Prediction of low flows in ungauged basins - Runoff Prediction in Ungauged Basins: Synthesis across Processes, Places and Scales - page 180

Figure 8.13. Regional low flow distributions
in Costa Rica estimated by the derived
distribution approach. The low flow
distributions relate to annual minima scaled
by a reference low flow. After Pacheco et al.
( 2006 ).

1.0

0.8

South
Pacific

0.6

Central
Pacific

North
Carribean

South
Carribean

0.4

Central
Valley

North
Pacific A

0.2

North
Pacific B

0.0

1.02 2 5 10 20 50 100

Return period (yrs)

apply to low flow data from humid tropical conditions.
They assumed dry spells to be exponentially distributed
and assumed non-linear low flow recession behaviour.
Their model allowed them to estimate physically interpret-
able parameters from the low flow distributions. These
include the average length of dry spells and the intensity
of events. They then grouped the gauged catchments for all
of Costa Rica according to these parameters and obtained
derived low flow distributions for each of these groups
from representative parameters. The derived distributions
are shown in Figure 8.13 . The distributions suggest that, in
Costa Rica, the Northern Pacific region is the driest region
and that the rivers on the Caribbean slopes have relatively
high low flows, even during the dry period.

(e.g., grid cells; Engeland et al., 2001 ). Some studies that
simultaneously address the estimation of regional param-
eters and modelling uncertainties for distributed models
apply multi-objective and Bayesian methods (Engeland
et al., 2006 ). A case study in south-western Norway (Enge-
land and Hisdal, 2009 ) suggested that a regional regression
may give better estimators of low flow characteristics
in ungauged catchments than a distributed hydrological
model.
Coupled groundwater surface water models that simu-

late water flow driven by potential gradients have been
tested by van Lanen et al.( 1997 ) for the purpose of
predicting low flows in ungauged basins. They emphasised
the strength of the simpler conceptual models over the
more complex coupled models and suggested that very
detailed information on the subsurface is needed for low
flow characteristics to be represented well without calibra-
tion. Coupled groundwater

-

8.4.2 Continuous models
Runoff time series simulated by continuous rainfall

surface water models are usu-
ally calibrated with groundwater level data in addition to
runoff. The strength of such coupled models lies in their
ability to incorporate management scenarios into low flow
predictions, such as the impact of human influences on low
flows (e.g., abstractions, land use change, climate change)
(Querner et al., 1997 ), the impacts of drought mitigation
measures (Querner and van Lanen, 2001 ) and the value of
indicators for assessing the regional groundwater resources
that will impact the low flow regime (Henriksen et al.,
2008 ).

-

runoff
models in ungauged basins can also be used to estimate
low flow characteristics. Rainfall

-

runoff models are dealt
with in Chapter 10 , so this section focuses on the specifics
of low flows. Since low flow characteristics very much
depend on the subsurface characteristics of the catchment
related to long flow paths ( Chapter 4 ) rainfall

-

-

runoff
models for low flow estimation need to represent
these well. For conceptual models that are calibrated
against runoff data this is usually ensured by using the
logarithm of runoff instead of runoff itself in the calibra-
tion (e.g., Seibert, 2005 ). The model parameters are then
transferred from gauged to ungauged basins as discussed in
Chapter 10 . Examples are multiple regressions used for
lumped models to relate model parameters to catchment
characteristics (e.g., Abdulla and Lettenmaier, 1997 ; Xu,
1999 ) and regional calibration procedures that relate model
parameters to the characteristics of each modelling unit

8.4.3 Proxy data on low flow processes
Explorative field surveys can provide useful, qualitative
information about low flows and their temporal variability.
Climate data (e.g., from high-resolution global databases
with long time series of climate data, see Chapter 3 )