Geoscience Reference
In-Depth Information
that was later used by John Hewlett in discussing fast subsurface stormflow in catchments where surface
runoff was observed to be rare (Beven, 2004c).
Analysis of residence times in catchments is fraught with difficulties of both observation and assump-
tions. This means that it has also proven rather difficult to make predictions that properly represent water
pathways and residence time distributions. This is partly because of the limitations of the available tracers
in hydrology. An ideal tracer would have a low background concentration, would be detectable in very
low concentrations, and would not be subject to sorption, fractionation, volatilization or other chemical
reactions during its residence time in the system so that it can be considered conservative in its behaviour.
Ideally also, the plot, hillslope or catchment under study would be water-tight so that the tracer is not lost to
pathways not being observed while input concentrations should be well defined and constant in space and
time. These characteristics are difficult to meet completely. Many of the commonly used tracers are either
not entirely conservative or, for the environmental tracers, have variations in space and time that make it
difficult to assess the true mass balance for the tracer (even if the hydrological system is really water-tight).
This means that, as in the case of modelling hydrographs, inferences are often made on the basis of
less than ideal observations and assumptions that are necessarily approximate. The nearest to the ideal
case is probably when artificial applications of either hydrogen or oxygen isotopes (tritium, deuterium
or
18
O) are made in short-term experiments (Rodhe
et al.
, 1996). This is because these isotopes can be
added as part of the water molecule and therefore should track the water pathways directly. They are
subject to fractionation due to freeze-thaw, evaporation and transpiration processes but, for short-term
experiments where snow and ice are not an issue, this should not be a significant factor. More important
is whether all the outputs of tracer are being measured or whether the tracer is retained in long-term
immobile
storage. Both result in not all the tracer that is input being measured, so that mass balance is
apparently not preserved. It is rather typical in such experiments (and more so with other tracers subject
to sorption or biogeochemical reactions) that only a low proportion of the input tracer can be accounted
for in the output observations. The tracer (or solute) is then said to be
non-conservative
, though this
might be just because not all the output fluxes or storage in the system have been detected, rather than
because of biogeochemical processes. Even with the isotopes of the water molecule, interpretation of
measurements may be difficult. It has been suggested by Brooks
et al.
(2010), for example, that the water
extracted from soil storage by trees for evapotranspiration may be different in isotopic concentration to
that in soil water discharge, even though no fractionation process would appear to be involved.
There have been relatively few comparative studies of using different tracers in the analysis of residence
times but it is known that different tracers might reveal different aspects of the catchment response. In
particular, the oxygen and deuterium isotopes cannot provide information about water ages over about
four years, whereas tritium can identify older waters (Stewart
et al.
, 2010). Some groundwater systems,
of course, contain waters that date back 25 000 years or more, having been replenished during the last
Pleistocene glaciation. The use of such waters for water supply is effectively mining of waters that are
not now being recharged. Other dating techniques (such as
14
C) can be used for such ages, but here we
are concerned much more with the short-term rainfall-runoff processes for which oxygen and deuterium
isotopes are useful. They are commonly assumed to be linearly related but might still provide somewhat
different types of information (e.g. Lyon
et al.
, 2009).
In the remainder of this chapter, we look at the use of mixing models to infer sources of runoff and
models for inferring distributions of
residence times
and
travel
(or
transit
)
times
in catchment systems
(see Section 11.9). A residence time distribution summarises the lengths of time that water molecules
have spent in the flow domain of interest (soil core, lysimeter, isolated plot, hillslope or catchment) but it
will be seen that the residence time distribution for an increment of input might be quite different from the
residence time distribution for an increment of output and both might be different from the residence time
distribution for water in storage in the catchment at any time (e.g. Rinaldo
et al.
, 2011). In the general
case we expect all of these distributions to vary over time as the catchment wets and dries. Thus, in
identifying either water sources or residence times, the reader should be aware of the strong assumptions
