ABSTRACT
Works such as Ledolter (1976, 1977) introduced the time-series analysis in hydrology. Salas et al. (1980) and Salas (1992) introduced key concepts and theories related to the modelling of time series in hydrology, with applications focused on engineering. According to Nourani et al. (2013), auto regressive integrated moving average (ARIMA) and seasonal ARIMA are widely used to predict hydrological time series. As a modelling technique, ARIMA has been useful in predicting hydro meteorological parameters (Boochabun et al., 2004, Chattopadhyay and Chattopadhyay, 2010 and Chattopadhyay et al., 2011). Pektas and Keren Cigizoglu (2013) used the ARIMA model to develop a more accurate and generalisable model to predict the monthly direct runoff coefficient time series. Narayanam et al. (2013) used the ARIMA to forecast rainfall in western India for the period of 2010-2030. Lohani et al. (2012) used auto regressive models and the fuzzy inference system to predict monthly flows. Wu and Chau (2010) used ARIMA models and neural networks for streamflow forecasting in different rivers basins in China. Birinci and Akay (2010) used ARIMA type models to predict rainfall, which in turn, are inputs into the artificial neural network models used to forecast daily streamflow. Koutroumanidis
1 INTRODUCTION
Fluvial navigation is the most important means of passenger and freight transport in the Amazon, linking communities and poles of production, as commercialisation and consumption are established along its vast waterways. Antagonistically, its economic dynamics, their operational peculiarities and the quantitative and qualitative information of activity are not well known and rarely systematised. By virtue of regional conditions, the vast Amazon River Basin is predominantly used as a means of access. Knowing the behaviour of the water levels and consequently knowing the depths of the Amazonian rivers is essential for satisfactory and safe navigation.
