Geoscience Reference
In-Depth Information
for small scale catchments with several hundreds square km, accumulations
of hydrologic data are usually quite insucient. Generally, the shape of a
flood hydrograph at small scale basins is sensitive to space and especially
time distributions of rainfall patterns. Therefore, flood runoff predictions at
small scale basins request more detailed information of rainfall distribution
patterns than at large scale basins with more than several thousand square
km. In addition, the flood data with the magnitude of a design flood level
or above the level does not exist in most situations. In this sense, rainfall-
runoff models are not validated for estimating a flood with a magnitude of
design level. To examine the performance of a rainfall-runoff model for the
historical 2004 Fukui flood under available hydrologic data is a good test to
understand the behaviors and predictability of a rainfall-runoff model and
this gives knowledge to improve the prediction of a runoff model.
In this paper, we use a physically based distributed rainfall-runoff model
based on topographic representations by grid based DEMs and kinematic
flow routing
1
-
3
and apply the rainfall-runoff model to the upper part of the
Asuwa River basin (351 km
2
); examine the predictability of the model for
the 2004 Fukui flood and the source of flood prediction uncertainty; then
discuss a direction to reduce the uncertainty and enhance the reliability of
flood discharge prediction.
2. Distributed Rainfall-Runoff Model
Figure 1 shows the topographic model of the Asuwa River basin (351 km
2
)
using a DEM processed with the algorithms by Shiiba
et al
.
1
Aslopeseg-
ment is represented by a rectangle formed by the adjacent two grid points
determined to have the steepest gradient. The catchment topography is
represented as connections of these slope segments. The spatial resolution
of a DEM used here is 50 m.
According to the flow directions shown in Fig. 1, the slope flow is routed
one dimensionally; the slope discharge is given to the river flow routing
model; then the river flow is routed to the outlet. In each slope segment,
the slope is assumed to be covered by a permeable soil layer composed of
a capillary layer and a noncapillary layer. In these layers, slow flow and
quick flow are modeled as unsaturated Darcy flow with variable hydraulic
conductivity and saturated Darcy flow. If the depth of water exceeds the
soil water capacity, overland flow happens. These processes are represented
with a kinematic wave model using the continuity equation and a function
