Geography Reference
In-Depth Information
Figure 12.5. Dependence of cross-validation performance of runoff prediction methods in ungauged basins on (a) aridity index and (b) mean air
temperature (Level 1 and Level 2 assessments from Chapters 5
-
10).
things, the comparative assessment aimed to understand
whether general patterns exist beyond individual case stud-
ies and how well we can predict runoff in ungauged basins.
Understanding the controls on performance will provide an
opportunity to generalise the conclusions drawn from indi-
vidual studies. Three climate-related controls
to predict in arid regions. Overall, the consistency between
the L1 and L2 assessments suggests that there exists a
strong and consistent control of aridity on all runoff signa-
tures, and that they can be generalised beyond the studies
examined in this topic.
Air temperature was expressed here as the long-term
mean, spatially averaged across the catchment. In cold
regions it is an indicator of the role of snow processes,
which will, again, affect all runoff signatures. Of course,
being strongly related to aridity, air temperature is not a
fully independent variable. Figure 12.5 suggests perform-
ance can decrease with increasing temperature, but it
depends on the signature examined. It is strongest for low
flows, floods and runoff hydrographs. It appears that the
existence of snow in a catchment leads to more regularity
in runoff, thereby improving runoff predictions. This is
particularly the case for hydrographs, as the model effi-
ciency measure indicates how well temporal patterns of
runoff are represented. This is also the case for floods. The
flood statistics indicate that snowmelt-driven floods tend to
be more predictable than other flood types. Similarly, for
low flows in cold places, winter low flows related to snow
storage are more predictable than summer low flows
related to the interplay of precipitation and evaporation.
For annual and seasonal runoff, the dependence is much
less pronounced due to the averaging over longer time
periods.
Topographic elevation, averaged over the catchment, is
a composite indicator that reflects a range of processes
related to elevation, such as long-term precipitation, soil
moisture availability and air temperature. In some environ-
ments there will also be a relationship between elevation
and aridity, and elevation and snow processes. The L2
assessments in Chapters 6
the aridity
index, air temperature and topographic elevation
-
-
were
examined. The figures in Chapters 5
10 were examined
visually, and the dependence of runoff prediction perform-
ance on climate and catchment characteristics was classi-
fied as (either positive or negative) strong dependence,
medium dependence, weak dependence or no dependence.
Figure 12.5 summarises the dependencies for aridity and
air temperature.
The aridity index (long-term ratio of potential evapor-
ation and precipitation, averaged across the catchment) is
an indicator of the relative availability of energy and water
affecting the water balance (and therefore all runoff signa-
tures). This dependence is clearly seen in Figure 12.5a . For
all signatures, regionalisation performance decreases with
increasing aridity index; this is true for both the L1 and L2
assessments. The dependence of performance on aridity is
related to a number of factors. Most importantly, runoff
processes tend to become more non-linear with increasing
aridity (Atkinson et al., 2002 ; Farmer et al., 2003 ; Harman
et al., 2011a ). Consequently, runoff processes in arid cli-
mates tend to be more spatially heterogeneous than in
humid or cold climates. Similarly, the temporal dynamics
of runoff tend to be more episodic in arid climates. The
relatively larger space-time variability of runoff then
results in lower predictability in ungauged basins. Low
flows have the strongest effect in the L2 assessment. This
is because the low end of the runoff spectrum is very
sensitive to climate, and low flows are particularly difficult
-
-
10 indicate that the effect of
Search WWH ::




Custom Search