CN115983511B - Precipitation prediction method and system based on improved statistical downscaling method - Google Patents
Precipitation prediction method and system based on improved statistical downscaling method Download PDFInfo
- Publication number
- CN115983511B CN115983511B CN202310285989.4A CN202310285989A CN115983511B CN 115983511 B CN115983511 B CN 115983511B CN 202310285989 A CN202310285989 A CN 202310285989A CN 115983511 B CN115983511 B CN 115983511B
- Authority
- CN
- China
- Prior art keywords
- precipitation
- data
- gcms
- matrix
- rainfall
- 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.)
- Active
Links
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
Abstract
The invention discloses a precipitation prediction method and a system based on an improved statistical downscaling method, wherein the method comprises the steps of collecting basic data of daily scales in a preset range and preprocessing; acquiring the preprocessed basic data, calculating precipitation parameters of observed precipitation data and contemporaneous GCMs precipitation data of each ground rainfall station, and acquiring an average value of the precipitation parameters; calling a CLIGEN model based on the average precipitation parameters to generate reduced-scale GCMs precipitation data; calculating a spatial correlation coefficient of the observed precipitation data; and (3) evaluating the precision of the GCMs downscaling precipitation data, if the precision meets the requirements, replacing the contemporaneous GCMs precipitation data with the GCMs precipitation data in future scenes, and downscaling to obtain the downscaling precipitation characteristics of the future period. The method can be used for carrying out downscaling treatment on the GCMs precipitation data under different future shared socioeconomic paths and typical concentration path scenes, and is high in prediction accuracy.
Description
Technical Field
The invention relates to a precipitation data prediction method, in particular to a precipitation prediction method and system based on an improved statistical scale reduction method.
Background
Global climate patterns (GCMs) are considered to be the most important and feasible method for researching the influence of climate change, can provide important data input for researching the influence of climate change on water circulation of a river basin, and has important significance for developing strategies for coping with climate risks. However, GCMs are limited to the complexity and computational effort of the global climate system, which typically provide grid-scale climate change information with a spatial resolution of hundreds of kilometers, whereas hydrologic models typically require basin-or site-specific scale climate data, resulting in input data that does not match the hydrologic model spatial scale.
Therefore, spatial downscaling of the GCMs data is required. The dynamic and statistical downscaling methods are used to make up the gap between the GCMs output and the hydrological model data requirements. The statistical downscaling method can correct systematic errors of global climate modes, and is widely used without considering the influence of boundary conditions on prediction results. As an effective downscaling method, random weather generators can reproduce daily sequences with similar statistical features, but most weather generators can only simulate spatially uncorrelated single site meteorological data, which will affect the accuracy of the simulation of watershed extreme flows and the output of the hydrological model.
Therefore, the inventor provides a multi-station statistical downscaling method, and the method simultaneously performs downscaling processing of time and space scales through steps of spatial downscaling, temporal downscaling, spatial correlation reconstruction and the like, converts data of a coarse grid daily scale into data of a fine grid daily scale, and has the problem that the processing precision of daily scale rainfall trend information of different stations in the same grid is insufficient during processing, namely, the problem of non-uniformity of a subgrid. Therefore, further research innovation is required.
Disclosure of Invention
The invention aims to: a precipitation prediction method based on an improved statistical downscaling method is provided, and a system for realizing the method is further provided, so as to solve the problems in the prior art.
According to one aspect of the present invention, there is provided a precipitation prediction method based on an improved statistical downscaling method, comprising the steps of:
step S1, acquiring basic data of a daily scale in a preset range, wherein the basic data comprise observed precipitation data of a ground rainfall station and at least two groups of GCMs precipitation data; preprocessing the basic data to obtain preprocessed observed precipitation data and GCMs precipitation data;
S2, acquiring preprocessed basic data, calculating precipitation parameters of observed precipitation data and contemporaneous GCMs precipitation data of each ground rainfall station, and acquiring an average value of the precipitation parameters; calling a CLIGEN model based on the average precipitation parameters to generate reduced-scale GCMs precipitation data;
s3, calculating spatial correlation coefficients of observed precipitation data of all sites, and performing spatial correlation reconstruction on the downscaled GCMs precipitation data by using a two-stage shuffle algorithm; reconstructing annual change trend of the contemporaneous GCMs downscaling data according to the annual precipitation change trend of the observed precipitation data;
and S4, evaluating the precision of the GCMs downscaling precipitation data, if the precision meets the preset requirement, replacing the synchronous GCMs precipitation data with the GCMs precipitation data in future scenes, downscaling to obtain downscaling precipitation characteristics in future periods, and drawing related images.
According to one aspect of the present application, the step S2 is further:
s21, calculating precipitation parameters of observed precipitation data and contemporaneous GCMs precipitation data of each ground rainfall station, wherein the precipitation parameters comprise a daily precipitation average value, a daily precipitation standard deviation, a deviation coefficient of the daily precipitation, a daily precipitation-precipitation transition probability and a daily non-precipitation transition probability;
S22, determining four GCM grid center points closest to the ground rainfall station according to an angular distance calculation formula; respectively calculating the CLIGEN precipitation parameter change rate of each period, and calculating the CLIGEN precipitation parameters of each period at the ground rainfall station position by using a bilinear interpolation method;
and S23, collecting and averaging CLIGEN precipitation parameters of each GCMs, and calling a CLIGEN model based on the average precipitation parameters to generate reduced-scale GCMs precipitation data.
According to one aspect of the application, the angular distance calculation formula is: d=cos -1 （cosθ d cosθ s （cosβ d cosβ s +sinθ d sinβ d ）+sinθ d sinθ s ）；θ d The target rainfall station latitude;β d a target rain station longitude;θ s the latitude of the central point of the GCM data grid;β s longitude is the center point of the GCM data grid;
the process of calculating the CLIGEN precipitation parameter change rate in each period is as follows: CP (control program) m =（CP m-p /CP m-c ）×CP o-c ；
CP m CLIGEN parameters for the calculated future period GCM precipitation data;CP m-p CLIGEN parameters for the future period GCM precipitation data;CP m-c CLIGEN parameters for historical period GCM precipitation data;CP o-c the CLIGEN parameter of precipitation data is observed for historical periods.
According to one aspect of the application, the step S1 is further:
s11, acquiring GCMs (graphic arts model) precipitation data, wherein the GCMs precipitation data adopt a preset number of CMIP6 mode historical period precipitation data and future period precipitation data; the historical period precipitation data are statistical test data, and the future period precipitation data are combined scene data of four shared socioeconomic paths and typical concentration paths, namely SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP 5-8.5; the preparation of the GCMs precipitation data in the historical period is required to be intercepted based on the period of the observed precipitation data, so that the period of the observed precipitation data and the period of the observed precipitation data of the GCMs are consistent;
Step S12, adopting a bilinear interpolation method to carry out statistical interpolation on each GCMs precipitation data to obtain a grid with preset precision;
s13, calculating and adopting a correlation coefficient, a standard root mean square error, a relative standard deviation and a TSS index to evaluate the precision of the GCMs precipitation data, and if the TSS index of a certain GCM is greater than a threshold value, using the precipitation data as input data of precipitation prediction and importing the precipitation data into a database;
and S14, performing deviation correction on GCMs precipitation data with the accuracy reaching a threshold value by adopting an equidistant cumulative distribution function method.
According to one aspect of the present application, the step S3 is further:
s31, arranging daily basic data according to month, and constructing 12 precipitation data matrixes n multiplied by kX]Wherein the matrix of observed precipitation data is recorded as [ [X obs ]CLIGEN precipitation data matrix is recorded as [X GEN ]N is the total number of days per month during the study period,kthe number of the rainfall stations;
step S32, CLIGEN precipitation data matrixX GEN ]Each column of data is ordered according to the size to obtain the rank order of each dataiVan der Waals scoring matrix [ S ] based on rank order calculation matrix]Where s=Φ -1 （i/m+1），SIs a van der Waals score;Φ -1 is an inverse function of normal distribution;irank order for each data in each column; mTotal data for each column;
step S33, calculating the observed precipitation data matrixX obs ]Is derived from the spearman rank correlation coefficient of (2)k×kMatrix of [ ofC S-obs ]And judges the matrixC S-obs ]Whether or not the matrix is positive, if so, obtaining the matrix by using Cholesky decomposition methodC S-obs ]Upper triangular matrix of (2)R]If not, then further matrix [C S-obs ]Decomposition into feature vectorsVAnd characteristic valueDSubstituting positive number for negative value in feature vector, and using corrected feature value diagonal matrixD’Generating matrix [C S-obs-m ]Finally, matrix [C S-obs-m ]After standardized treatmentMatrix obtained by Cholesky decomposition methodC S-obs ]Upper triangular matrix of (2)R]；
Step S34, reconstructing Van der Waals score matrixS * ]The calculation formula is as follows: [S * ]=[S][R]The method comprises the steps of carrying out a first treatment on the surface of the Calculating Van der Waals scoring matrixS * ]Is to matrix CLIGEN precipitation data [X GEN ]According to Van der Waals scoring matrixS * ]Adjusting the rank to obtain a reconstructed matrixX GEN-R ]；
Step S35, calculate matrix [X GEN-R ]Pearson correlation coefficient matrix of [C P-GEN-R ]Establishing Pearson correlation coefficient matrixC P-GEN-R ]And matrix [C S-obs ]Linear relation of (c): [C S-obs ]=a×[C P-GEN-R ]+b；
Step S36, calculating Pearson correlation coefficient matrix of observed precipitation data [C P-obs ]Using Pearson correlation coefficient matrixC P-obs ]Will be the formula [C S-obs ]=a×[C P-GEN-R ]In +b [C P-GEN-R ]Substitution is performed to obtain a new matrix C S-obs ’ ]；
Step S37, use matrix [C S-obs ’ ]Replacement of step S35C S-obs ]And repeating the steps S35 to S36 until the precision meets the requirement.
According to one aspect of the application, the method further comprises a precipitation trend recovery process:
step S38, calculating a correlation coefficient among stations based on rainfall station observation rainfall sequences, selecting one rainfall station as a control station, and taking the rainfall station with the largest average correlation coefficient with other rainfall stations as the control station;
step S39, rearranging the column data according to the rainfall generating structure of the control station in the rainy days, randomly arranging the non-rainfall days of the column, and adjusting the data in other columns; and calculating the annual precipitation rank of the observed precipitation data, and adjusting the histologic test data of the GCMs at the same site and at the same period according to the annual precipitation rank of the observed precipitation data so as to ensure that the annual precipitation trend is unchanged.
According to one aspect of the present application, the step S4 is further:
s41, calculating CLIGEN parameters, spatial correlation coefficients and average absolute errors, standard root mean square errors and decision coefficients of drainage basin precipitation of observed precipitation data and histologic test data of synchronous GCMs; evaluating the precision of a downscaling method based on a CLIGEN model and a two-stage shuffle algorithm;
Step S42, obtaining or calculating precipitation characteristics, wherein the precipitation characteristics comprise average daily, monthly and annual precipitation of each station and extreme precipitation indexes, and calculating average annual precipitation of a basin in a future period;
and step S43, drawing a comparison histogram, a scatter diagram, a box diagram and a frequency curve based on the rainfall characteristic.
According to an aspect of the application, the step S39 further includes:
step S39a, acquiring an observation rainfall sequence in a preset period of a control station, extracting daily rainfall data, dividing preset intervals according to a maximum value and a minimum value, and setting a marker value for each interval; constructing a mapping relation between daily precipitation data and the marker; converting the observed rainfall sequences of all rainfall stations into marker value sequences to form a marker value sequence set;
step S39b, acquiring histologic test data of contemporaneous GCMs, sequencing precipitation data according to the annual precipitation rank of the observed precipitation data, constructing daily precipitation and constructing an identifier numerical sequence set;
step S39c, calculating the sum of Euclidean distances between the identifier value sequence of the observed precipitation sequence of each site and the precipitation data acquired by the GCMs one by one, judging whether the sum is smaller than a preset value, and if the sum is smaller than the preset value, considering that the regulated annual precipitation trend accords with the expectation.
According to an aspect of the application, the step S22 further includes:
step S22a, reading the position parameters of each ground rainfall station and the position parameters of each GCM grid center point, determining four GCM grid center points nearest to each ground rainfall station by adopting an angular distance calculation formula, calculating the repetition times of each GCM grid center point, and constructing a weight coefficient matrix of each GCM grid;
step S22b, acquiring GCM precipitation data and CLIGEN precipitation parameters of observed precipitation data in future periods and historical periods, and calculating the CLIGEN precipitation parameter change rate of each period; judging whether the variation rate difference value of the rainfall parameters of each ground rainfall station in each grid exceeds a threshold value one by one, and if so, marking the first scale-reducing grid as the first scale-reducing grid;
and step S22c, judging whether abnormal points exist in the rainfall parameters of each ground rainfall station in each grid, and if so, calculating the CLIGEN rainfall parameters of each period at the position of each ground rainfall station by using a bilinear interpolation method.
According to another aspect of the present invention, there is provided a precipitation prediction system based on an improved statistical downscaling method, comprising:
at least one processor; and
a memory communicatively coupled to at least one of the processors; wherein, the liquid crystal display device comprises a liquid crystal display device,
The memory stores instructions executable by the processor for execution by the processor to implement the precipitation prediction method based on the improved statistical downscaling method according to any one of the above technical solutions.
The beneficial effects are that: according to the invention, the distance between the rainfall station and the adjacent grid is calculated and interpolation processing is carried out, so that the problem that the rainfall stations in the same GCM output grid can adopt the same change rate in the prior art is solved, and the respective change rate is calculated for the rainfall stations in each output grid, thereby providing assistance for accurately predicting precipitation. The application solves the technical problem existing in the prior art all the time.
Drawings
Fig. 1 is an overall flow chart of the present invention.
Fig. 2 is a flowchart of step S2 of the present invention.
FIG. 3 is a scatter plot of the spatial correlation coefficient of downscaled data versus the spatial correlation coefficient of observed precipitation data of the present invention.
FIG. 4 is a chart of rainfall comparison boxes of estimated and observed drainage basin rainfall data in the future period according to the present invention.
FIG. 5 is a schematic diagram of the invention prior to reconstruction of historical period basin annual precipitation data trend.
FIG. 6 is a schematic diagram of the invention after the trend of the historical period basin annual precipitation data is reconstructed.
FIG. 7 is a schematic representation of the invention prior to reconstruction of the trend of the basin annual precipitation data at a future time period.
FIG. 8 is a schematic representation of the present invention after reconstruction of the trend of the basin annual precipitation data at a future time period.
Detailed Description
In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the present invention. It will be apparent, however, to one skilled in the art that the invention may be practiced without one or more of these details. In other instances, well-known features have not been described in detail in order to avoid obscuring the invention.
Embodiment 1, as shown in fig. 1, provides a precipitation prediction method based on an improved statistical downscaling method, which comprises the following steps:
step S1, acquiring basic data of a daily scale in a preset range, wherein the basic data comprise observed precipitation data of a ground rainfall station and at least two groups of GCMs precipitation data; preprocessing the basic data to obtain preprocessed observed precipitation data and GCMs precipitation data;
s2, acquiring preprocessed basic data, calculating precipitation parameters of observed precipitation data and contemporaneous GCMs precipitation data of each ground rainfall station, and acquiring an average value of the precipitation parameters; calling a CLIGEN model based on the average precipitation parameters to generate reduced-scale GCMs precipitation data;
S3, calculating spatial correlation coefficients of observed precipitation data of all sites, and performing spatial correlation reconstruction on the downscaled GCMs precipitation data by using a two-stage shuffle algorithm; reconstructing annual change trend of the contemporaneous GCMs downscaling data according to the annual precipitation change trend of the observed precipitation data;
and S4, evaluating the precision of the GCMs downscaling precipitation data, if the precision meets the preset requirement, replacing the synchronous GCMs precipitation data with the GCMs precipitation data in future scenes, downscaling to obtain downscaling precipitation characteristics in future periods, and drawing related images.
Example 2 as shown in fig. 2, the step S2 is further:
s21, calculating precipitation parameters of observed precipitation data and contemporaneous GCMs precipitation data of each ground rainfall station, wherein the precipitation parameters comprise a daily precipitation average value, a daily precipitation standard deviation, a deviation coefficient of the daily precipitation, a daily precipitation-precipitation transition probability and a daily non-precipitation transition probability;
s22, determining four GCM grid center points closest to the ground rainfall station according to an angular distance calculation formula; respectively calculating the CLIGEN precipitation parameter change rate of each period, and calculating the CLIGEN precipitation parameters of each period at the ground rainfall station position by using a bilinear interpolation method;
And S23, collecting and averaging CLIGEN precipitation parameters of each GCMs, and calling a CLIGEN model based on the average precipitation parameters to generate reduced-scale GCMs precipitation data.
In embodiment 3, the angular distance calculation formula is: d=cos -1 （cosθ d cosθ s （cosβ d cosβ s +sinθ d sinβ d ）+sinθ d sinθ s ）；θ d The target rainfall station latitude;β d a target rain station longitude;θ s the latitude of the central point of the GCM data grid;β s longitude is the center point of the GCM data grid;
the process of calculating the CLIGEN precipitation parameter change rate in each period is as follows: CP (control program) m =（CP m-p /CP m-c ）×CP o-c ；
CP m CLIGEN parameters for the calculated future period GCM precipitation data;CP m-p CLIGEN parameters for the future period GCM precipitation data;CP m-c CLIGEN parameters for historical period GCM precipitation data;CP o-c the CLIGEN parameter of precipitation data is observed for historical periods.
In the aspect of data selection of future rainfall data prediction, the scheme uses CMIP6 global climate mode data, and compared with the traditional climate mode data, the data has more perfect climate change situation and higher simulation precision, and can effectively improve the precision of rainfall prediction results; in terms of downscaling model selection, the application of CLIGEN is limited because it is a single site weather generator and there is no sophisticated method to calculate the CLIGEN parameters for precipitation at future times. The invention solves the problem of poor spatial correlation of each site by using a two-stage Shuffles algorithm, and further provides a GLIGEN parameter generation method for precipitation in the future period, so that the application range of a CLIGEN model is greatly improved; in the aspect of annual rainfall trend, because the random weather generator randomly generates rainfall according to probability, it is difficult to ensure that the rainfall area after the downscaling is the same as the original rainfall sequence, and the recognition of the hydrologic process of the current area can be influenced.
Aiming at the problem, the invention provides a method for reconstructing annual precipitation trend of downscaling data based on original sequence rank order. The spatial correlation and annual precipitation trend reconstruction method provided by the invention can be used as a post-processing technology to be coupled with any single-site downscaling method, so that not only is the application range of other single-point downscaling methods enriched, but also a possible way for processing other meteorological data is provided.
The method solves the defects that the existing single-site statistical downscaling method ignores spatial correlation and annual precipitation trend cannot be accurately simulated, solves the problem of space-time mismatch between climate model output and hydrological model data demand, and improves universality of the single-site statistical downscaling method and accuracy of estimated precipitation data.
In embodiment 4, the step S39 further includes:
step S39a, acquiring an observation rainfall sequence in a preset period of a control station, extracting daily rainfall data, dividing preset intervals according to a maximum value and a minimum value, and setting a marker value for each interval; constructing a mapping relation between daily precipitation data and the marker; converting the observed rainfall sequences of all rainfall stations into marker value sequences to form a marker value sequence set;
Step S39b, acquiring histologic test data of contemporaneous GCMs, sequencing precipitation data according to the annual precipitation rank of the observed precipitation data, constructing daily precipitation and constructing an identifier numerical sequence set;
step S39c, calculating the sum of Euclidean distances between the identifier value sequence of the observed precipitation sequence of each site and the precipitation data acquired by the GCMs one by one, judging whether the sum is smaller than a preset value, and if the sum is smaller than the preset value, considering that the regulated annual precipitation trend accords with the expectation.
In the prior art, the future trend is not corrected, and the trend of the rainfall data generated by the GCM is opposite to the actual situation or has a large difference, so that the prediction cannot be better performed.
In embodiment 5, the step S22 further includes:
step S22a, reading the position parameters of each ground rainfall station and the position parameters of each GCM grid center point, determining four GCM grid center points nearest to each ground rainfall station by adopting an angular distance calculation formula, calculating the repetition times of each GCM grid center point, and constructing a weight coefficient matrix of each GCM grid;
Step S22b, acquiring GCM precipitation data and CLIGEN precipitation parameters of observed precipitation data in future periods and historical periods, and calculating the CLIGEN precipitation parameter change rate of each period; judging whether the variation rate difference value of the rainfall parameters of each ground rainfall station in each grid exceeds a threshold value one by one, and if so, marking the first scale-reducing grid as the first scale-reducing grid;
and step S22c, judging whether abnormal points exist in the rainfall parameters of each ground rainfall station in each grid, and if so, calculating the CLIGEN rainfall parameters of each period at the position of each ground rainfall station by using a bilinear interpolation method.
In order to solve the problems that in the prior art, the rainfall characteristic of the future can not be truly reflected due to insufficient spatial precision of the data of the lunar scale is reduced to the daily scale, the method for reducing the daily scale of the rainfall data of the lunar scale to the daily scale is provided, and the method and the data limit can only reduce the scale to the grid, so that more than 1 rainfall station in the grid can not reflect the real situation under the assumption that the change trend is consistent. In the method, the change trend of different rainfall stations in the same grid can be the same or different by downscaling daily rainfall data to the level of the daily rainfall stations, so that the real physical process can be reflected more accurately.
Example 6 the procedure of the invention is as follows:
step 1, collecting ground rainfall station observation rainfall data and various GCMs rainfall data in a preset range in a rainfall data preparation module, performing quality inspection and screening on the observation rainfall data, determining a using time range, further performing interpolation processing on the various GCMs rainfall data, and importing the rainfall data into a database;
step 2, in the GCMs data screening and processing module, evaluating the precision of GCMs precipitation data, screening the GCMs precipitation data, and correcting deviation of the selected climate mode precipitation data;
step 3, in the GCMs data downscaling module, CLIGEN precipitation parameters of observed precipitation data and contemporaneous GCMs precipitation data are calculated, further, the CLIGEN precipitation parameters of all the GCMs are averaged in a set, and a CLIGEN model is called based on the CLIGEN precipitation parameters averaged in the set to generate downscaling GCMs precipitation data;
step 4, calculating a spatial correlation coefficient of observed precipitation data in a GCMs downscaling data spatial correlation reconstruction module, reconstructing the spatial correlation of the GCMs downscaling data by using a two-stage shuffle algorithm, and further reconstructing annual change trends of the contemporaneous GCMs downscaling data according to annual precipitation change trends of original precipitation data;
Step 5, in the GCMs data downscaling precision evaluation module, calculating an average absolute Error (Mean Absolute Error, MAE), a standard root Mean Square Error (Standardized Root-Mean-Square Error, SRMSE) and a decision coefficient (Determination Coefficient), and evaluating the precision of the GCMs downscaling data;
step 6, in a precipitation data prediction module, replacing the 3 contemporaneous GCMs precipitation data in the step with the GCMs precipitation data under the conditions of four sharing social and economic paths and typical concentration paths in the future, repeating the steps 3-4, performing downscaling treatment on the precipitation data in the future period, and further reconstructing the estimated annual change trend of the precipitation data;
and 7, analyzing the rainfall characteristics of the future period in a future rainfall analysis module, drawing related images, and importing the estimated future rainfall data into a database.
Example 7, procedure of step 1 is further as follows:
step 11, adopting 20 CMIP6 mode histories and future period precipitation data for the GCMs data, further adopting history period data as history test data, wherein the future period precipitation data is combined scene data of four shared socioeconomic paths and typical concentration paths, namely SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP 5-8.5;
Step 12, intercepting the GCMs precipitation data preparation in the historical period based on the period of the observed precipitation data, and ensuring consistency of the period of the history test precipitation data of the GCMs and the period of the observed precipitation data;
and step 13, performing interpolation processing on all GCMs precipitation data by adopting a bilinear interpolation method to obtain a grid with the statistical interpolation of 0.1 degree multiplied by 0.1 degree.
Example 8, procedure of step 2 is further as follows:
step 2-1, evaluating the precision of the GCMs precipitation data by using a standardized Taylor diagram, wherein further, the standardized Taylor diagram adopts evaluation criteria comprising a correlation coefficient, a standard root mean square error and a relative standard deviation, and introducingTSS(Taylor Skill Score) the index comprehensively evaluates the accuracy of GCMs data,TSScloser to 1 means higher accuracy.
Step 2-2, wherein the judgment standard of the screening GCMs precipitation data isTSSValue of ifTSSIf the value is greater than 0.5, using the model as input data of precipitation prediction and importing the input data into a database;
step 2-3, the deviation correcting method is an equidistant accumulated distribution function method (EDCDFm);
further, the cumulative distribution function of the precipitation data adopts the cumulative distribution function of the gamma distribution with two mixed parameters, and the formula is as follows: G（x）=（1-P）·H（x）+P·F（x) The method comprises the steps of carrying out a first treatment on the surface of the In the method, in the process of the invention,Pis the percentage of the month of rain;H(x) Is a jump function, 0 when no rain, 1 when rain;F(x) A Cumulative Distribution Function (CDF) of precipitation data is fitted to use the two-parameter gamma distribution. Wherein the two parameter gamma distribution fits a probability density function of precipitation data.
Example 9, the procedure of step 3 is further as follows:
step 3-1, wherein the CLIGEN precipitation parameters comprise a monthly daily precipitation average value (Pw), a monthly daily precipitation standard deviation (Sd), a monthly daily precipitation bias coefficient (Shew) and a monthly daily precipitation transition probability: precipitation-precipitation probability (P (w|w)), non-precipitation probability (P (w|d)), the calculation formula is as follows:
（P（W|W））=N ww /（N wd +N ww ）；
（P（W|D））=N dw /（N dw +N dd ）
in the method, in the process of the invention,N ww is precipitation-days of precipitation;N wd days with precipitation-without precipitation;N dw days without precipitation-days with precipitation;N dd days without precipitation-without precipitation.
Step 3-2, the future precipitation parameters are calculated based on the Delta method and the bilinear interpolation method (Delta-BI method), because the Delta method does not consider the sub-network non-uniformity of the GCM, namely, when the Delta method is used for calculating the future precipitation, the rainfall stations in the same GCM output grid can adopt the same change rate. Further, the Delta-BI method calculation flow is as follows: firstly, determining four GCM grid center points closest to the position of a rainfall station according to an angular distance formula; secondly, calculating the CLIGEN rainfall parameter change rate in the future period by using a Delta method; thirdly, calculating a CLIGEN precipitation parameter of a future period at the position of the rainfall station by using a bilinear interpolation method; finally, the CLIGEN precipitation parameters of all GCMs were averaged together. The calculation formula is as follows:
Angular distance calculation formula: d=cos -1 （cosθ d cosθ s （cosβ d cosβ s +sinθ d sinβ d ）+sinθ d sinθ s ）；θ d The target rainfall station latitude;β d a target rain station longitude;θ s the latitude of the central point of the GCM data grid;β s longitude is the center point of the GCM data grid;
the process of calculating the CLIGEN precipitation parameter change rate in each period is as follows: CP (control program) m =（CP m-p /CP m-c ）×CP o-c ；
CP m CLIGEN parameters for the calculated future period GCM precipitation data;CP m-p CLIGEN parameters for the future period GCM precipitation data;CP m-c CLIGEN parameters for historical period GCM precipitation data;CP o-c the CLIGEN parameter of precipitation data is observed for historical periods.
Example 10, the procedure of step 4 is further as follows:
step 4-1, sorting daily precipitation data according to month to obtain 12 pieces of daily precipitation data with the size of 12 piecesn×kMatrix of [ ofX]Wherein the observed data is recorded as [ [X obs ]The CLIGEN data is recorded as [ [X GEN ]WhereinnRepresenting the total number of days per month during the study period,kthe number of the rainfall stations;
step 4-2, will [ [X GEN ]Each column of data is ordered according to the size to obtain the rank order of each dataiFurther, a matrix [ S ] of Van der Waals score (van der Waerden scores) of the matrix is calculated based on the rank order]The calculation formula is as follows: s=Φ -1 (i/m+1) wherein S is a Van der Waals score; phi -1 Is an inverse function of normal distribution; i is the rank of each data in each column; m is the total number of data per column.
Step 4-3, calculating the observation data matrix respectivelyX obs ]Is derived from the spearman rank correlation coefficient (Spearman rank correlation coefficient)k×kMatrix of [ ofC S-obs ]；
Step 4-4, judgment [ [C S-obs ]Whether the matrix is positive or not, if so, the matrix is obtained by using Cholesky decomposition methodC S-obs ]Upper triangular matrix of (2)R]If not, then further will [C S-obs ]Decomposition into feature vectorsVAnd characteristic valueDThe negative value in the eigenvector is replaced by a very small positive number, and further, the corrected eigenvalue diagonal matrix is usedD’Generation [C S-obs-m ]Finally, will [C S-obs-m ]After normalization, the matrix is obtained by Cholesky decomposition methodC S-obs ]Upper triangular matrix of (2)R]The normalized formula is:C S-obs =C S-obs-m /（diag（C S-obs-m ）diag（C S-obs-m ）’） 1/2 the method comprises the steps of carrying out a first treatment on the surface of the In the method, in the process of the invention,C S-obs to be normalized and to repair feature vectorsAnd (5) observing a Spearman rank correlation coefficient matrix of precipitation after the positive period.
Step 4-5, obtaining matrix by Cholesky decomposition methodC S-obs ]Upper triangular matrix [ R ]]；
Step 4-6, reconstructing Van der Waals score matrixS * ]The calculation formula is as follows: [S * ]=[S][R]；
Step 4-7, calculate ]S * ]Further, matrix [X GEN ]According to [S * ]Adjusting the rank to obtain a reconstructed matrixX GEN-R ]；
Step 4-8, calculate matrix [X GEN-R ]Pearson correlation coefficient [ ofC P-GEN-R ]Build up ]C P-GEN-R ]And [ with ]C S-obs ]Linear relation of (c): [C S-obs ]=a×[C P-GEN-R ]+b；
Step 4-9, calculating Pearson correlation coefficient of observed precipitation data [ C P-obs ]Use [ useC P-obs ]Step 4-8 in the formula [C P-GEN-R ]Substitution is performed to obtain a new matrixC S-obs ’ ]；
Step 4-10, use [ [C S-obs ’ ]Replacement of step 4-5 [C S-obs ]And repeating the steps 4-5 to 4-9 until the precision meets the requirement.
Embodiment 11 further includes a trend recovery process, specifically as follows:
step 4-11, selecting one rainfall station as a control station, calculating the correlation coefficient among stations based on the rainfall station observation rainfall sequence by using the selection standard, and taking the rainfall station with the largest average correlation coefficient with other rainfall stations as the control station;
step 4-12, rearranging the column data according to a rainfall generating structure of a control station in a rainy day, and randomly arranging the non-rainfall days of the column, and further, adjusting the data in other columns;
and 4-13, calculating annual precipitation rank of the observed precipitation data, and further, adjusting the history test data of the same-station and same-period GCMs according to the annual precipitation rank of the observed precipitation data so as to ensure that the annual precipitation trend is unchanged.
Example 12, the procedure of step 5 is further as follows:
step 5-1, calculating CLIGEN parameters, spatial correlation coefficients and average absolute errors, standard root mean square errors and decision coefficients of drainage basin precipitation of observed precipitation data and history experimental data of synchronous GCMs;
And 5-2, evaluating the accuracy of a downscaling method based on the CLIGEN model and a two-stage shuffle algorithm.
Example 13, the procedure of step 6 is further as follows:
step 6-1, the future four shared socioeconomic paths and typical concentration path scenarios include SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP5-8.5;
step 6-2, repeating the steps 3-1 to 4-12, and estimating precipitation data of different scenes in a future period;
and 6-3, calculating the annual precipitation rank of the original GCMs in the future period, and further, adjusting the annual precipitation data of the GCMs at the site after the scale is reduced according to the annual precipitation rank of the original GCMs so as to ensure that the annual precipitation trend is unchanged.
Example 14, procedure 7 is further as follows:
the precipitation characteristics of the step 7-1 comprise average daily precipitation, average monthly precipitation and annual precipitation of each station and extreme precipitation indexes, and average annual precipitation of a basin in the future period.
And 7-2, drawing a comparison histogram, a scatter diagram, a box diagram and a frequency curve based on the calculation result of the step 7-1.
Taking a basin as an example, the water resources of the basin are relatively scarce, and the spatial and temporal distribution of the basin is uneven. The ecological environment and socioeconomic development of the basin are very sensitive to climate changes, in particular precipitation changes. The watershed is not affected by resource development at present, and the hydrologic process is mainly controlled by natural factors. This example uses observed daily precipitation data for 10 sites in a basin and 20 GCMs precipitation data for CMIP 6.
The method mainly comprises the following steps:
s1, precipitation data preparation: observed daily precipitation data of 10 sites of a river basin from 1976 to 2014, histologic test precipitation data of 20 GCMs of CMIP6 from 1976 to 2014, and 7 combined scenario precipitation data of 20 GCMs of CMIP6 from 2015 to 2100 were collected, and all the GCMs precipitation data were interpolated into a grid of 0.1 ° x 0.1 °.
S2, respectively using a standardized Taylor diagram to evaluate the precision of 20 GCMs precipitation data, screening the GCMs precipitation data, using an equidistant cumulative distribution function method (EDCDFm method) to carry out deviation correction on the screened GCMs precipitation data, and further, calculating the average GCMs annual precipitation statistical characteristics of the multi-mode set for comparing the precision improvement condition before and after deviation correction. The data are specifically as follows: the mean value of the observed data is 378.56, and the standard deviation is 84.64;
GCMs data before bias correction: the mean value is 502.65, the standard deviation is 23.82, and the mean absolute error is: 131.40 the relative deviation was 0.45 and the root mean square error was 156.24.
Bias post-subscription pre-GCMs data: the mean value was 375.46, the standard deviation was 82.05, the mean absolute error was 1.51, the relative deviation was 0.005, and the root mean square error was 103. S3, calculating CLIGEN precipitation parameters of the observed precipitation data and the GCMs precipitation data, further generating downscaled contemporaneous GCMs precipitation data based on the precipitation parameters, and reconstructing spatial correlation of the GCMs downscaled data by using a two-stage shuffle algorithm.
As shown in fig. 3, fig. 3 is an embodiment of the present invention: a scatter plot of the spatial correlation coefficient of downscaled data and the spatial correlation coefficient of observed precipitation data.
And S4, calculating average daily, monthly and annual precipitation and extreme precipitation indexes of each station in the future period, calculating average annual precipitation of the basin in the future period, and analyzing the precipitation characteristics in the future period.
As shown in FIG. 4, FIG. 4 is a rainfall comparison bin graph of estimated and observed drainage basin rainfall data at a future time period. It can be seen from fig. 4 that the effect of the present application is better than the prior art.
In this embodiment, aiming at most weather generators at present, a single site is mostly used as a research object, so that the reduced scale data of each site lacks of spatial correlation, and the simulation precision of the hydrographic element extremum in the flow domain is poor. The method is based on a CLIGEN random weather generator and a two-stage Shuffle algorithm, firstly, a method for generating the CLIGEN rainfall parameters in the future period is provided, and the application of the CLIGEN random weather generator in rainfall estimation is expanded; secondly, the spatial correlation of precipitation at each site is reconstructed by using a two-stage Shuffle algorithm, and the method can be used as a post-processing technology to be combined with other statistical downscaling methods, so that the transformation from a single-site downscaling method to a multi-site downscaling method is realized, and the method has a wide application prospect; finally, a annual precipitation trend reconstruction method based on the original precipitation sequence rank order is provided, the problem that annual precipitation data generated by the statistical downscaling method cannot guarantee trend is solved, and a method system of the statistical downscaling field is expanded. In a specific embodiment, it is found that a conclusion which is not completely consistent with a single-site downscaling method can be obtained by adopting a new downscaling method, namely, the single-site downscaling method has higher simulation precision in terms of annual precipitation average values, but the space correlation of each site is not considered, so that the generated watershed precipitation range is too small, and the simulation of extreme precipitation is poor.
As shown in fig. 5 to 8, in the embodiment of the present invention: the historical period and the future period are reconstructed by the annual precipitation estimated data trend of the watershed. The Euclidean distances of the MME (multimode average), BCC-CSM2-MR and CMCC-ESM2 of the measured annual precipitation and duration period are 228, 322 and 310 respectively, so that reconstruction can be carried out according to the trend of the MME in the future.
The embodiment mainly relates to geographic information, rainfall station observation and precipitation products of 20 global climate modes in information utilization. It should be noted that the method has strong expansibility, and other single-site downscaling methods can be adopted by the method. As described above, although the present invention has been shown and described with reference to certain preferred embodiments, it is not to be construed as limiting the invention itself. Various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (5)
1. The precipitation prediction method based on the improved statistical downscaling method is characterized by comprising the following steps of:
step S1, acquiring basic data of a daily scale in a preset range, wherein the basic data comprise observed precipitation data of a ground rainfall station and at least two groups of GCMs precipitation data; preprocessing the basic data to obtain preprocessed observed precipitation data and GCMs precipitation data;
S2, acquiring preprocessed basic data, calculating precipitation parameters of observed precipitation data and contemporaneous GCMs precipitation data of each ground rainfall station, and acquiring an average value of the precipitation parameters; calling a CLIGEN model based on the average precipitation parameters to generate reduced-scale GCMs precipitation data;
s3, calculating spatial correlation coefficients of rainfall data observed by each rainfall station, and reconstructing spatial correlation of the downscaled GCMs rainfall data by using a two-stage shuffle algorithm; reconstructing annual change trend of the contemporaneous GCMs downscaling data according to the annual precipitation change trend of the observed precipitation data;
s4, evaluating the precision of the GCMs downscaling precipitation data, if the precision meets the preset requirement, replacing the synchronous GCMs precipitation data with the GCMs precipitation data in future scenes, downscaling to obtain downscaling precipitation characteristics of future periods, and drawing related images;
the step S1 is further:
s11, acquiring GCMs (graphic arts model) precipitation data, wherein the GCMs precipitation data adopt a preset number of CMIP6 mode historical period precipitation data and future period precipitation data; the historical period precipitation data are statistical test data, and the future period precipitation data are combined scene data of four shared socioeconomic paths and typical concentration paths, namely SSP1-2.6, SSP2-4.5, SSP3-7.0 and SSP 5-8.5; the preparation of the GCMs precipitation data in the historical period is required to be intercepted based on the period of the observed precipitation data, so that the period of the observed precipitation data and the period of the observed precipitation data of the GCMs are consistent;
Step S12, adopting a bilinear interpolation method to carry out statistical interpolation on each GCMs precipitation data to obtain a grid with preset precision;
s13, calculating and adopting a correlation coefficient, a standard root mean square error, a relative standard deviation and a TSS index to evaluate the precision of the GCMs precipitation data, and if the TSS index of a certain GCM is greater than a threshold value, using the precipitation data as input data of precipitation prediction and importing the precipitation data into a database;
s14, performing deviation correction on GCMs precipitation data with accuracy not reaching a threshold value by adopting an equidistant cumulative distribution function method;
the step S2 is further:
s21, calculating precipitation parameters of observed precipitation data and contemporaneous GCMs precipitation data of each ground rainfall station, wherein the precipitation parameters comprise a daily precipitation average value, a daily precipitation standard deviation, a deviation coefficient of the daily precipitation, a daily precipitation-precipitation transition probability and a daily non-precipitation transition probability;
s22, determining four GCM grid center points closest to the ground rainfall station according to an angular distance calculation formula; respectively calculating the CLIGEN precipitation parameter change rate of each period, and calculating the CLIGEN precipitation parameters of each period at the ground rainfall station position by using a bilinear interpolation method;
Step S23, collecting and averaging CLIGEN precipitation parameters of each GCMs, and calling a CLIGEN model based on the average precipitation parameters to generate reduced-scale GCMs precipitation data;
the angular distance calculation formula is as follows: d=cos -1 （cosθ d cosθ s （cosβ d cosβ s +sinθ d sinβ d ）+sinθ d sinθ s ）；θ d The target rainfall station latitude;β d a target rain station longitude;θ s the latitude of the central point of the GCM data grid;β s longitude is the center point of the GCM data grid;
the process of calculating the CLIGEN precipitation parameter change rate in each period is as follows: CP (control program) m =（CP m-p /CP m-c ）×CP o-c ；
CP m CLIGEN parameters for the calculated future period GCM precipitation data;CP m-p CLIGEN parameters for the future period GCM precipitation data;CP m-c CLIGEN parameters for historical period GCM precipitation data;CP o-c CLIGEN parameters for observing precipitation data for historical periods;
the step S3 is further:
s31, arranging daily basic data according to month, and constructing 12 precipitation data matrixes n multiplied by kX]Wherein the matrix of observed precipitation data is recorded as [ [X obs ]CLIGEN precipitation data matrix is recorded as [X GEN ]N is the total number of days per month during the study period,kthe number of the rainfall stations;
step S32, CLIGEN precipitation data matrixX GEN ]Each column of data is ordered according to the size to obtain the rank order of each dataiVan der Waals scoring matrix [ S ] based on rank order calculation matrix ]Where s=Φ -1 (i/m+1), S is a Van der Waals score; phi -1 Is an inverse function of normal distribution; i is the rank of each data in each column; m is the total number of data per column;
step S33, calculating an observed precipitation data matrixX obs ]Is derived from the spearman rank correlation coefficient of (2)k×kMatrix of [ ofC S-obs ]And judges the matrixC S-obs ]Whether or not the matrix is positive, if so, obtaining the matrix by using Cholesky decomposition methodC S-obs ]Upper three of (3)Angle matrix [R]If not, then further matrix [C S-obs ]Decomposition into feature vectorsVAnd characteristic valueDSubstituting positive number for negative value in feature vector, and using corrected feature value diagonal matrixD’Generating matrix [C S-obs-m ]Finally, matrix [C S-obs-m ]After normalization, the matrix is obtained by Cholesky decomposition methodC S-obs ]Upper triangular matrix of (2)R]；
Step S34, reconstructing Van der Waals score matrixS * ]The calculation formula is as follows: [S * ]=[S] [R]The method comprises the steps of carrying out a first treatment on the surface of the Calculating Van der Waals scoring matrixS * ]Is to matrix CLIGEN precipitation data [X GEN ]According to Van der Waals scoring matrixS * ]Adjusting the rank to obtain a reconstructed matrixX GEN-R ]；
Step S35, calculate matrix [X GEN-R ]Pearson correlation coefficient matrix of [C P-GEN-R ]Establishing Pearson correlation coefficient matrixC P-GEN-R ]And matrix [C S-obs ]Linear relation of (c): [C S-obs ]=a×[C P-GEN-R ]+b；
Step S36, calculating Pearson correlation coefficient matrix of observed precipitation data [ C P-obs ]Using Pearson correlation coefficient matrixC P-obs ]Will be the formula [C S-obs ]=a×[C P-GEN-R ]In +b [C P-GEN-R ]Substitution is performed to obtain a new matrixC S-obs ’ ]；
Step S37, use matrix [C S-obs ’ ]Replacing the matrix in step S35C S-obs ]Repeating the steps S35 to S36 until the precision meets the requirement;
the step S4 is further:
s41, calculating CLIGEN parameters, spatial correlation coefficients and average absolute errors, standard root mean square errors and decision coefficients of drainage basin precipitation of observed precipitation data and histologic test data of synchronous GCMs; evaluating the precision of a downscaling method based on a CLIGEN model and a two-stage shuffle algorithm;
step S42, obtaining or calculating precipitation characteristics, wherein the precipitation characteristics comprise average daily, monthly and annual precipitation of each station and extreme precipitation indexes, and calculating average annual precipitation of a basin in a future period;
and step S43, drawing a comparison histogram, a scatter diagram, a box diagram and a frequency curve based on the rainfall characteristic.
2. The precipitation prediction method based on the improved statistical downscaling method of claim 1, wherein the step S3 further comprises a precipitation trend recovery process:
step S38, calculating the correlation coefficient among stations based on rainfall station observation rainfall sequences, selecting one rainfall station as a control station, and taking the rainfall station with the largest average correlation coefficient with other rainfall stations as the control station;
Step S39, rearranging the column data according to the rainfall generating structure of the control station in the rainy days, randomly arranging the non-rainfall days of the column, and adjusting the data in other columns; and calculating the annual precipitation rank of the observed precipitation data, and adjusting the histologic test data of the GCMs at the same site and at the same period according to the annual precipitation rank of the observed precipitation data so as to ensure that the annual precipitation trend is unchanged.
3. The precipitation prediction method based on the improved statistical downscaling method of claim 2, wherein the step S39 further comprises:
step S39a, acquiring an observation rainfall sequence in a preset period of a control station, extracting daily rainfall data, dividing preset intervals according to a maximum value and a minimum value, and setting a marker value for each interval; constructing a mapping relation between daily precipitation data and the marker; converting the observed rainfall sequences of all rainfall stations into marker value sequences to form a marker value sequence set;
step S39b, acquiring histologic test data of contemporaneous GCMs, sequencing precipitation data according to the annual precipitation rank of the observed precipitation data, constructing daily precipitation and constructing an identifier numerical sequence set;
Step S39c, calculating the sum of Euclidean distances between the identifier value sequence of the observed precipitation sequence of each site and the precipitation data acquired by the GCMs one by one, judging whether the sum is smaller than a preset value, and if the sum is smaller than the preset value, considering that the regulated annual precipitation trend accords with the expectation.
4. The precipitation prediction method based on the improved statistical downscaling method of claim 3, wherein the step S22 further comprises:
step S22a, reading the position parameters of each ground rainfall station and the position parameters of each GCM grid center point, determining four GCM grid center points nearest to each ground rainfall station by adopting an angular distance calculation formula, calculating the repetition times of each GCM grid center point, and constructing a weight coefficient matrix of each GCM grid;
step S22b, acquiring GCM precipitation data and CLIGEN precipitation parameters of observed precipitation data in future periods and historical periods, and calculating the CLIGEN precipitation parameter change rate of each period; judging whether the variation rate difference value of the rainfall parameters of each ground rainfall station in each grid exceeds a threshold value one by one, and if so, marking the first scale-reducing grid as the first scale-reducing grid;
and step S22c, judging whether abnormal points exist in the rainfall parameters of each ground rainfall station in each grid, and if so, calculating the CLIGEN rainfall parameters of each period at the position of each ground rainfall station by using a bilinear interpolation method.
5. Precipitation prediction system based on improved statistical downscaling method, which is characterized by comprising:
at least one processor; and
a memory communicatively coupled to at least one of the processors; wherein, the liquid crystal display device comprises a liquid crystal display device,
the memory stores instructions executable by the processor for execution by the processor to implement the precipitation prediction method based on the improved statistical downscaling method of any one of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310285989.4A CN115983511B (en) | 2023-03-22 | 2023-03-22 | Precipitation prediction method and system based on improved statistical downscaling method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310285989.4A CN115983511B (en) | 2023-03-22 | 2023-03-22 | Precipitation prediction method and system based on improved statistical downscaling method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115983511A CN115983511A (en) | 2023-04-18 |
| CN115983511B true CN115983511B (en) | 2023-05-23 |
Family
ID=85970588
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310285989.4A Active CN115983511B (en) | 2023-03-22 | 2023-03-22 | Precipitation prediction method and system based on improved statistical downscaling method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115983511B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116449460B (en) * | 2023-04-23 | 2023-10-10 | 水利部交通运输部国家能源局南京水利科学研究院 | Regional month precipitation prediction method and system based on convolution UNet and transfer learning |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110597873A (en) * | 2019-08-23 | 2019-12-20 | 北京师范大学 | Precipitation data estimation method, precipitation data estimation device, precipitation data estimation equipment and storage medium |
| CN114677059A (en) * | 2022-05-26 | 2022-06-28 | 水利部交通运输部国家能源局南京水利科学研究院 | Method and system for comprehensively evaluating precision of inversion precipitation product by integrating time-space indexes |
Family Cites Families (3)
| 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 |
| CN110874630B (en) * | 2019-10-23 | 2022-12-13 | 中国人民解放军空军研究院战场环境研究所 | Deep learning-based numerical model product downscaling refinement method |
| CN115357847B (en) * | 2022-08-28 | 2023-09-19 | 水利部交通运输部国家能源局南京水利科学研究院 | Solar scale satellite-ground precipitation fusion method based on error decomposition |
-
2023
- 2023-03-22 CN CN202310285989.4A patent/CN115983511B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110597873A (en) * | 2019-08-23 | 2019-12-20 | 北京师范大学 | Precipitation data estimation method, precipitation data estimation device, precipitation data estimation equipment and storage medium |
| CN114677059A (en) * | 2022-05-26 | 2022-06-28 | 水利部交通运输部国家能源局南京水利科学研究院 | Method and system for comprehensively evaluating precision of inversion precipitation product by integrating time-space indexes |
Non-Patent Citations (1)
| Title |
|---|
| Land use and climate change impacts on runoff and soil erosion at the hillslope scale in the Brazilian Cerrado;Jamil A.A. Anache等;《Science of The Total Environment》;第622-623卷;第140-151页，全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115983511A (en) | 2023-04-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Nipen et al. | Adopting citizen observations in operational weather prediction | |
| Wolff et al. | Beyond the basics: Evaluating model-based precipitation forecasts using traditional, spatial, and object-based methods | |
| CN111027175B (en) | Method for evaluating social and economic influences of flood based on coupling model integrated simulation | |
| CN110555561B (en) | Medium-and-long-term runoff ensemble forecasting method | |
| Pagano et al. | Ensemble dressing for hydrological applications | |
| CN107563554B (en) | Screening method of statistical downscaling model forecasting factors | |
| Guyennon et al. | Benefits from using combined dynamical-statistical downscaling approaches–lessons from a case study in the Mediterranean region | |
| CN112861072B (en) | Satellite-ground multi-source rainfall self-adaptive dynamic fusion method | |
| CN115983511B (en) | Precipitation prediction method and system based on improved statistical downscaling method | |
| Segond et al. | Simulation and spatio-temporal disaggregation of multi-site rainfall data for urban drainage applications | |
| CN105974495A (en) | Method for pre-judging future average cloud amount of target area by using classification fitting method | |
| Zhao et al. | A framework for projecting future intensity-duration-frequency (IDF) curves based on CORDEX Southeast Asia multi-model simulations: An application for two cities in Southern Vietnam | |
| Shlyaeva et al. | Towards ensemble data assimilation for the environment Canada regional ice prediction system | |
| CN109272144B (en) | BPNN-based prediction method for NDVI (normalized difference of variance) in northern grassland area of China | |
| Smith et al. | Evaluation of the S (T) assimilation method with the Argo dataset | |
| Li et al. | A nonstationary runoff frequency analysis for future climate change and its uncertainties | |
| Monjo et al. | Estimation of future extreme rainfall in Barcelona (Spain) under monofractal hypothesis | |
| CN114254802A (en) | Prediction method of vegetation coverage space-time change under climate change drive | |
| Hu et al. | Approximating the internal variability of bias-corrected global temperature projections with spatial stochastic generators | |
| Zhou et al. | Daily rainfall model to merge TRMM and ground based observations for rainfall estimations | |
| CN116341841A (en) | Runoff forecast error correction method, apparatus, device, medium and program product | |
| Xu et al. | The impact of mean dynamic topography on a sea-level anomaly assimilation in the South China Sea based on an eddy-resolving model | |
| Holland et al. | Combination of physics based simulation and machine learning for PV power forecasting of large power plants | |
| CN109886497B (en) | Ground air temperature interpolation method based on latitude improved inverse distance weighting method | |
| Huang et al. | Hybrid Intra-hour Solar PV Power Forecasting using Statistical and Skycam-based Methods |
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 |