Summary:
This dataset is 0The optimal distribution and parameter set of the annual precipitation frequency model with 25° resolution (including Taiwan), a total of 15,355 data grids. There are six alternative distributions in the dataset: generalized extreme distribution (GEV), three-parameter gamma distribution (P3), generalized Pareto distribution (GP), lognormal distribution (LN), log-logistic distribution (ll), and Weber distribution (Weibull).
Key words: China Annual precipitation Frequency analysis
Data description
This dataset contains the optimal frequency model distribution and parameters of annual precipitation in China from 1948 to 2016 (including Taiwan), with a spatial resolution of 025°, a total of 15,355 precipitation grids. The research basis of this dataset is the "China 025° resolution dataset of annual precipitation". In this dataset, the genetic algorithm (GA) and the differential evolution algorithm (DE) are used to solve the six alternative distributions for 4 times to determine the optimal distribution. Main results: (1) The actual annual precipitation and frequency distribution model fit well, and the maximum root mean square error was 90 in all precipitation grids37（mm）。(2) In all precipitation grids, 497% of the precipitation grid belongs to the GEV distribution, 297% of the precipitation grid belongs to the P3 distribution, 13The 3% precipitation grid belongs to the LL distribution, 73% of the precipitation grids belong to the Weibull distribution, and a very small number of grids (01%) belong to the GP distribution and LN distribution.
There are three main parts to the dataset generation process:
Firstly, the alternative distribution was selected, and six models were selected as alternative distributions, including the generalized extreme distribution (GEV), the three-parameter gamma distribution (P3), the generalized Pareto distribution (GP), the lognormal distribution (LN), the log-logistic distribution (LL), and the Weber distribution (Weibull), which are common and applicable in the annual precipitation frequency in China. GEV, P3, and GP are three-parameter distributions, and LN, ll, and Weibull are two-parameter distributions.
Secondly, six frequency distribution models were used to fit 15,355 precipitation grids one by one according to the goal of minimum root mean square error, and the minimum root mean square error and optimal parameter value of each frequency distribution model were recorded. In the first step, the GA algorithm is used to solve the problem once, and the range of six distribution parameters is preliminarily obtained. Taking the preliminary range as the upper and lower limits of the parameter values, GA and DE were used to solve the problem twice, and the minimum root mean square error and optimal parameter values of the six frequency distribution models were obtained from the optimal values of the four times to the precipitation grid.
Finally, the distribution with the smallest root mean square error in the six frequency distribution models was selected as the optimal distribution of the grid, and the parameter values of the distribution were recorded.
The dataset is presented in CSV format, in the CSV file: ObjectID is the number of the data grid, LAT represents the latitude in the upper right corner of the cell, LON represents the longitude in the upper right corner of the cell, Index represents the parameter values of the optimal distribution, a, b, and c represent the optimal parameters, and Min RMSE represents the root mean square error value of the optimal distribution model and the actual precipitation. The optimal distribution was evaluated based on the root mean square error (RMSE) of the empirical and theoretical frequencies of precipitation.
The meaning of the six parameters is shown below**: