Abstract. Flash floods are experienced almost annually in the ungauged Mbire District
of the Middle Zambezi Basin. Studies related to hydrological modelling
(rainfall-runoff) and flood forecasting require major inputs such as
precipitation which, due to shortage of observed data, are increasingly using
indirect methods for estimating precipitation. This study therefore evaluated
performance of CMORPH and TRMM satellite rainfall estimates (SREs) for
30 min, 1 h, 3 h and daily intensities through hydrologic and flash flood
modelling in the Lower Middle Zambezi Basin for the period 2013–2016. On a
daily timestep, uncorrected CMORPH and TRMM show Probability of Detection
(POD) of 61 and 59 %, respectively, when compared to rain gauge
observations. The best performance using Correlation Coefficient (CC) was 70
and 60 % on daily timesteps for CMORPH and TRMM, respectively. The best
RMSE for CMORPH was 0.81 % for 30 min timestep and for TRMM was 2,
11 % on 3 h timestep. For the year 2014 to 2015, the HEC-HMS
(Hydrological Engineering Centre-Hydrological Modelling System) daily model
calibration Nash Sutcliffe efficiency (NSE) for Musengezi sub catchment was
59 % whilst for Angwa it was 55 %. Angwa sub-catchment daily NSE
results for the period 2015–2016 was 61 %. HEC-RAS flash flood modeling
at 100, 50 and 25 year return periods for Angwa sub catchment, inundated 811
and 867 ha for TRMM rainfall simulated discharge at 3 h and daily
timesteps, respectively. For CMORPH generated rainfall, the inundation was
818, 876, 890 and 891 ha at daily, 3 h, 1 h and 30 min timesteps. The
30 min time step for CMORPH effectively captures flash floods with the
measure of agreement between simulated flood extent and ground control points
of 69 %. For TRMM, the 3 h timestep effectively captures flash floods
with coefficient of 67 %. The study therefore concludes that satellite
products are most effective in capturing localized hydrological processes
such as flash floods for sub-daily rainfall, because of improved spatial and
temporal resolution.