ACCURACY OF DEM-BASED TOPOGRAPHIC DATA IN FLOOD INUNDATION MODELLING: A CASE OF WAMI RIVER, TANZANIA

Abstract

The objective of this paper is to evaluate and further comment on the accuracy of the DEM-
based topographic data as input in flood modelling. Recent studies have indicated that
hydraulic parameters of the modified cross sections based on high resolution DEM are
geometrically and hydraulically similar to the measured ones. In the current study hydraulic
modelling was intended to guide a development project whereby a floodplain protection
embankment is contemplated. Therefore, accuracy in terms of positioning/alignment of natural
features such as river banks and heights/elevations was imperative to attain. In order to achieve
these, accuracy of various topographic data sources were qualitatively and/or quantitatively
analysed before and after hydraulic modelling. As the study river reach is ungauged, the
hydraulic model was verified and validated with field observations, readily available
information and professional judgement. Based on error analysis, it was found that
uncalibrated 90-m resolution STRM DEM-based topographic data are biased, i.e., 80% of the
error mass is negative. The error increases with elevation. The final input to the hydraulic
model was a composite/hybrid DEM derived after merging geometric data from four
representative field-measured cross sections, a 2-m interval contour map covering part of the
river valley, and calibrated 90-m resolution DEM. Although, calibration improves the data, it
does not remove completely the inherent bias. Further to DEM calibration, it was therefore
recommended to include an elevation uncertainty value of -0.5 m, as derived from topographic
analysis, into simulated water surface profiles elevations. The hydraulic model satisfactorily
simulated flood inundation extent after considering the elevation uncertainty. In order to
improve the findings further studies should use adequate measured cross sections and DEM of
higher resolutions.