Abstract: The quantitative precipitation estimation accuracy of dual-polarization radar is affected by many factors. In order to better use dual-polarization radar to estimate precipitation and further improve the rainfall estimate accuracy, error analysis and modeling of radar precipitation estimation are needed. Based on observations of the C-band dual-polarization radar of Nanjing University of Information Science and Technology from 2015 to 2016, raindrop spectrometer observation data and rain gauge data in Nanjing, the error distribution of radar estimation precipitation is statistically analyzed, the gauge representativeness error variance is separated, and the error quantitative model based on random error and systematic error is established. Firstly, the data of dual polarization radar is preprocessed, and the radar rainfall formula is fitted by the data of raindrop spectrograph in the observation base of Nanjing University of Information Science and Technology. By comparing the four radar rainfall formula R (Z_H), R (Z_H, Z_DR), R (K_DP), R (K_DP, Z_DR) with the rain gauge, the estimation performance of each formula under different precipitation thresholds is analyzed. Then, the spatial correlation function of rain gauge data is estimated, the gauge representative error caused by spatial mismatch between radar and rain gauge is calculated, and the proportion of measurement error and parameter error in radar precipitation estimation error is analyzed. Based on the attribute and distribution rule of radar precipitation estimation errors, they are divided into random errors and systematic errors, and a quantitative model is established. Finally, based on the performance and error analysis of four radar rainfall formula, an optimized combination of dual polarization radar precipitation estimation algorithm is proposed. The results show that R (Z_H) and R (Z_H, Z_DR) have better performance for light precipitation estimation. When the rainfall threshold is greater than 2.5 mm/h, the advantage of K_DP becomes obvious. The gauge representative error caused by spatial mismatch cannot be ignored. Therefore, the point to area error should be eliminated when the radar resolution unit is large. The radar error is modeled according to the systematic error and random error, and it is found that the systematic error of radar near surface precipitation is proportional to the rainfall intensity in the form of linear function, and the double exponential model better represents the random error distribution. Through the performance analysis of radar rain measurement formula and dual polarization signal analysis, it is found that the optimized combination is better than the single rain algorithm in accuracy and stability.