Abstract
Rainfall is a natural process, which has a high degree of variability in both space and
time. Information on the spatial and temporal variability of rainfall plays an important
role in the process of surface runoff generation. Hence it is important for a variety of
applications in hydrology and water resources management. The spatial variability of
rainfall can be substantial even for very small catchments and an important factor in the
reliability of rainfall-runoff simulations. Catchments in urban areas usually are small,
and the management problems often require the numerical simulation of catchment
processes and hence the need to consider the spatial and temporal variability of rainfall.
A need exists, therefore, to analyse the sensitivity of rainfall-runoff behaviour of
catchment modelling systems (CMS) to imperfect knowledge of rainfall input, in order
to judge whether or not they are reliable and robust, especially if they are to be used for
operational purposes.
Development of a methodology for identification of storm events according to the
degree of heterogeneity in space and time and thence development of a detailed spatial
and temporal rainfall model within a hydroinformatic environment utilising real-time
data has been the focus of this project. The improvement in runoff prediction accuracy
and hence the importance of the rainfall input model in runoff prediction is then
demonstrated through the application of a CMS for differing variability of real storm
events to catchments with differing orders of scale.
The study identified both spatial and temporal semi-variograms, which were produced
by plotting the semi-variance of gauge records in space and time against distance and
time respectively. These semi-variograms were utilised in introducing estimators to
measure the degree of heterogeneity of each individual storm events in their space and
time scale. Also, the proposed estimators use ground based gauge records of the real
storm events and do not rely on delicate meteorological interpretations. As the results of
the investigation on the developed semi-variogram approach, real storm events were
categorised as being High Spatial-High Temporal (HS-HT); High Spatial-Low
Temporal; (HS-LT); Low Spatial-High Temporal (LS-HT); and Low Spatial-Low
Temporal variability.A comparatively detailed rainfall distribution model in space and time was developed
within the Geographical Information Systems (GIS). The enhanced rainfall
representation in both space and time scale is made feasible in the study by the aid of
the powerful spatial analytic capability of GIS. The basis of this rainfall model is an
extension of the rainfall model developed by Luk and Ball (1998) through a temporal
discretisation of the storm event. From this model, improved estimates of the spatially
distributed with smaller time steps hyetographs suited for especially the urban
catchments could be obtained.
The importance of the detailed space-time rainfall model in improving the robustness of
runoff prediction of CMS was investigated by comparing error parameters for
predictions from CMS using alternate rainfall models, for various degrees of spatiotemporal
heterogeneity events. Also it is appropriate to investigate whether the degree
of this improvement to be dependent on the variability of the storm event which is
assessed by the adopted semi-variogram approach. From the investigations made, it was
found that the spline surface rainfall model, which considered the spatial and temporal
variability of the rainfall in greater detail than the Thiessen rainfall model resulted in
predicted hydrographs that more closely duplicated the recorded hydrograph for the
same parameter set. The degree of this improvement in the predicted hydrograph was
found to be dependent on the spatial and temporal variability of the storm event as
measured by the proposed semi-variogram approach for assessing this feature of a storm
event.
The analysis is based on forty real events recorded from the Centennial Park Catchment
(1.3km2) and the Upper Parramatta River Catchment (110km2) in Sydney, Australia.
These two case study catchments were selected to ensure that catchment scale effects
were incorporated in the conclusions developed during the study.