A refined spatial analysis method for precipitation temporal variability characteristics

Abstract

The present invention discloses a method for fine-grained spatial analysis of precipitation temporal variation characteristics, comprising the following steps: pre-processing satellite precipitation data and ground station observation precipitation data respectively; statistical downscaling using quadratic multivariate regression equation; using station data to correct errors of the two downscaled data respectively; fusing the two downscaled data; sub-gridding precipitation; and analyzing the trend, volatility, and annual distribution changes of precipitation grid points. Compared with the traditional method of obtaining spatial distribution of precipitation temporal variation characteristics, the present invention can obtain more effective and fine precipitation spatial distribution conditions.

Description

Technical Field

[0001] The present invention relates to a spatial distribution analysis technology of temporal variation characteristics of precipitation data, and in particular to a refined spatial analysis method of temporal variation characteristics of precipitation based on multi-source data. Background Art

[0002] Climate factors are an important foundation for human survival and development, and are also important factors affecting social and economic development. Precipitation, as one of the important sources of surface water and groundwater resources, plays an important role in terrestrial water cycle and ecosystem evolution. However, precipitation is affected by many factors such as region, topography, and underlying surface, and its temporal and spatial distribution is relatively complex. Over the past half century, almost all regions in the world have experienced a warming process, and the continued rise in temperature has led to an acceleration of the water cycle at the global and regional scales. The temporal and spatial characteristics of precipitation events have changed significantly. Obtaining accurate information on the temporal and spatial changes in precipitation is not only crucial to studying climate trends and variability, but also to regional water resources planning and management, ecological and environmental protection, and so on.

[0003] There are four main ways to obtain precipitation information: precipitation site observation, precipitation radar observation, satellite remote sensing inversion, and precipitation reanalysis data. Precipitation site monitoring can obtain relatively accurate precipitation information. However, it is difficult to obtain complete precipitation spatial information through site precipitation data due to the number of precipitation sites, site spatial distribution, site representativeness, and precipitation spatial characteristics. Satellite remote sensing precipitation inversion mainly uses different wavelengths of signals emitted and received by different onboard sensors to invert precipitation. Satellite remote sensing precipitation inversion can achieve large-scale scanning. Spatial precipitation information can be obtained, but different sensors and their inversion methods have a great impact on the accuracy of satellite precipitation observations. In addition, the spatial resolution of currently commonly used precipitation remote sensing products is low and difficult to meet the needs of various work. The data assimilation system obtains precipitation spatial information by continuously simulating existing meteorological data and land surface process models, and can be used to predict future precipitation. However, the data assimilation system is restricted by the regional adaptability of the system and other ground data conditions, and the accuracy varies greatly. In summary, all types of precipitation data have obvious limitations in obtaining spatial estimates of precipitation. Especially for areas with large spatial variability in precipitation (hilly areas) or sparsely distributed precipitation stations, it is difficult to evaluate the spatial information of precipitation data using single-source precipitation data. Precipitation downscaling can largely reduce the deficiencies of station and satellite remote sensing precipitation products, and relatively accurate and high-resolution precipitation spatial data can be obtained in a single time period. However, satellite remote sensing precipitation products from different sources have different spatial expression information, and there are certain deviations when analyzing the spatial distribution of precipitation temporal variability.

[0004] Disadvantages of existing technology:

[0005] 1. All existing precipitation downscaling methods adopt the "single precipitation satellite remote sensing data" + "precipitation site" model, and use different downscaling statistical methods to estimate precipitation spatial information. However, due to the differences in sensor types and inversion methods, there are certain differences in precipitation spatial information for different precipitation satellite remote sensing data. As a result, the precipitation spatial information of different precipitation satellite remote sensing data is quite different.

[0006] 2. The temporal variation characteristics of existing precipitation data are mostly analyzed using station data. For areas with a high density of precipitation stations and a relatively uniform distribution, the regional precipitation variation distribution can be obtained through methods such as station interpolation. However, for areas with large spatial variations in precipitation, such as mountainous areas where the precipitation gradient is more complex due to the influence of orographic rainfall, and the distribution of precipitation stations in mountainous areas is limited due to conditions, it is difficult to obtain accurate spatial distribution of precipitation temporal variation characteristics through interpolation. The spatial distribution resolution of temporal variation characteristics obtained through satellite remote sensing data is low and cannot meet the needs of refined water resources management. Summary of the invention

[0007] The main purpose of the present invention is to provide a method for spatial analysis of the temporal variation characteristics of precipitation.

[0008] The technical solution adopted by the present invention is: A method for fine-tuning spatial analysis of precipitation temporal variation characteristics, comprising the following steps:

[0009] The satellite precipitation data and ground station observation precipitation data were preprocessed separately, and the Trmm 3B42 and CMORPH CRT data within the study area were extracted and the data format was unified; the terrain data, vegetation index NDVI and other data were resampled, and 1km data, data with the same resolution as Trmm and CMORPH, and 0.5km resolution DEM data were extracted respectively;

[0010] The quadratic multiple regression equation was used for statistical downscaling. The quadratic multiple regression coefficients were calculated for the two data of monthly scale Trmm and CMORPH with longitude, latitude, elevation, slope, aspect and vegetation within the research time range. The preliminary downscaling data of Trmm with a resolution of 1 km and the preliminary downscaling data of CMORPH precipitation were obtained according to the regression coefficients.

[0011] The station data pairs were used to correct the errors of the two downscaled data respectively. The errors of the preliminary downscaled data and the measured data of the station were analyzed according to the station location, and the errors were corrected using the trend surface method. The Trmm downscaling error correction data and the CMORPH precipitation downscaling error correction data were obtained.

[0012] Two types of downscaled data fusion;

[0013] precipitation subgridding;

[0014] Analysis of precipitation grid trend, volatility and annual distribution changes.

[0015] Furthermore, the statistical downscaling using a quadratic multivariate regression equation includes:

[0016] First, the regression coefficients are solved for Trmm and CMORPH data on a monthly basis using the following regression equation:

[0017]

[0018] Where: Lat, Lon, Ele, P, X, NDVI are the grid latitude, longitude, elevation, slope, aspect, NDVI index and other values, respectively, and g is the regression constant. The least squares method is used to solve the regression coefficients a1, a2…f1, f2 and the regression constant g in the above formula. Substituting the 1 km scale latitude, longitude, elevation, slope, aspect and NDVI index into the above formula, the preliminary precipitation downscaling results of Trmm and CMORPH data can be obtained respectively.

[0019] Furthermore, the using the station data pair to respectively correct the errors of the two downscaled data comprises:

[0020] The following formula is used to calculate the deviation between the initial downscaled data and the measured value of the station for each station corresponding grid:

[0021]

[0022] Where: Ps is the observation data of the precipitation station, Pd0 is the grid value corresponding to the preliminary downscaling, and ΔP is the deviation between the preliminary downscaling data and the actual value of the station;

[0023] In order to avoid the bull's eye error caused by conventional interpolation, this patent uses the trend surface method to interpolate the deviation:

[0024]

[0025] In the formula, a1, b1, c1, and g are residual regression coefficients and regression constants, which are obtained by the least squares algorithm of station residuals; the following formula is used to correct the preliminary downscaled data; the downscaled data of Trmm and CMORPH data are obtained respectively;

[0026] .

[0027] Furthermore, the two downscaling data fusions include:

[0028] The following formula is used to fuse the two types of precipitation statistical downscaling data:

[0029]

[0030] Where: Pd is the 1 km precipitation downscaling fusion data, PTrmm and PCMORPH are the downscaling results of Trmm and CMORPH satellite remote sensing precipitation data respectively.

[0031] Furthermore, the precipitation sub-gridding includes:

[0032] Based on the 1km grid precipitation data, the precipitation grid point data is sub-gridded according to the precipitation gradient information to obtain the 0.5km grid precipitation data;

[0033] The precipitation elevation gradient value is calculated using the following formula:

[0034]

[0035] Where: a1 is the precipitation elevation gradient, P is the 1km precipitation downscaling value, and Ele is the corresponding grid elevation;

[0036] On each grid of 1km data, the following formula is used to calculate precipitation downscaling:

[0037]

[0038] Where: Pd is the downscaled precipitation data with a resolution of 0.5 km, P0 is the downscaled precipitation data with a resolution of 1 km, Ele0 is the elevation of the central point of the 1 km downscaled data, and Eled is the average elevation of each grid in the 1 km downscaled data grid.

[0039] Furthermore, the precipitation trend and volatility include:

[0040] Extract the annual precipitation sequence Y on each grid, take time ti as the independent variable and annual precipitation yi as the dependent variable, and establish a univariate linear regression equation;

[0041]

[0042] Where: a is the regression constant; b is the regression coefficient, also known as the trend coefficient. The positive or negative sign of b reflects the linear change trend of the variable rising or falling; b < 0 means that the variable is decreasing with time during the calculation period, and b > 0 means that the variable is increasing with time. The following formula is used to calculate the fluctuation characteristic value:

[0043]

[0044]

[0045] Where: Zi is the residual absolute value sequence, d is the residual regression coefficient, also known as the variable fluctuation coefficient; d < 0 means that it shows a downward trend with time during the calculation period, that is, the fluctuation of the variable shows a weakening trend; d > 0 means that it shows an upward trend with time, that is, the fluctuation of the variable shows an increasing trend; the regional 0.5 km resolution fluctuation characteristic value and trend value can be obtained.

[0046] Furthermore, the annual distribution of precipitation includes:

[0047] The Gini coefficient and Lorenz asymmetry coefficient are used to express the annual distribution indicators, and the area method is used for calculation. The formula is as follows:

[0048]

[0049] In the formula, G is the Gini index, yi is the cumulative percentage of monthly precipitation in the year, and xi is the cumulative percentage value of time;

[0050] Calculation of the Lorenz asymmetry coefficient: For a certain year, the monthly precipitation data is sorted from small to large, that is, x1≤x2≤…≤x12. The Lorenz asymmetry coefficient LS can be calculated using the following formula:

[0051]

[0052]

[0053]

[0054] Where: n is the number of samples; m is the number of samples less than the mean; Lm is the sum of the sizes of the m samples; Ln is the sum of the sizes of the n samples; LS < 1 indicates that the reason for the uneven distribution of precipitation within the year is that the months with less precipitation account for a larger proportion of the annual precipitation; LS > 1 indicates that the months with more precipitation cause the uneven distribution of precipitation within the year; the GiNi index and Lorenz asymmetry coefficient of precipitation at each grid point can be calculated, and the trend coefficient of the GiNi index and Lorenz asymmetry coefficient of each grid point can be calculated respectively;

[0055] The trend coefficient of the GINI index is >0, indicating that the contribution of months with less precipitation to the uneven distribution of precipitation within the year is large, that is, the proportion of months with less precipitation in the total precipitation gradually increases, and the annual distribution of precipitation tends to be uniform; conversely, it indicates that the annual distribution of precipitation tends to be uneven.

[0056] Advantages of the present invention:

[0057] The present invention uses statistical downscaling methods, data fusion methods, and precipitation data sub-gridding to obtain 0.5km high-precision monthly-scale precipitation spatial information, which can greatly improve the accuracy of spatial precipitation sequence data.

[0058] Based on high-precision precipitation grid time series data, the present invention calculates the annual trend and volatility of precipitation data at each grid point, as well as the GINI index and other information within the year. It can obtain more accurate spatial distribution information of the temporal variation characteristics of precipitation elements.

[0059] The present invention can obtain continuous high-resolution spatial precipitation data by integrating CMORPH and TRMM precipitation.

[0060] The present invention can obtain the spatial distribution of precipitation time variation characteristics. Compared with the time series analysis of precipitation station data, the spatial distribution of precipitation time variation characteristic values ​​obtained by this patent has more temporal continuity. Compared with direct acquisition of remote sensing products, the spatial distribution of precipitation time variation characteristic values ​​obtained by this patent has more advantages in accuracy and resolution.

[0061] The continuous high-resolution precipitation data obtained by the present invention are of great significance for climate analysis, regional water cycle, water balance, water resources evaluation, drought and flood disaster management, and ecological environment governance.

[0062] The spatial distribution of continuous high-resolution precipitation temporal variation characteristics obtained by the present invention can provide reliable data support for refined water resources management and ecological environmental protection.

[0063] Compared with the traditional method of obtaining the spatial distribution of precipitation time variation characteristics, the present invention can obtain more effective and precise precipitation spatial distribution conditions.

[0064] In addition to the above-described purposes, features and advantages, the present invention has other purposes, features and advantages. The present invention will be further described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The drawings constituting a part of this application are used to provide a further understanding of the present invention. The illustrative embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0066] Figure 1 It is a flow chart of the refined spatial analysis method of precipitation temporal variation characteristics of the present invention. DETAILED DESCRIPTION

[0067] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0068] A method for spatial analysis of precipitation temporal variation characteristics based on multi-source data includes the following steps:

[0069] 1. Preprocess the satellite precipitation data and ground station observation precipitation data, extract the Trmm 3B42 and CMORPH CRT data within the study area, and unify the data format. Resample the terrain data (elevation, slope, slope aspect), vegetation index NDVI and other data, and extract 1km data, data with the same resolution as Trmm and CMORPH, and 0.5km resolution DEM data.

[0070] 2. Use quadratic multiple regression equation for statistical downscaling. The quadratic multiple regression coefficients of the monthly scale Trmm and CMORPH data and longitude, latitude, elevation, slope, aspect, and vegetation within the research time range are statistically analyzed, and the 1km resolution Trmm preliminary downscaling data and CMORPH preliminary precipitation downscaling data are obtained according to the regression coefficients.

[0071] First, the regression coefficients are solved for Trmm and CMORPH data on a monthly basis using the following regression equation:

[0072]

[0073] Where: Lat, Lon, Ele, P, X, NDVI are the latitude, longitude, elevation, slope, aspect, NDVI index, etc. of the grid point, and g is the regression constant. The least squares method is used to solve the regression coefficients a1, a2…f1, f2 and the regression constant g in the above formula. Substituting the 1km scale latitude, longitude, elevation, slope, aspect, and NDVI index into the above formula, the preliminary precipitation downscaling results of Trmm and CMORPH data can be obtained respectively.

[0074] 3. Use the station data to correct the errors of two downscaled data (Geographical Different Analysis, GDA) respectively, obtain the error analysis of the preliminary downscaled data and the measured data of the station according to the station location, and use the trend surface method to correct the errors. Obtain the Trmm downscaling error correction data and the CMORPH precipitation downscaling error correction data.

[0075] The following formula is used to calculate the deviation between the initial downscaled data and the measured value of the station for each station corresponding grid:

[0076]

[0077] Where: Ps is the observation data of the precipitation station, Pd0 is the grid value corresponding to the preliminary downscaling, and ΔP is the deviation between the preliminary downscaling data and the actual value of the station.

[0078] In order to avoid the bull's eye error caused by conventional interpolation, this patent uses the trend surface method to interpolate the deviation:

[0079]

[0080] Where a1, b1, c1, g are the residual regression coefficients and regression constants, obtained by the least squares algorithm of the station residuals, and the other parameters have the same meanings as above. The following formula is used to correct the preliminary downscaled data. The downscaled data of Trmm and CMORPH data are obtained respectively.

[0081]

[0082] 4. Fusion of two downscaled data. Take the average value of two precipitation products on each grid, so as not to lose the spatial information of each precipitation downscaled data.

[0083] The following formula is used to fuse the two types of precipitation statistical downscaling data:

[0084]

[0085] Where: Pd is the 1 km precipitation downscaling fusion data, PTrmm and PCMORPH are the downscaling results of Trmm and CMORPH satellite remote sensing precipitation data respectively.

[0086] 5. Sub-gridding of precipitation. Based on the 1km grid precipitation data, sub-grid the precipitation grid data according to the precipitation gradient information. Obtain 0.5km grid precipitation data.

[0087] The precipitation elevation gradient value is calculated using the following formula:

[0088]

[0089] Where a1 is the precipitation elevation gradient, P is the 1 km precipitation downscaling value, and Ele is the corresponding grid elevation.

[0090] On each grid of 1km data, the following formula is used to calculate precipitation downscaling:

[0091]

[0092] Where: Pd is the downscaled precipitation data with a resolution of 0.5 km, P0 is the downscaled precipitation data with a resolution of 1 km, Ele0 is the elevation of the central point of the 1 km downscaled data, and Eled is the average elevation of each grid in the 1 km downscaled data grid.

[0093] Repeating steps 2-5 above can obtain monthly precipitation raster data with a spatial resolution of 0.5 km from 1998 to 2017.

[0094] 6. Analysis of precipitation grid trend, volatility and annual distribution changes.

[0095] (1) Precipitation trend and volatility

[0096] The annual precipitation sequence Y is extracted on each grid, and a univariate linear regression equation is established with time ti as the independent variable and annual precipitation yi as the dependent variable.

[0097]

[0098] Where: a is the regression constant; b is the regression coefficient, also known as the trend coefficient. The positive or negative sign of b reflects the linear change trend of the variable rising or falling. b<0 means that the variable is decreasing with time during the calculation period, and b>0 means that the variable is increasing with time. Use the following formula to calculate the fluctuation characteristic value:

[0099]

[0100]

[0101] Where: Zi is the residual absolute value sequence, d is the residual regression coefficient, also known as the variable fluctuation coefficient. d < 0 means that it is decreasing with time during the calculation period, that is, the fluctuation of the variable is weakening; d > 0 means that it is increasing with time, that is, the fluctuation of the variable is increasing. In summary, the regional 0.5 km resolution fluctuation characteristic value and trend value can be obtained.

[0102] (2) Annual distribution of precipitation

[0103] This patent uses the Gini coefficient and the Lorenz asymmetry coefficient to express the annual distribution index, and uses the area method to calculate. The formula is as follows:

[0104]

[0105] Where G is the Gini index, yi is the cumulative percentage of monthly precipitation in the year, and xi is the cumulative percentage value over time.

[0106] Calculation of the Lorenz asymmetry coefficient: For a certain year, the monthly precipitation data is sorted from small to large, that is, x1≤x2≤…≤x12. The Lorenz asymmetry coefficient LS can be calculated using the following formula:

[0107]

[0108]

[0109]

[0110] Where: n is the number of samples; m is the number of samples less than the mean; Lm is the sum of the sizes of the m samples; Ln is the sum of the sizes of the n samples. LS < 1 indicates that the reason for the uneven distribution of precipitation within the year is that the months with less precipitation account for a larger proportion of the annual precipitation; LS > 1 indicates that the months with more precipitation cause the uneven distribution of precipitation within the year. In summary, the GiNi index and Lorenz asymmetry coefficient of precipitation within the year of each grid point can be calculated, and the trend coefficient of the GiNi index and Lorenz asymmetry coefficient of each grid point can be calculated respectively.

[0111] The GINI index trend coefficient is > 0, indicating that the contribution of months with less precipitation to the uneven distribution of precipitation within the year is large, that is, the proportion of months with less precipitation in the total precipitation gradually increases, and the annual distribution of precipitation tends to be uniform. On the contrary, it means that the annual distribution of precipitation tends to be uneven.

[0112] Embodiment 1:

[0113] This case uses 73 precipitation station data, SPOT NDVI data, and SRTM3 DEM data in the Qilian Mountains to downscale the precipitation data for CMORPH CRT and Trmm 3B43 data from January 1998 to December 2017, and obtains the monthly precipitation data with a resolution of 0.5 km in the Qilian Mountains from January 1998 to December 2017. The determination coefficient, bias coefficient, and root mean square error are used to verify the effectiveness of the downscaling method. It is found that the downscaling results are significantly improved compared with the original data. The results are shown in the following table:

[0114]

[0115] Downscaling was used to analyze the trend, volatility and intra-annual distribution changes in the Qilian Mountains, and the spatial distribution of the temporal variability characteristics of precipitation in the Qilian Mountains from 1998 to 2017 was obtained.

[0116] The present invention downscales the precipitation of two precipitation satellite remote sensing products, CMORPH and TRMM, respectively, and fuses the two precipitation downscaling results, effectively reducing the temporal variation and spatial deviation caused by different spatial expressions of different satellite remote sensing products. Based on the fused data, this patent performs precipitation sub-gridding calculation on precipitation data to obtain a higher-precision precipitation spatial distribution sequence.

[0117] The present invention provides a precipitation time variation characteristic analysis method based on spatial refinement of multi-source precipitation data analysis, selects precipitation satellite remote sensing products CMORPH and TRMM data, respectively constructs satellite remote sensing products and geographical location factors, terrain factors (slope, slope aspect), elevation factors, and underlying surface vegetation index factors (SPOT NDVI data) at a monthly scale to establish a multivariate regression model, and obtains 1km scale precipitation data. Using station data and a correction method of geographic difference analysis (GDA), precipitation downscaling data of two satellite remote sensing products CMORPH and TRMM are obtained respectively. A data fusion method is used to obtain precipitation spatial raster data of two fused satellite remote sensing products with a resolution of 1km. A relationship is established between 1km precipitation spatial remote sensing data and elevation, and a sub-gridding method is used within the grid to refine the precipitation spatial raster data again, and precipitation spatial raster data with a resolution of 0.5km is obtained. And the station data are used to perform correlation analysis and error testing. Based on the spatially refined time series data of precipitation data, the trend value and fluctuation characteristic value of the annual precipitation value in the selected time period are extracted on each precipitation spatial grid, as well as the trend of the GINI index and the Lorenz asymmetry coefficient of precipitation within the year. The spatial distribution of accurate regional precipitation temporal variation information can be obtained through the present invention, which is not only crucial for studying climate trends and variability, but also crucial for regional water resources planning and management, and ecological environmental protection.

[0118] The present invention has the advantages of fusion downscaling of multiple satellite precipitation products and spatial analysis of precipitation temporal variation characteristics.

[0119] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for fine-grained spatial analysis of precipitation temporal variation characteristics, characterized in that: The following steps are involved: The satellite precipitation data and ground station observation precipitation data were preprocessed, and Trmm3B42 and CMORPHCRT data within the study area were extracted and the data format was unified; the terrain data and vegetation index NDVI data were resampled, and 1km data, data with the same resolution as Trmm and CMORPH, and 0.5km resolution DEM data were extracted; The quadratic multiple regression equation was used for statistical downscaling. The quadratic multiple regression coefficients were calculated for the two data of monthly scale Trmm and CMORPH with longitude, latitude, elevation, slope, aspect and vegetation within the research time range. The preliminary downscaling data of Trmm with a resolution of 1 km and the preliminary downscaling data of CMORPH precipitation were obtained according to the regression coefficients. The station data pairs were used to correct the errors of the two downscaled data respectively. The errors of the preliminary downscaled data and the measured data of the station were analyzed according to the station location, and the errors were corrected using the trend surface method. The Trmm downscaling error correction data and the CMORPH precipitation downscaling error correction data were obtained. Two types of downscaled data fusion; precipitation subgridding; Analysis of precipitation grid trend, volatility and annual distribution changes; The statistical downscaling using the quadratic multiple regression equation includes: First, the regression coefficients are solved for Trmm and CMORPH data on a monthly basis using the following regression equation: P ij =a1*Lat+a2*Lat 2 +b1*Lon+b2*Lon 2 +c1*Ele+c2*Ele 2 +… d1*P+d2*P 2 +e1*X+e2*X 2 +f1*NDVI+f2*NDVI 2 +g Where: Lat, Lon, Ele, P, X, NDVI are the grid point latitude, longitude, elevation, slope, aspect, and NDVI index values, respectively, and g is the regression constant. The least squares method is used to solve the above equations for the regression coefficients a1, a2…f1, f2 and the regression constant g. Substituting the 1km scale latitude, longitude, elevation, slope, aspect, and NDVI index into the above equations, the preliminary precipitation downscaling results of Trmm and CMORPH data can be obtained respectively. The precipitation sub-gridding includes: Based on the 1km grid precipitation data, the precipitation grid point data is sub-gridded according to the precipitation gradient information to obtain the 0.5km grid precipitation data; The precipitation elevation gradient value is calculated using the following formula: P=a1*Ele+g Where: a1 is the precipitation elevation gradient, P is the 1km precipitation downscaling value, and Ele is the corresponding grid elevation; On each grid of 1km data, the following formula is used to calculate precipitation downscaling: Pd=P0+a*(He0-He d ) Where: Pd is the downscaled precipitation data with a resolution of 0.5 km, P0 is the downscaled precipitation data with a resolution of 1 km, Ele0 is the elevation of the central point of the 1 km downscaled data, and Eled is the average elevation of each grid in the 1 km downscaled data grid.

2. The method for spatial analysis of the time variation characteristics of precipitation according to claim 1 is characterized in that: The use of the station data pair to respectively correct the errors of the two downscaled data comprises: The following formula is used to calculate the deviation between the initial downscaled data and the measured value of the station for each station corresponding grid: ΔP=P s -P d0 Where: Ps is the observation data of the precipitation station, Pd0 is the grid value corresponding to the preliminary downscaling, and ΔP is the deviation between the preliminary downscaling data and the actual value of the station; In order to avoid the bull's eye error caused by conventional interpolation, this patent uses the trend surface method to interpolate the deviation: ΔP ij =a1*Lat+b1*Lon+c1*Ele+g Where a1, b1, c1, g are the residual regression coefficients and regression constants, which are obtained by the station residual least squares algorithm; the following formula is used to correct the preliminary downscaled data; the downscaled data of Trmm and CMORPH data are obtained respectively; P d1 =P d0 +ΔP ij 。 3. The method for spatial analysis of the time variation characteristics of precipitation according to claim 1 is characterized in that: The two types of downscaling data fusion include: The following formula is used to fuse the two types of precipitation statistical downscaling data: P d =(P Trmm +P CMORPH ) / 2 Where: Pd is the 1 km precipitation downscaling fusion data, PTrmm and PCMORPH are the downscaling results of Trmm and CMORPH satellite remote sensing precipitation data respectively.

4. The method for spatial analysis of precipitation temporal variation characteristics according to claim 1 is characterized in that: The precipitation grid trend and volatility include: Extract the annual precipitation sequence Y on each grid, take time ti as the independent variable and annual precipitation yi as the dependent variable, and establish a univariate linear regression equation; the i =a+bt i Where: a is the regression constant; b is the regression coefficient, also known as the trend coefficient. The positive and negative signs of b reflect the linear change trend of the variable rising or falling; b < 0 means that the variable is decreasing with time during the calculation period, and b > 0 means that the variable is increasing with time. The following formula is used to calculate the fluctuation characteristic value: <h2 style=";text-align:left;direction:ltr">Z<h2 style=";text-align:left;direction:ltr"> i <h2 style=";text-align:left;direction:ltr"> =y<h2 style=";text-align:left;direction:ltr"> i <h2 style=";text-align:left;direction:ltr"> -a+bt<h2 style=";text-align:left;direction:ltr"> i Z i =c+dt i Where Zi is the residual absolute value sequence, d is the residual regression coefficient, also known as the variable fluctuation coefficient; d < 0 means that it shows a downward trend with time increasing during the calculation period, that is, the fluctuation of the variable shows a weakening trend; d > 0 means that it shows an upward trend with time increasing, that is, the fluctuation of the variable shows an increasing trend; the regional 0.5 km resolution fluctuation characteristic value and trend value can be obtained.

5. The method for spatial analysis of the time variation characteristics of precipitation according to claim 1 is characterized in that: The annual distribution of precipitation includes: The Gini coefficient and Lorenz asymmetry coefficient are used to express the annual distribution indicators, and the area method is used for calculation. The formula is as follows: In the formula, G is the Gini index, yi is the cumulative percentage of monthly precipitation in the year, and xi is the cumulative percentage value of time; Calculation of the Lorenz asymmetry coefficient: For a certain year, the monthly precipitation data is sorted from small to large, that is, x1≤x2≤…≤x12. The Lorenz asymmetry coefficient LS can be calculated using the following formula: Where: n is the number of samples; m is the number of samples less than the mean; Lm is the sum of the sizes of the m samples; Ln is the sum of the sizes of the n samples; LS < 1 indicates that the reason for the uneven distribution of precipitation within the year is that the months with less precipitation account for a larger proportion of the annual precipitation; LS > 1 indicates that the months with more precipitation cause the uneven distribution of precipitation within the year; the GiNi index and Lorenz asymmetry coefficient of precipitation within the year at each grid point can be calculated, and the trend coefficient of the GiNi index and Lorenz asymmetry coefficient of each grid point can be calculated respectively; The trend coefficient of the GINI index is >0, indicating that the contribution of months with less precipitation to the uneven distribution of precipitation within the year is large, that is, the proportion of months with less precipitation in the total precipitation gradually increases, and the annual distribution of precipitation tends to be uniform; conversely, it indicates that the annual distribution of precipitation tends to be uneven.

Images