Abstract:
A number of semi-distributed hydrological models have been developed to model the
hydrology of watersheds. However, it becomes difficult to potential model users to identify
the most economic and efficient hydrological models for specific watershed. This study aimed
to compare stream flow prediction efficiency of HEC-HMS and SWAT models and their
associated uncertainty in Bilate and Gidabo watersheds. The model approach started with
input data preparation, sensitivity analysis, calibration, validation and uncertainty
assessment to test the capability in predicting stream flow at the outlet of Bilate and Gidabo
watersheds. The parameter uncertainty is taken in to account and MCMC (HEC-HMS) and
SUF-2 (SWAT) analysis are employed to analyze the parameter uncertainties. During
sensitivity analysis, the results show that constant loss rate (CR) is most sensitive followed by
lag time (LT) in case of HEC-HMS for both watersheds. Whereas, the most sensitive
parameter detected by SWAT is ALPHA_BF in Bilate watershed and CN_2 in Gidabo
watershed. The models were calibrated and validated from 2001 to 2010 and 2011 to 2015 of
Bilate and from 1998 to 2003 and 2004 to 2006 Gidabo watershed respectively. The model
efficiency on daily time step during calibration HEC-HMS (NSE=0.55 and0.65) and SWAT
(NSE=0.53 and 0.58) were obtained for Bilate and Gidabo watersheds respectively.
Similarly for validation period HEC-HMS (NSE=0.55 and 0.63) and SWAT (NSE=0.52 and
0.56) were also obtained for Bilate and Gidabo watersheds respectively. Both models are
able to simulate the hydrology of both watersheds in acceptable way. Statistical indicators
reveal that in both watersheds, the models performances in monthly time step are higher than
in daily time step. Despite their similar modeling abilities, a comparison analysis revealed
that the HEC-HMS model was performed slightly better at simulating streamflow in both
watersheds based on common performance measures. Based on both plot of uncertainty and
results obtained for P factor and R factor indicates that the parameter uncertainty estimated
for both watersheds were found to be good. Because HEC-HMS model (with DCL method),
with less input data is easy to calibrate and therefore gives better statistical result with goodcertainty values compared to SWAT model.
