PUB in practice: case studies - Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales - page 316

Figure 11.47. Example of the use of
the multi-method approach:
overview of data sets used in the soil
moisture study, along with their
temporal and spatial resolution, main
results and synergetic effects.
(L: low, M: medium, H: high,
S: spatial, T: temporal; if not
otherwise specified the data sets in
column 2 refer to different aspects of
the soil moisture data set). From
Blume et al.( 2009 ).

datasets

resolution results

LSHT time series

binary maps

HSMT

HSMT dm scale variability

dye tracer

HSLT

LSLT hydrophobicity

event dynamics

LSHT

LSHT dye tracer & dynamics

Legend
higher spatial than temporal variability
low temporal variability
preferential flow
preferential flow (inferred)

persistent patterns
seasonal effects
seasonal effects (inferred)
fast response

confirmed by
explained by

The model implementing the longer preferential flow
paths resulted in a hillslope response that mimicked catch-
ment response quite well ( Figure 11.49 ), on the other hand,
reducing the depth of preferential flow paths to 0.65 m
resulted in a dampened and unrealistic event response
( Figure 11.50 ), thus confirming the importance of fast flow
processes such as preferential subsurface flow (note: sur-
face runoff is unlikely due to the very high porosities and
permeabilities of the volcanic ash soils).

12%

10%

8%

6%

4%

Discussion
The multi-method approach of gathering data under time

-

and

-

financial constraints and the general focus on one typical

hillslope proved to be successful in the data-scarce Malalca-
huello catchment, delivering much deeper insights than could
be expected from rainfall and discharge time series alone.
Rainfall and runoff are generally the first parameters
to be measured in previously ungauged catchments
(see Chapter 3 ). In order to investigate the catchment

2%

0%

0 5 10 15 20

Event Number

s
response to rainfall, event runoff coefficients are naturally
one of the first parameters to be extracted from these short
time series. The best predictors for these runoff coefficients
can be determined with linear statistical models and can be
used to infer runoff processes. The more additional data
(e.g., on soil physics, hydrogeology or soft data such as
observations of local residents) are gathered on targeted
field campaigns, the better the results of the statistical
model analysis that can be interpreted (Blume et al., 2007 ).
The linear statistical model can also be used as an
additional catchment descriptor. Event runoff coefficients

'

Observed Linear model LM Jackknifed data

Figure 11.48. Data-based and predicted event runoff coefficients for
17 events. From Blume et al.( 2007 ).

hypothesis. Preferential flow was parameterised in two
different scenarios: (i) depth of preferential flow ¼ 0.65m

( Figure 11.49 ), (ii) depth of preferential flow 1.3 m

( Figure 11.50 ), which corresponds to the approximate
rooting depth and is also in the same range as the observed
length of flow paths of the dye-tracer experiments (1.15 m)
(Blume et al., 2008b ).

¼