Water level forecasting and navigability conditions of the Tapajós River - Amazon - Brazil

Abstract

In the Amazon, hydrological modelling is of great importance to the safety and operation of inland waterway transport. In this paper, a stochastic model named ARIMA (Autoregressive Integrated Moving Average) was used to forecast water levels and seaworthiness 24 months in advance in the Tapajos River. Additionally, a time series analysis aimed to model and predicting water levels in the Tapajos river basin was completed. During validation, the model had an average coefficient of Nash and Sutcliffe R² of 0.95 and an average RMSE of 0.06. Drought forecast for the model predicted depth gains of 30 cm and a load capacity gain of 2,800 tons per trip. In the analysis of the two seasonal periods, the model predicted annual savings in freight cost of U.S. $350,000.00. The performance of the model allows us to conclude that ARIMA models should be used to forecast water levels and navigation conditions in the Amazon.