Application of recursive digital filter (RDF) methods for baseflow separation: study at Brantas watershed

Abstract

Baseflow is an important component affecting the availability of water in the river during the dry season. Availability of water in the dry season is useful for water resources management. This research aims to test and to compare six recursive digital filters (RDF) methods for calculating baseflow and baseflow index. This research was conducted in Brantas Watershed. Two outlets (sub-watersheds) located at Kertosono and Ploso were used. Daily discharge from 1996 to 2015 of the two outlets above was used as main input for this study. While rainfall data were used to determine the calibration period. The sequence procedures of this research, consist of: (1) inventory of daily discharge and rainfall data, (2) data processing, (3) calibration, (4) validation, and (5) evaluation of models’ performances. Six (6) methods of baseflow separation based on recursive digital filters were evaluated. The calibration process was carried out for periods 1996 to 2005. The periods from July to September was assumed to be the peak of the dry season and then selected for calibration process. The parameter values were calibrated using the data from dry season for each year. Furthermore, the average value of parameters obtained from calibration period then used to separate baseflow in validation process (periods 2006 to 2015). The result of separation both in calibration and validation are then evaluated using root mean square error (RMSE), coefficient of determination (R²) and FDC. This research shows that the Lyne-Hollick and EWMA filters perform better than other methods. In Brantas Kertosono sub-watershed, the optimal parameter value for Lyne Hollick algoritmh (αly) = 0.995 dan for EWMA filter (αew) = 0.003 and in Brantas Ploso sub-watershed (αly ) = 0.99 dan (αew) = 0.003.