This research aimed to evaluate nine (9) algorithmes for baseflow separation. The six (6) recursive digital filter (RDF) and three (3) graphical separation algoritms were evaluated in this study. Four watersheds located at the administratuve boundary of UPT PSDA Madiun were used. The methodology consists of (1) data inventory, (2) data processing, (3) calibration and validation, and (4) evaluating of models performance. The algorithms were calibrated using daily discharge data (from 1996 to 2015) for each year. The average values of the parameters obtained from calibration process were used to separate baseflow for the whole period record. Calibration watershed was determined by considering the completenes of recording discharge data (in this case: we use Nambangan sub-watershed). Validation process to others three (3) sub-watersheds conducted using parameter value from the calibrated watersheds. Then, RMSE, scatter plot, FDC and hydrograph were used to evaluate model performance during the dry season (July to September) both on calibrated and validated watersheds. The result shows that pincipally all of the nine (9) algorithms can be used to separate baseflow from the total flow. Howerver, Lyne & Hollick and EWMA algortimes show the best performance compared to others algorithms. This conclusion based on the calculation of RMSE and R Squared (r2). The average value of RMSE =0,036 and R2=0,8 for Lyne &Hollick. Similarly, RMSE = 0,041 and r2= 0,83 for EWMA filter.