Geoscience Reference
In-Depth Information
and those in which the flows were heavily modified or regulated. These are types of catchment that,
elsewhere, have generally been excluded from regionalisation studies but which might be important in
specific practical applications.
These results emphasise the real difficulty in addressing the ungauged catchment problem in hydrology.
The final report of the PUB initiative has still to appear but, while considerable progress has been
made, it will certainly not declare that the problem has been solved either by improved hydrological
theorising or by common agreement on a regionalisation strategy. Approaches based on regionalisation
of hydrograph indices perhaps show the greatest promise, particularly as they provide ameans for allowing
for uncertainty in the models that might be appropriate for an ungauged site. However, it might well be
that the best near-term solution to the ungauged catchment problem is a short field campaign of discharge
measurements (see Section 10.11).
10.10 HRUs and LSPs as Models of Ungauged Basins
There is a quite different approach to model parameterisation in ungauged catchments that has arisen
out of the use of hydrological models as land surface parameterisations (LSPs) in global atmospheric
models. The requirement is that the whole of the land surface of the globe must have some form of
representation of the surface water and energy balances, albeit at the rather coarse resolution of global
or regional atmospheric models. This includes all those large parts of the global land surface that are
effectively ungauged at these scales. The parameters of the LSPs must therefore be derived in other ways,
depending on what information might be available.
In some cases, models are provided with default parameter sets that have been derived from calibration
exercises in some gauged catchments. This is the case for the VIC model of Box 2.2 (e.g. Abdulla and
Lettenmaier, 1997b; Li
et al.
, 2009) and the SWAT model of Box 6.2. The default parameters depend on
defining soil type and land management classes from other information, such as remote sensing imaging
(e.g. Lakshmi, 2004). LSPs are generally used in a way that reflects the fractions of different classes
(or “tiles”) in a larger atmospheric model grid square. There is, therefore, some commonality with the
hydrological response unit (HRU) models discussed in Section 6.6, albeit that the areas represented by
particular parameter sets as homogeneous tend to be much larger when used as LSPs. In fact, the only
recognition of spatial heterogeneity in hydrological responses recognised by operational LSPs is in terms
of these fractional tiles. Thus, the representation of surface hydrology in operational LSP models can be
considered totally inadequate (e.g. Beven, 2011).
More interesting is the question of whether this formof prior estimation of default model parameters can
be successful in reproducing the response of ungauged catchments at finer HRU scales, especially when
most such studies totally neglect the uncertainty inherent in estimating such parameters (see the discussion
of pedotransfer functions in Box 5.5). Effectively, this is similar to a classification approach, where the
classification is based on the soil and land use classes used to define the default parameters. Most studies
have shown that some form of local conditioning results in improved performance over using default
estimates of parameter values, even in ungauged catchments (Dornes
et al.
, 2008; Parada and Liang,
2010). Buytaert and Beven (2009) argue that this should be part of the learning process. They showed
how prior estimates of parameter values taken fromnearby catchments, with some (uncertain) speculation
about the impacts of land use change, could provide surprisingly good predictions but that uncertainties
could be significantly constrained as discharge observations became available for the new site.
10.11 Gauging the Ungauged Basin
One of the options in a learning process to improve the predictions for an ungauged site is always to
take some discharge measurements at the site of interest. With the availability of the new generation of
