Climate mode data calibration method based on spatio-temporal variation transfer function

Abstract

The invention discloses a climate mode data calibration method based on a spatio-temporal variation transfer function, which comprises the following steps of: taking a single observation station as an independent individual, solving a transfer function between actual measurement data and mode data of the corresponding observation station by utilizing a quantile mapping method, and comparing the transfer functions at different mode data grid points to obtain a climate mode data calibration result; and analyzing the spatial distribution characteristics of the model data and the change trend of the underlying surface, calculating transfer functions in different observation periods, establishing a space-variant unsteady transfer function, and applying the space-variant unsteady transfer function to calibration of the model data. According to the method, variables can be effectively adjusted to be matched with observation data, and the error correction effect under different landforms and altitude conditions is stable.

Description

Technical Field

[0001] The present invention belongs to the technical field of meteorological prediction, and particularly relates to a calibration method for climate model data based on a spatio-temporal variation transfer function. Background Art

[0002] Climate models can provide meteorological data with significantly higher spatio-temporal resolution than observational methods and can conduct future climate research. However, affected by many factors such as systematic errors and constraint conditions, there are often significant deviations between the simulation results of climate models and the measured data, resulting in uncertainties in the assessment of extreme climates and affecting subsequent climate prediction and analysis of evolutionary trends. Error correction is an important method to solve the problem of model deviation. Commonly used error correction methods include total correction and frequency correction. The total correction method is mainly applicable to the correction of the average state, is difficult to process daily-scale data, and cannot correct errors in the frequency distribution. The frequency correction method (such as the quantile mapping method) can effectively make up for the deficiencies of the total correction method and has obvious advantages in the simulation of temporal variability and extreme climate events. However, in plateau areas with obvious elevation differences and complex topographical and geomorphic conditions, due to the lack of effective boundary condition constraints (low spatio-temporal resolution of observational data), coupled with the complexity of climate and terrain, the simulation effect of the RCM is unstable, and the resulting systematic errors often show the characteristics of spatial variation (spatial variation). The traditional quantile mapping method is difficult to effectively adjust variables to match the observational data, and there are obvious differences in the error correction effects under different topographical and geomorphic conditions and elevation conditions. Summary of the Invention

[0003] The purpose of the present invention is to provide a calibration method for climate model data based on a spatio-temporal variation transfer function, which can effectively adjust variables to match the observational data, and the error correction effect is stable under different topographical and geomorphic conditions and elevation conditions.

[0004] To achieve the above purpose, the present invention provides a calibration method for climate model data based on a spatio-temporal variation transfer function, including the following steps:

[0005] Taking a single observation station as an independent individual, and using the quantile mapping method to obtain the transfer function between the measured data and the model data of the corresponding observation station:

[0006] x obs_ref =F ref (x sim_ref ) (1)

[0007] Wherein, F ref is the transfer function, and x obs_ref , x sim_ref are respectively the measured data of the observation station used in the solution process and its corresponding model data;

[0008] Compare the transfer functions at different model data grid points, analyze their spatial distribution characteristics and the changing trend of the underlying surface, calculate the transfer functions in different observation periods, establish the space-varying non-steady-state transfer function, and apply it to the calibration of model data:

[0009] x bc =F upd (x sim ,S,T) (2)

[0010] where F upd is the updated space-variable unsteady transfer function, x bc 、x sim are the corrected output data and the model input data respectively, S is the space-variable factor of the transfer function, and it is proposed to use F in formula (1) ref It is obtained after interpolation and reconstruction of the remaining grids. T is the non-steady-state factor of the transfer function, which is obtained by fitting the time-varying trend of the transfer function in different periods.

[0011] As a further solution of the present invention: the spatial variation factor S in formula (2) is mainly determined by the underlying surface factors of the study area. During the calculation process, the higher the altitude, the more drastic the change of the spatial variation factor. At the same time, the spatial variation factors of the surface covered by vegetation and bare land are also different. The spatial variation factor of bare land cover is more obvious than that of vegetation cover, and the corresponding S is more drastic. Its mathematical expression is as follows:

[0012] S=E Norm *C Norm (3)

[0013] Where E Norm and C Norm They represent the normalized altitude spatial variation factor and the normalized land cover spatial variation factor respectively, and the spatial variation factor S is the product of the two;

[0014] The non-steady-state factor T in formula (2) is obtained by linear fitting based on the average observation results according to the observation period, and its mathematical expression is as follows:

[0015] T=f lin (M obs ,t period ) (4)

[0016] Among them, M obs is the observed data, t period is the selected observation period, f lin It is the mathematical expression of linear fitting.

[0017] As a further solution of the present invention: Normalized altitude space-variable factor E Norm is based on observation data from different observation sites with the same vegetation cover. obs_CIt is calculated that its mathematical expression is:

[0018] E Norm = f lin (M obs_C , E) (5)

[0019] where E is the altitude of the observation site, M obs_C is the observation data of the observation site, and f lin is the linear fitting mathematical expression.

[0020] As a further solution of the present invention: The normalized surface cover spatial variation factor C Norm is calculated based on the observation data M of different stations with the same altitude obs_E and its mathematical expression is:

[0021] C Norm = f lin (M obs_E , C) (6)

[0022] where C is different surface covers, M obs_E is the observation data of different stations, and f lin is the linear fitting mathematical expression.

[0023] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0024] It can perform adaptive error calibration for specific underlying surface conditions of different grid points, and at the same time can fit the time-varying law under the background of global climate change, so as to effectively adjust variables to match the observation data, and the error correction effect under different periods, different topographies and altitudes is more stable than the existing methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is the spatial variation and non-steady state factor calculated by the present invention using the data of some basic meteorological observation stations in the research area.

[0026] Figure 2 This is the comparison of the results before and after the spatial variation non-steady state error correction of the present invention taking Xining as an example for temperature (upper), wind speed (middle) and water vapor pressure (lower). DETAILED DESCRIPTION OF THE INVENTION

[0027] The present invention will be further described below through embodiments.

[0028] A climate model data calibration method based on a spatio-temporal variation transfer function includes the following steps:

[0029] Taking a single observation site as an independent individual, using the quantile mapping method to obtain the transfer function between the measured data and the model data of the corresponding observation site:

[0030] x obs_ref =F ref (x sim_ref ) (1)

[0031] Among them, F ref is the transfer function, x obs_ref 、x sim_ref are the measured data of the observation stations and their corresponding model data used in the solution process;

[0032] Compare the transfer functions at different model data grid points, analyze their spatial distribution characteristics and the changing trends of the underlying surface (topography, altitude, vegetation cover), calculate the transfer functions within different observation periods (such as 1 year as a period), explain their non-steady-state characteristics, establish a space-variable non-steady-state transfer function, and apply it to the calibration of model data:

[0033] x bc =F upd (x sim ,S,T) (2)

[0034] where F upd is the updated space-variable unsteady transfer function, x bc 、x sim are the corrected output data and the model input data respectively, S is the space-variable factor of the transfer function, and it is proposed to use F in formula (1) ref It is obtained after interpolation and reconstruction of the remaining grids. T is the non-steady-state factor of the transfer function, which is obtained by fitting the time-varying trend of the transfer function in different periods.

[0035] Furthermore, the spatial variation factor S in formula (2) is mainly determined by the underlying surface factors of the study area. During the calculation process, the higher the altitude, the more drastic the change of the spatial variation factor. At the same time, the spatial variation factors of the surface covered by vegetation and bare land are also different. The spatial variation factor of bare land cover is more obvious than that of vegetation cover, and the corresponding S is more drastic. Its mathematical expression is as follows:

[0036] S=E Norm *C Norm (3)

[0037] Where E Norm and C Norm They represent the normalized altitude spatial variation factor and the normalized land cover spatial variation factor respectively, and the spatial variation factor S is the product of the two;

[0038] The non-steady-state factor T in formula (2) is obtained by linear fitting based on the average observation results according to the observation period, and its mathematical expression is as follows:

[0039] T=f lin(M obs ,t period ) (4)

[0040] where M obs is the observed data, t period is the selected observation period, and f lin is the mathematical expression of linear fitting.

[0041] Furthermore, the normalized elevation spatial variation factor E Norm is calculated based on the observed data M of different observation stations with the same vegetation cover obs_C , and its mathematical expression is:

[0042] E Norm = f lin (M obs_C ,E) (5)

[0043] where E is the elevation of the observation station, M obs_C is the observed data of the observation station, and f lin is the mathematical expression of linear fitting.

[0044] Furthermore, the normalized land cover spatial variation factor C Norm is calculated based on the observed data M of different stations with the same elevation obs_E , and its mathematical expression is:

[0045] C Norm = f lin (M obs_E ,C) (6)

[0046] where C is different land cover, M obs_E is the observed data of different stations, and f lin is the mathematical expression of linear fitting.

[0047] To verify the effectiveness of the above calibration method, the Qinghai-Tibet Plateau with severe elevation fluctuations is selected as the research area, and the CORDEX-EAST model, which is widely used in alpine regions and has good adaptability, is chosen as the climate model. Figure 1 is the spatial variation and non-steady state factor calculated using the data of some benchmark meteorological observation stations in this area. It can be seen from Figure 1 that the spatial variation and time-varying factors at different stations are different, which reflects the possible uncertainty in the simulation results of the model data for different underlying surfaces and different periods. The climate model data calibration method based on the spatio-temporal variation transfer function can perform adaptive error calibration for the specific underlying surface conditions of different grid points, and at the same time can fit the time-varying law under the background of global climate change, so as to effectively adjust the variables to match the observed data and achieve error correction under different periods, different topographies and elevation conditions.

[0048] Figure 2 Shows the comparison of the results before and after the correction of the air-variable non-steady error taking Xining as an example, including temperature (upper), wind speed (middle), and water vapor pressure (lower). Through analysis, it can be seen that there are relatively obvious deviations between the original simulation results of the climate model and the observed data, especially in terms of air temperature (upper). After being corrected by the error correction method based on the spatio-temporal variation transfer function, the model data is very close to the observed results, thus improving the reliability of the model simulation results.

Claims

1. A climate model data calibration method based on spatiotemporal variation transfer function, characterized in that: The following steps are involved: Taking a single observation station as an independent individual, the transfer function between the measured data and the model data of the corresponding observation station is obtained using the quantile mapping method: x obs_ref =F ref (x sim_ref ) (1) Among them, F ref is the transfer function, x obs_ref 、x sim_ref are the measured data of the observation stations and their corresponding model data used in the solution process; Compare the transfer functions at different model data grid points, analyze their spatial distribution characteristics and the changing trend of the underlying surface, calculate the transfer functions in different observation periods, establish the space-varying non-steady-state transfer function, and apply it to the calibration of model data: x bc =F upd (x sim ,S,T) (2) where F upd is the updated space-variable unsteady transfer function, x bc 、x sim are the corrected output data and the model input data respectively, S is the space-variable factor of the transfer function, and it is proposed to use F in formula (1) ref It is obtained after interpolation and reconstruction of the remaining grids. T is the non-steady-state factor of the transfer function, which is obtained by fitting the time-varying trend of the transfer function in different periods.

2. The method for calibrating climate model data based on spatiotemporal variation transfer function according to claim 1, characterized in that: The spatial variation factor S in formula (2) is mainly determined by the underlying surface factors of the study area. During the calculation process, the higher the altitude, the more drastic the change of the spatial variation factor. At the same time, the spatial variation factors of the surface covered by vegetation and bare land are also different. The spatial variation factor of bare land cover is more obvious than that of vegetation cover, and the corresponding S is more drastic. Its mathematical expression is as follows: S=E Norm *C Norm (3) Where E Norm and C Norm They represent the normalized altitude spatial variation factor and the normalized land cover spatial variation factor respectively, and the spatial variation factor S is the product of the two; The non-steady-state factor T in formula (2) is obtained by linear fitting based on the average observation results according to the observation period, and its mathematical expression is as follows: T=f lin (M obs ,t period ) (4) Among them, M obs is the observed data, t period is the selected observation period, f lin It is the mathematical expression of linear fitting.

3. The method for calibrating climate model data based on spatiotemporal variation transfer function according to claim 2, characterized in that: Normalized altitude spatial variation factor E Norm is based on observation data from different observation sites with the same vegetation cover. obs_C The mathematical expression is calculated as follows: HAVE BEEN Norm =f lin (M obs_C ,E) (5) Where E is the altitude of the observation station, M obs_C is the observation data of the observation site, f lin It is the mathematical expression of linear fitting.

4. A climate model data calibration method based on spatiotemporal variation transfer function according to claim 2 or 3, characterized in that: Normalized land cover spatial variation factor C Norm is based on observation data from different stations at the same altitude. obs_E The mathematical expression is calculated as follows: C Norm =f lin (M obs_E ,C) (6) Among them, C is different surface coverage, M obs_E is the observation data at different stations, f lin It is the mathematical expression of linear fitting.

Images