See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/229739643

Prediction in Ungauged Basins: A Grand Challenge for Theoretical Hydrology

Article  in  Hydrological Processes · October 2003

DOI: 10.1002/hyp.5155

CITATIONS READS

426 1,118

1 author:

Murugesu Sivapalan
University of Illinois, Urbana-Champaign
483 PUBLICATIONS   28,065 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

NSF sustainable, resilient and reliable infrastructure system for biofuel View project

Flow duration curve View project

All content following this page was uploaded by Murugesu Sivapalan on 15 October 2017.

The user has requested enhancement of the downloaded file.

 INVITED COMMENTARY
HYDROLOGICAL PROCESSES
Hydrol. Process. 17, 3163–3170 (2003)
Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/hyp.5155

Prediction in ungauged basins: a grand challenge for

theoretical hydrology

Murugesu Sivapalan* Introduction

Centre for Water Research, The Owing to the critical role of water in the Earth system—as global
University of Western Australia, heat exchanger, climatic thermostat, land-forming agent, life-blood
Crawley, WA, Australia and lymphatic fluid (National Research Council, 1991)—the science
of hydrology holds a unique and central place in the field of envi-
*Correspondence to: ronmental Earth science, being intimately connected to the sister
Murugesu Sivapalan, Centre for
Water Research, The University of disciplines of meteorology, climatology, geomorphology, hydrogeol-
Western Australia, 35 Stirling ogy and ecology. As an applied science, hydrology is also highly
Highway, Crawley WA 6009, relevant to day-to-day human experiences, focusing, as it does, on
Australia. floods, droughts, drainage basins, material storages and fluxes, and
E-mail: sivapalan@cwr.uwa.edu.au effects of large-scale engineering projects such as streamflow modi-
fication, soil conservation, and channel modification (Dunne, 1998).
Thus, hydrology should be an exciting field of study; instead,
it appears to be fragmented, deeply rooted in empiricism, and
struggling to realize its full potential. Why? Part of the explanation
that is bandied about is the tension between the expectations of
hydrology as a natural Earth science and as an applied engineering
science, and the difficulties and inertia this imposes against efforts
aimed at making significant scientific advances. There have been
frequent calls in recent times for a new scientific theory to move the
field of hydrology to a modern science rooted firmly in improved
holistic understanding of how water interacts with the entire Earth
system, reconciling these requirements with the societal expectations
of hydrology as an applied science (Dunne, 1998). These have come
about with the gradual recognition of the effects of multiscale
heterogeneities encountered in hydrology, and of the clear need for
an integrative, rather than piece-meal or fragmented, description
of the Earth system, including the role of biota, for solving many
contemporary environmental problems (Gupta, 2000). But what
shape will a new theory take? How will we develop it? There is
a lot of philosophizing, but, as yet, no clear answers.
Perhaps focusing on a grand hydrological problem or on a
question that draws heavily on new fundamental and theoretical
advances in the natural and Earth sciences to benefit practical
applications may be the best way to revitalize hydrology towards
a modern and holistic science, since it will then be in a position to
meet both its scientific and societal obligations. In this commentary
I will argue that the new IAHS decadal initiative on ‘Predictions in
Ungauged Basins’ (PUB) is an example of such a grand challenge, a

Received 25 May 2003

Copyright  2003 John Wiley & Sons, Ltd. 3163 Accepted 4 June 2003
 M. SIVAPALAN

practical problem that, nevertheless, demands ma- ungauged or poorly gauged, especially in many
jor breakthroughs in theoretical, observational developing countries, and in some cases the mea-
and computational hydrology. surement networks have been actually declining in
recent decades. In fact, it is in those parts of the
Predictions in Ungauged Basins (PUB) world that are ungauged or poorly gauged that
we also see the greatest human impacts occur-
‘Observation is the foundation of all learning’, yet
ring.
almost every process of interest to hydrology is
Three thousand years after the Nilometer, the
difficult to observe—evaporation, infiltration and
subsurface flow, even rainfall, have remained dif- model, was first conceived, in spite of their increa-
ficult to measure routinely and unambiguously. sed sophistication and process complexity, the best
Water level in the river, and by inference flow in existing models remain calibrated ones. Indeed,
the river, is the one quantity that we can measure the ‘success’ of these models in mimicking past
with some confidence and which, in addition, has data may even have engendered a certain degree
the advantage of being an integrated measure at of complacency, and a strong reliance on calibra-
the basin scale. This is probably why human civi- tion, preventing the pursuit of real understanding,
lizations have relied so heavily on river gauging which should be the basis of any predictions. Pre-
to understand and quantify the response of the diction in ungauged basins (PUB), sans calibra-
land surface to rainfall inputs. The Nilometer of tion, therefore, remains a difficult, unsolved prob-
the ancient Egyptians is a classic example. lem, demanding urgent resolution, and requiring
The Nilometer may, in fact, be considered as significant new breakthroughs in data collection,
the earliest hydrological model, except that it was process knowledge and understanding.
based on a historical calibration between water
level in the Nile and measures of human comfort
and prosperity (National Research Council, 1991). Main Obstacle to Hydrologic Predictions
However, predictions based on such a model would
Why is it so difficult to predict basin responses?
not have been based on real understanding of the
The root cause of our difficulties is the tremendous
processes that lead to flow in the river. Therefore,
heterogeneity of the land surface condition, soils,
it is doubtful whether it could have predicted the
vegetation, land use, etc., and the space–time vari-
changes in river behaviour resulting from human
impacts or climatic changes that may have con- ability of climatic inputs, occurring over a wide
tributed to the eventual collapse of Egyptian, and range of space and time scales. The heterogene-
many other ancient, civilizations. ity is pervasive, and never seems to disappear at
It is fair to say that, in many parts of the world, whatever scale we observe, ranging from large-
our current understanding of basin responses is scale geological formations, the soil catena at hill-
inadequate even to extrapolate, with a quantifiable slope and basin scales, down to the macropores
level of confidence, from a gauged to an ungauged that are ubiquitous at the plot scale. Considerable
basin. In addition, the global scale of human activ- advances have been made in developing sophis-
ities and the historically unprecedented magnitude ticated descriptions of heterogeneities, yet there
of human-induced land use and climatic changes remain a number of serious problems. For one
mean that past data may not be a reliable guide thing, we do not have sufficient field data at all
to predictions in the future. This is due to the scales to decide unambiguously between one pos-
presence of strong nonlinearities and feedbacks sible description and another. Because of changes
in the way that basins respond to these changes. in structure, these descriptions may not yet be uni-
These are circumstances that demand continued, versal over the wide range of scales that we are
and even enhanced, gauging to monitor and under- interested in. Use of new measurement techniques
stand the impacts of such human-induced changes. in the future may assist, but may yet reveal further
Yet, over 3000 years after river gauging began complexities than we know now. Also, based on the
on the Nile, the Earth’s land surface remains measurements presently available, we are not yet

Copyright  2003 John Wiley & Sons, Ltd. 3164 Hydrol. Process. 17, 3163–3170 (2003)
 INVITED COMMENTARY

able to develop representations of hydrologic pro- models that helped revolutionize the descriptions
cesses resulting from the heterogeneities of land- of precipitation fields and turbulent fluid motions.
scape properties and climatic inputs. Finally, in I believe that the new paradigm underpinning a
spite of any significant advances in the descrip- new theory of hydrology must satisfy the following
tions of climatic and landscape heterogeneities, criteria:
and the parameterization of their effects on hydro-
logic processes, there remains the problem that
we may never have the measurement techniques 1. acceptance of multiscale heterogeneity, to be
to characterize fully the actual properties in spe- seen as a true reflection of nature that, indeed,
cific basins in which predictions are required, the we can benefit from, and not as a problem to be
very essence of the PUB problem (Beven, 2000, endured;
2002). 2. an integrated, holistic description of heterogene-
Regardless of the natural heterogeneities in the ity that respects the hierarchy of elements nested
basin system, the classical theories that we have for within each other, and the interconnectedness of
describing the constituent hydrologic processes are different landscape elements arising from the co-
based on the assumption of homogeneity, derived evolution of climate, soils, vegetation and topog-
as they are from experiments carried out at lab- raphy, including actions of recent landscape-
oratory or small field scales (field plot, hillslopes, forming events such as extreme hydrologic
small basins). Lying at the heart of our difficulties events;
in making predictions in ungauged basins is that 3. an Earth-centric view that considers basins pri-
the homogeneity assumption underlying current marily as creations of nature (not of humans,
theories and process descriptions is incompatible except for recent dramatic human interven-
with the tremendous multiscale heterogeneity that tions), and emphasizes a culture of continual
nature exhibits. learning from field observations;
The current dominant paradigm in hydrology 4. learning from observed data, and from patterns
is reductionist: it involves using an a priori set behind the data, through the testing of falsifiable
of small-scale theories and process descriptions, models of climate–soil–vegetation–topography
and the splitting of basins into small enough ele- interactions at the basin scale; and
ments that are deemed sufficiently uniform for 5. acknowledgment that we will never have full
these theories to work. This approach has been knowledge of the heterogeneities and complex-
enriched by the most sophisticated process descrip- ities present in specific basins, and a realistic
tions that cognate fields of soil physics, meteorol- accounting of this lack of knowledge in terms of
ogy and hydraulics have been able to offer, the its impact on predictions.
advent of powerful computers further accelerat-
ing the trend. Absence of measured values for the
In terms of the philosophy of science, this new
numerous model parameters that arise means that
paradigm contains elements of the reductionist or
the only way the models can be tied to reality is
mechanistic (e.g. Newtonian) and the holistic or
through calibration with gauged data. Invariably,
ecologic (e.g. Darwinian) worldviews (Capra, 1996;
over-parameterization is the order of the day, lead-
Harte, 2002). In terms of recent hydrologic think-
ing to considerable uncertainty of predictions in
ing, it will be seen as a synthesis of the upward and
ungauged basins.
downward routes to conceptualizations (Klemes,
1983; Sivapalan et al., 2003), consistent with the
Paradigm Shifts Heralding a New Theory of view of basins as ‘complex systems with some
Hydrology degree of organization’ (Dooge, 1986). In the con-
To overcome the almost insurmountable prob- text of PUB, it also acknowledges and incorporates
lems described above, we require nothing short concepts of ‘uniqueness of place’ and ‘landscape
of a paradigm change, as radical as the cascade space to model space mapping’ enunciated recently
paradigm and the corresponding multiplicative by Beven (2000, 2002).

Copyright  2003 John Wiley & Sons, Ltd. 3165 Hydrol. Process. 17, 3163–3170 (2003)
 M. SIVAPALAN

While the search of the new theory of hydrol- on information from disparate sources: (experi-
ogy at the basin scale will be based on the ele- mental plots, field surveys, weather radar reflec-
ments of the new paradigm as outlined above, it tivities, satellite data, tree rings and palaeobio-
is not clear yet what shape it will take eventu- logical and palaeohydrological data) and repre-
ally. Nevertheless, I believe that its development sent variability over a wide range of space and
might involve the following challenges for theoret- time scales. Fractals, multifractals, and random
ical hydrology. cascades are modern stochastic techniques that
exploit concepts such as geometric and statistical
Heterogeneity of landscapes and climatic inputs: self-similarity to quantify the relationship between
search for patterns and laws the variabilities present at different scales, and
they have the power to describe complex forms
At the most fundamental level, climate acts as
of heterogeneity occurring over a wide range of
the unifying global force in the co-evolution of scales with a minimum of number of parame-
landscapes and vegetation. An illustration of this ters. Multifractal concepts have already been used
is the fact that the world’s broad vegetation successfully to characterize space–time variabil-
classes can be predicted by the combination of ity of rainfall fields and the structure of stream
just two climatic variables: temperature and rain- networks, and they are beginning to be used
fall. Observed precipitation patterns demonstrate to describe spatial patterns of soils and veg-
space–time variability over a wide range of scales, etation, soil moisture, and a number of other
including but not limited to such definable units geophysical and biophysical phenomena. How-
as cells, small mesoscale areas, large mesoscale ever, it should be noted that these are essen-
areas, synoptic areas, etc., as they develop, mature, tially descriptive methods, and they do not yet
move and dissipate. Rivers carve the landscape have the power to explain, physically, the observed
into beautiful shapes called river networks, which space–time patterns. Therefore, it is not clear how
appear to embody a deep sense of symmetry, and they can be used for predictions, or for extrapo-
some general principles such as minimum energy lating between measurement sites and beyond to
expenditure are said to be at work. Geological other sites.
facies tend to be discrete units associated with spe- The field of hydrogeomorphology is attempt-
cific formational processes, sometimes organized in ing to discover the underlying order or symme-
certain sequences. Soils tend to develop in response try behind quantitative measures of drainage net-
to state controls such as topography, with differ- work composition, and the mechanisms that have
ent parts of a basin (nose, slope, hollow) being contributed to these. Similarly, the new field of
formed by different processes and having dif- ecohydrology is attempting to discover the rules
ferent functions. The resulting soil patterns (soil governing the patterns of vegetation density and
catena) exhibit a common form of organization type, linking these to the underlying water and
and symmetry, with water movement clearly being energy balances, principles of land formation and
an active agent in their formation. The interac- erosional stability, giving rise to natural variabil-
tions between climate, soils, vegetation and topog- ity exhibited at all scales, and consistent with a
raphy thus contribute to the most amazing pat- Darwinian natural selection process that is optimal
terns that we see in natural basins, which must for growth and reproduction within the prevail-
contain valuable information about the way they ing climate and geology (Eagleson, 2002). Similar
function. A fundamental aim of theoretical hydrol- theories and associated models of soil formation,
ogy must be to identify such patterns, decipher shallow landsliding and erosion are being pur-
the underlying order or symmetry occurring over sued to explain and predict the observed natu-
a wide range of scales and the rules that govern ral patterns of soil profiles and soil properties.
them, and explore the mechanisms that generate These advances in the quantification of landscape
them. and vegetation patterns, and the exploration of
Observed patterns of heterogeneity of climatic the underlying causal mechanisms, will ultimately
inputs and landscape elements have been based become the basic elements of the new theory

Copyright  2003 John Wiley & Sons, Ltd. 3166 Hydrol. Process. 17, 3163–3170 (2003)
 INVITED COMMENTARY

of hydrology at the basin scale. They will cer- reliable and universally applicable regionalizations
tainly contribute to our ability to predict ungauged across to ungauged basins.
basins.

Field experiments in nested river basins: process Derivation of new balance equations for nested
studies, perceptual models and generalizations river basins
Hydrology is an Earth science and field observa- The new theory of hydrology should include a
tions must play a substantial role in the devel- consistent set of coupled, scale-dependent, bal-
opment of any new theory. A number of highly ance equations expressible directly at the sub-basin
focused and detailed field experiments must be scale for mass (water), momentum and energy bal-
carefully designed and carried out in different ances, explicitly respecting and being able to incor-
regions of the world, in nested fashion, in order to porate within it, in a nested manner, all of the
identify and quantify the heterogeneity of hydro- natural heterogeneity that is present within this
logic processes at the plot, hillslope and basin system.
scales, and to obtain the necessary internal and Heterogeneity of basin processes occurs over a
surrogate data needed to make inferences about wide range of space and time scales; therefore,
the underlying mechanisms at work, and of their extrapolations to ungauged basins within a region
connections to landscape and climatic heterogenei- require basic understanding of the physical and
ties. An understanding of the spatial variability statistical relationships that connect the compo-
among the plots and hillsides within the basin and nents of the hydrological cycle across these scales,
of their spatial interactions is necessary for link- and the data requirements in any basin to enable
ing process conceptualizations across scales, and this to be done. Therefore, a primary requirement
to making predictions at basin scales. The assess- is that the modelling framework be distributed,
ment of such variability is not an easy problem,
explicitly respecting the river network that con-
especially in view of the status of current flux mea-
nects different parts of the basin, and which also
surement techniques, and poses significant chal-
influences much of the spatial heterogeneities that
lenges for the design of future field experiments
exist within it (features such as topography, soil
and for the subsequent step of testing hypotheses
catena, vegetation). Natural building blocks for the
about the characterisation and exploration of the
derivation of the balance laws should, therefore, be
physical controls on the variability at sub-basin
scales. the nested set of sub-basins associated with indi-
Detailed field experiments, combined with detail- vidual stream links, accompanied by a strategy for
ed soil surveys, etc., can also be used to form or the estimation of parameters at the sub-basin scale.
update the ‘perceptual models’ we have of basin Gupta and Waymire (1998) presented the cou-
responses. However, these only apply to the one pled set of mass balance equations for flow within
or a few hillslopes of highly instrumented basins, the stream network, and for runoff processes
and must be generalized to other hillslopes and/or within the associated set of sub-basins, expressed
basins. Since field experiments are expensive and in the form of a set of hierarchical difference
of a short—term nature, emphasis must also be equations. Reggiani et al. (1999) derived a more
given to inferring the perceptual models using only complete set of coupled mass, momentum and
the rainfall and runoff data measured at the basin energy balance equations (also in the form of dif-
scale, without the support of any process con- ference equations) for the stream network and
ceptualizations generated in detailed field exper- associated sub-basins. However, considerable work
iments (Sivapalan, 2003), and to reconciling the remains to be done to develop efficient parame-
differences with those obtained from the latter. terizations of the effects of within-sub-basin het-
These will help us to focus on common concepts erogeneity, to develop efficient numerical models
or emergent features that could form the thread based on these governing equations to account for
required to connect between multiscale variabil- the large range of time scales associated with com-
ities and process descriptions, and to formulate plete basin responses, and to test the predictive

Copyright  2003 John Wiley & Sons, Ltd. 3167 Hydrol. Process. 17, 3163–3170 (2003)
 M. SIVAPALAN

ability of these models with the limited data avail- Such falsification exercises based on patterns in
able in gauged basins. observed data are a necessary requirement for any
new theory of hydrology, and predictions based on
Learning from patterns: data-based approaches the theory must be refuted or confirmed by obser-
to making inferences about basin responses vations (Dunne, 1998). In most current hydrologi-
As mentioned before, in basin-scale hydrology cal models, since gauged data are used to estimate
we encounter the fundamental difficulty that the their parameters by calibration, the opportunity
internal process dynamics, together with the state to learn from wrong predictions is eliminated and,
variables and landscape properties, are often un- therefore, the models are not falsifiable. There-
known and/or unobservable. Hence, traditional fore, for our theory development to succeed, we
approaches to making predictions by recourse to need: (1) to form rigorous and testable hypothe-
small-scale process dynamics are fraught with sig- ses; (2) to carry out field or numerical experiments
nificant difficulties. However, the overall basin with outcomes that are capable of falsifying or
response is itself easily observable, and when prop- confirming these hypotheses; and (3) to repeat the
erly analysed often reveals spatial and temporal above, making further sub-hypotheses or sequen-
patterns that are richly complex and difficult to tial hypotheses.
understand from a purely deterministic point of There are already some successful examples in
view.
hydrology of simple models being used to test
Theory development will advance if we can
hypotheses, and to falsify them. An example is the
develop simple models (which may be mere carica-
tures of the basin system but, nevertheless, contain work of Gupta and Waymire (1998) and Gupta
within them the basic properties of actual basins), (2003), and work cited therein, in which simple,
provided, importantly, that they can be falsified distributed models built around the stream net-
with large-scale patterns extracted from observed work have been used to test hypotheses about the
data. An example in this context is the notion of scaling behaviour of flood peaks, especially the
‘stochastic coherence’, requiring model predictions issue of ‘stochastic coherence’. No calibration of
to exhibit the same scaling behaviour as the data physical parameters has been necessary in this
(Lovejoy and Schertzer, 1995). Understanding the falsification exercise. In this sense, the scaling the-
physical bases of such patterns will deepen our ory of floods provides a new mathematical frame-
understanding of the physics of the system, since work for interpreting empirical scaling parameters
many of these observed space–time patterns are in terms of numerical and analytical solutions of
often emergent properties (Rundle et al., 2000) and physical equations and thereby testing different
may not be reproduced by traditional models of hypotheses.
the reductionist type. Another example is the work of Reggiani et al.
Model falsification exercises allow us to develop (2000), who worked with a simple, but physically
new intuitions and insights through computations based model of water balance based on coupled
that address simple, but critical, hypotheses about
mass and momentum balance equations for a sim-
general basin behaviour, and to test them against
ple hillslope element. They used it to reproduce
a wide range of datasets. As theory advances, the
the Budyko curve (Budyko, 1974), which encap-
simple models will give way to the development
and testing of more complex ones that capture sulates the climate (precipitation and radiation)
more of the essential characteristics of real basins. controls on annual water balance. In particular,
Such tests can also provide the basis for new their work opened up questions regarding how
experiments and observations. The continual inter- climate, soils and vegetation have evolved over
actions among theory, data and models (already time, the rules underlying their mutual interac-
described in detail before) will lead to new physical tions, and how these can be tested with available
understandings and an improved basis for predic- data. Other notable examples include the work on
tions in ungauged basins. physical controls (basin form and function) of the

Copyright  2003 John Wiley & Sons, Ltd. 3168 Hydrol. Process. 17, 3163–3170 (2003)
 INVITED COMMENTARY

instantaneous unit hydrograph (IUH), and ques- problem, of immediate relevance to society, deal-
tions regarding the shape of the IUH, its depen- ing with questions such as the impacts of land use
dence on basin size, and the effect of nonlinear- and climatic changes, biodiversity and sustainable
ity, which are all issues that remain unresolved to development. Consequently, PUB forces us to chal-
this day. lenge and critically evaluate existing approaches to
In a global and practical context, there exist making hydrologic predictions, and acts as a cat-
numerous other signatures or general patterns alyst for triggering significant theoretical break-
that can form the necessary testing grounds for throughs.
model falsification. McMahon et al. (1992) pre-
sented a compendium of inter-hemispherical and
inter-regional comparisons of inter-annual vari-
ability of runoff volumes and maximum flood References
peaks. Since runoff always refers to an area
Beven KJ. 2000. Uniqueness of place and the representation of
(basin), many of the compiled statistics relating to hydrological processes. Hydrology and Earth System Science 4:
runoff (volume, flood peak) are area dependent. 203– 213.
McMahon et al. (1992) found significant global Beven K. 2002. Towards a coherent philosophy for modelling the
and regional differences in their inter-annual vari- environment. Proceedings of the Royal Society of London, Series
A 458: 1–20.
ability and the dependence on basin area. They
Budyko MI. 1974. Climate and Life. Academic Press.
proposed a number of hypotheses about the con-
nections between the inter-hemispheric differences Capra F. 1996. The Web of Life: A New Scientific Understanding
of Living Systems. Anchor Books, Doubleday: New York.
and key physical characteristics (land and sea dis-
Dooge JCI. 1986. Looking for hydrologic laws. Water Resources
tributions, latitudinal distribution of land masses, Research 22(9): 46S–58S.
receipt of solar energy, extent of glaciation, age of Dunne T. 1998. Wolman Lecture: Hydrologic science . . . in land-
landscapes, extent of coverage of mountain ranges, scapes . . . on a planet . . . in the future. In Hydrologic Sciences:
impact of El Niño–southern oscillation effects, Taking Stock and Looking Ahead. National Academy Press:
Washington, DC.
etc.). When it comes to differences between con-
Eagleson PS. 2002. Ecohydrology, A Darwinian Expression of
tinental regions and climatic zones, the differences Vegetation Form and Function. Cambridge University Press:
in runoff behaviour may be attributed to differ- Cambridge, UK.
ences between climate, soils and vegetation, includ- Gupta VK. 2000. A framework for reassessment of basic research
ing the way native vegetation adapts to water and educational priorities in hydrologic sciences. A report to
stress, such as leaf shedding and deep rooting, the National Science Foundation, USA. http://cires.colorado.edu/
hydrology. Accessed 4 September 2003.
and how, over long periods of time, such adap-
Gupta VK. 2004. Emergence of statistical scaling in floods on
tive responses engender an evolutionary selection channel networks from complex runoff dynamics. Chaos, Solitons
of dominant vegetation types. All of these observed and Fractals 19: 357–365.
patterns are ripe for further analyses and repre- Gupta VK, Waymire E. 1998. Spatial variability and scale invari-
sent excellent challenges for the development and ance in hydrologic regionalization. In Scale Dependence and Scale
Invariance in Hydrology, Sposito G (ed.). Cambridge University
falsification of generalized models. Such analysis Press: Cambridge; 88–135.
and modelling are at the heart of predictions of
Harte J. 2002. Toward a synthesis of the Newtonian and Dar-
ungauged basins, especially in a global context. winian worldviews. Physics Today 55(10): 29– 34.
Klemeŝ V. 1983. Conceptualization and scale in hydrology. Jour-
nal of Hydrology 65: 1–23.
Summary
Lovejoy S, Schertzer D. 1995. Multifractals and rain. In New
It is clear that PUB forces us to deal with questions Uncertainty Concepts in Hydrology and Water Resources, Kundze-
wicz ZW (ed.). Cambridge University Press: 62–103.
that are at once deep, grand, and practical. On the
one hand, PUB is a challenging and exciting sci- McMahon TA, Finlayson BL, Haines AT, Srikanthan R. 1992.
Global Runoff — Continental Comparisons of Annual Flows and
entific problem, requiring nothing short of major Peak Discharges. Catena-Verlag: Cremlingen-Destedt.
breakthroughs in its theoretical foundations. On National Research Council. 1991. Opportunities in the Hydrologic
the other hand, PUB addresses an urgent practical Sciences. National Academy Press: Washington, DC.

Copyright  2003 John Wiley & Sons, Ltd. 3169 Hydrol. Process. 17, 3163–3170 (2003)
 M. SIVAPALAN

Reggiani P, Hassanizadeh SM, Sivapalan M, Gray WG. 1999. A Sivapalan M. 2003. Process complexity at hillslope scale, process
unifying framework for watershed thermodynamics. Constitutive simplicity at the watershed scale: is there a connection? Hydro-
relationships. Advances in Water Resources 23(1): 15–39. logical Processes 17: 1037– 1041, doi: 10.1002/hyp.5109.
Reggiani P, Sivapalan M, Hassanizadeh SM. 2000. Conservation Sivapalan M, Blöschl G, Zhang L, Vertessy R. 2003. Downward
equations governing hillslope responses: exploring the physi- approach to hydrologic prediction. Hydrological Processes 17:
cal basis of water balance. Water Resources Research 36(7): 2101– 2111, doi: 10.1002/hyp.1425.
1845– 1864.
Rundle JB, Klein W, Tiampo K, Gross S. 2000. Linear pattern
dynamics in nonlinear threshold systems. Physical Review E
61(3): 2418– 2431.

Copyright  2003 John Wiley & Sons, Ltd. 3170 Hydrol. Process. 17, 3163–3170 (2003)

View publication stats