A data acquisition framework for runoff prediction in ungauged basins - Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales - page 49

Figure 3.19. Change in annual water
balance components (mean and
standard deviation) during the three
modelling stages for the Chicken
Creek. The variability in
precipitation is caused by
precipitation correction by a few
models. From Bormann et al.
( 2011b ).

subsurface flow with little direct runoff. In reality, surface
runoff was a major flow component despite the fairly
coarse soil texture. The actual evaporation (AE) and the
ratio between actual and potential E was systematically
overestimated by nine of the ten models. None of the
model simulations came even close to the observed water
balance for the entire 3-year study period

of the differences in the model results. The model para-
meterization and choice of initial conditions depended on
the modeller's judgment and were therefore a result of the
modeller's experience in terms of model types and case
studies

'

(Bormann et al., 2011b ). The study therefore
confirms the findings of previous studies (e.g., Diekkrüger
et al., 1995 ) on the importance of the modeller's subjective
decisions, particularly in the case of a-priori prediction.
'

' (Holländer

et al., 2009 ).
The spread of the model simulations narrowed during
stage two, after modellers had discussed their results and
had visited the catchment. As a consequence of the discus-
sions and the field visit, modellers tended to change the
model setup in the same direction (resulting from a common
process understanding). All modellers tended to reduce the
total runoff generation while increasing surface runoff gen-
eration, since a biological soil crust had been identified
(Fischer et al., 2010 ). Some modellers also adapted the
representation of subsurface storage behaviour and changed
initial conditions because it had emerged from discussions
that the catchment was dry after construction.
In the third modelling stage, modellers were asked to
select the required data from an available data pool con-
sidering hypothetical costs they would be willing to pay for
the data. Most modellers asked for soil hydraulic and soil
physical data as well as for soil moisture and infiltration
rates, while only a few modellers used the extended vege-
tation data set, the new digital elevation model and the new
aerial photo. Most of the modellers used the data for
reassessing model parameters and adjusting initial condi-
tions. However, the spread of the models after these adjust-
ments remained similar to that of the second modelling
step. The additional observations available during the third
step led to smaller changes in the model simulations than
those due to initial data, joint discussion and actual visits to
the catchment (e.g., water balance terms; Figure 3.19 ).
Overall, the study participants concluded:

The most important parameters to be presumed were the
soil parameters and the initial soil water content while
plant parameterization had, in this particular case of sparse
vegetation, only a minor influence on the results

' (Hol-

länder et al., 2009 ).
The study further showed that the use of soft as well as
hard data is valuable in the case of sparsely gauged catch-
ments. Soft data, e.g., obtained from field visits or even aerial
photos, can inform the modeller about dominant or at least
important hydrological processes in a catchment that will
help improve hydrological process understanding. The mod-
eller can then decide how to use such information in the
modelling process. In this study, additional data predomin-
antly only confirmed the modeller's assumptions that were
based on field visits and discussion. They did, however,
assist in improving the adequate choice of initial and bound-
ary conditions. After carrying out the fourth modelling stage,
consisting of model calibration against observed event runoff
froma subcatchment, further analysis of the predictive uncer-
tainty of the a-priori modelling steps will be feasible.

3.7.3 Forensic analysis of magnitude and causes
of a flood (Sel š ka Sora, Slovenia)
Observations of traces left by water and sediments during
flood events provide an opportunity for developing spatially
detailed estimates of peak runoff along the stream network
( Figure 3.20 ). This information is helpful for better under-
standing the role of rainfall accumulation rates and of soil
and land use properties in runoff generation in the context of

the compari-
son indicates that, in addition to model philosophy,
the personal judgment of the modellers was a major source

'