Fluviometric Index Prediction Using a Neuro-Fuzzy Approach

This work presents a proposal for monitoring and forecasting the level of the Amazon River in the city of Santar m, using an ANFIS approach and historical data on the annual fluviometric cycle. The data was collected in the cities of Santar m, Manaus and Itaituba, and was filtered to the period from 01/03/03 to 28/02/17. The data was analysed in order to establish threshold values for alerts of emerging situations for monitoring the river level, comprising the times of the water volumes: drought, drought-normal, normal, normal-full and flood. The historical series can be considered reliable, as it is possible to clearly identify the annual hydrological cycles of the rivers. Predictive solutions using statistical and computational techniques are capable of automating such forecasts with relatively low rates of river level error. With this in mind, it is considered that the proposed solution can be used to guide actions that minimise the disruption caused to the city of Santar m in extreme situations of the Amazon River's water volumes. The proposed approach has also been shown to be applicable in other environments with hydrological cycles and, therefore, possible global applicability.