Prediction of runoff hydrographs in ungauged basins - Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales - page 246

Figure 10.19. Conceptualisation of
dominant hydrological processes
based on a landscape classification
into wetlands, hillslopes and
plateaus. From Savenije ( 2010 ).

hillslopes, subsurface drainage through preferential path-
ways and storage excess subsurface flow (SSF) are domin-
ant. After subtraction of interception, the rainfall is
partitioned by a beta function into fast subsurface runoff
and soil moisture storage. Part of the fast runoff may be
directed to the groundwater reservoir through preferential
recharge or percolation. This groundwater reservoir is con-
nected to the plateau, where it receives recharge by deep
percolation (DP), which is the balance between precipita-
tion and evaporation. On the plateaus, the unsaturated
reservoir that constrains evaporation is key to determining
the percolation. During extreme rainfall events, the plat-
eaus may also trigger infiltration excess or Hortonian over-
land flow (HOF) towards the drainage system, represented
by a threshold in the model structure.
In essence, this type of model reflects the natural organisa-
tion and co-evolution of the landscape. The same can be said
about the role of geology, and models tailored to particular
geological formations could also be developed; for example
in the UK, catchments vary between chalk and clay domin-
ated ones, each requiring very different model structures
(Lee et al., 2005 ). The hydrological landscape units in
Figure 10.19 are another example of how the similarity
between landscape units could be used for deciding on
the structure of a conceptual rainfall

catchments, some or all of the model parameters are
usually calibrated to observed runoff in order to reduce
bias in the runoff hydrograph predictions. Calibration
can correct for biases in the inputs such as precipitation.
Calibration can also correct for errors due to empirical
elements in the model equations and processes that are
not accounted for such as macropore flow in a particu-
lar hydrological setting. Finally, calibration can correct
for the effects of heterogeneity of the media properties
(both soil and vegetation), which are never known very
well. In ungauged catchments, however, calibration to
observed runoff

is not an option. Alternatives are

needed.
Methods for obtaining suitable model parameters in
ungauged basins depend on the nature of the parameter
and on the information that is available in a particular
case. The nature and meaning of parameters differs
depending on the type of model used. The more
process-based a model is, the more the parameters reflect
measurable landscape characteristics relating to the catch-
ment. In contrast, the more conceptual the model is, the
more the parameters reflect the functional aspect of the
entire catchment, and the less they are related to measur-
able landscape characteristics. Since the models used for
estimating runoff hydrographs in ungauged basins span
the entire range from physics-based to conceptual, a var-
iety of methods for estimating model parameters in
ungauged basins have been proposed. The methods can
be grouped into four main categories ( Figure 10.20 ): (a)
a-priori estimation of model parameters from catchment
characteristics; (b) transfer of calibrated model parameters
from gauged catchments; (c) constraining model param-
eters by regionalised runoff characteristics; (d) constrain-
ing model parameters by dynamic proxy data.

runoff model. This
approach of selecting conceptual model structures is emerging
and new research on this is expected in the near future.

-

10.4.2 Parameters of rainfall - runoff models in
ungauged basins: overview
Once a suitable model structure has been chosen for the
catchment in question, the important next step is to
estimate a set of model parameters. In gauged