Modeling of an ungauged watershed with the associated uncertainties of the input data is
presented. The MapWindow versions of the Soil and Water Assessment Tool (SWAT)
have been applied to a complex and ungauged watershed of about 248,000ha in an area
close to the Niger River, Nigeria. The Kwara State Government of Nigeria in
collaboration with the newly relocated former Zimbabwean farmers now occupied the
largest portion of this watershed for an “Agricultural Estate Initiative ”. The government
and these farmers are decision makers who need to take appropriate actions despite little
or no data availability. SWAT being a physically based model, allow the use of
Geographical Information System (GIS) inputs like the Digital Elevation Model(DEM),
landuse and soil maps. The MapWindow-SWAT(MSWAT) involves processes like the
Watershed Delineation, Hydrological Response Units (HRUs) Process and the SWAT
run. The watershed was delineated into 11 subbasins and 28 HRUs. There were 8 landuse
classes and 5 soil types. The model was able to simulate and forecast for several
years(1990-2016). The results look 'reasonable' since there is no observed data from the
watershed for statistical validation. However, using the Water Balance equation as a
validation criteria, the correlation coefficient between the simulated rainfall and runoff
was 0.84 for the subbasin 11 (outlet). Thereafter, the uncertainties in the continuous
numerical input (i.e. rainfall) was examined using the Data Uncertainty Engine (DUE).
One parameter exponential probability model was used for the daily rainfall amount
based on the histogram. 700 realizations were generated from this uncertain input.
Randomly selected numbers of the realizations were prepared and used as inputs into the
MWSWAT model. It was surprising that there were no changes in the results when
compared to the initial 'real' value (outflows from outlet) although other parameters of the
model were kept constant.