Downscaling processing method and device for climate data

Abstract

The invention discloses a downscaling processing method and device for climate data, and relates to the technical field of climate data processing, and the method comprises the steps: cutting the climate data with a unified resolution in a research area range; creating first and second datasets of climate data evaluation using the clipped climate data; determining multi-year daily average climate data of each grid pixel in a historical period in the research area according to the first data set; constructing a geographic weighting matrix so as to describe local spatial heterogeneity among different grid pixels in spatial dimension; according to the matrix and the multi-year daily average climate data, constructing a space-time weighted regression model to obtain a multi-year daily average regression coefficient; and performing deviation correction on the second data set by using the coefficient after smoothing processing to obtain downscaling climate data. According to the method, the time-space heterogeneity of the climate data can be accurately captured by dynamically estimating the daily average regression coefficient for many years and combining space weighting and time dynamic change, and the precision of climate data simulation is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of climate data processing, and particularly to a method and device for downscaling climate data. Background Art

[0002] This section aims to provide background or context for the technical features stated in the claims. The description herein is not admitted to be prior art merely by virtue of its inclusion in this section.

[0003] Downscaling refers to a method of converting large-scale climate model data into higher-resolution regional climate data. It aims to improve the spatial resolution or temporal resolution of climate model data so that it can better reflect local climate characteristics, especially climate variables such as precipitation.

[0004] In recent years, with the continuous in-depth study of climate change, accurate regional climate prediction and disaster warning have become the focus of research. As an important data source in global climate research, climate model data provides various climate scenarios and future climate predictions. However, the spatial resolution of climate model data is relatively low, making it difficult to meet the needs of regional climate research and disaster prediction. Therefore, it is of great research significance to study more accurate downscaling methods.

[0005] Current downscaling methods include physical methods and statistical methods. Traditional statistical downscaling methods, such as regression analysis and interpolation techniques, mainly perform downscaling based on the simple relationship between observational data and climate model data. Common statistical downscaling methods include linear regression, least squares method, distribution mapping method, etc. Although these methods can improve the spatial resolution of climate model data to a certain extent, they usually ignore the complex changes of climate variables in the spatio-temporal dimension, resulting in the inability to comprehensively and accurately reflect the spatio-temporal heterogeneity of climate. Summary of the Invention

[0006] The technical problem existing in the present invention is: the problem of insufficient accuracy in dealing with the spatial heterogeneity and temporal dynamic characteristics of precipitation in traditional downscaling methods, and the low accuracy of climate data simulation.

[0007] To solve the above technical problems, in a first aspect, the present invention provides a method for downscaling climate data, which is used to accurately capture the spatio-temporal heterogeneity of climate variables and improve the simulation accuracy of climate data by dynamically estimating the daily regression coefficients of multiple years and combining spatial weighting and temporal dynamic changes. The method includes: Obtain climate model data within the scope of the research area and perform unit conversion on the climate model data; Perform interpolation processing on the climate model data after unit conversion to unify the resolution; The climate model data after being unified in resolution are cropped using the research area range to obtain climate model data with the same range size and corresponding spatial positions. Using the climate model data with the same range size and corresponding spatial positions, the first data set and the second data set for climate data evaluation are produced; the first data set is the experimental data into which the daily precipitation data in the historical period of the climate model data are divided, and the second data set is the verification data into which the daily precipitation data in the historical period of the climate model data are divided. Based on the first data set, the multi-year average daily climate data of each grid cell in the research area in the historical period are determined. A geographically weighted matrix is constructed to describe the local spatial heterogeneity among different grid cells in the spatial dimension; based on the geographically weighted matrix and the multi-year average daily climate data, a spatio-temporal weighted regression model is constructed to obtain the multi-year average daily regression coefficients. The multi-year average daily regression coefficients are smoothed to obtain the smoothed multi-year average daily regression coefficients. Using the smoothed multi-year average daily regression coefficients, the second data set is corrected for bias to obtain the downscaled climate model data.

[0008] In a second aspect, the present invention also provides a device for downscaling climate data, which is used to accurately capture the spatio-temporal heterogeneity of climate variables and improve the simulation accuracy of climate data by dynamically estimating the multi-year average daily regression coefficients and combining spatial weighting and temporal dynamic changes. The device includes: A conversion unit, which is used to obtain the climate model data within the research area range and perform unit conversion on the climate model data. A spatial resolution processing unit, which is used to perform interpolation processing on the climate model data after unit conversion to unify the resolution. A cropping processing unit, which is used to crop the climate model data after being unified in resolution using the research area range to obtain climate model data with the same range size and corresponding spatial positions. A data set production unit, which is used to produce the first data set and the second data set for climate data evaluation using the climate model data with the same range size and corresponding spatial positions; the first data set is the experimental data into which the daily precipitation data in the historical period of the climate model data are divided, and the second data set is the verification data into which the daily precipitation data in the historical period of the climate model data are divided. A multi-year average daily climate data determination unit, which is used to determine the multi-year average daily climate data of each grid cell in the research area in the historical period based on the first data set. The multi-year average daily regression coefficient determination unit is used to construct a geographically weighted matrix to describe the local spatial heterogeneity among different grid cells in the spatial dimension; according to the geographically weighted matrix and the multi-year average daily climate data, a spatio-temporal weighted regression model is constructed to obtain the multi-year average daily regression coefficient; The smoothing processing unit is used to smooth the multi-year average daily regression coefficient to obtain the smoothed multi-year average daily regression coefficient; The correction unit is used to perform bias correction on the second data set by using the smoothed multi-year average daily regression coefficient to obtain the downscaled climate model data.

[0009] In a third aspect, the present invention also provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the above-mentioned downscaling method for climate data is implemented.

[0010] In a fourth aspect, the present invention also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the above-mentioned downscaling method for climate data is implemented.

[0011] In a fifth aspect, the present invention also provides a computer program product. The computer program product includes a computer program. When the computer program is executed by a processor, the above-mentioned downscaling method for climate data is implemented.

[0012] Compared with the prior art solution for downscaling based on the simple relationship between observational data and climate model data, the beneficial technical effect of the downscaling solution for climate data provided by the present invention is that it can accurately capture the spatio-temporal heterogeneity and non-linear characteristics of climate variables by dynamically estimating the multi-year average daily regression coefficient, combining spatial weighting and temporal dynamic changes, and improving the accuracy of climate data simulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. In the drawings: Figure 1 It is a flowchart of the downscaling method for climate data in an embodiment of the present invention; Figure 2 It is a precipitation data graph before and after a set of unit conversions in an embodiment of the present invention. Figure 2 In (a), it is the original data before the unit conversion. Figure 2In (b) are the data after the transformation unit; Figure 3 This is an interpolated precipitation data map in an embodiment of the present invention; Figure 4 This is a time series change diagram of precipitation before and after calibration of observed data and model data within a preset time period (e.g., 14 years) in the study area in an embodiment of the present invention; Figure 5 This is a broken line diagram of the correlation coefficient of the downscaling result in an embodiment of the present invention; Figure 6 This is an error analysis diagram of climate model data and observed data in an embodiment of the present invention; Figure 7 This is a schematic structural diagram of a downscaling processing device for climate data in an embodiment of the present invention. Detailed implementation manners

[0014] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer and more understandable, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Herein, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0015] In the technical solution of this application, the acquisition, storage, use, processing, etc. of data all comply with the relevant regulations of laws and regulations.

[0016] Figure 1 This is a schematic flow diagram of a downscaling processing method for climate data in an embodiment of the present invention. As Figure 1 shown, the method includes the following steps:

[0017] Step 101: Obtain climate model data within the scope of the study area and perform unit conversion on the climate model data;

[0018] Step 102: Perform interpolation processing on the climate model data after unit conversion to unify the resolution; So that the climate model data and the observed data have the same spatial resolution;

[0019] Step 103: Crop the climate model data with unified resolution using the scope of the study area to obtain climate model data with the same scope size and corresponding spatial positions;

[0020] Step 104: Use the climate model data with the same scope size and corresponding spatial positions to produce a first data set and a second data set for climate data evaluation; the first data set is the experimental data into which the daily precipitation data in the historical period of the climate model data is divided, and the second data set is the verification data into which the daily precipitation data in the historical period of the climate model data is divided;

[0021] Step 105: Determine the multi-year daily average climate data of each grid cell in the study area during the historical period according to the first data set;

[0022] Step 106: Construct a geographically weighted matrix to describe the local spatial heterogeneity among different grid cells in the spatial dimension; according to the geographically weighted matrix and the multi-year daily average climate data, construct a spatio-temporal weighted regression model to obtain the multi-year daily average regression coefficients;

[0023] Step 107: Smooth the multi-year daily average regression coefficients to obtain the smoothed multi-year daily average regression coefficients; To reduce the impact of the fluctuations of the multi-year daily average regression coefficients and abnormal data on the model;

[0024] Step 108: Use the smoothed multi-year daily average regression coefficients to correct the bias of the second data set to obtain the downscaled climate model data.

[0025] When the downscaling method of climate data provided by the embodiment of the present invention works: obtain the climate model data within the scope of the study area, perform unit conversion on the climate model data so that the precipitation unit is converted from kilograms per square meter per second to millimeters per day; perform interpolation processing on the unit-converted climate model data to unify the resolution so that the climate model data has the same spatial resolution as the observed data; crop the climate model data with unified resolution using the scope of the study area to obtain climate model data with the same scope size and corresponding spatial positions; use the climate model data with the same scope size and corresponding spatial positions to produce the first data set and the second data set for climate data evaluation; determine the multi-year daily average climate data of each grid cell in the study area during the historical period according to the first data set; construct a geographically weighted matrix to describe the local spatial heterogeneity among different grid cells in the spatial dimension; according to the geographically weighted matrix and the multi-year daily average climate data, construct a spatio-temporal weighted regression model to obtain the multi-year daily average regression coefficients; smooth the multi-year daily average regression coefficients to obtain the smoothed multi-year daily average regression coefficients to reduce the impact of the fluctuations of the multi-year daily average regression coefficients and abnormal data on the model; use the smoothed multi-year daily average regression coefficients to correct the bias of the second data set to obtain the downscaled climate model data.

[0026] Compared with the prior art solution for downscaling based on the simple relationship between observed data and climate model data, the downscaling method of climate data provided by the embodiment of the present invention can accurately capture the spatio-temporal heterogeneity and non-linear characteristics of climate variables by dynamically estimating the multi-year daily average regression coefficients, combining spatial weighting and temporal dynamic changes, and improves the accuracy of climate data simulation. The following combines Figures 2 to 6 The details are introduced as follows.

[0027] In view of the current problem that CMIP6 climate model data is difficult to meet the needs of regional climate research, an embodiment of the present invention provides a downscaling method for climate data. This method relates to the downscaling method in climate modeling. It is a CMIP6 climate data downscaling method that combines Bayesian ridge regression and a spatial weight matrix (which can be called Bayesian ridge spatio-temporal weighted regression in the embodiment of the present invention). By dynamically estimating the daily regression coefficients over the years and combining spatial weighting and temporal dynamic changes, it can more accurately capture the spatio-temporal heterogeneity of climate variables and improve the simulation accuracy of climate variables such as precipitation. The following is a detailed introduction.

[0028] For ease of understanding, the following takes climate data as climate model precipitation data as an example to illustrate the downscaling method for climate data provided by the embodiment of the present invention.

[0029] Step 101: Obtain climate model data within the research area and perform unit conversion on the climate model data: Obtain CMIP6 climate model data (such as climate model precipitation data) and perform unit conversion on it. The precipitation unit is converted from kilograms per square meter per second to millimeters per day to eliminate the unit impact. The climate data in the embodiment of the present invention can be CMIP6 (Coupled Model Intercomparison Project Phase 6) climate model data.

[0030] Furthermore, the above step 101 of obtaining CMIP6 climate model precipitation data and performing unit conversion on it is specifically as follows: 1.1. Select climate (such as precipitation) model data with a time scale of days and a long time series range.

[0031] 1.2. Multiply the original climate model data (such as precipitation model data) by 86400. Taking precipitation climate as an example, the precipitation unit changes from kilograms per square meter per second to millimeters per day, and precipitation data with the same unit as the measured data is obtained. Figure 2 This is a precipitation data graph before and after unit transformation in the embodiment of the present invention. Figure 2 In (a) is the original data before unit transformation. Figure 2 In (b) is the data after unit transformation.

[0032] Step 102: Perform interpolation processing on the climate model data after unit conversion to unify the resolution: Perform interpolation processing on the climate model data after the above unit transformation to unify the resolution so that it has the same spatial resolution as the observed data (in the embodiment of the present invention, the CN05.1 dataset is used as the observed data). Figure 3 This is a precipitation data graph after interpolation in the embodiment of the present invention.

[0033] The above step 102 of unifying the spatial resolution of the climate model is specifically as follows: The spatial resolution of climate model data is improved to regular grid cells of 0.25°×0.25° by bilinear interpolation method. The calculation formula is as follows:

[0034]

[0035]

[0036]

[0037] Among them, P1(x1, y1), P2(x2, y1), P3(x1, y2) and P4(x2, y2) are known coordinate points, and P(x, y) is the point to be interpolated and the intermediate result of interpolation between R1 and R2.

[0038] Step 103: Crop the climate model data with the unified resolution using the research area range to obtain climate model data with the same range size and corresponding spatial positions: Crop the interpolated high-resolution climate model data with the research area range to obtain precipitation data with the same range size and corresponding spatial positions.

[0039] This step 103 is specifically as follows: Reproject the research area boundary file to equal-angle equal-area, then vectorize and rasterize the file to convert it into a raster image with a resolution of 0.25, and finally use the exported raster file as a mask to crop the climate model data. That is: 3.1. Obtain the boundary range of the research area, check the projection type of the boundary file, and reproject the research area boundary file to equal-angle equal-area.

[0040] 3.2. Rasterize the equal-angle equal-area vector boundary to obtain a research area raster image with a resolution of 0.25 degrees.

[0041] 3.3. Use the exported raster image file as a mask to crop the climate model data, thereby generating a new raster dataset, that is, generating climate model data (such as precipitation data) with the same range size and corresponding spatial positions.

[0042] Step 104: Use the climate model data with the same range size and corresponding spatial positions to produce the first dataset and the second dataset for climate data evaluation, that is, produce the first dataset and the second dataset for precipitation data evaluation.

[0043] This step 104 is specifically as follows: Divide the daily precipitation data of the climate model data from 1970 to 2014 in the historical period into experimental data (the first dataset) and verification data (the second dataset). The experimental data (the first dataset) from 1970 to 2000 is used to calculate the regression coefficient, and the verification data (the second dataset) from 2001 to 2014 is used to verify the established downscaling method.

[0044] Step 105: Determine the multi-year daily average climate data of each grid cell in the study area based on the first data set: Calculate the daily average data of the multi-year precipitation of each grid cell in the above first data set. Specifically, for Step 105, calculate the multi-year daily average values of the data during the experimental period from 1970 to 2000, and calculate the multi-year daily average data of 30 years for each grid cell respectively.

[0045] Step 106: Construct a geographically weighted matrix to describe the local spatial heterogeneity among different grid cells in the spatial dimension; Based on the geographically weighted matrix and the multi-year daily average climate data, construct a spatio-temporal weighted regression model to obtain the multi-year daily average regression coefficients: Construct a geographically weighted matrix to describe the local spatial heterogeneity among different grid cells in the spatial dimension; Construct a spatio-temporal weighted regression model for the calculated multi-year daily average precipitation data in the study area to obtain a multi-year daily average regression coefficient model, and the multi-year daily average regression coefficients can be obtained through this model.

[0046] In specific implementation, estimate the multi-year daily average regression coefficients day by day based on the multi-year daily average data in the time dimension to reveal the dynamic changes of the climate model data and the observed data over time; In the spatial dimension, through the geographically weighted matrix w i describe the local spatial heterogeneity among different grid cells. This "local spatial heterogeneity" refers to the degree of change or distribution difference of precipitation between different regions or grid units, emphasizing the non-uniformity in space. For example: In some regions, the precipitation may be relatively uniform, while in other regions, the precipitation may have large fluctuations. Through the geographically weighted matrix, when calculating the regression coefficient of the current grid cell, the precipitation values of the surrounding cells can be considered. Specifically:

[0047] 6.1. In the spatial dimension, use the geographically weighted matrix to describe the local spatial heterogeneity among different grid cells. This weighted matrix is constructed based on the Gaussian kernel function to reflect the influence of spatially adjacent grid cells on the regression coefficients. The weighted matrix w i is a diagonal matrix. In one embodiment, the geographically weighted matrix can be: w i =diag(w i1 , w i2 ,···,w in ) ; The weighted matrix w i The element w ij represents the weighting coefficient of the i-th grid cell to the j-th grid cell, and the calculation formula is:

[0048] Wherein: w ij is the weight of grid cell i for grid cell j; h is the bandwidth parameter that controls the rate at which the weight decays with distance; d ij is the distance between the i-th grid cell and the j-th grid cell, .

[0049] In specific implementation, the above "bandwidth" is a key parameter in the Gaussian kernel function, which controls the smoothness and influence range of the Gaussian kernel function. In the embodiments of the present invention, it determines how the similarity between sample points decays as the distance increases. When the bandwidth is small, only grid cells with a relatively close distance can have a greater impact on the target point. When the bandwidth is large, the influence range of the Gaussian kernel increases, and sample points with a relatively far distance will also have a greater impact on the target point.

[0050] In specific implementation, n is the total number of samples. Through this weighted matrix, the contribution of spatially adjacent points to the estimation of the regression coefficient is greater, while the influence of distant points is weakened, thereby effectively capturing the heterogeneity of the observed data and the simulated data in the spatial dimension.

[0051] 6.2. Use the multi-year daily precipitation data to estimate the regression coefficient for each day. At each time point t, calculate the regression relationship between the training period climate model data (GCM simulation data, i.e., the first data set) and the observed data. According to the regression coefficient formula, sequentially obtain the regression coefficients of each position of each day's multi-year daily data.

[0052] In one embodiment, according to the geographically weighted matrix and the multi-year daily climate data, construct a spatio-temporal weighted regression model to obtain the multi-year daily regression coefficients, which may include: obtaining the multi-year daily regression coefficients according to the following multi-year daily regression coefficient model (regression model):

[0053]

[0054] Wherein: β i (t) is the regression coefficient of time t and grid cell i, Y(t) is the observed data vector at time t, X(t) is the climate model data matrix at time t, X(t) and Y(t) are the multi-year daily climate data; λ is the regularization parameter; I is the identity matrix, w i is the geographically weighted matrix.

[0055] Step 107: Smooth the multi-year daily average regression coefficients to obtain the smoothed multi-year daily average regression coefficients: Smooth the calculated multi-year daily average regression coefficients through a sliding time window to reduce the influence of the fluctuations of the regression coefficients and abnormal data on the model.

[0056] In specific implementation, when calculating the regression coefficient of a certain day, consider the data of several days before and after it, and smooth the fluctuations within a short period of time through weighted average. The setting of the weights can be adjusted according to the time distance. The days closer to the current time have larger weights, while the days farther from the current time have smaller weights. Specifically:

[0057] 7.1. Select a suitable size of the sliding time window for calculating the smoothed result of the regression coefficient. The range of the sliding window is from t - w to t + w, where t is the current time point, and the window size 2w + 1 will determine the number of days involved in the smoothing calculation.

[0058] 7.2. For each time point t, perform a weighted average of the regression coefficients within the sliding window to obtain the smoothed regression coefficient.

[0059] In one embodiment, smoothing the multi-year daily average regression coefficients to obtain the smoothed multi-year daily average regression coefficients may include: obtaining the smoothed multi-year daily average regression coefficients according to the following formula:

[0060]

[0061] Where: β i (t) is the regression coefficient smoothed through the sliding time window, β i (k) is the original regression coefficient on the k th day, w k is the weight of the sliding time window, w is the half-width of the sliding time window, and the range of the sliding time window is t - w , t + w .

[0062] 7.3. Determine the weights and window range. The weights are generally assigned according to the distance from the current time. The days closer to the current time will obtain larger weights, and the days far from the current time have smaller weights. By experimenting with different weight assignments and window sizes, and optimizing according to the performance on the first data set, select the weights and windows with the best effects. That is, in one embodiment, the above method for downscaling climate data may further include:

[0063] Perform sliding time window weight allocation according to the distance from the current time. The weights of the days closer to the current time are larger, and the weights of the days farther from the current time are smaller.

[0064] Optimize by experimenting with different sliding time window weights and their performance in the first dataset to obtain the sliding time window weight allocation and range with the best effect.

[0065] Step 108: Use the smoothed multi-year daily average regression coefficients to perform bias correction on the second dataset to obtain downscaled climate model data. Apply the smoothed multi-year daily average regression coefficients to perform bias correction on the second dataset to obtain downscaled climate model data, which can be used for subsequent regional climate change assessment, climate change prediction, disaster warning, etc. Multiply the precipitation data of each day in the second dataset by the corresponding multi-year daily average regression coefficient to obtain the bias-corrected result for that date. Calculate for all days in turn to obtain the bias-corrected second dataset. Figure 4 This is the time series change diagram of precipitation before and after correction of the observed data and model data in the preset time period (e.g., 14 years) of the study area in the embodiment of the present invention.

[0066] In one embodiment, the above method for downscaling climate data may further include: evaluating the accuracy and reliability of the downscaled climate model data.

[0067] Specifically, when implementing, evaluate the accuracy and reliability of the downscaled climate model data, which can be achieved through existing technologies. Compare the downscaled climate model data in the verification period with the actual climate (e.g., precipitation) data (CN05.1 dataset) of the study area, and use the correlation coefficient, standard deviation, and root mean square error as evaluation indicators to evaluate the accuracy of this downscaling method. Figure 5 This is the line graph of the correlation coefficient of the downscaling result in the embodiment of the present invention. Figure 6 This is the error analysis diagram (i.e., Taylor diagram, drawn based on the correlation coefficient, standard deviation, and root mean square error) of the climate model data and the observed data in the embodiment of the present invention.

[0068] The beneficial effects of the embodiments of the present invention are as follows: By proposing the Bayesian ridge spatio-temporal weighted regression CMIP6 climate data downscaling method, the embodiments of the present invention significantly improve the climate simulation accuracy of CMIP6 climate model data in regional climate research. By dynamically estimating daily regression coefficients and combining spatio-temporal weighting, the present invention can accurately capture the spatio-temporal heterogeneity and non-linear characteristics of climate variables such as precipitation, overcoming the problem of insufficient accuracy in traditional downscaling methods when dealing with precipitation spatial heterogeneity and temporal dynamic characteristics. The method also smooths the regression coefficients through a sliding time window, reduces the fluctuations of the regression coefficients, and improves the stability and reliability of long-term climate change prediction. This downscaling method improves the spatial and temporal accuracy of CMIP6 climate model data, enabling climate data to more accurately reflect local climate characteristics at a higher resolution. The method provides more accurate data support for subsequent regional climate change assessment, climate change prediction, and disaster warning, has broad application prospects, and can provide a more reliable basis for relevant research and decision-making.

[0069] An apparatus for downscaling climate data is also provided in the embodiments of the present invention, as described in the following embodiments. Since the principle of the apparatus for solving problems is similar to that of the method for downscaling climate data, the implementation of the apparatus can refer to the implementation of the method for downscaling climate data, and the repeated parts will not be described again.

[0070] Figure 7 is a schematic structural diagram of the apparatus for downscaling climate data in the embodiments of the present invention, as Figure 7 shown. The apparatus includes: A conversion unit 01, configured to obtain climate model data within the scope of the research area and perform unit conversion on the climate model data; A spatial resolution processing unit 02, configured to perform interpolation processing on the climate model data after unit conversion to unify the resolution; A cropping processing unit 03, configured to crop the climate model data after unifying the resolution with the scope of the research area to obtain climate model data with the same scope size and corresponding spatial positions; A data set production unit 04, configured to use the climate model data with the same scope size and corresponding spatial positions to produce a first data set and a second data set for climate data evaluation; the first data set is the experimental data obtained by dividing the daily precipitation data in the historical period of the climate model data, and the second data set is the verification data obtained by dividing the daily precipitation data in the historical period of the climate model data;

[0071] A multi-year daily average climate data determination unit 05, configured to determine the multi-year daily average climate data of each grid cell in the historical period within the research area according to the first data set;

[0072] The multi-year average daily regression coefficient determination unit 06 is used to construct a geographically weighted matrix to describe the local spatial heterogeneity among different grid cells in the spatial dimension; according to the geographically weighted matrix and the multi-year average daily climate data, a spatio-temporal weighted regression model is constructed to obtain the multi-year average daily regression coefficient;

[0073] The smoothing processing unit 07 is used to perform smoothing processing on the multi-year average daily regression coefficient to obtain the smoothed multi-year average daily regression coefficient;

[0074] The correction unit 08 is used to perform bias correction on the second data set by using the smoothed multi-year average daily regression coefficient to obtain the downscaled climate model data.

[0075] In one embodiment, the multi-year average daily regression coefficient determination unit is specifically used to: obtain the multi-year average daily regression coefficient according to the following multi-year average daily regression coefficient model:

[0076]

[0077] Where: β i (t) is the regression coefficient at time t and grid cell i, Y(t) is the observation data vector at time t, X(t) is the model data matrix at time t, X(t) and Y(t) are the multi-year average daily climate data; λ is the regularization parameter; I is the identity matrix, w i is the geographically weighted matrix.

[0078] In one embodiment, the geographically weighted matrix w i is a diagonal matrix, and its form is: w i =diag(w i1 , w i2 ,···,w in ) ; the element w i of the geographically weighted matrix w ij represents the weighting coefficient of the i-th grid cell to the j-th grid cell, and the calculation formula is:

[0079] Where: h is the bandwidth parameter, which is used to control the attenuation speed of the weight with distance;d ij is the distance between the i-th grid cell and the j-th grid cell.

[0080] In one embodiment, the smoothing unit is specifically configured to: obtain the smoothed multi-year daily regression coefficient according to the following formula:

[0081] ;

[0082] Where: β i (t) is the regression coefficient after being smoothed by a sliding time window, β i (k) is the k original regression coefficient on the w k day, w is the weight of the sliding time window, and the range of the sliding time window is t - w , t + w .

[0083] In one embodiment, the downscaling processing device for the above climate data further includes:

[0084] A weight allocation unit, configured to perform weight allocation for the sliding time window according to the distance from the current time, where the weights of the days closer to the current time are larger and the weights of the days farther from the current time are smaller;

[0085] An optimization unit, configured to optimize by testing the performance of different sliding time window weights and ranges in the first data set to obtain the optimal sliding time window weight allocation and range.

[0086] In one embodiment, the downscaling processing device for the above climate data further includes: an evaluation unit, configured to evaluate the accuracy and reliability of the downscaled climate model data.

[0087] An embodiment of the present invention further provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the above method for downscaling climate data is implemented.

[0088] An embodiment of the present invention further provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above method for downscaling climate data is implemented.

[0089] An embodiment of the present invention also provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the above-mentioned downscaling method for climate data.

[0090] Compared with the prior art's downscaling solution based on the relationship between observed data and climate model data, the downscaling solution for climate data provided by the embodiment of the present invention can accurately capture the spatio-temporal heterogeneity and non-linear characteristics of climate variables by dynamically estimating the daily regression coefficients over the years, combined with spatial weighting and temporal dynamic changes, improving the accuracy of climate data simulation.

[0091] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0092] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0093] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0094] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide means for implementing the functions specified in Figure 1One or more processes and / or boxes Figure 1 Steps of functions specified in one box or multiple boxes.

[0095] In the specific embodiments described above, the objectives, technical solutions and beneficial effects of the present invention have been further described in detail. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A downscaling method for climate data, characterized in that: include: Obtain climate model data within the study area and convert the units of climate model data; Interpolate the climate model data after unit conversion to unify the resolution; The climate model data with uniform resolution are clipped with the scope of the research area to obtain climate model data with consistent scope and corresponding spatial location; Using climate model data with the same range and corresponding spatial position, a first data set and a second data set for climate data evaluation are produced; the first data set is experimental data in which the daily precipitation data in the historical period of the climate model data is divided into, and the second data set is verification data in which the daily precipitation data in the historical period of the climate model data is divided into; Based on the first data set, the multi-year daily average climate data of each grid pixel in the study area in the historical period is determined; A geographically weighted matrix is ​​constructed to describe the local spatial heterogeneity between different grid pixels in the spatial dimension; a spatiotemporal weighted regression model is constructed based on the geographically weighted matrix and multi-year daily average climate data to obtain the multi-year daily average regression coefficient; Smoothing the multi-year daily average regression coefficients to obtain the multi-year daily average regression coefficients after smoothing; The second data set is bias-corrected using the smoothed multi-year daily average regression coefficient to obtain downscaled climate model data.

2. The method according to claim 1, characterized in that According to the geographical weighted matrix and the multi-year daily average climate data, a spatiotemporal weighted regression model is constructed to obtain the multi-year daily average regression coefficient, including: obtaining the multi-year daily average regression coefficient according to the following multi-year daily average regression coefficient model: in: β i (t) is the regression coefficient of time t and grid pixel i, Y(t) is the observed data vector at time t, X(t) is the pattern data matrix at time t, X(t) and Y(t) It is the daily average climate data for many years; λ is the regularization parameter; I is the identity matrix, w i is the geographically weighted matrix.

3. The method according to claim 2, characterized in that The geographical weighted matrix is: w i =diag(w i1 ,w i2 ,···,w in ) ; Geographically weighted matrix w i Elements w ij It represents the weight coefficient of the i-th grid pixel to the j-th grid pixel, and the calculation formula is: in: h is the bandwidth parameter, which is used to control the speed at which the weight decays with distance; d ij is the distance between the i-th grid pixel and the j-th grid pixel.

4. The method according to claim 1, characterized in that Smoothing the multi-year daily average regression coefficient to obtain the multi-year daily average regression coefficient after smoothing includes: obtaining the multi-year daily average regression coefficient after smoothing according to the following formula: ； in: β i (t) is the regression coefficient after smoothing through the sliding time window, β i (k) For the k The original regression coefficient of day, w k is the sliding time window weight, w is the half width of the sliding time window, and the range of the sliding time window is [ tw , t+w ].

5. The method according to claim 4, characterized in that Also includes: The sliding time window weights are assigned based on the distance from the current time. Days close to the current time have a larger weight, while days far from the current time have a smaller weight. By testing different sliding time window weights and ranges, the performance of the first data set is optimized to obtain the best sliding time window weight allocation and range.

6. The method according to claim 1, characterized in that Also includes: Assess the accuracy and reliability of downscaled climate model data.

7. A downscaling processing device for climate data, characterized in that: include: Conversion unit, used to obtain climate model data within the study area and perform unit conversion on climate model data; A spatial resolution processing unit is used to interpolate the climate model data after unit conversion to a uniform resolution; A clipping processing unit is used to clip the climate model data with uniform resolution using the scope of the research area to obtain climate model data with consistent scope size and corresponding spatial position; The data set making unit is used to make a first data set and a second data set for climate data evaluation by using climate model data with the same range size and corresponding spatial position; the first data set is experimental data in which the daily precipitation data in the historical period of the climate model data is divided into, and the second data set is verification data in which the daily precipitation data in the historical period of the climate model data is divided into; A multi-year daily average climate data determination unit is used to determine the multi-year daily average climate data of each grid pixel in the study area in the historical period according to the first data set; The multi-year daily average regression coefficient determination unit is used to construct a geographically weighted matrix to describe the local spatial heterogeneity between different grid pixels in the spatial dimension; based on the geographically weighted matrix and multi-year daily average climate data, a spatiotemporal weighted regression model is constructed to obtain the multi-year daily average regression coefficient; A smoothing processing unit is used to smooth the multi-year daily average regression coefficient to obtain the multi-year daily average regression coefficient after smoothing; The correction unit is used to perform bias correction on the second data set by using the multi-year daily average regression coefficient after smoothing to obtain downscaled climate model data.

8. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images