Introduction - Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales - page 4

runoff predictions, at least some of the model parameters
need to be calibrated to runoff data. However, this is of
course not possible in ungauged basins.
A number of alternative methods have therefore been
developed and have been the methods of choice in prac-
tical applications for a long time. These alternatives
involve the use of runoff data from gauged catchments in
a region, and models for ungauged catchments that
strongly build on these runoff data. These can be statistical
models or simple process models of a conceptual kind,
without recourse to Newtonian physics. However, these
models are centred on the notion of similarity between
the gauged and the ungauged catchments. These types of
models acknowledge that, even though there are no runoff
data in the catchment of interest, runoff data do exist in
other, similar catchments, and these can be transferred in
some way in space to help make runoff predictions in the
ungauged basins.

Processes: Different processes in hydrology have often
been dealt with separately, and therefore hydrologists have
often looked at flow characteristics at different time scales
in an independent way. The annual water yield is usually
studied independently of knowledge of low flows of the
catchment of interest; floods are often studied independ-
ently of knowledge of the seasonal flow patterns within
catchments; and flow duration curves are studied separ-
ately. Is there a deeper connection between these pro-
cesses? What is needed is a simultaneous treatment of
these processes at different time scales.
Places: As each research group has tended to analyse
their own catchments, over the years tremendous under-
standing of runoff processes has been developed for indi-
vidual places, but transferring this to other places has been
hard. Models are often tailor-made to a particular catchment
and it is hard to reason why a particular model structure or
model parameters should be preferred over others. Different
schools of thought have developed their own favourite
methods for different environments and purposes, e.g., stat-
istical versus causal methods or physically based versus
conceptual models. Generalising the findings of how well
the models work and why has been notoriously difficult.
Scales: Research has been performed over a huge range
of scales, and connecting them has caused tremendous
difficulties. This is known as the scale problem in hydrol-
ogy (Blöschl and Sivapalan, 1995 ). When upscaling
laboratory-scale infiltration equations to the catchment
scale, assumptions need to be made about the natural
hydrological variability and how it is organised (Blöschl,
2001 ). Similarly, routing equations at a plot scale may
differ from those at the hillslope scale. This situation has
been exacerbated by the spectrum of disciplines involved,
including engineers, geologists, soil scientists and meteor-
ologists, each of them with different worldviews of at what
scales processes should be conceptualised.
Current textbooks on hydrology propagate the same
fragmented vision of hydrology, organised by process,
and written in the form of recipes, e.g., ten different for-
mulas for estimating infiltration, potential evaporation, and
so on. The situation is literally analogous to

1.3 Fragmentation in hydrology
Because distributed process-based hydrological models are
not the only method of making runoff predictions in
ungauged basins, a plethora of other methods have been
developed that are based on the notion of similarity. There is
no one standard method of runoff predictions in ungauged
basins, rather there are literally hundreds of different
methods. They differ by their model structure, their param-
eters, and by the inputs they use. They also differ in what
processes they represent. Depending on the environments,
the relative role of snow processes, runoff generation pro-
cesses and transpiration processes may differ, as may the
factors that control them. Some of the differences between
the models are directly related to the differences in climate
and catchment characteristics. Also, historically, hydrolo-
gists have had less incentive than researchers in other dis-
ciplines in the earth sciences to collaborate with colleagues
around the world, as the land surface is organised into
separate river basins, and there is little water exchange
across them. Unlike meteorology, for example, a single
catchment can be studied with much success in isolation.
As hydrologists we do not have a single object of study as,
say, a physicist who studies the structure of a particular
atom. All physicists around the world may study the hydro-
gen atom and the models they come up with relate to the
same common object

'

a cacophony

of noises (Sivapalan, 1997 ;

Sivapalan et al., 2003b ). This fragmentation can be best
illustrated by the famous Indian legend of the

… not a harmonious melody '

' six blind

men and the elephant . By touching different parts of the

body of an elephant, these blind men are trying to figure
out for themselves what an elephant may look like, but
have no other way of

'

one hydrogen atom. In contrast,
every hydrological research group around the world is
studying a different object, i.e., a different catchment with
different response characteristics. This is a fundamental
difference that hydrology must face up to.
All of these factors, collectively, have contributed to the
fragmentation of hydrology at various levels.

-

it. Each of them tries to
make inferences about the elephant by touching one body
part of the elephant: it seems like a wall to the blind man
that touches the side of the elephant, a spear to the one who
feels the tusk, a snake to the one who handles the trunk, a
tree to the one who feels the leg, a fan to the one who

' seeing '