Statistical correction method for land water resource availability simulation data

Abstract

The invention discloses a statistical correction method for land water resource availability simulation data, and belongs to the technical field of climate data analysis. Constructing a space-time distribution simulation capability evaluation model, screening out a global climate system mode with a relatively good simulation effect, and determining historical simulation data # imgabs0 # and future simulation data # imgabs1 #; a cumulative distribution function is established, and the variable quantity and the change rate from historical simulation data # imgabs2 # of the same quantile to future simulation data # imgabs3 # are quantized; selecting a mode simulation trend storage mode, and obtaining deviation-corrected future simulation data # imgabs4 # by using a quantile mapping method; and sequencing the future simulation data # imgabs5 # subjected to deviation correction to obtain future simulation data # imgabs6 # based on a time sequence, and performing deviation correction on the multi-period simulation data to form a perfect historical simulation data set and a future simulation data set. According to the method, the WA simulation precision is fundamentally improved by taking the WA as a correction target variable, the statistical correction effect of the quantile mapping method on the WA variable is analyzed, and most climate scale research requirements are met.

Description

Technical Field

[0001] The present invention belongs to the technical field of climate data analysis, and particularly relates to a statistical correction method for simulated data of terrestrial water availability. Background Art

[0002] Terrestrial water availability (WA) is defined as the difference between precipitation and evapotranspiration in the field of climate research. On monthly and longer time scales, according to the atmospheric water vapor balance equation, WA is in balance with the divergence of the water vapor flux in the whole atmosphere layer; according to the terrestrial water balance equation, WA can generally be regarded as the main contributing term of local runoff and terrestrial water storage. The change of WA is closely related to the water cycle, is an important indicator of water resource change, and has a significant impact on human social activities and the natural ecological environment. Global climate system models are objective tools and means for carrying out climate prediction and projection. However, the simulated values of precipitation and evapotranspiration output by current models usually have large deviations compared with the observed values, resulting in poor simulation effects of the estimated terrestrial WA. Therefore, the model output data should be post-processed before analyzing the WA simulation data.

[0003] Existing studies on the spatio-temporal distribution characteristics of global WA often directly use the simulation results of equal-weighted multi-model ensemble means for analysis, or perform different corrections on the simulation data based on research purposes. For example, in the comparative study of the future changes in land-sea WA characteristics, only the evaporation in the ocean area is corrected to ensure the global land-sea total water balance; another example is that when studying the impact of the combined changes in precipitation and evapotranspiration in different climate regions on WA, the Bayesian model averaging method is used to perform multi-model weighted average correction on the precipitation and evapotranspiration output by the models respectively, and analyze the contributions and dominant situations of precipitation and evapotranspiration to the change of WA. In the study of the spatio-temporal evolution of water resources dominated by meteorological conditions and its ecological environment effects, it is urgent to correct the bias of the global gridded future simulation data of WA. However, there is still relatively little discussion on the grid-by-grid correction method for WA simulation data at present. Summary of the Invention

[0004] The present invention provides a statistical correction method for simulated data of terrestrial water availability to solve the technical problems existing in the above background art.

[0005] The present invention is achieved through the following technical solutions: A statistical correction method for simulated data of terrestrial water availability, comprising the following steps: Calculating the historical observed value of WA based on the observed data of precipitation and evapotranspiration in the historical period, and using the historical observed value of WA as the historical reference data ; Calculate the historical simulation value of WA based on the precipitation and evapotranspiration data in the historical period by simulation, construct a spatio-temporal distribution simulation ability evaluation model, screen out the global climate system models with better simulation effects of WA, and use the historical data of WA simulated by them as historical simulation data and the future data of WA as future simulation data ; Establish the cumulative distribution functions of the historical simulation data and the future simulation data to quantify the change amount from the historical simulation data to the future simulation data and the change rate , where is the cumulative probability; Establish the cumulative distribution function of the historical reference data According to the corresponding relationship of the quantile function values between the historical reference data and the historical simulation data , select an appropriate mode simulation trend preservation method; use the quantile mapping method to obtain the bias-corrected future simulation data containing the historical reference data and the future trend information of the model simulation ; Based on the quantile sequence of the future simulation data sort the bias-corrected future simulation data to obtain the future simulation data based on the time series .

[0006] In a further embodiment, the construction process of the spatio-temporal distribution simulation ability evaluation model is as follows: Create a scoring model, and use the scoring model to calculate the Taylor score and the interannual variability score corresponding to the global climate system model; among them, the higher the Taylor score , the better the spatial distribution simulation effect of the corresponding global climate system model on the WA value, and the lower the interannual variability score , the better the interannual variability simulation effect of the corresponding global climate system model on the WA value; Based on the Taylor score and the interannual variability score , use the comprehensive score to obtain the spatio-temporal distribution simulation ability of the global climate system model on the WA value, and the calculation formula of the comprehensive score is as follows: ; In the formula, is the number of global climate system models participating in the scoring, is the number of scoring methods, , then is the descending order of the scoring methods from the best, the larger the value, the stronger the simulation ability of the corresponding global climate system model for the spatio-temporal distribution characteristics of WA. The larger the value, the stronger the simulation ability of the corresponding global climate system model for the spatio-temporal distribution characteristics of WA.

[0007] In a further embodiment, the solution formula of the change amount is as follows: ; where represents the quantile function value of the future simulation data at the cumulative probability , is the quantile function value of the historical simulation data at the cumulative probability ; The calculation formula of the change rate is as follows: .

[0008] In a further embodiment, the mode simulation trend preservation methods at least include: additive simulation trend preservation method, multiplicative simulation trend preservation method, and hybrid simulation trend preservation method; correspondingly, the steps for selecting and adapting the mode simulation trend preservation method are as follows: Step 101, obtain the quantile function value of the historical reference data and the quantile function value of the historical simulation data at the same cumulative probability . If the signs of the quantile function value and the quantile function value are the same, then execute Step 102; otherwise, select the additive simulation trend preservation method; Step 102, if , then select the multiplicative simulation trend preservation method, otherwise execute Step 103; Step 103, if , then select the hybrid simulation trend preservation method, otherwise select the additive simulation trend preservation method.

[0009] In a further embodiment, the process of obtaining the bias-corrected future simulation data is as follows: Based on the quantile mapping method, couple the change trend simulated by the model to the historical reference data through the adapted mode simulation trend preservation method, and its coupling form is as follows: , is to select the adapted mode simulation trend preservation method; Correspondingly, the future simulation data based on time series is expressed as: , where represents the year of the future simulation data, is the empirical cumulative distribution function of the future simulation data, is the quantile function of the bias-corrected future simulation data, represents the year of the bias-corrected future simulation value.

[0010] In a further embodiment, the quantile mapping method of the additive simulation trend preservation method is as follows: ; in the formula, is the historical reference data of the year , where is the cumulative probability, represents the bias-corrected future simulation data at the cumulative probability .

[0011] In a further embodiment, the quantile mapping method of the multiplicative simulation trend preservation method is as follows: ; in the formula, is the historical observation data of the year , is the cumulative probability, represents the bias-corrected future simulation data at the cumulative probability , .

[0012] In a further embodiment, the quantile mapping method of the hybrid simulation trend preservation method is as follows: ; where is the historical observation data of the year , is the cumulative probability, represents the bias-corrected future simulation data at the cumulative probability , is the hybrid correction operator.

[0013] In a further embodiment, the formula for calculating the hybrid correction operator is as follows: , in the formula, represents the historical reference data at the cumulative probability The quantile function value, is the historical simulation data at the cumulative probability of the quantile function value.

[0014] In a further embodiment, the following steps are further included: Keeping the historical period unchanged as the modeling period, taking the future period as the calibration period, and using a sliding window to select different calibration periods to calibrate the future simulation data, so as to form a long time series of the bias-corrected historical simulation data and future simulation data; Performing the above calibration process on the global grid point, monthly, and individual global climate system future simulation data to form a bias-corrected historical simulation data set and a future simulation data set, and obtaining a perfect historical simulation data set and a future simulation data set.

[0015] Advantages of the present invention: The present invention directly uses WA as the calibration target variable, skips the process of calibrating each variable one by one for a single component, and improves the WA simulation accuracy fundamentally by globally optimizing the distribution characteristics of the difference. This design avoids the cumulative error problem of traditional step-by-step calibration, and is especially suitable for the physical mechanism in which WA is affected by the combined changes of precipitation and evapotranspiration, ensuring that the calibrated data can more truly reflect the actual state of water resource surplus / deficit.

[0016] This article describes the statistical calibration steps of the quantile mapping method for the WA variable, expands the multiplication and mixed simulation trend preservation methods in the quantile mapping method into algorithms applicable to both positive and negative values, analyzes and compares the calibration effects of the addition, multiplication, and mixed simulation trend preservation methods on WA, selects the optimal simulation trend preservation method to form a globally gridded WA simulation calibration data set on a monthly time scale, and the data distribution meets the requirements of most climate scale studies. Description of the Drawings

[0017] Figure 1 is the flowchart of the statistical calibration method for the land water resource availability simulation data in Embodiment 1.

[0018] Figure 2 is the schematic diagram of the cumulative distribution function of the historical reference data, historical simulation data, future simulation data, and the quantile mapping method bias-corrected future simulation data of various simulation trend preservation methods.

[0019] Figure 3 is the extended mixed calibration operator in the mixed simulation trend preservation method of the schematic diagram. Detailed Embodiment

[0020] The present invention will be further described below in conjunction with the drawings of the specification and the embodiments.

[0021] Embodiment 1 This embodiment discloses a statistical correction method for simulated data of terrestrial water resource availability, as Figure 1 shown, including: Calculating the historical observed value of WA based on the precipitation and evapotranspiration observed data in the historical period, and using the historical observed value of WA as the historical reference data ; calculating the historical simulated value of WA based on the precipitation and evapotranspiration data in the historical period simulated by the model, constructing a spatio-temporal distribution simulation ability evaluation model, screening out the global climate system models with better WA simulation effects, and using the historical WA data simulated by them as the historical simulation data and the future WA data as the future simulation data .

[0022] Establish the cumulative distribution functions of the historical simulation data and the future simulation data , quantify the change from the historical simulation data at the quantile to the future simulation data and the change rate , , where is the cumulative probability; Establish the cumulative distribution function of the historical reference data , select the appropriate mode simulation trend preservation method according to the corresponding relationship of the quantile function values between the historical reference data and the historical simulation data ; use the quantile mapping method to obtain the bias-corrected future simulation data containing the historical reference data and the future trend information of the model simulation ; Sort the bias-corrected future simulation data based on the quantile sequence of the future simulation data to obtain the future simulation data based on the time series .

[0023] It should be noted that the precipitation and evapotranspiration observed data in the historical period described in this embodiment are all from the ERA5-Land reanalysis land surface meteorological data set provided by the European Centre for Medium-Range Weather Forecasts, which has the characteristics of high precision, grid-based, and coherent change of multiple variables, such as the precipitation and evapotranspiration data with a monthly accuracy of 0.1° from 1950 to 2023. Further, , where is the precipitation, is the evapotranspiration in the same period. Therefore, the historical reference data of WA can be directly calculated from the precipitation and evapotranspiration observed data in the historical period.

[0024] In addition, the global climate system model described in this embodiment may be 29 global climate system models in the Sixth Coupled Model Intercomparison Project (CMIP6). These 29 global climate system models contain precipitation and evapotranspiration simulation data for the historical simulation experiment from 1850 to 2014 and the future scenario model comparison plan (ScenarioMIP) from 2015 to 2100. Based on the above example, a global climate system model with better WA simulation effect is selected from 29 global climate system models using the spatio-temporal distribution simulation ability evaluation model.

[0025] Furthermore, the construction process of the spatio-temporal distribution simulation ability evaluation model is as follows: Create a scoring model, and use the scoring model to calculate the Taylor score corresponding to the global climate system model and the interannual variability score ; among them, the higher the Taylor score , the better the spatial distribution simulation effect of the corresponding global climate system model on the WA value. The highest score of the Taylor score is 1; the lower the interannual variability score , the better the interannual variability simulation effect of the corresponding global climate system model on the WA value. The lowest score of the interannual variability score is 0.

[0026] Based on the Taylor score and the interannual variability score , use the comprehensive score to obtain the spatio-temporal distribution simulation ability of the global climate system model on WA. The calculation formula of the comprehensive score is as follows: ; In the formula, is the number of global climate system models participating in the scoring, is the number of scoring methods, , then is the descending order of the scoring method . The larger the value, the stronger the spatio-temporal distribution characteristic simulation ability of the corresponding global climate system model on WA. Combining this embodiment, , .

[0027] In this embodiment, the calculation formulas of the Taylor score and the interannual variability score are as follows respectively: ; ; In the formula, For the Taylor score, R represents the centered spatial correlation coefficient between the future simulation data and the historical reference data of the global climate system of WA. is the maximum value in the spatial correlation coefficient. The superscript ref represents the reference data, and sim represents the simulation data. represents the standard deviation of the spatial dimension corresponding to the simulation data. is the standard deviation of the spatial dimension corresponding to the reference data. is the interannual variability score. is the standard deviation of the WA simulation data in the time dimension. is the standard deviation of the WA reference data in the time dimension.

[0028] Table 1 Information and score rankings of five models

[0029] Based on this, in this embodiment, among 29 global climate system models, the top 13 models with strong simulation capabilities of the spatio-temporal distribution of WA from 1951 to 2014 are selected for subsequent calibration work. Table 1 provides five of the 13 selected global climate system models (CESM2-WACCM, CMCC-ESM2, CMCC-CM2-SR5, NorESM2-MM, TaiESM1, MIROC6), as well as the corresponding comprehensive scores.

[0030] All data in this embodiment are processed to a precision of 1° using the bilinear interpolation method.

[0031] In the prior art, theoretical distribution functions with parameters are often used to determine the distribution of variables. However, more and more studies have shown that these theoretical distribution functions are not applicable to all spatial grid points and global climate system models. On the contrary, an empirical cumulative distribution function without parameters can be easily obtained through the cubic spline function smoothing method, which has a better matching degree with the reference data. Therefore, this embodiment uses the cubic spline function smoothing method to obtain the cumulative distribution function of WA data.

[0032] Historical simulation data of the same quantile to future simulation data The change trend can be quantified using the change amount or the change rate. Correspondingly, in this embodiment, the change amount is solved by the following formula: ; In the formula, represents the quantile function value of the future simulation data at the cumulative probability , is the quantile function value of the historical simulation data at the cumulative probability ; The change rate The calculation formula is as follows: ; In the formula, represents the quantile function value of the future simulated data at the cumulative probability , is the quantile function value of the historical simulated data at the cumulative probability .

[0033] Furthermore, using the quantile mapping method of the trend preservation method for multiple model simulations, the quantified model-simulated future change trend is coupled to the historical observed data to form the bias-corrected future simulated data. For example, only the additive trend preservation method of the model-simulated future change amount can be used for quantile mapping correction. However, even if the WA observations and simulation values at the same grid point are both of the same sign, the additive trend preservation method can still cause the sign of the corrected WA simulation data to change, confusing the local water surplus or deficit situation. And only using the multiplicative trend preservation method of the model-simulated future change rate in the quantile mapping method, although it ensures the consistency of the sign of the correction result in this case, when the historical observed value is much larger than the historical simulated value , even if the amplitude of the future change rate is limited to a reasonable range, the multiplicative trend preservation method will still result in false extremely large values. In addition, when the WA historical simulation and observed values are of different signs, that is, the model historical simulation result fails to accurately simulate the WA sign, the multiplicative trend preservation method will transmit the wrong simulation sign to the subsequent correction value.

[0034] Therefore, this embodiment introduces a hybrid trend preservation method that simultaneously includes the model-simulated change amount and the change rate , and adds a selection and adaptation step for the trend preservation method to form a quantile-adapted trend preservation method. By comparing the corresponding situations of the model simulations and observed data in the historical period at the same quantile, the trend preservation method of the model simulation is refined under different historical simulation bias conditions, and the additive, multiplicative, and hybrid trend preservation methods are comprehensively used for the bias correction processing of the quantile mapping method.

[0035] Therefore, the trend preservation method of the model simulation in this embodiment at least includes: the additive trend preservation method, the multiplicative trend preservation method, and the hybrid trend preservation method.

[0036] Furthermore, taking the subscript hist as the historical period and fut as the future period, the steps for the selection and adaptation of the trend preservation method of the model simulation are as follows: Step 101, obtain the same cumulative probability Quantile function values of historical reference data and quantile function values of historical simulation data , if the signs of the quantile function value and the quantile function value are the same, then step 102 is executed; otherwise, the addition simulation trend preservation method is selected; Step 102: If , then the multiplication simulation trend preservation method is selected, otherwise step 103 is executed; Step 103: If , then the hybrid simulation trend preservation method is selected, otherwise the addition simulation trend preservation method is selected.

[0037] Furthermore, the process of obtaining the bias-corrected future simulation data is as follows: Based on the quantile mapping method, the changing trend of the pattern simulation is coupled to the historical reference data through the adapted pattern simulation trend preservation method , and its coupling form is as follows: , is to select the adapted pattern simulation trend preservation method; To better understand the mathematical principles of each simulation trend preservation method and its selection and adaptation steps in the quantile mapping method, taking the CESM2-WACCM future simulation data and ERA5-Land observation data of WA in July at the grid point of (30°S, 70°W) as an example, the addition simulation trend preservation method, multiplication simulation trend preservation method, and hybrid simulation trend preservation method of the quantile mapping correction method are further elaborated.

[0038] Combined with Figure 2 , the historical reference data is represented by a black solid line, the historical simulation data is represented by a blue solid line, the future simulation data is represented by a blue dashed line, the correction result of the quantile mapping method using only the addition simulation trend preservation method is represented by an orange dashed line, the correction result of the quantile mapping method using only the multiplication simulation trend preservation method is represented by a red dashed line, and the correction result of the quantile mapping method using the quantile-adapted simulation trend preservation method is represented by a red solid line, respectively indicating the corresponding cumulative distribution functions. Figure 2 The dark blue and light blue curves show that there are significant differences in the distribution characteristics of the historical and future simulation values of WA at the same grid point, so a correction method for preserving the pattern simulation trend is required.

[0039] As Figure 2 shown by the green arrow in (a), the form of the quantile mapping method of the addition simulation trend preservation method is as follows: ; in the formula, is the year Historical reference data, is the cumulative probability, indicating the cumulative probability of the corrected future simulation data.

[0040] Figure 2 (a) shows that the simulated future change calculated at the cumulative probability is linearly superimposed on the historical observation data, Figure 2 and the orange dashed line in

[0041] is the correction result using only the additive simulation trend preservation method. Figure 2 The quantile mapping method of the multiplicative simulation trend preservation method is as shown by the green arrow in (b) of Figure 2 (b), however, it should be noted that the cases where negative signs appear in historical observations and simulation values are not shown in ; in the formula, is the historical observation data for year , is the cumulative probability, indicating the corrected future simulation data at the cumulative probability . Figure 2 The red dashed line in

[0042] is the correction result of the quantile mapping method using only the multiplicative simulation trend preservation method. The quantile mapping method of the mixed simulation trend preservation method is as follows: where is the historical reference data for year is the cumulative probability, indicating the corrected future simulation data, and

[0043] is the mixed correction operator. Furthermore, the calculation formula of the mixed correction operator is as follows: where represents the quantile function value of the historical reference data at the cumulative probability is the quantile function value of the historical simulation data at the cumulative probability .

[0044] Furthermore, the mixed correction operator By taking the value of , the selection adaptation step of the described simulated trend preservation method can be integrated into the formula of the hybrid simulated trend preservation method to form a simulated trend preservation method with quantile adaptation. The extended hybrid correction operator The formula is as follows: ; In the formula, represents the quantile function value of the historical reference data at the cumulative probability , is the quantile function value of the historical simulation data at the cumulative probability .

[0045] Figure 3 is the distribution characteristic of the extended hybrid correction operator , showing the selection steps of different simulated trend preservation methods in the quantile mapping method and the value of when using the hybrid trend preservation method . Among them, the horizontal coordinate QQ in Figure 3 is the abbreviation of . The purple circle is the operator value corresponding to the relationship between the ERA5-Land value and the CESM2-WACCM simulation value at the example grid point (30°S, 70°W) from July 1951 to 2014. There is a situation where the historical simulation and observed data at the same quantile have different signs at this grid point (i.e., ), at this time is 0, that is, the additive simulated trend preservation method is used, and the absolute value of the observed value at the remaining quantiles is appropriately greater than the simulated value ( ), at this time that is, the hybrid simulated trend preservation method is used.

[0046] It can be seen that this method couples the future change trend simulated by the physical equations of the global climate system model on the basis of historical observation data, and the calibration idea is intuitive and reasonable.

[0047] Furthermore, the expression form of the future simulation data based on the time series is: , where represents the future simulation data of the year , is the empirical cumulative distribution function of the future simulation data, is the quantile function of the bias-corrected future simulation data, represents the bias-corrected future simulation value of the year .

[0048] In another embodiment, the following steps are further included: Keep the historical period as the modeling period unchanged, use the future period as the calibration period, and select different calibration periods using a sliding window to calibrate the future simulation data, forming a long time series of bias-corrected historical simulation data and future simulation data; Perform the above calibration process on the global grid-point, monthly, and individual global climate system future simulation data to form a bias-corrected historical simulation data set and a future simulation data set, obtaining a complete historical simulation data set and a future simulation data set.

[0049] For example, for the global climate system simulation data of the widely used Sixth Coupled Model Intercomparison Project (CMIP6), the 36-year period from 1979 to 2014 can be set as the modeling period, and the calibration period is set as a 36-year window of the same time length. The historical simulation data is calibrated in segments, and the period division is as follows: 1850 - 1885, 1886 - 1921, 1922 - 1957, 1958 - 1993, and 1994 - 2029. Among them, 1994 - 2029 is the combination of the historical simulation data from 1994 to 2014 and the future simulation data under the SSP-5.85 scenario from 2015 to 2029, thus forming the bias-corrected historical simulation data from 1850 to 2014; the calibration period division for the future simulation data is as follows: 2015 - 2050 (retain the calibration results from 2015 to 2043), 2040 - 2075 (retain the calibration results from 2044 to 2072), 2065 - 2100 (retain the calibration results from 2073 to 2100), thus forming the bias-corrected future simulation data from 2015 to 2100.

[0050] In another embodiment, it further includes: quantitatively evaluating and comparatively analyzing the calibration results of the quantile mapping method for the additive simulation trend preservation method, the multiplicative simulation trend preservation method, and the quantile adaptation simulation trend preservation method.

[0051] Furthermore, three indicators, namely the centered spatial correlation coefficient (R), the mean absolute error (MAE; mm day -1 ), and the sign agreement rate (SAR; %), are used to test the bias correction effect of the quantile mapping method of different simulation trend preservation methods on WA.

[0052] The analysis period is set to 64 years from 1951 to 2014. This period is divided into 4 sub-periods with a sliding window of 10 years. Among them, 32 consecutive years are the modeling periods for the statistical calibration algorithm, namely 1951 - 1982, 1961 - 1992, 1971 - 2012, and 1981 - 2012. The remaining 32 years are the test periods. The cross-validation results composed of these 4 groups of data can reduce the influence of the selection of modeling and test periods on the test results.

[0053] R represents the spatial similarity between the simulation and the reference data, and the optimal value is 1; MAE measures the magnitude of the average difference between the simulated value and the reference value, and the optimal value is 0 mm day -1 ; SAR depicts the same-sign rate of the simulated value and the reference value, and the optimal value is 100%. Calculate the values of the three indicators for each sub-period and each global climate system model (where MAE and SAR first calculate the evaluation indicators for each grid point), and then take the average to obtain the overall indicators. At the same time, it is noted that each annual average parameter has 52 samples (4 sub-periods × 13 models). Therefore, the non-parametric Mann-Whitney U test can be used to compare whether there are significant differences in each evaluation indicator under different simulation trend preservation methods. The significance level tested in this paper is p < 0.05.

[0054] While evaluating the global average simulation effect of the WA variable, the WA correction effect under different drought conditions is evaluated. The degree of terrestrial drought is quantified by the average drought index (Aridity Index, AI) from 1951 to 2014. The definition of AI is the ratio of annual precipitation to annual potential evapotranspiration: AI < 0.05 is the extreme arid area, 0.05 ≤ AI < 0.20 is the arid area, 0.20 ≤ AI < 0.50 is the semi-arid area, 0.50 ≤ AI < 0.65 is the dry sub-humid area, and AI ≥ 0.65 is the humid area.

[0055] Example 2 This example discloses a statistical calibration system for simulated data of terrestrial water availability, which is used to implement the statistical calibration method for simulated data of terrestrial water availability as described in the example, including: The first module is configured to calculate the historical observed value of WA based on the observed data of precipitation and evapotranspiration in the historical period, and use the historical observed value of WA as the historical reference data ; calculate the historical simulated value of WA based on the simulated precipitation and evapotranspiration data in the historical period, construct a spatio-temporal distribution simulation ability evaluation model, screen out the global climate system models with better WA simulation effects, and use the simulated historical data of WA as the historical simulation data and the future data of WA as the future simulation data ; The second module is configured to establish historical simulation data and future simulation data of the cumulative distribution function, and quantify the historical simulation data of the quantiles to the change amount from historical simulation data and the change rate , as the cumulative probability; The third module is configured to establish the cumulative distribution function of historical reference data , and select an appropriate mode simulation trend preservation method according to the corresponding relationship of the quantile function values between the historical reference data and the historical simulation data ; use the quantile mapping method to obtain bias-corrected future simulation data containing historical reference data and mode simulation future trend information ; The fourth module is configured to sort the bias-corrected future simulation data based on the quantile sequence of the future simulation data to obtain future simulation data based on the time series ; The fifth module is configured to keep the historical period unchanged as the modeling period, use the future period as the correction period, and use a sliding window to select different correction periods to correct the future simulation data, forming a long time series of bias-corrected historical simulation data and future simulation data; The sixth module is configured to perform the above correction processing on the global grid-point, monthly, and individual global climate system future simulation data, form bias-corrected historical simulation data sets and future simulation data sets, and obtain complete historical simulation data sets and future simulation data sets.

Claims

1. A statistical correction method for simulated data of terrestrial water resource availability, characterized in that, It includes the following steps: The historical observed values of WA are calculated based on the observed precipitation and evapotranspiration data in the historical period, and the historical observed values of WA are used as historical reference data ; The historical simulated values of WA are calculated based on the simulated precipitation and evapotranspiration data in the historical period, a spatio-temporal distribution simulation ability evaluation model is constructed, the global climate system models with better simulation effects of WA are selected, and the historical data of WA simulated by them are used as historical simulation data , and the future data of WA are used as future simulation data ; Establish historical simulation data and future simulation data of the cumulative distribution function, and quantify the historical simulation data of the quantile to the future simulation data of the change amount and the change rate , where the cumulative probability is; Establish historical reference data The cumulative distribution function is based on the historical reference data and the historical simulation data Based on the quantile function value correspondence between them, select an appropriate mode to simulate the trend preservation method; Using the quantile mapping method, obtain the bias-corrected future simulation data that contains historical reference data and information on the future trends simulated by the model ; Based on future simulated data The quantile sequence of the bias-corrected future simulated data is sorted to obtain future simulated data based on the time series .

2. The statistical calibration method for terrestrial water resource availability simulation data according to claim 1, wherein The construction process of the spatio-temporal distribution simulation ability evaluation model is as follows: Create a scoring model and calculate the Taylor score corresponding to the global climate system model using the scoring model and the interannual variability score ; among them, the higher the Taylor score , the better the simulation effect of the corresponding global climate system model on the spatial distribution of the WA value, and the lower the interannual variability score , the better the simulation effect of the corresponding global climate system model on the interannual variability of the WA value; Based on the Taylor score and the interannual variability score , the simulation ability of the global climate system model for the spatio-temporal distribution of the WA value is obtained using a comprehensive score . The calculation formula of the comprehensive score is as follows: ； In the formula, is the number of global climate system models participating in the scoring, is the number of scoring methods, , then is the descending order of the scoring method from the best, and the larger the value, the stronger the simulation ability of the corresponding global climate system model for the spatio-temporal distribution characteristics of WA.​ 3. A statistical correction method for simulated data of land water resource availability according to claim 1, characterized in that The amount of change The solution formula is as follows: ; where, represents the quantile function value of future simulated data at the cumulative probability , is the quantile function value of historical simulated data at the cumulative probability ; The rate of change is calculated as follows: 。 4. A statistical correction method for simulating data on the availability of land water resources according to claim 1, characterized in that The pattern simulation trend preservation methods at least include: addition simulation trend preservation method, multiplication simulation trend preservation method, and mixed simulation trend preservation method; correspondingly, the steps for selecting and adapting the pattern simulation trend preservation methods are as follows: Step 101: Obtain the same cumulative probability The quantile function value of the historical reference data and the quantile function value of historical simulation data , if the quantile function value and quantile function values If the signs of are the same, then step 102 is executed; otherwise, the additive simulation trend saving mode is selected; Step 102, if , then select the multiplication simulation trend preservation method, otherwise execute Step 103; Step 103, if , then select the hybrid simulation trend saving method; otherwise, select the addition simulation trend saving method.

5. A statistical correction method for simulating data on the availability of land water resources according to claim 1, characterized in that The future simulation data of the deviation correction is obtained as follows: Based on the quantile mapping method, the changing trend simulated by the model is coupled to the historical reference data through an adapted pattern simulation trend preservation method , and its coupling form is as follows: , Save the mode simulation trend for selection of the adapted mode Correspondingly, future simulation data based on time series The expression form is as follows: , where represents the year of future simulation data, is the empirical cumulative distribution function of future simulation data, is the quantile function of the bias-corrected future simulation data, represents the year of the bias-corrected future simulation values.

6. A statistical correction method for simulated data of terrestrial water resource availability according to claim 4, characterized in that The form of the quantile mapping method for the addition simulation trend preservation method is as follows: ; wherein, is the historical reference data of the year , where is the cumulative probability, represents the future simulation data of bias correction at the cumulative probability .

7. A statistical correction method for simulated data of land water resource availability according to claim 4, characterized in that The form of the quantile mapping method for the multiplication simulation trend preservation method is as follows: ; wherein, is the historical observation data of the year , is the cumulative probability of , indicating the bias-corrected future simulation data at the cumulative probability , .

8. A statistical correction method for simulated data of terrestrial water resource availability according to claim 4, characterized in that, The form of the quantile mapping method for the mixed simulation trend preservation method is as follows: ; wherein, is the historical observation data of the year , is the cumulative probability, represents the future simulation data of bias correction at the cumulative probability , and is the hybrid correction operator.

9. A statistical correction method for simulated data of land water resource availability according to claim 8, characterized in that The hybrid correction operator has the following calculation formula: , where represents the quantile function value of the historical reference data at the cumulative probability , is the quantile function value of the historical simulation data at the cumulative probability .

10. A statistical correction method for simulated data of land water resource availability according to claim 1, characterized in that, It also includes the following steps: Taking the historical period as the modeling period without change, using the future period as the calibration period, and using a sliding window to select different calibration periods to calibrate the future simulation data, forming a long time series of bias-corrected historical simulation data and future simulation data; Performing the above calibration processing on the global grid-point, monthly, and individual global climate system future simulation data to form a bias-corrected historical simulation data set and future simulation data set, and obtaining a complete historical simulation data set and future simulation data set.

Images