Andrew Frost

Joint probability and design storms at the crossroads

Abstract The joint probability problem inherent in flood estimation is complex. Although the design storm approach has a long tradition it lacks the fundamental rigour of joint probability analysis. The use of average values for random inputs other than rainfall intensity and duration can be justified from a joint probability perspective provided variations in the input affect the peak flow density in a linear fashion. However, the assignment of the average value for initial conditions is problematic. A case study involving a detention basin demonstrates large biases arising from mis-specification of initial conditions in volume-sensitive systems. It is suggested that the current revision of ARR needs to articulate the shortcomings of the design storm approach, identify calibration strategies that ensure closure and give guidance about its reliability in different applications. Looking to the future, ARR needs to move towards event and total joint probability approaches that are underpinned by a rigorous joint probability framework. Continuous simulation is emerging as a practical tool and remains the most rigorous tool available. Event joint probability methods based on Monte Carlo simulation are computationally less demanding but require specification of the probability distribution of initial conditions. Stochastic rainfall models are on the verge of practical application to service Monte Carlo methods.

Comparison of Statistical Downscaling Techniques for Multisite Daily Rainfall Conditioned on Atmospheric Variables for the Sydney Region

Abstract Predictions of rainfall spatial and temporal variability (including climate change effects) on a catchment basis are urgently required by water resource planners within Australia. Large spatial scale predictions of (typically 300 to 500 km grids) global scale climate scenarios output by General Circulation Models (GCMs) are inadequate for such use as they do not capture the large degree of spatial variability over smaller distances, which is inherent in rainfall. Multisite daily rainfall - a common requirement within many hydrological models - is a required input for modelling complex multi-catchment systems, as small scale spatial variability due to factors such as topography has a large bearing on how much rainfall falls in a given area. Statistical downscaling is a technique that can produce such fine spatial scale rainfall pattern predictions conditional on the larger scale climate scenarios output by a GCM. The GLIMCLIM (Generalised Linear Model for daily Climate time series) software package (Chandler, 2002) has been used to analyse and simulate spatial daily rainfall given natural climate variability influences in the UK, and further to predict the influence of various future climate scenarios on regional rainfall by downscaling larger spatial scale GCM simulations. This paper describes the comparison of this method to the non-parametric, non-homogeneous hidden Markov model - kernel probability density estimation (NNHMM-KDE) downscaling technique of Mehrotra & Sharma (2006), a method which has found application in Australia previously.

The Impact of Rainwater Tanks in the Upper Parramatta River Catchment

Abstract This study investigates the extent to which rainwater tanks reduce the amount of on-site stormwater detention (OSD) storage required to satisfy the Upper Parramatta River Catchment Trust’s (UPRCT’s) OSD policy. In view of the limitations of the design storm approach, a continuous simulation approach was adopted. The DRIP stochastic rainfall model was linked with an allotment water balance model to evaluate different allotment scenarios using a 1000-year synthetic pluviograph record. The DRIP model was calibrated to a 53-year pluviograph located at Ryde. Comparison with statistics not used in calibration showed that DRIP performed satisfactorily. In particular, good agreement with observed intensity-frequency-duration (IFD) curves was obtained, whereas AR&R IFD curves consistently underestimated the observed IFDs. Scenarios involving combinations of OSD, using 10kL rainwater tanks with 0 and 5 kL of detention storage were examined. For allotments with single dwellings between 50 to 70% of the tank volume can be counted towards the allotment’s OSD volume. For a townhouse development this percentage varied between 36% and 53%. Rainwater tanks used in the single dwelling and townhouse scenarios are expected to reduce mains water consumption by 39% – 30% and 32% – 27% respectively. The variation depends on the number of occupants and the amount of tank airspace reserved for detention storage and the fraction of allotment drained by the rainwater tank(s).

Incorporating Long-term Climate Variability into a Short-timescale Rainfall Model Using a Hidden State Markov Model

Abstract Inter-annual persistence, a characteristic feature of Australian hydroclimatological series, is often difficult for event-based rainfall models to reproduce. This persistence can lead to poor reproduction of variables important in design such as Intensity-Frequency-Duration (IFD) curves and drought risk. This paper outlines the conditioning of a short-timescale rainfall model DRIP on the output series of a two-state Hidden Markov Model (HSM). The HSM model assumes that the rainfall in a particular year is in either a wet state or a dry state, with different rainfall distributions depending on the state. The inclusion of HSM into DRIP has produced a model that has improved capability to adequately reproduce rainfall characteristics including persistence for a variety of sites throughout Australia for timescales ranging from six minutes up to a year.