distance in summer and the relative tendency of the former winter Nino3.4 distance and the former winter Nino3.4 abnormality as prediction factors in 2011-2015 in the embodiment of the present invention;
FIG. 7 is a graph showing the return results obtained by using the European Asian rainfall distance in summer 2016-2020 and the relative tendency of the former winter Nino3.4 distance and the former winter Nino3.4 abnormality as prediction factors respectively in the embodiment of the present invention.
Detailed Description
The invention is further described with reference to the following figures and specific examples, which are intended to illustrate the invention and are not intended to limit the scope of the invention. Various equivalent modifications of the invention, which fall within the scope of the appended claims of this application, will occur to persons skilled in the art upon reading this disclosure.
As shown in fig. 1, the abnormal relative tendency generating system for climate prediction disclosed by the present invention includes the following 4 modules, and the original data is input into the system and then sequentially passes through the modules, so that the abnormal relative tendency can be generated and output:
(1) a time smoothing module: and acquiring original data, performing time smoothing on the original data, and removing high frequency variability.
For example: when the original data is the daily average value of a certain climate variable and the prediction target is the climate average state in a certain period of the next seasonal scale in the future, the daily data is preferably averaged in ten days (10 days) or 5 days), and the interference of the high-frequency variability of the weather scale on the prediction is removed; when the original data is the monthly average value of a certain climate variable and the prediction target is the climate average state of a certain season in the future, the seasonal average of the monthly data is preferably calculated, and the interference of the sub-seasonal variability is removed.
(2) A pitch calculation module: and subtracting the climate state of the smoothed data to obtain the pitch.
When the smooth data obtained by the time smoothing module has time units in years such as climate, ten days, season and the like, calculating the climate average value of each climate, ten days and season by using the long-term historical data to form a climate state annual cycle, and subtracting the climate annual cycle from the smooth data to obtain the interval level; when the time unit of the smoothed data obtained by the time smoothing module is year, the climate average is calculated by using the long-term historical data, and is subtracted from the smoothed data to obtain the pitch. And recording the data at any n time in the obtained range time sequence as x [ n ].
(3) A recent anomaly background calculation module: and selecting the length of the early-stage average time period at each moment according to the target frequency band, and defining a corresponding recent abnormal background.
The essence of calculating the recent anomaly background is to low-pass filter the sequence of time-series of moments. And (3) taking r [ n ] to represent the recent abnormal background from any n time data in the flat time sequence x, and defining the recent abnormal background as the average value of x on B continuous times in the early stage of the n time:
wherein w is a weight coefficient satisfyingWhen the weighting coefficients are allAnd taking the arithmetic mean value of x at the previous stage of the n moments and B continuous moments as the recent abnormal background. When the weight coefficient is determined based on mathematical rules such as the binomial coefficient, and the like, the weighted average value of x at the previous B continuous moments of the n moments is used as the recent abnormal background, for example, when B is 3, the weight coefficient determined based on the binomial coefficient is
(4) An abnormal relative tendency calculation module: and subtracting the recent abnormal background from the pitch, and removing the low-frequency variability to obtain the abnormal relative tendency.
Subtracting the recent anomaly background from the pitch yields the relative propensity of the anomaly, removing the low frequency variability, retaining the higher frequency components of the pitch, which can be considered as high pass filtering the pitch. And representing the time data of the abnormal relative trend time sequence n by y [ n ], wherein y [ n ] is equal to the difference between the horizontal x [ n ] and the recent abnormal background r [ n ]:
(2) the discrete convolution form of equation:
the coefficient h defines a causal finite impulse response filter of order B, called the anomalous relative trend filter, and h is the impulse response of the filter. h discrete time fourier transform to the frequency response of the filter, denoted h (f):
the argument f denotes the normal frequency in "cycles/time", Δ T denotes the sampling interval in "time/sample".
The modulus of H (f) is called the amplitude response of the filter, reflecting the output sequence y [ n ]]And the input sequence x [ n ]]The ratio of the amplitudes of the frequency components in (3) indicates that the operation of the formula (3) will make x [ n ]]Which frequencies (periods) are enhanced or attenuated. Fig. 2 is an amplitude response function image of the abnormal relative inclination filter when the parameter B in the equation (3) is 1,2, and 3, respectively, and the weight coefficient w is determined based on the binomial coefficient. Independent variable in image is normalized periodRepresents the number of time-series samples required to constitute one cycle, in units of "sample/cycle", and takes a real number in the range of [2, ∞). The real time unit such as year, ten days, and waiting time can be arbitrarily specified. Three horizontal reference lines in the figure respectively represent a double power line, a unit power line and a half power line from top to bottom, the intersection points of the three lines and the amplitude response curve are represented by solid dots, and the standardized period corresponding to the intersection points is marked by a vertical reference line.
Fig. 2 shows that when B is 1, the amplitude response of the anomaly relative trend monotonically decreases with increasing period, the shortest period component of 2 samples/cycle becomes twice as large in amplitude and the power correspondingly becomes four times as large; the power of 4 samples/cycle is multiplied, the power of 6 samples/cycle is unchanged, and the power of 9 samples/cycle is attenuated to less than half of the original power. If the half power line is taken as a boundary for determining whether or not each period component is retained after filtering, when B is taken to be 1, the abnormal relative tendency retains the period component smaller than 9 samples/cycle, and acts as a reinforcement for the component smaller than 6 samples/cycle, and the shorter the period, the larger the reinforcement degree. When B is 2, the amplitude response of the anomalous relative trend has a maximum value, the periodic component of less than 13 samples/cycle is retained, the component of less than 9 samples/cycle is enhanced, and the component of about 3 to 5 samples/cycle is enhanced to more than twice the original power. When B is 3, the abnormally-relatively-inclined amplitude response has a minimum and a maximum, retains the periodic components less than 17 samples/cycle, maintains the power of 2 samples/cycle unchanged, enhances the components of about 3 to 11.5 samples/cycle, and enhances the power of the components of about 4 to 7 samples/cycle more than twice as high.
According to the frequency band concerned by climate prediction, selecting a proper early-stage average period length parameter B, determining a proper abnormal relative tendency algorithm, pertinently enhancing the main period of the signal as a prediction factor, and highlighting the change of the corresponding period in the prediction object.
As shown in FIG. 1, the climate prediction system based on abnormal relative tendency disclosed by the invention comprises 4 modules of an abnormal relative tendency generation system and a modeling prediction module, and the number of the modules is 5. The method comprises the steps of firstly inputting a prediction factor and original data of a prediction object into an abnormal relative tendency generation system, outputting the abnormal relative tendency of the prediction factor and the prediction object, then inputting the abnormal relative tendency of the prediction factor and the prediction object into a modeling prediction module, establishing a climate prediction model by using the abnormal relative tendency of the prediction factor and the prediction object and adopting methods such as linear regression or artificial neural network and the like, predicting the future abnormal relative tendency of the prediction object, then substituting and adding known abnormal backgrounds to realize the prediction of the distance to the horizon, further substituting and adding climate states to realize the prediction of an original field. The system finally outputs the prediction results of the predicted object future abnormal relative tendency, the distance and the original field.
The specific implementation process and effect of the abnormal relative tendency generation method are explained by taking the application of the abnormal relative tendency generation method for climate prediction disclosed by the invention to the annual variability highlighting the Nino3.4 index in winter as an example, a unitary linear regression climate prediction model is established by utilizing the abnormal relative tendency of the Nino3.4 in winter and the abnormal relative tendency of the European-Asia rainfall in the next summer to predict the rainfall level in the next summer, and compared with the traditional method for modeling and predicting by utilizing the rainfall level, the method for modeling and predicting by utilizing the abnormal relative tendency has obvious improvement effect on the accuracy and stability of the climate prediction.
(1) And carrying out time smoothing on the climate variable original data, and removing the high frequency variability.
The Nino3.4 index is the regional average of the surface temperature of the east Pacific (5N-5S, 170W-120W) sea in the tropics, and is commonly used for characterizing the states of Erlenno and southern billow (ENSO), wherein ENSO is the strongest annual scale signal in a climate system, and the period is 2-7 years. The obtained Nino3.4 index raw data is monthly average data of 1950-2020 and 71 years. The average of the winter of each year (12 months before the year and 1 and 2 months in the year) is calculated, the sub-seasonal variability is removed, a time series with the length of 71 is formed, and the time unit is converted from month to year.
(2) And subtracting the climate state of the smoothed data to obtain the pitch.
Calculating the climate average value of the Nino3.4 index in winter by using the data of 1981-2010 in the time series obtained in the step (1), and subtracting the average value from the original sequence to obtain the Nino3.4 distance time series in winter of 1950-2020, wherein the time series is shown as a dot-dash line in FIG. 3.
(3) And selecting the length of the early-stage average time period at each moment according to the target frequency band, and defining a corresponding recent abnormal background.
According to fig. 2, when the time unit is year and the parameter B is 2, that is, when the average of 2 years in the early stage of each year is used as the recent abnormal background, the relative tendency of abnormality can be highlighted for a period of 2 to 9 years, and the power of the apparatus is enhanced to more than twice of the original power for a period of 3 to 5 years, so that the apparatus has the function of highlighting the annual variability and can pertinently enhance the ENSO signal, and therefore, the average of 2 years in the early stage of each year is defined as the recent abnormal background.
(4) And subtracting the recent abnormal background from the pitch, and removing the low-frequency variability to obtain the abnormal relative tendency.
The recent abnormal background of each year is subtracted from the Nino3.4 winter interval to obtain the relative trend time series of Nino3.4 winter abnormalities in 1950-2020, which is shown as a solid line in FIG. 3.
Fig. 3 shows that the winter nino3.4 pitch exhibits mainly the characteristics of the annual variation in some periods, while the relative inclination of the abnormality of nino3.4 in the same period exhibits mainly the characteristics of the annual variation, i.e. highlights the annual variability. Fig. 4 is a power spectrum density diagram of nino3.4 pitch and abnormal relative trend in the winter of 1950 to 2020, and it can be known that the power spectrum density of the annual change rate of nino3.4 pitch in the winter is equivalent to the power spectrum density of the annual change rate, and if the annual change rate and the annual change rate of the climate are taken into consideration when the winter nino3.4 pitch is directly used as a predictor to perform climate prediction; compared with the Nino3.4 in winter, the chronologic rate of the abnormal relative tendency of the Nino3.4 in winter is weakened, the chronologic rate is obviously enhanced, and the annual scale ENSO signal is highlighted.
(5) A climate prediction model is established by utilizing the prediction factors and the abnormal relative tendency of the prediction object, the future abnormal relative tendency of the prediction object is predicted, the known recent abnormal backgrounds are substituted and added to realize the prediction of the distance, and the climate state is further substituted and added to realize the prediction of the original field.
Respectively taking the Nino3.4 pitch and the abnormal relative tendency in winter as prediction factors, taking the European-Asia rainfall pitch or the abnormal relative tendency in summer in the next year as prediction objects, establishing a unitary linear regression climate prediction model, predicting the European-Asia rainfall pitch in summer in the next year, and comparing the prediction effects of the two modeling methods.
Carrying out the same data preprocessing on the Nino3.4 index in winter and the European-Asia rainfall field in summer of the next year, wherein the horizontal distance is represented by a parameter B which is 0, namely no recent abnormal background exists; the parameter B of the relative tendency of the anomaly is 2, and the weight coefficient of the recent anomaly background is takenThe modeling and cross-checking period of the unitary linear regression climate prediction model is 30 years in 1981-2010, the independent return period is 10 years in 2011-2020, and grid points in a Chinese range are selected to calculate the distance correlation coefficient (ACC).
FIG. 5 is a diagram of the pitch-flat correlation coefficients of the actual rainfall pitch and the return of China in summer in each year of cross-checking and independent return periods, the results obtained by pitch-flat modeling are represented by boxes and broken lines in the diagram, and the results obtained by abnormal relative tendency modeling are represented by columns in the diagram. In the cross-examination period, the average ACC obtained by abnormal relative tendency modeling is 0.05, the average ACC obtained in 30 years is a positive value, the average ACC obtained in 30 years is-0.35, and the average ACC obtained in 30 years is a positive value only in 4 years; in the independent return period, the average ACC obtained by abnormal relative tendency modeling is 0.24, 8 years in 10 years are positive values, the average ACC obtained by flat modeling is 0.07, and 6 years in 10 years are positive values; the average ACC obtained by modeling abnormal relative inclination is 0.10, and 29 years in 40 years are positive and account for 72.5% in the whole period of cross-examination and independent return, and the average ACC obtained by modeling is-0.25, and only 10 years in 40 years are positive and account for 25.0%. From the distance modeling to the abnormal relative tendency modeling, the long-term average ACC is converted from a negative value to a meaningful positive value, and the prediction accuracy is obviously improved; the positive ACC of the cross test and the independent return is changed from minority to majority in year, and the prediction stability is obviously improved.
FIGS. 6 and 7 show the independent results of the above reports obtained by using the relative trends of the former winter Nino3.4 distance and the former winter Nino3.4 abnormality as predictors, in summer European rainfall range in 2011-2015 and in summer 2016-2020. The image shows that the return magnitude obtained by the distance modeling is low in systematicness, and the return result equivalent to the live magnitude can be obtained by the abnormal relative tendency modeling.
In the embodiment, on the annual scale, the abnormal relative tendency of the Nino3.4 index in winter is used as a prediction factor, the abnormal relative tendency of European-Asia rainfall in the next year is used as a prediction object, the influence of an ENSO signal of the annual scale is highlighted, the annual variability and the long-term tendency are not predicted, but are directly substituted as the known recent abnormal background and added with the prediction result of the abnormal relative tendency to form the prediction of the European-Asia rainfall range in the next year, the error caused by the introduction of the predicted annual variability and the long-term tendency is avoided, and the accuracy and the stability of climate prediction are obviously improved.
Claims (10)
1. An abnormal relative tendency generation method for climate prediction is characterized by comprising the following steps:
(1) carrying out time smoothing on the climate variable original data, and removing high-frequency variability;
(2) subtracting the climate state of the smooth data to obtain the pitch;
(3) selecting the length of the early-stage average time period at each moment according to the target frequency band, and defining a corresponding recent abnormal background;
(4) and subtracting the recent abnormal background from the pitch, and removing the low-frequency variability to obtain the abnormal relative tendency.
2. The method for generating abnormal relative inclination for climate prediction according to claim 1, wherein when the climate variable raw data obtained in step (1) is a daily average, a weather scale rate is removed by averaging day-by-day data in ten days (10 days) or in 5 days); and when the acquired climate variable original data is a monthly average value, carrying out seasonal average on the monthly data, and removing the sub-seasonal variability.
3. The abnormal relative inclination generation method for climate prediction according to claim 1, wherein said step (3) is specifically calculated as: and (3) representing the data at any time n in the horizontal time sequence obtained in the step (2) by using x [ n ], representing the recent abnormal background of x [ n ] by using r [ n ], and defining the average value of the data at the previous B continuous times at the previous stage of the n times as:
4. The abnormal relative inclination generation method for climate prediction according to claim 3, wherein said step (4) is specifically calculated as: and representing data at any time n in the abnormal relative trend time sequence by y [ n ], wherein y [ n ] is equal to the difference between the distance x [ n ] and the recent abnormal background r [ n ]:
5. the abnormal relative inclination generating method for climate prediction according to claim 3, wherein the length parameter B of the early-stage averaging period for calculating the recent abnormal background is selected according to the characteristic of the amplitude response curve of the abnormal relative inclination filter, and the characteristic of the amplitude response curve for different values of B is as follows:
when B is 1, the amplitude response is monotonously reduced along with the increase of the period, the generated abnormal relative tendency retains the period component less than 9 samples/cycle, and plays a role in enhancing the component less than 6 samples/cycle, the shorter the period is, the greater the enhancement degree is, the amplitude of the component with the shortest period of 2 samples/cycle is doubled, and the power is correspondingly quadrupled;
when B is 2, the amplitude response has a maximum value, the generated abnormal relative tendency retains a periodic component less than 13 samples/cycle, the component less than 9 samples/cycle is enhanced, and the power of the component from 3 to 5 samples/cycle is enhanced to more than twice of the original power;
when B is 3, the amplitude response has a minimum value and a maximum value, the generated abnormal relative tendency retains the periodic component less than 17 samples/cycle, the power of 2 samples/cycle is kept unchanged, the enhancing effect is exerted on the component of 3 to 11.5 samples/cycle, and the power of the component of 4 to 7 samples/cycle is enhanced to be more than twice of the original power.
6. A climate prediction method based on abnormal relative tendencies, comprising: adopting the abnormal relative tendency generation method for climate prediction according to any one of claims 1-5 to obtain the abnormal relative tendency of the prediction factor and the prediction object;
a climate prediction model is established by utilizing the prediction factors and the abnormal relative tendency of the prediction object, the future abnormal relative tendency of the prediction object is predicted, the known recent abnormal backgrounds are substituted and added to realize the prediction of the distance, and the climate state is further substituted and added to realize the prediction of the original field.
7. An abnormal relative tendency generating system for climate prediction, characterized by comprising:
the time smoothing module is used for performing time smoothing on the climate variable original data and removing high frequency variability;
the pitch calculation module is used for subtracting the climate state of the smooth data to obtain the pitch;
the recent abnormal background calculation module is used for selecting the length of the early-stage average time period at each moment according to the target frequency band and defining a corresponding recent abnormal background;
and the abnormal relative tendency calculation module is used for subtracting the recent abnormal background from the pitch, removing the low-frequency variability and obtaining the abnormal relative tendency.
8. A climate prediction system based on abnormal relative tendencies, comprising:
the time smoothing module is used for performing time smoothing on the climate variable original data and removing high frequency variability;
the pitch calculation module is used for subtracting the climate state of the smooth data to obtain the pitch;
the recent abnormal background calculation module is used for selecting the length of the early-stage average time period at each moment according to the target frequency band and defining a corresponding recent abnormal background;
the abnormal relative tendency calculation module is used for subtracting the recent abnormal background from the distance to the horizontal plane and removing the low-frequency variability to obtain the abnormal relative tendency;
and the modeling prediction module is used for establishing a climate prediction model by utilizing the prediction factor and the abnormal relative tendency of the prediction object, predicting the future abnormal relative tendency of the prediction object, substituting and adding the known recent abnormal backgrounds to realize the prediction of the pitch, and further substituting and adding the climate states to realize the prediction of the original field.
9. A computer system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the computer program, when loaded into the processor, implements the abnormal relative tendency generation method for climate prediction according to any of claims 1-5.
10. A computer system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the computer program when loaded into the processor implements the climate prediction method based on relative propensity to anomalies according to claim 6.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111570454.9A CN114330850B (en) | 2021-12-21 | 2021-12-21 | Abnormal relative trend generation method and system for climate prediction |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111570454.9A CN114330850B (en) | 2021-12-21 | 2021-12-21 | Abnormal relative trend generation method and system for climate prediction |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114330850A true CN114330850A (en) | 2022-04-12 |
CN114330850B CN114330850B (en) | 2023-11-17 |
Family
ID=81055227
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111570454.9A Active CN114330850B (en) | 2021-12-21 | 2021-12-21 | Abnormal relative trend generation method and system for climate prediction |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114330850B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117909888A (en) * | 2024-03-20 | 2024-04-19 | 南京大学 | Intelligent artificial intelligence climate prediction method |
CN117950087A (en) * | 2024-03-21 | 2024-04-30 | 南京大学 | Artificial intelligence downscale climate prediction method based on large-scale optimal climate mode |
Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1744185A1 (en) * | 2005-07-13 | 2007-01-17 | CIMA-Università degli Studi di Genova | Method for estimating a variable geophysical field particularly for forecasting atmospheric precipitation |
CA2820129A1 (en) * | 2010-12-15 | 2012-06-21 | Eni S.P.A. | Medium-long term meteorological forecasting method and system |
US20170176640A1 (en) * | 2014-03-28 | 2017-06-22 | Northeastern University | System for Multivariate Climate Change Forecasting With Uncertainty Quantification |
CN107103396A (en) * | 2017-06-13 | 2017-08-29 | 南京大学 | The Chinese seasonal climate Forecasting Methodology modeled based on main SVD mode |
WO2018214060A1 (en) * | 2017-05-24 | 2018-11-29 | 北京质享科技有限公司 | Small-scale air quality index prediction method and system for city |
CN109583653A (en) * | 2018-11-30 | 2019-04-05 | 南京信息工程大学 | The extended peroid forecasting procedure of NORTHWESTERN PACIFIC TYPHOON based on statistical model |
CN109670540A (en) * | 2018-12-04 | 2019-04-23 | 华南理工大学 | It is resident number variation tendency Forecasting Approach for Short-term in Passenger Transport Hub region based on kNN algorithm |
KR102011689B1 (en) * | 2019-03-06 | 2019-08-19 | 주식회사 위엠비 | Method for monitoring time-series data, System for monitoring time-series data and Computer program for the same |
KR20190138716A (en) * | 2018-06-05 | 2019-12-16 | 동국대학교 산학협력단 | Disease prediction apparatus using climate change data and operating method thereof |
CN111260111A (en) * | 2020-01-08 | 2020-06-09 | 中国科学院上海技术物理研究所苏州研究院 | Runoff forecast improvement method based on meteorological big data |
AU2020104000A4 (en) * | 2020-12-10 | 2021-02-18 | Guangxi University | Short-term Load Forecasting Method Based on TCN and IPSO-LSSVM Combined Model |
CN113052430A (en) * | 2020-01-10 | 2021-06-29 | 国网江苏省电力有限公司 | Analysis system and analysis method for omnibearing power economy prosperity index |
CN113270901A (en) * | 2021-06-16 | 2021-08-17 | 华能澜沧江水电股份有限公司 | Multi-energy power supply control method giving consideration to battery state and active power dynamic stability |
CN113592165A (en) * | 2021-07-23 | 2021-11-02 | 南京大学 | Method and system for predicting sub-season climate based on ten-day tendency and physical modal modeling |
CN113706035A (en) * | 2021-08-31 | 2021-11-26 | 北京伟杰东博信息科技有限公司 | Prediction calling method and system thereof |
-
2021
- 2021-12-21 CN CN202111570454.9A patent/CN114330850B/en active Active
Patent Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP1744185A1 (en) * | 2005-07-13 | 2007-01-17 | CIMA-Università degli Studi di Genova | Method for estimating a variable geophysical field particularly for forecasting atmospheric precipitation |
CA2820129A1 (en) * | 2010-12-15 | 2012-06-21 | Eni S.P.A. | Medium-long term meteorological forecasting method and system |
US20170176640A1 (en) * | 2014-03-28 | 2017-06-22 | Northeastern University | System for Multivariate Climate Change Forecasting With Uncertainty Quantification |
WO2018214060A1 (en) * | 2017-05-24 | 2018-11-29 | 北京质享科技有限公司 | Small-scale air quality index prediction method and system for city |
CN107103396A (en) * | 2017-06-13 | 2017-08-29 | 南京大学 | The Chinese seasonal climate Forecasting Methodology modeled based on main SVD mode |
KR20190138716A (en) * | 2018-06-05 | 2019-12-16 | 동국대학교 산학협력단 | Disease prediction apparatus using climate change data and operating method thereof |
CN109583653A (en) * | 2018-11-30 | 2019-04-05 | 南京信息工程大学 | The extended peroid forecasting procedure of NORTHWESTERN PACIFIC TYPHOON based on statistical model |
CN109670540A (en) * | 2018-12-04 | 2019-04-23 | 华南理工大学 | It is resident number variation tendency Forecasting Approach for Short-term in Passenger Transport Hub region based on kNN algorithm |
KR102011689B1 (en) * | 2019-03-06 | 2019-08-19 | 주식회사 위엠비 | Method for monitoring time-series data, System for monitoring time-series data and Computer program for the same |
CN111260111A (en) * | 2020-01-08 | 2020-06-09 | 中国科学院上海技术物理研究所苏州研究院 | Runoff forecast improvement method based on meteorological big data |
CN113052430A (en) * | 2020-01-10 | 2021-06-29 | 国网江苏省电力有限公司 | Analysis system and analysis method for omnibearing power economy prosperity index |
AU2020104000A4 (en) * | 2020-12-10 | 2021-02-18 | Guangxi University | Short-term Load Forecasting Method Based on TCN and IPSO-LSSVM Combined Model |
CN113270901A (en) * | 2021-06-16 | 2021-08-17 | 华能澜沧江水电股份有限公司 | Multi-energy power supply control method giving consideration to battery state and active power dynamic stability |
CN113592165A (en) * | 2021-07-23 | 2021-11-02 | 南京大学 | Method and system for predicting sub-season climate based on ten-day tendency and physical modal modeling |
CN113706035A (en) * | 2021-08-31 | 2021-11-26 | 北京伟杰东博信息科技有限公司 | Prediction calling method and system thereof |
Non-Patent Citations (1)
Title |
---|
王昱: "一种基于全球动力模式和SMART 原理结合的 统计降尺度区域季节气候预测方法", 《JOURNAL OF THE METEOROLOGICAL SCIENCES》, vol. 41, no. 5, pages 569 - 583 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117909888A (en) * | 2024-03-20 | 2024-04-19 | 南京大学 | Intelligent artificial intelligence climate prediction method |
CN117909888B (en) * | 2024-03-20 | 2024-05-28 | 南京大学 | Intelligent artificial intelligence climate prediction method |
CN117950087A (en) * | 2024-03-21 | 2024-04-30 | 南京大学 | Artificial intelligence downscale climate prediction method based on large-scale optimal climate mode |
Also Published As
Publication number | Publication date |
---|---|
CN114330850B (en) | 2023-11-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11888316B2 (en) | Method and system of predicting electric system load based on wavelet noise reduction and EMD-ARIMA | |
CN114330850B (en) | Abnormal relative trend generation method and system for climate prediction | |
CN110376457B (en) | Non-invasive load monitoring method and device based on semi-supervised learning algorithm | |
CN103576060A (en) | Partial discharge signal denoising method based on wavelet adaptive threshold | |
CN108399434B (en) | Analysis and prediction method of high-dimensional time series data based on feature extraction | |
CN110212592B (en) | Thermal power generating unit load regulation maximum rate estimation method and system based on piecewise linear expression | |
CN117434153B (en) | Road nondestructive testing method and system based on ultrasonic technology | |
CN117236084B (en) | Intelligent management method and system for woodworking machining production | |
CN111931983A (en) | Precipitation prediction method and system | |
CN114977216A (en) | Parameter identification method of oscillation signal and terminal | |
CN115902528B (en) | Method for identifying oscillation and short-circuit faults of direct-current traction network | |
CN107732940B (en) | Power system stabilizer parameter optimization test method based on ADPSS | |
CN113375065B (en) | Method and device for eliminating trend signal in pipeline leakage monitoring | |
CN110969238A (en) | Method and device for calibrating electricity consumption data | |
CN115859048A (en) | Noise processing method and device for partial discharge signal | |
CN117668496B (en) | Geogrid strain testing method and system | |
CN113052938A (en) | Method and device for constructing boiler energy efficiency curve | |
Alfaouri et al. | Quality evaluation techniques of processing the ECG signal | |
CN117613905B (en) | Power demand mid-term prediction method and system based on multidimensional component decomposition | |
CN117454778B (en) | Water storage amount data processing method, device and storage medium | |
CN113470674B (en) | Voice noise reduction method and device, storage medium and computer equipment | |
CN117349661B (en) | Method, device, equipment and storage medium for extracting vibration signal characteristics of plunger pump | |
CN117056665A (en) | Non-invasive load identification method and system for high-frequency aliasing data | |
CN112966851A (en) | Short-term load change trend prediction method | |
CN116026881A (en) | Test section construction method and system based on measured temperature data |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |