Abstract:Water demand forecasts are needed for the design, operation and management of urban water supply systems. In this study, the relative performance of regression, time series analysis and artificial neural network (ANN) models are investigated for short-term peak water demand forecasting. The significance of climatic variables (rainfall and maximum air temperature, in addition to past water demand) on water demand management is also investigated.Numerical analysis was performed on data from the city of Ottawa, Ontario, Canada. The existing water supply infrastructure will not be able to meet the demand for projected population growth; thus, a study is needed to determine the effect of peak water demand management on the sizing and staging of facilities for developing an expansion strategy. Three different ANNs and regression models and seven time-series models have been developed and compared. The ANN models consistently outperformed the regression and time-series models developed in this study. It has been found that water demand on a weekly basis is more significantly correlated with the rainfall amount than the occurrence of rainfall.

Abstract:Information on intensity-duration-frequency of rainfall is commonly required for a variety of hydrologic applications. In this study, trends are estimated for different durations of annual extreme rainfall using the regional average Mann-Kendall S trend test. The method of L-moments was employed to delineate homogeneous regions. The trend test was modified to account for observed autocorrelation, and a bootstrap methodology was used to account for the observed spatial correlation.Numerical analysis was performed on 44 rainfall stations from the province of Ontario, Canada, for a 20 year time frame. This was done using data from homogeneous regions established using the L-moments procedure for the annual maximum observations for the following durations: 5, 10, 15 and 30 min, and 1, 2, 6 and 12 h. Depending on different rainfall durations, four or five homogeneous regions were delineated. Based on a 5% significance level, approximately 23% of the regions tested had a significant trend, predominantly for short-duration storms. Serial dependency was observed in 2Ð3% of data sets and spatial correlation was found in 18% of the regions. The presence of serial and spatial correlation had a significant impact on trend determination.

No abstract

Abstract:The detection and estimation of trends in the presence of noise, periodicities, or discontinuous patterns is important in hydrology and climate research studies. The basic idea of currently available trend estimation techniques (tests) is that the trends should be smooth and monotonic; however, hydro-climatologic variables contain multiple signals, and have segments of increasing and decreasing trends. As a result, estimating trends in time series is an essential but arcane art and it is therefore important to continue developing the theory and practice of trend analysis.In this paper, a new technique is proposed based on the continuous wavelet transform (CWT). CWT permits the transformation of observed time series into wavelet coefficients according to time and scale simultaneously. These coefficients can be used to detect and estimate trends or to reconstruct signals that are of interest. The proposed CWT method was first tested on computer-generated data exhibiting both periodic and noise components. It was then applied to observed monthly minimum streamflow observations extracted from the Reference Hydrometric Basin Network (RHBN) for five different eco-zones in Canada.It was concluded that the proposed wavelet transform (WT) based method provides a very flexible and accurate tool for detecting and estimating complicated signals. The results from monthly minimum observations indicate that short period fluctuations are decreasing, while multi-annual variability is increasing in Canada. And finally, a persistent ¾55-year signal is well correlated with the Pacific Decadal Oscillation in all records, which indicates that trends are not controlled by a single factor.

A currently used approach to flood frequency analysis is based on the concept of parametric statistical inference. In this analysis the assumption is made that the distribution function describing flood data is known, for example, a log-Pearson type III distribution. However, such an assumption is not always justified and often leads to other difficulties; it could also result in considerable variability in the estimation of design floods. A new method is developed in this article based on the nonparametric procedure for estimating probability distribution function. The results indicate that design floods computed from the different assumed distribution and from the nonparametric method provide comparable results. However, the nonparametric method is a viable alternative with the advantage of not requiring a distributional assumption, and has the ability of estimating multimodal distributions. INTRODUCTION Determination of flood frequencies and other hydrologic events are based on statistical probability distributions estimated from sample data. In hydrology, the log-normal (two and three parameters) and the log-Pearson type III (recommended by the U.S. Water Resources Council [WRC, 1982] as a base method) have been extensively studied. Other distributions, such as Wakabe and two-component distributions have been recently introduced. The contending distributions that fit the observed data satisfactorily, usually differ significantly in the extreme tail of the distribution. Some of the situations that cause problems with parametric methods are selection of a particular distribution, parameter estimation (especially for skewed data), and most of the commonly used distributions are unimodal [Greis, 1983]. Recent developments in statistical theory have provided a new approach of alternative nonparametric density estimates. Such a method does not require assumption of any functional form of density; it is particularly suited for multimodal distributions and therefore, can be very attractive in hydrologic applications, and is well worth considering. Yakowitz [1985] recognized the need and possibility of applying the nonparametric method in hydrology. Investigations by Unny et al. [1981] also considered the nonparametric approach but in the more general context of pattern recognition as it pertains to time series analysis.To the author's knowledge, the nonparametric density estimation method has not been applied to the estimation of flood frequencies. This paper is an attempt to introduce this method into hydrology.The statistical literature in this area is available elsewhere [Tapia and Thompson, 1978]; thus only condensed relevant material will be presented.