Investigation of Different Base flow Separation Methods ‎ Using Flow Duration Indices, case study: Khazar region

Objectives and Background
Extraction of the base flow index (BFI) has been one of the most important issues in hydrology and plays a key role in the optimal water resources management .Development of the base flow analysis from the hydrograph has a long-term history which date back to the theoretical and experimental research of Horton, (5). In this regard, numerous and useful literature review is presented to show progress, (14, 25, 27). Different methods of hydrograph separation were examined using correlation analysis by Nathan and McMahon (1990).they suggested a simple one parameter digital filter with 0.925 coefficients (18). Different graphical, digital and tracing methods were studied in the catchments of the lowlands of the Netherlands by Gonzalez et al (2009). Results showed that Eckhart algorithm is a good method in the lowlands (12). Literature review showed that In relation to the comparison of different base flow separation methods on the basis of different climates, enough research has not been done in the country. The aim of this study is investigation of different base flow separation methods using flow duration curve indices In order to select the most appropriate method.

Materials and methods
In this research, the base flow and BFI were extracted using daily flows of twenty gauging station located in the Khazar region. Seven algorithms were selected, four based on recursive digital filter (One-parameter algorithm, Boughton two parameter algorithm, BFLOW algorithm, Chapman algorithm) and the others on simple smoothing (Local minimum, Fixed interval, Sliding interval).Filter coefficients selected in the range required for each method according to the best response to changes in the coefficients and using the graphical display of Hydro Office software. Flow duration curve of all the stations traced and their indices were extracted. Results were evaluated using different statistical methods (Mean absolute error, Standard deviation, and Standard error of estimate, Coefficient of determination and Correlation Matrix) with respect to flow duration indices. Since the parameters extracted from the flow duration curve is calculated using the measured and observational data and this index known as an indicator of the groundwater contribution to the steam flow. Therefore, as representative of the actual flow were used for comparison of hydrograph separation models.

Results
The results showed that the average annual of BFI, fluctuate between 0.56 and 0.91, which related to one parameter recursive digital filter and B-Flow algorithms. This value represents a high ground water contribution in stream flow in the study area. Correlation matrix between different set of BFI values and flow duration curve indices showed high value of coefficient of determination between B-Flow method and Q90/Q50 and the lowest were identified between Chapman and Fixed intervals. Comparison of Standard division of different methods showed that B-FLOW has the lowest standard deviation. In addition, this method has the highest correlation with the flow duration curve index. On the basis of these two criteria’s, B-Flow recursive digital filter compared to other, was diagnosed as a better option. According to the Mean absolute error, and Standard error of estimate, one parameter method had the best estimate.

Conclusion
The overall conclusion obtained using different statistical tests were found to yield suitable estimation of base flow index by one parameter algorithm because of low Mean absolute error, the lowest Standard error of estimate and 0.71 coefficient of determination with flow duration curve index. This method in the absence of tracing methods, were diagnosed as the most appropriate method in the study area.