Prediction of runoff hydrographs in ungauged basins - Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales - page 252

transposition typically involves the following steps suitable similarity measure to select the donor catch-

ment. Proximity is usually defined on the basis of dis-
tances to the catchment outlets or catchment centroids
(Zvolensky et al., 2008 ;Liet al., 2009 ). It is also
possible to use the geostatistical distances (or Ghosh
distances) that account for the nestedness of the catch-
ments (e.g., Skøien and Blöschl, 2007 ; Gottschalk et al.,
2011 ).

(Blöschl, 2005 ):

(a) Delineation of homogeneous regions and/or identifica-
tion of one or more gauged catchments, termed donor
catchments, based on any of the similarity measures
discussed in Section 10.2 .
(b) Estimation of model parameters for the donor catch-
ments by manual or automatic calibration on observed
runoff data.
(c) Selection of catchment characteristics that are
deemed to affect catchment response to rainfall. This
is either based on an a-priori understanding of what
catchment characteristics may be relevant to a particu-
lar model parameter or on some goodness-of-fit
measure.
(d) Setting up models relating each rainfall

Similarity: An alternative is to choose the donor on the
basis of the similarity of the climate and catchment
characteristic in the two catchments. Similarity is usu-
ally measured by the root mean square difference of all
the characteristics in a pair of catchments. The charac-
teristics are usually standardised by their standard devi-
ation or transformed in another way to make them
comparable. Studies that chose a donor on the basis of
this method use a wide range of climate and catchment
characteristics. Kokkonen et al.( 2003 ) transferred the
complete parameter set from the catchment with the
most similar elevation of the catchment outlet. McIntyre
et al.( 2004 ) defined the most similar catchment in terms
of the catchment area, standardised annual average pre-
cipitation and baseflow index. Other studies used a
larger number of characteristics, such as Parajka et al.
( 2005 ), who defined the similarity by mean catchment
elevation, stream network density, lake index, areal pro-
portion of porous aquifers, land use, soils and geology,
and Zhang and Chiew ( 2009 ), who identified the most
similar catchments in terms of area, mean elevation,
slope, stream length, aridity, woody vegetation fraction
and plant available water holding capacity.

runoff model
parameter to a set of catchment characteristics. In this
step, commonly, multiple linear regressions are used
and some or all of the catchment characteristics are
(e.g., logarithmically) transformed.
(e) Testing the strength of the relationship of (d), e.g., by
some goodness-of-fit measure such as a correlation
coefficient.
(f) Estimating each parameter of the rainfall

-

- runoff model

for the ungauged catchment from the (regression)

model.
(g) Simulating runoff for the ungauged catchment of inter-
est by applying the same model as in (b), using the
regionally transposed model parameters.
(h) Testing the transposition by cross-validation.
A gauged catchment is assumed to be ungauged, run-
off is simulated as in (g) and then compared with the
locally observed runoff.

Model averaging: Sometimes a weighted combination of
the parameter sets frommore than one donor catchment is
used, where the catchments are selected based on either
proximity, catchment characteristics or both ( Goswami
et al., 2007 ; Kim and Kaluarachchi, 2008 ; Seibert and
Beven, 2009 ). One can either assume a fixed subdivision
of the region into groups of catchments or, alternatively,
allow each catchment to have its own group of donor
catchments (Burn and Boorman, 1993 ; Young, 2000 ).

Some of these steps can be skipped depending on data
availability and the regionalisation method chosen.

Spatial proximity, similarity and model averaging
The most straightforward approach for transferring cali-
brated model parameters to ungauged basins is to identify
one or more similar gauged catchments in the region
(termed donor or analogue catchments), and assume that
the entire parameter set is also valid in the ungauged basins.
The justification for the approach is that if two catchments
are similar in their catchment characteristics one would
hope that their hydrological response should be similar
too, so the model parameter values should be similar. There
are three main ways the entire parameter set is transferred
from the donor catchment(s) to the catchment of interest:

In all three methods it is important that the catchments
selected as donors are indeed hydrologically similar,
i.e., have similar flow systems. It is possible that
some of the donors are disinformative even though their
climate/catchment characteristics are similar to the
ungauged basins. Transferring parameters from such
catchments decreases the model performance in the
ungauged basin compared to using, say, the mean par-
ameter set of all the catchments in the region. Boldetti
et al.( 2010 ) addressed this issue and proposed an
approach to detect potentially undesirable donor catch-
ments by an iterative approach.

Spatial proximity: If one assumes that climate and catch-
ment characteristics vary only smoothly in space then
spatial proximity between the catchments may be a