Geoscience Reference
In-Depth Information
The PUB decade has caused a significant rethink about how the ungauged basin problem should be
addressed and the development of some new methodologies (Wagener and Montanari, 2011). They are
outlined in Section 10.6.
10.3 The MOPEX Initiative
A recent international experiment has attempted to search for generalities in regionalisation approaches
using a large database of long-term (10 years or more) catchment rainfall-runoff observations from
different countries (although mostly daily time step data were used in this case). The primary aim of
the model parameterisation experiment (MOPEX, see www.weather.gov/oh/mopex/mo welcome.htm)
is to develop techniques for the
a priori
estimation of the parameters used in land surface parame-
terisation schemes of atmospheric and hydrological models that can be applied widely in ungauged
basins. A good overview of the project is provided by Duan
et al.
(2006). Particular attention was
paid to having adequate precipitation data for the chosen basins, in terms of raingauge densities in
a catchment.
Within the MOPEX project a number of different rainfall-runoff and land surface parameterisation
models were applied to different groups of catchments. In general it was found that using parameter values
estimated
a priori
none of the models could reproduce the observed discharges for all the catchments
when treated as ungauged. Results were greatly improved if the model parameters were calibrated and
tested in prediction using a split sample approach, although in the study reported in Duan
et al.
(2006) no
uncertainty in the estimated parameter values was considered. This suggested that methods for the prior
estimation of parameter values could be greatly improved given more hydrologically relevant information
about catchment characteristics. It is worth noting that the performance of the rainfall-runoff models were
mostly much better than the models that had been developed only as land surface parameterisations, and
that the performance of a combined model output was generally better than the performance of any of
the individual models.
A number of other studies have used the MOPEX data set as a test bed for estimating the responses
of ungauged basins using different rainfall-runoff models and methodologies for estimating parameters
(Ao
et al.
, 2006; Huang and Liang, 2006).
10.4 Ways of Making Predictions in Ungauged Basins
The choice of a way of making predictions in ungauged catchments in part depends on the purpose of
a particular study. Estimates of discharges for ungauged catchments are required for a wide variety of
purposes, including flood defence design, land planning, water resources, ecological, hydropower and
other decision-relevant applications. Prediction of the impacts of future change also involve aspects of
the ungauged basin problem since, by definition, future responses under different inputs or different land
surface characteristics cannot be gauged (see also Chapter 8).
Regionalisation methods have been developed primarily for the estimation of the characteristics of
flood frequency distributions, for estimation of flow duration curves, and for estimation of the parameters
of hydrological models at ungauged sites. The first two are also, in essence, a question of estimating
parameter values: in both frequency estimation and flow duration curve estimation the parameters are
those of a frequency distribution. Recent work has started to look at the estimation of mean residence times
for water in catchments (see alsoChapter 11). There is also significant interest in estimating the parameters
of
land surface parameterisations
(LSPs) of atmospheric models. The land surface hydrology can have
an important effect on latent and sensible heat fluxes as a lower boundary condition for atmospheric
models and, as noted in Section 8.9, has generally been represented rather poorly from a hydrological
