Abstract

DIAZ, Teófilo; PUIG, Pedro; DAMILANO, Gabriela  and  MERELES, Juan. Application of models for time series to an ionic component (total alkalinity) indicative of water quality in the yacyreta reservoir. Rep. cient. FACEN [online]. 2018, vol.9, n.1, pp.9-23. ISSN 2222-145X.  https://doi.org/10.18004/rcfacen.2018.9.1.9.

The Yacyreta reservoir provides electric energy to Paraguay and Argentina on the Parana river. It is important to verify the impact that the formation of the reservoir has generated on the quality of the water evaluating its ionic components. This assignment consists in verifying the changes occured in one of the ionic components, Total Alkalinity, in space-time, with models for time series, during the period among February 2001 and November 2010. The specific aims were: to analyze and interpret the data and in a descriptive way, to build a univariate model that describes the behavior, to select the best model and predict the future values. Data corresponding to three stations were used, from the EBY/FACEN´s water quality monitoring program. New series that correspond to the differences among points were created with the original data of sampling. The Box-Jenkins model and the principle of parsimony were used for the choosing of models. The original series and their transformed was analyzed through ARIMA stationary models (SARIMA(p,d,q)(P,D,Q)s=12) The dates present absent values that occupy only 2.5% of the total group and that are not excluded according the procedure established by Vo-Day (1978) for ARIMA models with incomplete data. Both the original series and theier transforms were analysed and comparized those models with lower AIC. The best univariate modelof the original series resulted to be m =SARIMA (1,1,1)(0,1,1) and for the best univariate model for the transformed series data was m =sarima(2,0,2)(0,1,1). The original series presents similar behaviors, with a slight constant linear tendency with constant pendient, with annual stationarity and constant variance. The transformed series present identical traits, but without the tendency. There are no spatial variations, but there are temporal variations for the original series; observing the values are increase with time. The transformed series present a similar behavior althought without tendency. Predictions are feasible for original series up to 20 months and for transformed series up to 12 months later. Using the values from December 2010 to August 2015, both for the original and for the transformed series, it was found that the selected models are reliable, since most of the data fell within the prediction intervals.

Keywords : Time Series; Ionic Components (Total Alkalinity), Water Quality; Yacyreta reservoir.

        · abstract in Spanish     · text in Spanish     · Spanish ( pdf )