Prediction of runoff hydrographs in ungauged basins - Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales - page 249

can be obtained either from field surveys or from analyses
of satellite data. LiDAR is an attractive alternative that has
recently received a lot of attention. There are a number of
methods to obtain roughness estimates from a large
number of data points at a higher spatial resolution than
the resolution that is used for digital elevation models.
Hollaus et al.( 2011 ) estimated roughness as the standard
deviation of subgrid data points. Casas et al.( 2010 ) esti-
mated roughnesses using an equation based on mixing
layer theory resulting in a co-varied relationship between
roughness height and topographic content. LiDAR can be
combined with other remote sensing data such as multi-
spectral images (Forzieri et al., 2011 , 2012 ). Typically the
roughness estimates are then used in hydrodynamic mod-
elling of ungauged basins (Smith et al., 2004 ).
The hydraulic geometry of streams can be obtained from
field surveys, possibly assisted by LiDAR. The photos in
Figure 10.22 show cross-sectional profiles surveyed in the
fiel d. They were included in a distributed hydrological
model of the catchment ( Rogger et al., 2012a ). The photos
illustrate that, even without quantitative measurements, cre-
ating a photo documentation of the catchment can be
extremely important. This, in fact, applies to all the a-priori
parameters. Field visits can provide extremely useful infor-
mation that supports understanding of the dominant hydro-
logical processes. Information from field visits has been
used to determine estimates of when the stream exceeds
bankfull discharge and retention in the floodplain starts
( Figure 10.22 ). In addition, understanding flow processes
on the hillslopes can be greatly assisted by such field visits,
e.g., by judging where overland flow may occur as assessed
by erosion marks. There is much information that cannot
easily be quantified numerically that can be very useful for
the modelling of runoff in ungauged basins. Sections 3.7.2
and 3.7.3 provide examples where the understanding of
flow processes in the catchment is improved during field
visits, which is complementary to remotely sensed data or
large-scale spatial databases. Sections 11.13 and 11.14 p ro-
vide further examples of the value of field visits for estimat-
ing runoff in ungauged basins.
There have been a number of inter-comparison studies
that have examined how well runoff can be predicted in
ungauged basins on the basis of a-priori model parameters,
i.e., without calibrating the parameters to the catchment of
interest or neighbouring catchments. One such inter-
comparison project was the MOPEX (Model Parameter
Estimation Experiment; Schaake et al., 2006 ; Duan et al.,
2006 ), which was performed in 12 selected catchments in
south-eastern USA. The hydrologists participating in the
inter-comparison used different methods to obtain these
a-priori parameters, such as those proposed by Koren
et al.( 2000 ) and Anderson et al.( 2006 ). They were asked
to simulate daily runoff at a number of locations without

having access to local runoff data, and their predictions
were later compared with the runoff measurements. They
were also compared to a different set of simulations with
the same models where calibration to local runoff data was
allowed. Over the catchments studied, the median NSE of
daily runoff using a-priori parameters was 0.2 0.6,

depending on the model ( Figure 10.23 ). When calibrated
to local runoff, the median NSE increased to 0.4

-

0.75. The
performance of monthly runoff is somewhat higher, par-
ticularly for the calibration period ( Figure 10.23 ). These
comparisons suggest that there is a clear role for calibration
in order to improve model performance over the use of
a-priori values. The project outcomes also suggest that
even with a reasonable specification of soil characteristics,
the look-up tables relating soil texture classes to soil
hydraulic properties (e.g., Koren et al., 2000 ) may not be
applicable at large spatial scales because those tables were
created under laboratory conditions and are applicable to
point or plot scales. To study the transferability of model
parameters to other catchment conditions, data from a wide
range of climatic conditions should be used. Such com-
parisons are important, perhaps globally, in the context of
comparative hydrology to get a better understanding of
what parameter

-

ranges are applicable in different

environments.
As mentioned above, the meaning of the parameters
differs depending on the type of model. However, even
for process-based models there are a number of difficulties
with using measured parameters in rainfall

runoff models
that are related to scale, as suggested in the MOPEX study
above and elsewhere (Blöschl and Sivapalan, 1995 ). The
measurement volume is usually much smaller than the
model element size, and in most cases there are only a
few measurement locations within a catchment and the
spacing between the measurements is therefore large. This
means that the measured parameter is not defined in
exactly the same way as in the model even if it shares the
same name (Beven, 1989 ). In principle, both scale dispar-
ities can be addressed by upscaling procedures (Blöschl,
2005c ). In practice, one often neglects the incompatibility
related to the support and addresses the incompatibility
related to the spacing by some sort of interpolation proced-
ure. Also, model parameters that are usually considered
static, such as the saturated hydraulic conductivity, may in
fact be dynamic and depend on a range of processes
occurring in the catchment that are not captured by
physics-based models. This is illustrated by data from
sprinkling experiments in Figure 10.24 . Simulated rainfall
of a constant intensity was applied onto plots until equilib-
rium was achieved and surface runoff occurred. The ratio
of surface runoff (at equilibrium) and rainfall intensity is
the runoff coefficient, which is not dependent on initial soil
moisture. The experiments were performed twice at the

-