Development of genetic algorithm-based optimization module in WHAT system for hydrograph analysis and model application
Introduction
Significant areas have been converted into impervious areas with human induced development activities in recent years. This "urban sprawl" has been a dominant phenomenon in urbanized regions worldwide according to the U.S. EPA (2001) due to social and economic benefits from development. However, the negative impacts of "urban sprawl" on hydrology and water quality have been recognized only recently. Thus, many computer simulation models have been developed and utilized to assess the impacts of urban sprawl to assist in environment-friendly land use planning.
The Long-Term Hydrologic Assessment Tool (L-THIA) (Harbor, 1994, Bhaduri et al., 2001; Lim et al., 2001, Lim et al., 2006), Soil and Water Assessment Tool (SWAT) (Arnold et al., 1995), and Hydrological Simulation Program—Fortran (HSPF) (Bicknell et al., 1997) models have been frequently used for this purpose. The accuracies of these models should be validated prior to their application in land use planning. The water quality components of these models rely on their corresponding hydrologic components. Thus, researchers and modelers typically first calibrate and validate the hydrologic component of models. In most cases, these models simulate direct runoff and baseflow components separately; thus it would be desirable to calibrate and validate the direct runoff and baseflow modules separately.
For accurate model calibration and validation, the direct runoff and baseflow components from stream flow have to be first separated. There are numerous methods, called "hydrograph analysis" or “baseflow separation”, available to separate baseflow from measured stream flow hydrographs. The traditional hydrograph analysis methods are not very efficient because these subjective techniques do not provide consistent results. Thus, the U.S. Geological Survey (USGS) has developed and distributed an automated hydrograph analysis program called HYSEP (Sloto and Crouse, 1996). Digital filtering methods have also been used in baseflow separation because they are easy to use and provide consistent results (Lyne and Hollick, 1979, Chapman, 1987, Nathan and McMahon, 1990, Arnold et al., 1995, Arnold and Allen, 1999, Eckhardt, 2005). The BFLOW (Lyne and Hollick, 1979, Arnold and Allen, 1999) and the Eckhardt (Eckhardt, 2005) digital filters are widely used for hydrograph analysis. Although a new parameter was introduced in the Eckhardt digital filter to reflect local hydrogeological situations (BFImax) and representative values were proposed for various aquifers, the use of a BFImax value specific to local conditions is strongly recommended instead of using the proposed representative BFImax values.
Lim et al. (2005) developed the Web GIS-based Hydrograph Analysis Tool (WHAT) (https://engineering.purdue.edu/∼what) to provide fully automated functions for baseflow separation. Lim et al. (2005) compared the filtered baseflow for 50 gaging stations in Indiana, USA using the Eckhardt filter with a default BFImax value of 0.80 and the BFLOW filter. The Nash–Sutcliffe coefficient values were 0.91 for 50 Indiana gaging stations. However, the use of the default BFImax value of 0.80 for 50 Indiana gaging stations is not recommended because the filtered baseflow data were compared with the results from the BFLOW digital filter which did not reflect the physical characteristics in the aquifer and watershed.
Lim et al. (2006) validated the L-THIA model accuracy by comparing L-THIA daily direct runoff with daily direct runoff values estimated using BFLOW and observed flow. The Nash–Sutcliffe coefficient values were 0.60 for both calibration and validation periods. Although the studies by Lim et al. (2005, 2006) provided higher statistics for comparisons, none of studies by Lim et al. (2005, 2006) was compared with the direct runoff values separated using digital filters reflecting the local hydrogeological conditions. There is a need for an automated module to separate baseflow accurately from stream flow in calibration and validation of hydrologic and water quality models, such as L-THIA and SWAT.
The objectives of this study are to: (1) develop an automated module in the WHAT system for determination of optimum Eckhardt BFImax value and filter parameter values using Genetic Algorithm (GA) techniques by comparing the filtered baseflow with baseflow from recession curve analysis; (2) assess the hydrologic and water quality impacts of using optimized BFImax values, rather than a default BFImax value of 0.80 provided by Eckhardt (2005).
Section snippets
Long-Term Hydrologic Impact Assessment (L-THIA) ArcView GIS
The L-THIA model was developed to estimate direct runoff using the CN method (Harbor, 1994). It utilizes daily rainfall depth, land use, and hydrologic soil group data. An ArcView GIS interface was developed and enhanced over the last several years to provide a user-friendly interface to the L-THIA model (Bhaduri et al., 2001; Lim et al., 2001), and a Web-based L-THIA was developed and is accessible from http://www.ecn.purdue.edu/runoff/. The L-THIA model estimates NPS pollutant loadings by
Study area
In this study, the Little Eagle Creek (LEC) watershed near Indianapolis, Indiana was chosen for daily direct runoff comparison because it was used in the Lim et al. (2006) study for the comparison of L-THIA estimated direct runoff with the filtered direct runoff using the BFLOW digital filter. The LEC watershed is 70.5 km2 in size (Fig. 2(a)). Fig. 2(b) shows the 1991 land uses for the LEC watershed. Urbanized land area in the LEC watershed was approximately 68% of the total land area in 1991 (
Baseflow separation using BFLOW and Eckhardt filters using WHAT system
The default BFImax value of 0.80 is provided by Eckhardt (2005) for perennial streams with porous aquifers. Thus, the default BFImax value of 0.80 is commonly used in separating baseflow from observed flow by most users. The filtered direct runoff values using the BFLOW filter and the Eckhardt filter (with a default BFImax value of 0.80 and a default filter parameter value of 0.98) were compared. The R2 value was 1.00 and the Nash–Sutcliffe coefficient was 0.99 for the comparison of the BFLOW
Conclusions and discussion
In this study, the optimum BFImax value and filter parameter value of the Eckhardt digital filter for the LEC watershed were determined with the BFImax GA-Analyzer module developed in this study. Although the comparison between the L-THIA estimates and the Eckhardt filtered direct runoff with the default BFImax value of 0.80 and a default filter parameter value of 0.98 provided the same accuracy (Nash–Sutcliffe coefficient values of 0.63 in both cases) compared with those with the optimized BFI
Acknowledgement
This work was supported by a grant (code: 2-2-3) from Sustainable Water Resources Research Center of 21st Century Frontier Research Program.
References (24)
- et al.
A hybrid genetic algorithm for train sequencing in the Korean railway
Omega
(2009) - et al.
Global optimization based on novel heuristics, low-discrepancy sequences and genetic algorithms
European Journal of Operational Research
(2009) - et al.
Comparing state-of-the-art evolutionary multi-objective algorithms for long-term groundwater monitoring design
Advances in Water Resources
(2006) - et al.
Effects of calibration on L-THIA GIS runoff and pollutant estimation
Journal of Environmental Management
(2006) A genetic algorithm for the proportionate multiprocessor open shop
Computers & Operation Research
(2009)- et al.
A comparative study of Monte Carlo simple genetic algorithm and noisy genetic algorithm for cost-effective sampling network design under uncertainty
Advanced in Water Resources
(2006) - et al.
Validation of automated methods for estimating baseflow and groundwater recharge from stream flow records
Journal of American Water Resources Association
(1999) - et al.
Automated baseflow separation and recession analysis techniques
Groundwater
(1995) - et al.
Hydrology and Floodplain Analysis
(2002) - et al.
Long-term hydrologic impact of urbanization: a tale of two models
Journal of Water Resources Planning and Management
(2001)
Hydrological Simulation Program—Fortran, User's Manual for Version 11
Unit hydrograph identification using only streamflow data
Transactions of the Institution of Engineers, Australia
Cited by (57)
Impact of irrigation reservoirs on budget of the watershed-scale water cycle under climate change
2023, Agricultural Water ManagementThe role of climate conditions and groundwater on baseflow separation in Urmia Lake Basin, Iran
2023, Journal of Hydrology: Regional StudiesBase flow simulation using a physically based subsurface model – The case of a tropical basin in the Western Ghats, India
2022, Journal of HydrologyCitation Excerpt :Though perfect for long term base flow estimation, these methods have been criticised for lacking a physical basis (Chapman, 1991). Though default filter parameter values have been suggested for the many widely used filters, various studies have brought out the need to calibrate the parameters specific to the region of application (Eckhardt, 2005; Lim et al., 2010). Tracer based methods have been used in various studies to calibrate the filters (Kouanda et al., 2018; Zhang et al., 2017).
Groundwater recharge rate estimation using remotely sensed and ground-based data: A method application in the mesoscale Thur catchment
2021, Journal of Hydrology: Regional StudiesImpacts of land-use and land-cover change on stream hydrochemistry in the Cerrado and Amazon biomes
2018, Science of the Total EnvironmentCitation Excerpt :Ss was computed as the flow change in response to event precipitation and ending at the point separating the stormflow components, i.e. the surface and subsurface stormflow, from the baseflow recession. These flows were determined using a recursive digital filter (Eckhardt, 2005) implemented in the Web GIS-based Hydrograph Analysis Tool (WHAT) for baseflow separation (Lim et al., 2005, 2010). Using this information, we calculated the ratio of Ss to total streamflow (St) discharge.
- 1
Tel.: +82 33 250 6468; fax: +82 33 251 1518.
- 2
Tel./fax: +82 31 910 0256.
- 3
Tel./fax: +82 2 450 3749.
- 4
Tel./fax: +82 54 820 5093.