Outcomes of synthesis - Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales - page 378

Figure 12.11. Comparative hydrology based on
Darwinian concepts consists of exploiting the
differences between regions to learn from them.
(Left) Differences in the Budyko curve between
regions. (Right) Differences of flood frequency
scaling with area between regions.

interested in such questions for some time, and several
candidate principles have been explored in different con-
texts, most of which take the form of optimality principles.
Examples include minimum energy expenditure
(Rodriguez-Iturbe et al., 1992 ), ecological optimality
(Eagleson, 1982 ), vegetation optimality (Schymanski
et al., 2009 ), maximum entropy production (Kleidon and
Schymanski, 2008 ) and maximum energy dissipation
(Zehe et al., 2010 ). The search for the organising (Darwin-
ian) principles is especially crucial to dealing with change,
both climate change and human-induced land use and land
cover changes (Schaefli et al., 2011 ). Change predictions
must account for not only simple (short-term) process
changes but also more complex process interactions and
feedbacks, such as metabolic processes. Simple Newtonian
models will not be able to account for the likely trajectories
of co-evolutionary change unless we constrain the degrees
of freedom through some external principles. A search for
organising

on the concept of a super-population. It can lead to substan-
tial improvements in performance over site-specific
methods if the region is homogeneous in some sense. There
has been significant progress during the PUB decade in
regionalisation approaches that pool hydrologically mean-
ingful information in the form of runoff signatures from
neighbouring catchments to constrain parameter estimates
and reduce predictive uncertainty (see e.g., Yadav et al.,
2007 ; Merz and Blöschl, 2008a , b; Bulygina et al., 2009 ;
Wagener and Montanari, 2011 ). Ultimately, the focus of
such regionalisation efforts is improved prediction.
The synthesis of the Newtonian and Darwinian
approaches based on comparative hydrology that we advo-
cate here is different. Let us consider the Budyko curve
( Figure 12.11a ). A comparative hydrology approach would
focus on how the Budyko curve comes about and why it
differs between regions. By contrasting the Budyko curves
between different regions their differences can be recog-
nised (e.g., Wolock and McCabe, 1999 , Chapter 5 ) and
Newtonian approaches (such as process-based modelling
of the atmosphere

principles must therefore underpin the

Newtonian - Darwinian synthesis.

land interactions) may be
brought to bear to help understand these differences in
terms of co-evolutionary processes (e.g., drainage density,
vegetation patterns etc.). Another example is the scaling of
flood frequency with area ( Figure 12.11b ). A comparative
hydrology approach would focus on how the flood fre-
quency scaling comes about the way it does, and why it
differs between regions (e.g., rain-fed and snowmelt-fed
regimes). Again, by contrasting different regions the dif-
ferences in the scaling can be detected and Newtonian
approaches (such as derived flood frequency) may be
invoked to help understand these differences in terms of
co-evolutionary processes. Therefore,

- vegetation -

Moving from regionalisation to comparative hydrology
Regionalisation approaches to improve hydrological pre-
dictions using spatial data and concepts of similarity have
been in use for decades, and are illustrated in this topic
through hundreds of examples. The Darwinian approach
applied in this topic also uses spatial data and similarity
concepts, but the two approaches do this in a very different
way. Regionalisation exploits the similarity of catchments
within a region to improve predictions. In contrast, com-
parative hydrology exploits the differences between catch-
ments in different regions to develop generalised
understanding about the causes and controls that lead to
these differences.
Regionalisation aims to combine local (at site) informa-
tion and regional information to advance predictions.
A typical example of the combination of local and regional
information for the case of flood frequency analysis is the
empirical Bayes approach (Kuczera, 1982 ). The method
uses information from hydrologically similar basins to
improve upon inference at a particular basin and is based

the goal is to

develop generalised understanding.

Studies from the literature to illustrate synthesis
of Newtonian and Darwinian approaches
While the synthesis of Newtonian and Darwinian approaches
is a new paradigm, there are already a number of early
examples that have led to improved understanding of co-
evolutionary processes, and give us confidence that this is an