Due to limitation of short time series of hydrological observations, it is hard to know the centennial timescale chance of extreme floods. The present study simulated catchment runoff and discharge of the 1889 flood year in Lake Taihu, on basis of historical flood records and by means of the catchment hydrology model. Three experiments were performed in the simulations:firstly control-run simulations for 1988-2002 to calibrate and validate the modeling, secondly simulations of daily processes for the 1889 flood year driven by observed climate records during the end of 19th century; thirdly simulations of 15 year-Monte Carlo Bootstrap resampling for the special flood year to decrease the 1-year sample uncertainty. Results showed that the 1889 flood peaks occurred during the end of 1889 June and the greater than 99th percentile of monthly runoff reached 8.6 mm/d (the 95%CI errors between -2.94 and 3.26 mm/d). The discharge peak to Lake Taihu also occurred in the same month, and the greater than 99th percentile of daily discharge reached 1286.9 m³/s (the 95%CI errors between-128.3 and 165.7 m³/s). Following the Log-Normal distribution that has been tested in the control run simulations, return period of the 1889 flood year is 149 years. After simulations for the error intervals by the Bootstrap confidence interval method, the return period during 70-175 years is much confident. The study will provide scientific basis of hydrological dynamics for extending the time series of the 20th century hydrology and knowing the centennial timescale chance of extreme floods.