Authors: A.A. Mamoon, A. Rahman, and H. Qasem
Abstract: In stormwater system design, flood insurance studies and flood protection works, hydrological models are adopted to estimate design flows. Design flows are referred to as a runoff discharge associated with a given average recurrence interval (ARI) or annual exceedance probability (AEP). These models require design rainfall as the most important input, among other inputs such as catchment characteristics representing runoff routing behaviour and losses. The design rainfall, often known as intensity-duration-frequency (IDF) data, is generally derived using a regional frequency analysis approach based on a group of rainfall stations that form a homogeneous region. In Australia, design rainfall is known as intensity-frequency-duration (IFD) data. A large degree of uncertainty is associated with IDF data, which often is not quantified and considered in the majority of hydrologic modelling applications.
This paper presents a modelling framework to quantify uncertainty in design rainfall for Qatar due to uncertainties arising from limited data length and parameters of the adopted probability distribution model. Qatar is situated in an arid region, which has limited rainfall data in terms of the number of stations, data resolution (e.g. only daily rainfall data is available for most of the stations) and record length of the available data. The proposed modelling framework accounts for the uncertainty in the rainfall data using a Monte Carlo simulation technique, where a multivariate normal distribution is adopted to account for the uncertainty in the parameters of the log Pearson Type 3 (LP3) distribution. A bootstrapping method is adopted to estimate the mean and standard error values and the correlations among the three parameters of the LP3 distribution to define the multivariate normal distribution. A total of 10,000 simulations are carried out to develop the 90% confidence intervals for the 24-hour duration rainfall quantile. It has been found that uncertainty in IDF curves is quite high; to reduce the uncertainty band in estimated rainfall quantiles, a higher record length is needed, which, however, is not currently available in the Qatar region.
The proposed modelling framework is in the developmental stage, which is applied in this paper to a single station and for one rainfall duration (24-hour). The proposed method is being enhanced by adding other sources of uncertainty in design rainfall estimation e.g. uncertainty due to data quality and climate change. Furthermore, other rainfall durations from a large number of stations will be considered, which will enable better quantification of the uncertainty in the design rainfalls in Qatar.
Keywords: Arid regions, floods, IDF, Monte Carlo simulation, Qatar, rainfalls
21st International Congress on Modelling and Simulation, MODSIM 2015, Gold Coast, Australia, 29 Nov to 4 Dec 2015 www.mssanz.org.au/modsim2015