Results - Response of Flood Events to Land Use and Climate Change - page 128

Upstream Upstream

(a)

(b)

Main chennel

Main chennel

Fig. 6.2 a and b the agreement between modeled and observed drainage network; blue line
represents the observed drainage networks; yellow line represents drainage network based on
DEM created by interpolation methods (as a sample: IDW); red line represents the drainage
network based original DEM (5 m resolution)

of the drainage network in the main channel (upstream) is missing in IDW DEM.
Then again, this part of the drainage network is obviously shown in the original
DEM and was more consistent with the observed drainage network. As an overall
result, it is concluded that the interpolation techniques were not successful in
creating a perfect DEM even more accurate than the original DEM. Consequently,
the first assumption is rejected.
Therefore, in the hydrodynamic simulation the original DEM was applied. The
quality of DEM was also tested hydrologically using the drainage enforcement
algorithm to recognize that each spurious sink is surrounded by a drainage divide
containing at least one saddle point. If the sink is associated with an elevation data
point, then the lowest such saddle, provided it is not also associated with an
elevation data point, identifies the region of the grid, which is modified in order to
remove the spurious sink. Figure 6.2 represents the comparison between modeled
and observed drainage networks, for IDW compared with the original DEM.
In this part, the second assumption was that the original DEM is included in
observed elevation point data with 100 % accuracy. Therefore, it was used as a
correct source to test the different interpolation methods. The idea behind this
assumption was to find the best interpolation technique which represents the best
agreement with the observed elevation of the data point set. Under this premise,
the interpolation process which was applied to the first assumption was repeated.
Finally, for testing the best method, the elevations for the left-out 30 % data point
set, which were created as unknown data points by using interpolation techniques,
were compared with the same observed data to see how the accuracy of the DEM
decreases as the spatial coverage of high data is reduced. Figure 6.3 represents the
process of creating DEM based on different interpolation techniques.
The production of large elevation error estimates, evenly distributed over the
DEM, makes it possible to explore some general questions, such as how to identify
the statistical characteristics of DEM error, and how much this differs among
interpolation methods and the relationship between RMSE of elevation and RMSE