Prediction of runoff hydrographs in ungauged basins - Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales - page 242

in-depth representation of local structures and
patterns and the adequacy of adopting simpler modelling
approaches given the dominant processes and
their physical controls. Examples of such a rapid assess-
ment at different scales are presented in Blume et al.
( 2008a , b ).

control runoff. For example, groundwater level data can be
directly used in these types of models to check the internal
state variables.
However, they have the disadvantage that they do need
the highest amount of information, and are computation-
ally demanding. The next class of physics-based models,
such as CATFLOW (Zehe and Blöschl, 2004 ), hillslope-
storage Boussinesq (hsB) model (Troch et al., 2003 ) and
THALES (Grayson et al., 1995 ), reduce the dimensionality
of the modelling problem by neglecting flows that are
perpendicular to the main downslope flow paths, and con-
centrate on downslope flows on hillslopes. Thus they have
the advantage of being less computationally demanding,
and need much less information for model setup, and can
therefore be more easily applied at large scales. However,
clearly they have the disadvantage that they work most
effectively where their assumptions hold best and cannot,
for example, explore three-dimensional effects such as
deep groundwater flows. They are therefore most applic-
able in headwater catchments. In the marsh of the Palo
Alto Baylands, California, where the flow system is con-
trolled by the interactions of the surface water in the tidal
channels and the groundwater, Moffett et al. 2012 )
selected a three-dimensional representation of the coupled
system ( Figure 10.16 ) with spatially variable hydraulic
conductivity and evaporation estimated from field data.
Because the soils are clay, the saturation values are very
high. The spatial patterns of soil moisture represent the
interplay of surface and groundwater flow for a given
topography, forced by the tidal signal and evaporation
( Western et al., 1999 ).

Model structure from similar gauged catchments:An
alternative approach is to use the same model structure
as in similar, gauged catchments where the structure is
identified from runoff data. To select suitable donor
catchments, similarity measures such as those in Section
10.2.2 may be used.

Additional considerations in selecting a model structure are
the modelling purpose (e.g., operational vs. investigative
models), data availability (more complex models require
larger data availability), resource constraints (simpler
models with lower budgets), and the modeller

'

s experience
(choosing models one has experience with, see Section
3.7.2 ). The following sections discuss model structure
selection for physics-based models, index-based models
and conceptual models, expanding on the information in
Table 10.1 .

Physics-based models
Physics-based models are physically consistent and expli-
citly account for the potential gradients and resistances that
determine water flows along the multiple flow paths, based
on balance equations (e.g., Richards equation, St Venant
equations). The main information one uses to guide model
structure are (i) a-priori perception of processes in the
catchment and (ii) field data from the catchment and read-
ing the landscape ( Table 10.1 ). Runoff data are not usually
used to determine the model structure, so model selection
in ungauged basins is no different from model selection in
gauged basins.
The model structure choices one has are the dimension-
ality of the flow system (one, two or three dimensions),
and which processes to include (e.g., macropores). In both
instances the decisions may be guided by the modellers

Processes to include Based on available information
from hydrogeological maps and pedo-geomorphological
reasoning, one may construct the likely subsurface archi-
tecture and decide on the relevant processes to include.
Examples are the presence of macropore flow, runon
infiltration and stream aquifer interactions. These choices

can be assisted by taking distributed observations within
the catchment, for instance using soil and geophysical
surveys, drilling exercises, installation of soil moisture
sensors, groundwater level recorders etc. Similarly, the
land surface can be mapped to document the nature of
soil cover as it impacts on infiltration characteristics of
soils and roughness properties that govern overland flow.
Kollet and Maxwell ( 2008 ) were particularly interested in
the linkage between groundwater dynamics and the mass
and energy balance at the land surface. They therefore
explicitly represented this linkage via shallow soil mois-
ture in a model that simulates three-dimensional variably
saturated subsurface flow as well as overland flow
( Figure 10.17 ).

-

'
experience in other, similar catchments. Process similarity
in terms of the perceptual model of the flow system is
applied here.

The dimensionality of the flow system Fully physics-
based models, such as Hydrogeosphere or Hydrus-3D,
solve the governing equations at the highest resolution
(three-dimensional subsurface, two-dimensional surface)
and require two- or three-dimensional information on the
time-invariant controls on gradients and flow resistances.
They have the advantage that they can accommodate the
highest amount of information and thus explore in greater
detail how structural properties of the catchment system