Marabá-PA city, located in the Amazon region, Southeast of Pará State, suffer annually with flood events, caused by periodic increase of Tocantins River and vulnerability of the population living in risk areas. State and municipal civil defense plans and annually prepare teams for actions to defend the city. At this stage the monitoring and prediction of flood events are importants. Therefore, in order to reduce errors in hydrological forecasts for Marabá city, developed a stochastic model to predict the level of Tocantins River, based on the methodology of Box and Jenkins. Used data from daily levels observed in the hydrological stations of National Water Agency (ANA) at Marabá, Carolina and Conceição do Araguaia, from 01/12/2008 to 31/03/2011. We conducted the adjustment of three models (Mt, Nt e Yt), using different statistical applications: SAS and Gretl, using different interpretations of the behavior of the series to generate models equations. The main difference between the applications is that the SAS uses the model of transfer function modeling. There was a sort of variable water level, through the technique of quantiles for the period 1972 to 2011, examining only levels categorizations and MUCH ABOVE ABOVE normal. For analysis of socioeconomic impacts were used the data of the shares of Civil Defense State of Pará in the floods of 2009 and 2011. The results showed that the number of events filled with much higher levels than normal, generally, may be associated with La Niña events. Another important result: generated models represented well the level of the river for seven days (from 04/01/2011 to 04/07/2011). The multivariate model Nt (with small errors) represented the behavior of original series, underestimating the real values on days 3, 4 and 5 April 2011, with a maximum error of 0.28 on day 4. The univariate model (Yt) had good results in simulations with absolute errors of around 0.12 m. The model with the lowest absolute error (0.08 m) for the same period was the model Mt, application developed by SAS, who plays the original series as nonlinear and non stationary. Quantitative analysis of the impacts fluviometric, floods occurred in 2009 and 2011 at Marabá city, revealed that on average more than 4000 families suffer from these events, implicated in high financial costs. Therefore, concludes that the levels forecast models are important tools that the Civil Defense, uses in the planning and preparation of preventive actions for the city of Marabá.
|
