Prediction of runoff hydrographs in ungauged basins - Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales - page 269

the signatures of runoff variability discussed so far, and
more. In particular, it retains the rise of runoff during
and after events (wet phase) and the recession following
events (dry phase), which are not included in any of the
previous signatures. The runoff hydrograph is the net
result of runoff generation processes during events, run-
off routing (on both hillslopes and the river network),
evaporation between events, including interactions
amongst all of these processes, and reflects the net effect
of the multiplicity of pathways and storages that water
has passed through on its way to a point in the network.

there should be a renewed search for new kinds of
predictors for rainfall runoff relationships that reflect

landscape organisation and/or regional climate patterns.

-

It would be timely for catchment hydrology to transition
away from traditional lumped models, and use the enor-
mous amounts of spatial information becoming avail-
able as proxy data, such as snow, remotely sensed soil
moisture and inundation patterns, as well as vegetation
patterns, river network structure and soil catena etc., to
build new and better models and constrain model pre-
dictions. The use of large-scale dynamic patterns should
be supplemented with the use of field visits to assemble
local data, through the fostering of a culture of reading
the landscape. Equally, there should be a transition in
modelling from just displaying and quantifying predic-
tion uncertainty to understanding uncertainty hydro-
logically and to making efforts to reduce it.

Measures that characterise the similarity of hydrographs
include all of the similarity indices discussed in Chapters
5

9 in respect of the other signatures. However, to be
effective they need to be organised hierarchically,
starting with the aridity index, which determines annual
runoff. Because the runoff hydrograph, as a signature, is
itself an emergent pattern that reflects the co-evolution
of the catchment structure along with climate, other co-
evolutionary indices that might reflect the similarity of
runoff behaviour include the way the landscape is organ-
ised in terms of landscape units (otherwise known as
hydrological response units or HRUs), river network
structure, the hypsometric curve etc.

-

There is considerable scope to use the new generation of
spatially distributed models to interrogate the differ-
ences between catchments in different hydroclimatic
regions of the world, and in this way gain new insights
into the types of model structure that will be needed in
different climates, and lead towards a synthesis, a har-
monisation of models and modelling approaches appro-
priate to different situations. Both statistical and process-based methods have been used

for predictions, although the trend is for increasing use of
process-based methods. Nevertheless, in data-rich regions
of the world, geostatistical methods that account for the
network structure are seen to work much better in ungauged
basins than process-based (rainfall

In the topic, the comparative assessments of runoff hydro-
graph predictions could only be done for lumped concep-
tual models. There have been too few distributed
modelling studies available for the assessment, a situation
that we hope will improve in the future, using the
examples of DMIP (Smith et al., 2004b ) and DMIP2
(Smith et al., 2012 ). The comparative assessment of these
model predictions indicated that predictive performance
gets worse with increasing aridity (both Level 1 and Level
2 assessments), and gets better with increasing catchment
size (Level 2). The differences between parameter estima-
tion methods did not impact markedly on model perform-
ance (Level 1). In humid catchments spatial proximity and
similarity methods perform best, while in arid catchments
similarity and parameter regression methods perform
slightly better (Level 2 assessment).

- runoff) methods.

When it comes to process-based methods, the choice of
model structure is a key step, and it is always guided by
prior knowledge of the hydrological system, the avail-
ability of data, and prior experience of the practitioner.
This has led to a plurality of models being used. To avoid
fragmentation and duplication, it might be valuable to
group the world into classes of similar behaviour, based
on some kind of classification scheme, and then to narrow
down the number of models adopted. This would
increase the experience with all such models, and through
the sharing of this experience it could lead to improve-
ment of the models themselves, and also improved pre-
dictive performance. This activity could take the form of
a synthesis across processes, places and scales.

Runoff hydrograph predictions should increasingly
focus on (i) distributed models, aimed at generating
and interpreting space-time patterns, and (ii) compara-
tive hydrology, aimed at understanding the differences
between places. The generation of dynamic patterns and
juxtaposing them against patterns of surrogate or proxy
data (through field observations or remote sensing)
should generate new questions to be answered through
further research, instead of just becoming grist to opti-
misation schemes aimed at curve fitting, or to data
assimilation techniques aimed at optimal predictions
for operational use.

Estimation of model parameters of process-based
models can be accomplished in a number of ways,
depending on data availability and model type: a-priori
estimation, through transfer from gauged catchments,
through regionalisation of runoff characteristics or sig-
natures, through the use of dynamic proxy data, and
their combinations. All of these are being used currently,
and have their advantages and disadvantages. Given the
co-evolutionary nature of catchment systems, however,