Geoscience Reference
In-Depth Information
compared regardless of changes in rainfall over time). McIntyre and Marshall (2010) have used a similar
approach for the Pontbren catchments in Wales to compare subcatchment characteristics for a shorter
data set.
Thus there are difficulties in detecting change in the analysis of historical data. The major problem
in predicting the effects of change in the future is that at least some of the model parameters should be
expected to change. This is evidently the case if there is a change of land use, but changes may also
be induced by a change in inputs as a result of climate change. It has been argued, for example, that
an increase in CO
2
concentrations in the atmosphere will cause a decrease in the density of stomata on
leaf surfaces because a plant can achieve the necessary exchanges of CO
2
for photosynthesis. The result
would be an increase in the effective canopy resistance, leading to smaller evapotranspiration losses as a
result of climate change. The argument is plausible and is supported by stomatal density measurements
for some plants in the historical collections at Kew Gardens, but it is not clear if this will be a general
response or a significant factor in the response of plants to climate change.
It has long been suggested that one of the most important reasons to develop a process-based dis-
tributed modelling capability is precisely that it might be easier to estimate such changes (Abbott
et al.
,
1986a; Ewen and Parkin, 1996; Dunn and Ferrier, 1999). It is also rare for the whole of a catchment
to change suddenly so that the changes in characteristics may be gradual and often quite local. Thus a
spatially distributed model can implement any changes in parameter values in their correct spatial con-
text, if we can be secure in estimating changes in the effective values of the parameters required by such
models. Some models include components that attempt to represent the growth of vegetation communi-
ties and their interactions with hydrological processes (e.g. RHESSys Band
et al.
, 1993; TOPOG-IRM
Dawes
et al.
, 1997; and MACAQUE Watson
et al.
, 1999) but with the consequent need to estimate even
more parameters.
The advantages of process-based models in this context are still widely supported and for good reasons
since the only alternative would seem to be to try to reason about potential changes in bulk catchment
scale parameters, such as the mean residence time of a transfer function, that necessarily integrate the
impacts of change on a variety of interacting processes. This would appear to be much more difficult
than reasoning about the impacts on the parameters of individual processes. However, this belief might
be a little naive since it has not yet been consistently demonstrated that the parameters of process-based
models can be estimated
a priori
and produce successful simulations under current conditions. It is
actually not clear that such models are always able to reproduce the behaviour of vegetation communities
at individual sites (see, for example, Mitchell
et al.
, 2009, 2011).
The one area where significant advances have been made in predicting catchment changes is in mod-
elling the impacts of urbanisation, driven by regulations that require that peak runoff rates from new
developments should not be any greater than the natural state (the concept of “sustainable urban drainage
systems” (SUDS). Although modern equivalents of the rational method for predicting peak flows are still
often used in such calculations, there are also sophisticated commercial packages that represent every im-
permeable area, drain, channel and detention basin in making the predictions needed (e.g. HydroWorks
from Wallingford Software, UK, and MOUSE from DHI, Denmark). Such packages can be used for
design purposes as well as simulating existing urban drainage systems. The representation of runoff gen-
eration in urban areas is still difficult, however, and Burges
et al.
(1998) argue that continued monitoring
of small catchments subject to change remains necessary to resolve some of the uncertainties associated
with the prediction of the hydrological impacts of development.
Catchment change is clearly an important issue. There is no doubt that land management can have
an important impact on runoff generation processes. There was much more on predicting different types
of catchment change in the first edition of this topic but, even after another decade, understanding the
nature and impacts of catchment change is still very much in its early stages, limited by the short length
of hydrological records which can make it difficult to detect change given the natural variability of
hydrological systems. Experimental work at the plot scale has shown that land management practices
