To evaluate the accuracy of extreme runoff simulation by combining the satellite remote sensing and the reanalysis data, the precipitation and temperature data of TRMM, PERSIANN-CDR, and CFSR, and measured data were selected to obtain eight combination schemes of precipitation and temperature. Hydrological models were utilized to evaluate the accuracy of different combinations in each extreme flow sequence. The results show that: among the combinations of remote sensing data without measured data, TRMM precipitation & CFSR air temperature and PERSIANN-CDR precipitation & CFSR air temperature are the best combinations for simulating the extreme high and low runoff, respectively; in addition to the reference combination (actually measured precipitation & measured temperature), measured precipitation & CFSR temperature and TRMM precipitation & measured temperature are the combinations with the highest simulation accuracy of the high and low runoff simulation, respectively; among the extreme flow series, CFSR precipitation & CFSR air temperature and CFSR precipitation & measured temperature seriously overestimate the extreme flow with the maximum overestimation reaching 651%, and the simulation error increases with the decrease of discharge. The other schemes underestimate the runoff with the maximum underestimation reaching 67.48%, and the error decreases with the decrease of runoff; The characteristics of extreme precipitation vary in different extreme value series. Compared with the other two precipitation products, TRMM can better reflect the frequency of extreme precipitation and the duration of no rain in the high flow series, while the precipitation of PERSIANN-CDR in the low flow series can best reflect the intensity, total amount, and characteristics of consecutive wet days.