Geoscience Reference
In-Depth Information
averaging log hydraulic conductivity values at smaller scales, had little effect on runoff amounts and
timing; geometric simplification, by using larger slope elements, had a somewhat greater effect but was
still not significant; the effects were significant when predicting sediment yield from the catchment.
The most recent application of KINEROS to Walnut Gulch, by Yatheendradas et al. (2008), has been
concerned with parameter estimation and uncertainty estimation by fitting observed discharge data using
the generalised likelihood uncertainty estimation (GLUE) methodology (see Section 7.10). Yatheendradas
et al. were concerned with flash flood forecasting using radar rainfall estimates as inputs. It reflects some
of the difficulties of doing rainfall-runoff modelling in this kind of environment, in that the derived
distributions of the most sensitive parameters were shown to vary with event. This could be for a variety
of reasons, including lack of knowledge of the true pattern of antecedent conditions and patterns of rainfall
for each event, as well as the limitations of representing the response as an infiltration excess phenomenon
and overland flow as a sheet flow. They suggest that it might be better to use data assimilation during an
event to adapt parameter values as the event proceeds, but in the case of flash floods this might reduce
the time available to issue warnings to the population at risk. More details about the KINEROS model
can be found in Appendix A.
Other models that have been used at Walnut Gulch include the Hortonian Infiltration and Runoff/on
(HIRO 2) model, which allows for spatial heterogeneity in soil parameters (Meng et al. , 2008), and
THALES (Grayson et al. , 1992a), which includes both infiltration excess and subsurface runoff compo-
nents. Grayson et al. showed that the goodness of fit of the model and the runoff generation mechanisms
simulated were both very sensitive to the parameters of the model, which were difficult to estimate from
the information available. By calibrating parameters, catchment outflows could be adequately predicted
using both Hortonian infiltration excess and partial area runoff generation mechanisms. They suggest
that the difficulties of validating such models are acute (see also the discussion in Grayson et al. , 1992b).
Finally, Houser et al. (1998) have applied the TOPLATS variant of the simplified distributed model
TOPMODEL (see Chapter 6) to Walnut Gulch. This study is included here because it is one of the
few studies in rainfall-runoff modelling that has attempted to include measurements of a distributed
variable within a data assimilation or updating framework. Data assimilation is well established in other
distributed modelling fields, such as numerical weather forecasting, but has not been widely used in
distributed hydrological modelling. In this study, the measured variable was surface soil moisture which
was available from the Push Broom Passive Microwave Radiometer (PBMR) carried on an aircraft
platform on six days during the MONSOON90 intensive field campaign, including a dry initial condition
and the dry down period following a rainstorm of 50 mm. The resulting PBMR soil moisture images
showed a very strong correlation with the pattern of rainfall volume interpolated for this storm, resulting
in a strong spatial correlation. There are a number of difficulties in trying to make use of such data
including the conversion from PBMR brightness temperature to surface soil moisture, the dependence
of predicted surface soil moisture on numerous model parameters (the model has some 35 soil and
vegetation parameters in all) and the choice of a method for updating the model given the PBMR images
over only part of the catchment area. The study compared different methods of varying complexity (and
computational burden), suggesting that there is a trade-off between the complexity of the method used
and the ability to make use of all the data available. The study confirmed the importance of the rainfall
forcing on the hydrologic response, not only for runoff generation but also for the spatial pattern of
evapotranspiration and sensible heat fluxes back to the atmosphere.
5.7 Case Study: Modelling the R-5 Catchment at Chickasha, Oklahoma
The R-5 catchment at Chickasha, Oklahoma (9.6 ha) has been the subject of a series of modelling
papers using versions of the Quasi-Physically Based Rainfall-Runoff Model (QPBRRM) originally
developed by Engman (1974) and then later the Integrated Hydrology Model (InHM) (VanderKwaak
Search WWH ::




Custom Search