Recommendations - Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales - page 384

13 Recommendations

Contributors: K. Takeuchi,* G. Blöschl, H. H. G.
Savenije, J. C. Schaake, M. Sivapalan, A. Viglione,
T. Wagener and G. Young

This topic is devoted to predicting runoff in ungauged
basins (PUB), i.e. predicting runoff at those locations
where no runoff data are available. It aimed at a synthesis
of research on predictions of runoff in ungauged basins
across processes, places and scales as a response to the
dilemma of fragmentation in hydrology. It takes a com-
parative approach to learning from the differences and
similarities between catchments around the world. The
book also provides a comparative performance assessment
(in the form of blind testing) of methods that are being used
for predictions in ungauged basins, interpreted in a hydro-
logically meaningful way. It therefore throws light on the
status of PUB at the present moment and can serve as a
benchmark against which future progress on PUB can be
judged. In so doing, the topic has also come out with a new
scientific framework that can guide the advances that are
needed to underpin PUB and to advance the science of
hydrology as a whole. The synthesis presented in the topic
is built on the collective experience of a large number of
researchers around the world inspired by the PUB initiative
of the International Association of Hydrological Sciences,
which makes it truly a community effort. It has provided
insights into the scientific, technical and societal factors
that contribute to PUB.
Drawing from the lessons learned from the topic, we are
now in a position to make recommendations on the pre-
dictive, scientific and community aspects of PUB, and to
offer best practice guidelines for runoff predictions in
ungauged basins.

focus needs to be on the hydrological interpretation of
both the method and the results. The level of detail of
the interpretation may vary with model complexity
(from statistical to physics-based), yet the hydrological
interpretation is still essential, informed by photographs
of the dynamics taking place on the landscape. The inter-
pretation may be in terms of physical causality or in terms
of the co-evolution of catchments. For example, a phys-
ical

-

flux relationships,
whereas a co-evolutionary interpretation may be the
Budyko curve.

interpretation may be gradient

13.1.2 Exploiting runoff signatures and linking them
Predicting runoff in ungauged basins requires a targeted
approach where the focus is on a specific signature of
interest. For example, if the interest is on the flood fre-
quency curve the focus should be on processes that are
relevant to floods, not on the hydrographs as a whole. The
targeted approach benefits from connecting to other rele-
vant signatures. For example, estimation of flood fre-
quency can benefit from the study of the seasonal flow
regime. The seasonality is a fingerprint of the catchment
and therefore may assist in predicting all the runoff signa-
tures (annual runoff, flow duration curve, low flows,
floods), including the runoff hydrographs. The modelling
of the signatures will benefit from following a hierarchical
analysis from annual runoff down to seasonal runoff, the
flow duration curve and the extremes (floods and low
flows).

13.1 Advancing runoff predictions in
ungauged basins
13.1.1 Understanding as the key to better predictions
Catchments must be treated as real objects in real places
with real processes operating rather than as abstract con-
cepts. Whatever methods are used for predictions,

13.1.3 Addressing uncertainty from a process
perspective
Estimating the uncertainty of runoff predictions in
ungauged basins is essential. When doing this the focus
should be on the hydrological interpretation of both the
uncertainty method and the uncertainty results. Methods
that draw on numerical experiments, such as Monte
Carlo analyses, may be complemented by comparative

the