Floods kill more people and cause greater devastation than any other natural calamity. However, the severity of floods can be minimized if their occurrence is foretold. The early warning should allow concerned authorities to take necessary flood-fighting measures. Most of the previous works reported in literature on flood forecasting considered complicated models, either physical or conceptual, requiring intensive data and complex differential equations for their implementation. As field engineers are wary of using such models, there is a need for newer and simpler models capable of efficiently forecasting high discharges in rivers. In recent times, data-based forecasting models, e.g., statistical methods, neural networks (NNs) and wavelet techniques have become popular in hydrological applications due to their rapid development, minimum information requirements and ease of real-time implementation (Adamowski, 2008). However, their difficulty in handling transient data makes statistical methods unsuitable for time series analysis. NNs have the ability to learn complex and nonlinear relationships between inputs and outputs which conventional methods find difficult to establish (Sahay, 2011), yet, their structures are hard to determine. Moreover, the surface of the objective function of a neural network is non-convex and contains multiple local optima where the network solution can easily get trapped.
In the last decade, the wavelet transform has become a useful technique for analyzing variations, periodicities and trends in a time series (Smith et al., 1998; Labat et al., 2000 and 2005; Chou and Wang, 2002; Lu, 2002; Wang and Ding, 2003; Xingang et al., 2003; Coulibaly and Burn, 2004; Kucuk and Agiralioglu, 2006; Zhou et al., 2008; Wang et al., 2009; Kisi, 2009; and Rajaee et al., 2010). Labat et al. (2000) applied wavelet methods to model rainfall rates and runoffs measured at different sampling rates from daily to half-hourly. Smith et al. (1998) used discrete wavelet transform for quantifying stream flow variability, in which they classified stream flows into distinct hydroclimatic categories. Coulibaly and Burn (2004) used wavelet analysis to identify variability in annual flows in Canadian rivers. Labat (2005) reviewed the recent applications of wavelet techniques in the field of earth sciences and illustrated new wavelet analysis methods (combined multi-resolution-continuous wavelet analysis method and wavelet entropy) in the field of hydrology. Partal and Kucuk (2006) used Discrete wavelet transform (DWT) for determining possible trends in the annual precipitation in Turkey, in which they concluded that DWT clearly explained the trend structure of the time series. Zhou et al. (2008) proposed a wavelet predictor-corrector model for the simulation of the monthly discharge time series.
|
