Assessing hydrological impact of potential land use change through hydrological and land use change modeling for the Kishwaukee River basin (USA)
Post on 26-Jun-2016
Embed Size (px)
Journal of Environmental Management 88
Available online 8 August 2007
viousness in the form of roofs, sidewalks, roads, parkinglots, and turf grass can dramatically increase the speed andmagnitude of runoff (Dunne and Leopold, 1978; Cheng
ARTICLE IN PRESS
Corresponding author. Tel.: +1204 474 6337; fax: +1 204 474 7513.E-mail addresses: email@example.com (W. Choi), firstname.lastname@example.org
and Wang, 2002). Understanding these complex socio-hydrologic dynamics is imperative for planning a moresustainable future.
0301-4797/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
1Present address: Department of Civil Engineering, University of
Manitoba, Winnipeg, MB, Canada R3T 5V6.2Tel.: +1 217 333 5172.1. Introduction
Streamow plays an important role in establishing someof the critical interactions that occur between physical orecological processes and social or economic processes.Socio-economic processes including population dynamics,land use transformation, migration, transportation andagricultural practices closely interact with and greatlyaffect ecological processes, such as vegetative growth,ecological succession, habitat formation and maintenance(Voinov et al., 1999a). In both cases, hydrology and
hydrologic dynamics can work as a medium or canvas forunderstanding both the conditions for interactions to takeplace and the consequences that such interactions some-times elicit. One of the most important socio-economicprocesses for establishing far-reaching and long-termecological effects is land use transformation, especiallythe human-induced variety termed urbanization. The far-reaching effects of urbanization can best be described by itsenormous impacts on basin hydrology and water quality(Ferguson, 1996; Bertrand-Krajewski et al., 2000; Valeoand Moin, 2000). Large proportional increases in imper-Abstract
We connected a cellular, dynamic, spatial urban growth model and a semi-distributed continuous hydrology model to quantitatively
predict streamow in response to possible future urban growth at a basin scale. The main goal was to demonstrate the utility of the
approach for informing public planning policy and investment choices. The Hydrological Simulation ProgramFortran (HSPF) was set
up and calibrated for the Kishwaukee River basin in the Midwestern USA and was repeatedly run with various land use scenarios
generated either by the urban growth model (LEAMluc) or hypothetically. The results indicate that (1) the land use scenarios generated
by LEAMluc result in little changes in total runoff but some noticeable changes in surface ow; (2) the argument that low ows tend to
decrease with more urbanized areas in a basin was conrmed in this study but the selection of indicators for low ows can result in
misleading conclusions; (3) dynamic simulation modeling by connecting a distributed land use change model and a semi-distributed
hydrological model can be a good decision support tool demanding reasonable amount of efforts and capable of long-term scenario-
r 2007 Elsevier Ltd. All rights reserved.
Keywords: Urban growth modeling; Hydrological modeling; HSPF; Environmental impact assessmentAssessing hydrological impacthrough hydrological and la
the Kishwaukee R
aDepartment of Geography, University of Illinois at UrbanabDepartment of Urban and Regional Planning, University of
Lorado Taft Dr., Cha
Received 23 September 2006; received in re(2008) 11191130
of potential land use changed use change modeling forver basin (USA)
Brian M. Dealb,2
mpaign, 607 S. Mathews Avenue, Urbana, IL 61801, USA
nois at Urbana-Champaign, 111 Temple Buell Hall, 611 E.
aign, IL 61820, USA
form 26 May 2007; accepted 2 June 2007
specically, this study focuses on how a local hydrologicalsystem responds to alternative land use scenarios generatedby a dynamic urban growth model and what substantivescientic, planning, and policy related implications emerge.Eventually, we intend to demonstrate the utility of theapproach as a decision support tool for regional planners.
2. The study region
The area of interest is the Kishwaukee River basin(KRB) in the Midwestern United States (Illinois andWisconsin); its drainage area is 3258 km2 (Fig. 1). Locatedroughly between the Metropolitan Chicago (ranked thethird in the US in population) and Rockford (ranked thethird in Illinois in population) areas, the KRB is underdevelopment pressure from both sides. Agriculture is thepredominant land use in the region. The 1992 NationalLand Cover Data (NLCD) from the US Geological Survey(USGS) indicates that row crops (mostly corns andsoybeans) cover more than 70% of the KRB. Urbanizedareas account for around 3% of the local land uses aroundthe year 2000.The annual mean precipitation and annual mean
temperature measured at Rockford, Illinois (National
ARTICLE IN PRESSnmental Management 88 (2008) 11191130The complexity and magnitude of each system (socio-economic and eco-hydrologic) require a systematic ap-proach for improving our understanding of each. In ouropinion, this is best accomplished through the adoptionand application of dynamic simulation modeling techni-ques. Explicit modeling of complex environmental pro-blems is essential for developing realistic descriptions ofpast behavior and the possible impacts of alternativemanagement policies (Costanza, 1991). Dynamic models ofcomplex and interconnected ecosystems enable scientists toexperiment with and come to understand the interactionsof dynamic system components (Forrester, 1969; Costanzaand Maxwell, 1991; Sklar and Costanza, 1991; Ruth andHannon, 1997). Modeling can provide assistance inmanaging uncertainty, developing feedbacks and lags,improving group decision-making techniques, and under-standing comprehensive learning tasks. The process ofmodeling can also help facilitate communication throughboth model results and model structure to nd possibleemergent properties of a system.Although a plethora of modeling and analysis has been
done in both hydrology and land use modeling, it is stillnovel to adopt (and connect) both hydrological models andland use change models for the purpose of informing publicplanning policy or investment choices. Good examples ofthis approach do exist and appear in several articles (e.g.,Beighley et al., 2003; Voinov et al., 1999a, b; Niehoff et al.,2002; Arthur-Hartranft et al., 2003). Voinov et al.(1999a, b) adopted a distributed cellular automata (CA)approach for both hydrological and land use changemodeling, where modeled land use change results weredynamically incorporated into the hydrological model.Niehoff et al. (2002) also utilized a distributed hydrologicalmodel with modeled land use change scenarios. Thesedistributed approaches have been described by others to becapable of a more accurate assessment of the hydrologiceffects of land use change, because their parameters includea physical interpretation of data and their model structuresgenerally allow for an improved representation of spatialvariability (Nandakumar and Mein, 1997). These models,however, are also very data-intensive and demand sub-stantial computing capability, which prevents them frombeing used for long-term assessments by planners.On the other hand, Beighley et al. (2003) applied a
lumped-parameter hydrological model with a future landuse scenario from a CA-based dynamic urban growthmodel to a basin in southern California. In such anapproach, a basin is partitioned into subbasins, and theoutput from an urban growth model is incorporated intoeach subbasin to run a hydrological model. A similarapproach is adopted in this paper with a different researchdesign in a different geographical setting.This paper presents the results of our work connecting a
CA, dynamic, spatial urban growth model and a semi-distributed continuous hydrological model to quantita-
W. Choi, B.M. Deal / Journal of Enviro1120tively predict hydrological variables in response to futureurban growth patterns in northeastern Illinois. MoreFig. 1. Subbasins and reaches of the Kishwaukee River basin along withWeather Service Cooperative Station ID 117382) over theperiod 19712000 are 930mm and 8.9 1C, respectively.Monthly mean temperatures vary from below 0 1C inDecember, January and February to above 20 1C in thesummer months (June, July and August). About 10% ofweather and streamow gauging stations. Metropolitan Rockford (left)
and Chicago (right) areas are shown along with the KRB in the inset map.
After delineating subbasins and calculating parameters,the BASINS data including the basin model and land coverdata were exported to HSPF to calculate streamow withinthe KRB. For simplicity, the USGS NLCD land covercategories were aggregated as in Table 2. Percent pervious
ARTICLE IN PRESS
Open water W/W 100
Low-intensity residential LR 63
High-intensity residential HR/C/I/T 45
Commercial/industrial/transportation HR/C/I/T 45
Road Road 1
Bare rock/quarries/transitional Barren 50
Forest Forest 100
Shrubland G/P 100
Orchards/vineyards/other G/P 100
Grasslands/herbaceous G/P 100
Pasture/hay G/P 100
Row crops Ag 100
Small grains Ag 100
Urban/recreational grasses G/P 100
Woody wetlands W/W 100
Emergent herbaceous wetlands W/W 100
nmethe annual precipitation falls as snow (Allen Jr. andCowan, 1985).There are ve active streamow gauging stations in the
KRB, along with a weather station whose data were usedin this study (see Fig. 1). Table 1 briey describes relevantstreamow and weather stations information. Althoughadditional weather stations (besides Rockford) exist in thenorth and east of the KRB, they were considered toodistant to be useful for this analysis.
3. Semi-distributed hydrological modeling
3.1. Delineating subbasins with BASINS
For this work, we used three primary tools in ourhydrological modeling efforts: (1) the Better AssessmentScience Integrating Point and Nonpoint Sources, orBASINS (US Environmental Protection Agency, 2001);(2) land cover, elevation, and hydrography data from theUSGS; and (3) Hydrologic Simulation ProgramFortran,or HSPF (Bicknell et al., 2001).BASINS was developed as a geographic information
system (GIS) based catchment assessment tool. It utilizesan ArcView GISs 3.x as a software platform to preprocessinput data for several hydrological models such as HSPFand Soil and Water Assessment Tool (SWAT). BASINS isuseful for delineating subbasins, reaches and outlets;
Weather and streamow gauging stations in and around the KRB
Station name Type (ID)
Rockford Greater Rockford Airport Weather (117382
Kishwaukee River near Perryville, IL Streamow (0544
Kishwaukee River at Belvidere, IL Streamow (0543
South Branch Kishwaukee River near Fairdale, IL Streamow (0543
South Branch Kishwaukee River at DeKalb, IL Streamow (0543
Piscasaw Creek near Walworth, WI Streamow (0543
Source: USGS and National Climatic Data Center.aa.s.l.: above sea level.
W. Choi, B.M. Deal / Journal of Enviroprocessing land use and digital elevation datasets; andcalculating related subbasin parameters for the notedhydrological models.The 1992 NLCD, National Elevation Dataset (NED)
and National Hydrography Dataset (NHD) were obtainedfrom the USGS (http://seamless.usgs.gov). With NED andNHD, subbasins were automatically delineated in BASINSusing the default threshold BASINS suggested consideringthe basin size. Relevant physiographic parameters such asmean elevation, area, slope, etc. were subsequently calcu-lated and saved in an attribute table by BASINS. Fig. 1shows that twenty subbasins have been delineated withinthe KRB. The 20th subbasin was delineated by manuallyadding an outlet point at the location of a streamowgauging station (USGS 05440000). The average subbasinsize is 157 km2, with the standard deviation of 127km2.Location Elevation
421120N/891060W 222.5 N/A0) 4211104000N/8815905500W 211.0 2848.720) 4211502200N/8815104700W 225.0 1394.550) 4210603800N/8815400200W 223.7 1003.140) 4115505200N/8814503400W 253.6 201.413) 4213101800N/8813903900W 285.0 24.83
Land cover categories aggregated for HSPF and associated percent
Land cover category in NLCD Aggregated
ntal Management 88 (2008) 11191130 1121values were estimated for each aggregated land usecategory, based on previous works in the literature (Brunand Band, 2000; Choi and Ball, 2002).
3.2. HSPF modeling
HSPF is a semi-distributed hydrological model em-bedded in BASINS. A previous work has shown it to beuseful for analyzing long-term hydrological effects, espe-cially in largely urbanized areas (Borah and Bera, 2003).HSPF has been developed to be readily applicable to mostbasins in the United States using readily-available weather,hydrologic, topographic, and land use informationgenerally supplied with BASINS program (Rahman andSalbe, 1995). It has three modules, PERLND, IMPLND,and RCHRES, which represent pervious land segments,
ARTICLE IN PRESSnmeimpervious land segments and reaches/reservoirs, respec-tively. A river basin is divided into subbasins which haveboth pervious land segments and impervious land segmentsin each of them. Hydrological and water quality processesare simulated by both PERLND and IMPLND modules ineach subbasin, and routing to downstream subbasins issimulated by RCHRES module.HSPF requires hourly meteorological data sets and a
watershed data set. In BASINS, hourly meteorologicaldata are stored in a WDM le for each state (http://www.epa.gov/waterscience/ftp/basins/wdm_data/).IL.WDM le was used in this study, which contains themeteorological data set measured at Rockford, Illinois, aspreviously noted. A watershed data set as a WSD le iscreated by exporting the subbasin boundaries and para-meters in BASINS for HSPF. A single WSD le containsall the subbasins information in it as text. An HSPF projectis created by combining the WDM le and the WSD le,and saved as *.UCI in a text format. A UCI le contains aset of all the parameters and paths to other les.
4. Model calibration
The HSPF model was calibrated against the observedstreamow data from the USGS gauging station 05440000Kishwaukee River near Perryville, IL (Table 1). Then thecalibrated model was applied to a small subbasin, forwhich observed streamow data are available, to show thatthe calibrated model parameters properly work in differentlocations in the KRB. This is the so-called modelconrmation conducted by Wicklein and Schiffer (2002),and Cao et al. (2006) adopted a similar approach withSWAT model. For model conrmation, a subbasindraining at the gauging station USGS 05439000 SouthBranch Kishwaukee River at DeKalb, IL (Table 1) wasselected. A separate HSPF model (UCI le) was built forthe subbasin, called the South Branch Kishwaukee RiverBasin (SBKRB) hereafter. The SBKRB has the area of201.41 km2 and located in the southeastern part of theKRB.In our analysis, the years 19881989 were used...