R

Im

Ca

b

c

U

h

•••••

a

ARRAA

KBGWULS

1

tadH

km

h0

Landscape and Urban Planning 140 (2015) 29–41

Contents lists available at ScienceDirect

Landscape and Urban Planning

j our na l ho me pa g e: www.elsev ier .com/ locate / landurbplan

esearch paper

dentifying priority sites for low impact development (LID) in aixed-use watershed

helsea J. Martin-Miklea,∗, Kirsten M. de Beursa, Jason P. Julianb, Paul M. Mayerc

The University of Oklahoma, Department of Geography and Environmental Sustainability, Norman, OK, United StatesTexas State University, Department of Geography, San Marcos, TX, United StatesUSEPA, Office of Research and Development, National Health and Environmental Effects Research Lab, Western Ecology Division, Corvallis, OR 97333,nited States

i g h l i g h t s

We present a spatially-explicit approach for placing LID in urban watersheds.The approach is based on publicly-available data to facilitate wide-spread use.Placing LID at prioritized sites is cost-effective and ecologically beneficial.Implementing LID across less than 1% of sub-catchment land area can reduce nutrient/sediment by 15%.We use a case study (mixed-use watershed) to test the efficacy of our siting tool.

r t i c l e i n f o

rticle history:eceived 3 July 2014eceived in revised form 23 February 2015ccepted 1 April 2015vailable online 24 April 2015

eywords:est management practicereen infrastructure

a b s t r a c t

Low impact development (LID), a comprehensive land use planning and design approach with the goalof mitigating land development impacts to the environment, is increasingly being touted as an effectiveapproach to lessen runoff and pollutant loadings to streams. Broad-scale approaches for siting LID havebeen developed for agricultural watersheds, but are rare for urban watersheds, largely due to greater landuse complexity. Here, we introduce a spatially-explicit approach to assist landscape architects, urbanplanners, and water managers in identifying priority sites for LID based exclusively on freely availabledata. We use a large, mixed-use watershed in central Oklahoma, the United States of America, as acase-study to demonstrate our approach. Our results indicate that for one sub-catchment of the Lake

ater qualityrban runoffow impact development (LID)tormwater

Thunderbird Watershed, LID placed in 11 priority locations can facilitate reductions in nutrient andsediment loading to receiving waters by as much as 16% and 17%, respectively. We had a high rate ofcorrectly identified sites (94 ± 5.7%). Our systematic and transferable approach for prioritizing LID siteshas the potential to facilitate effective implementation of LID to lessen the effects of urban land use onstream ecosystems.

© 2015 Published by Elsevier B.V.

. Background

Although urban land use covers a disproportionately small frac-ion of the United States (3%) (USCB, 2012), rapid urbanization

nd associated activities have in some cases contributed to streamegradation more than any other type of land use (Fuhrer, 1999;offman, Capel, & Larson, 2000; Omernik, 1976). Urban land area

∗ Corresponding author. Tel.: +1 715 587 2651.E-mail addresses: chelseamikle@gmail.com (C.J. Martin-Mikle),

debeurs@ou.edu (K.M. de Beurs), Jason.julian@txstate.edu (J.P. Julian),ayer.paul@epa.gov (P.M. Mayer).

ttp://dx.doi.org/10.1016/j.landurbplan.2015.04.002169-2046/© 2015 Published by Elsevier B.V.

in the United States quadrupled from 1945 to 2010 (USCB, 2012;USDA, 2011), a period during which water quality of streams drain-ing urban areas declined considerably (Paul & Meyer, 2001; Walsh,Fletcher, & Ladson, 2005). When urban areas expand, the increasingimpervious surface coverage and stream burial alters watershedhydrology (Kaushal & Belt, 2012). Catchments with large areasof impervious surface typically display lower water infiltrationand flashier hydrographs characterized by shorter response times,higher flood magnitudes and shorter flood durations (Paul & Meyer,

2001; Wolman & Schick, 1967). Less infiltration results in less filter-ing of pollutants by soil and vegetation (Gold et al., 2001; Osborne& Kovacic, 1993). Intense and high runoff volumes due to impervi-ous surfaces erode stream beds/banks, increase sediment/pollutant

3 and U

l&2

oa(1amrbirdraRtoFs(ttPa

e2woRow(wLmt(aSobs

pcHtpNhptQl

acoutLs

0 C.J. Martin-Mikle et al. / Landscape

oads, degrade stream ecosystems, and displace organisms (Julian Torres, 2006; Palmer, 2009; Paul & Meyer, 2001; Walsh et al.,005).

Given that it is often not practical to reverse or stop devel-pment, low impact development (LID) techniques are becoming

popular means to improve water quality in urban watershedsDietz, 2007; Pyke et al., 2011; Roy et al., 2008; Urbonas & Stahre,993). LID is a comprehensive land use planning and designpproach with the goal of mitigating urban impacts to the environ-ent at the sub-catchment level. LID techniques work by reducing

unoff from localized impervious source areas (e.g., by using rainarrels, green roofs and porous pavement), by slowing and filter-

ng overland water runoff, sediment, and pollutants before theyeach the main stream network (e.g., via grassed swales, rain gar-ens and detention/retention ponds), and by slowing and filteringunoff in or adjacent to the main stream network (e.g., protectionnd/or restoration of riparian buffers) (Craig et al., 2008; Mayer,eynolds, McCutchen, & Canfield, 2007). Effective LID implementa-ion is influenced by several variables such as placement, selectionf technique, design, construction, and upkeep (Muthukrishnan &ield, 2004). The location of LID implementation within a water-hed can be the most important factor determining effectivenessPasseport et al., 2013). For example, placement of LID determineshe volume of runoff, thereby directly influencing the benefits perhe associated cost (Agnew et al., 2006; Berry, Delgado, Khosla, &ierce, 2003; Qiu, 2009). Thus, a need exists for a spatially-explicitpproach for siting LID.

Despite the increased awareness and promotion of LID as anffective approach to reduce runoff and pollutant loads (Dietz,007; Van Roon, 2005), most LID techniques applied in urbanatersheds have been largely experimental, opportunistic, and

ften implemented to remedy local stormwater runoff issues (Vanoon, 2005; Walsh & Kunapo, 2009). Advantages to considering LIDn a watershed-scale include: (1) the effect that LID has on receivingaters is more easily quantifiable within watershed boundaries,

2) it is a more efficient use of resources to place LID where itill be most hydrologically effective, (3) fewer strategically placed

ID allows for more affordable and consistent maintenance andanagement of LID projects and, (4) improved opportunity for LID

echniques to provide connected recreational and habitat benefitsClar, Barfield, & Yu, 2002; Urbonas, 2000). The few tools currentlyvailable to site LID in urban watersheds are complex models (e.g.,USTAIN, SWMM) (USEPA, 2013) that require significant amountsf time, money, and expertise, which make them largely inaccessi-le to most planners and watershed managers, especially those inmaller cities without sufficient resources.

Recent LID-siting methods that target non-point source (NPS)ollutants from smaller sub-catchments of watersheds have beenreated based on variable source area (VSA) hydrology (Hewlett &ibbert, 1967), a process-based concept that identifies areas prone

o saturated overland runoff and thus increased potential to trans-ort pollutants (Agnew et al., 2006; Gburek, Drungil, Srinivasan,eedelman, & Woodward, 2002; Qiu, 2009). The related concept ofydrologically sensitive areas (HSAs) is based on the probability ofollution transport risk (Walter et al., 2000). HSAs have been usedo inform conservation buffer placement (Delgado & Berry, 2008;iu, 2009), but have not yet been utilized to prioritize site-specific

ocations for LID in urban watersheds.In order to address the need of a spatially-explicit and mech-

nistic (i.e., based on physical processes) LID siting tool thatonsiders multiple land uses across entire watersheds, we devel-ped a geographic information system (GIS)-based framework

sing publicly-available data intended to assist landscape archi-ects, urban planners, and watershed managers in making informedID-placement decisions. The approach outlined here prioritizesites where LID would be most effective based on: (1) identification

rban Planning 140 (2015) 29–41

of HSAs based on a multi-variable topographic index, and (2) cal-culation of suitability for LID application based on land use, spatialscale, position in the stream network, and effectiveness in imper-vious areas. We applied this approach to a large watershed withdiverse landscapes in order to demonstrate its flexibility and broadapplicability.

2. Methods

2.1. Study area

We developed and tested the LID siting framework on the666-km2 Lake Thunderbird Watershed in central Oklahoma, USA(Fig. 1). The Lake Thunderbird Watershed is a mixed-use watershed,encompassing portions of four cities: Midwest City to the north,Oklahoma City to the northwest, Moore to the west and Norman tothe southwest. Over forty percent of the watershed is consideredresidential, resulting in high impervious surface coverage in theseareas (i.e., roads, buildings and parking lots). Most of the housingand development infrastructure is relatively new (within the last40 years) and is increasing rapidly (ODEQ, 2010).

The watershed is dominated by intermittent surface waterrunoff because of its semi-arid climate, deep clayey soils, andhigh drainage density, particularly in the headwaters (Wilgruberet al., unpublished data). The Central Oklahoma Aquifer underl-ies the watershed, with a median depth to water table of 10 m(Mashburn, Ryter, Neel, Smith, & Magers, 2013). The central Okla-homa region relies on the Lake Thunderbird Reservoir for watersupply and recreation; however, the reservoir is experiencing sig-nificant water quality problems due primarily to urban runoffduring heavy rainfall events (Oklahoma Conservation Commission(OCC), 2008). Indeed, this watershed experiences intense rain-fall events, with a 24-h 2-year intensity of 87 mm/h. Excessiveamounts of phosphorus, nitrogen and sediment are transportedlargely through urban runoff into headwater streams, ultimatelyresulting in excessive turbidity and algae growth in Lake Thun-derbird, both of which exceed total maximum daily load (TMDL)regulations (ODEQ, 2008). Effective implementation of LID acrossthe watershed could mitigate urbanization impacts on Lake Thun-derbird by reducing pollutant loadings to its receiving waters.

2.2. Data

Publicly available GIS datasets were compiled from varioussources (Table 1). The resulting database was used for the integra-tion, management and development of data layers used to calculatea topographic index and to model basic land use and land coverrequirements for LID. A 10-m digital elevation model (DEM) wasprocessed using ESRI’s ArcGIS hydrology toolset to develop slope,flow direction, flow accumulation, and stream network raster lay-ers, and stream order from the stream network raster using theHorton–Strahler method (Strahler, 1957). All data layers, includ-ing soil conductivity and soil depth to restrictive (confining) layerfrom the Soil Survey Geographic database (SSURGO) were con-verted from polygon layers to 10-m raster layers with the sameextent as the DEM. Additionally, the road centerline vector layerwas buffered 3.9 m on each side in accordance with local street-width guidelines, and the stream network was buffered at 30-m oneither side of the stream in accordance with typical riparian bufferrequirements (Mayer, 2005).

2.3. Approach

While we mapped and considered all suitable LID sites, we focushere on priority sites, which we define as the 140 most sensitiveHSAs across the Lake Thunderbird Watershed. The prioritization

C.J. Martin-Mikle et al. / Landscape and Urban Planning 140 (2015) 29–41 31

F the Uw Fry etb

osauasl

TD

ig. 1. The Lake Thunderbird Watershed study area, located in central Oklahoma,

est and forest in the east, as demonstrated by the National Land Cover Database (oundary.

f sites where LID would be most effective was divided into twoteps (Fig. 2): (1) identification of HSAs using a topographic index,nd (2) suitability for application of LID techniques based on land

se, spatial scale and, the ability to construct LID in imperviousreas (Table 2).We chose 140 sites (approximately 6% of the water-hed) so as to yield a manageable number of priority LID sites forandscape architects, urban planners, and watershed managers.

able 1ata layers used to prioritize locations for low impact development (LID) at a watershed

GIS layer Resolution/format Source

Land cover 30-m/raster National LanImpervious Surface 10-m/raster National LanDEM 10-m/raster National EleSoil conductivity 6 ha/polygon Soil Survey

Soil depth to Restrictive layer 6 ha/polygon Soil Survey

Roads 1:100,000/Vector Topological(TIGER) 201

Zoning Vector City GIS DepBuilding Footprint Vector City GIS DepFloodplain Vector City GIS DepWater bodies 1:24,000/Vector National Hy

nited States of America, is a mixed-use watershed, dominated by grassland in the al., 2011). Major cities of central Oklahoma are labeled adjacent to the watershed

2.4. Deriving the topographic index and HSAs

Watersheds are heterogeneous with different soil types, slopes,

and land cover/use, all which influence how water flows acrossthe landscape. Therefore, we used a topographic index to modelpatterns of surface runoff according to VSA hydrology based on (1)a wetness index (sensu Beven & Kirkby, 1979; Moore, Grayson, &

scale.

Citation

d Cover Database (NLCD) 2006 Fry et al. (2011)d Cover Database (NLCD) 2006 Fry et al. (2011)vation Dataset (NED) Gesch (2007)Geographic Database (SSURGO) NRCS (2012)Geographic Database (SSURGO) NRCS (2012)ly Integrated Geographic Encoding and Referencing2, US Census Bureau

USCB(2013)

artmentsartmentsartmentsdrography Dataset (NHD), Version 2.0 USGS (2012)

32 C.J. Martin-Mikle et al. / Landscape and Urban Planning 140 (2015) 29–41

Fig. 2. Conceptual diagram for identification of the most effective locations for implementing LID at a watershed scale. A topographic index approach was applied. Equation

�∗ = ln(

˛/tan ˇ)

− ln (KsDISA)) based on slope, drainage area, depth to restrictive layer and soil hydraulic conductivity, to define hydrologically sensitive areas (HSAs). We

then factor in local land use and land cover conditions in four sub-models, which prioritize locations where groups of LID with similar sites-requirements would be mosteffective.

Table 2Common LID techniques grouped according to generalized site suitability1.

LID type Land use characteristics Scale of technique Effective in impervious areas

Rain barrel Ideal for collecting rooftop runoff Local Yes, ideally suitedGreen roof Ideal for collecting rooftop runoff Local Yes, ideally suitedPorous pavement Ideal for highly developed areas: parking

lots, driveways and low-volume roadsLocal Yes, intended to replace

impervious surfaceRain garden Ideal for collecting rooftop runoff and

runoff from yards and sidewalks. Also goodfor collecting runoff from roads and smallparking lots

Intermediate Intercepts runoff from imperviousareas but requires land forconstruction

Vegetated swale Ideal for collecting sheet flow runoff fromroads and highways. Also ideal forcollecting runoff from subdivisions

Intermediate Intercepts runoff from imperviousareas but requires land forconstruction

Detention pond Ideal for detaining runoff from largecatchment areas (i.e., as large as 75 acres)

Catchment Intercepts runoff from impervioussurfaces but requires large piece ofland for construction

Retention pond Ideal for retaining water from parking lotsand residential areas

Catchment Intercepts runoff from impervioussurfaces but requires large piece ofland for construction

Riparian buffer Ideal for land directly adjacent to streamsand rivers

Reach Intercepts runoff from impervioussurfaces but requires large landadjacent to stream forimplementation

1 Table adapted from Table 5-2. Generalized BMP Suitability for Relevant Drainage Area Characteristics. Chapter 5, Effective Use of BMP in Muthukrishnan, S., 2004. TheUse of Best Management Practices (BMPs) in Urban Watersheds.

and U

Lwt

(b

�

wimasth

�

wimAdoboidsTtvti

flcnpsa

D

woof�

�

tw2viisvttsi

C.J. Martin-Mikle et al. / Landscape

adson, 1991) derived from drainage area and slope and, (2) soilater storage, based on soil hydraulic conductivity and soil depth

o a restrictive layer.The wetness index � models surface runoff contributing areas

Beven & Kirkby, 1979) by assigning each cell a unitless index valueased on the equation:

= ln(

˛

tan ˇ

)(1)

here ̨ is the contributing drainage area per unit contour lengthn meters, and ̌ is the slope in radians. Note that all slope angles ˇ

ust be greater than zero (Qiu, 2009). The wetness index does notccount for soil water storage capacity. To account for soil watertorage, Walter et al. (2002) added a soil water storage componento the wetness index. The combination of these two components isereafter referred to as the topographic index (�) and is defined as

= ln(

˛

tanˇ

)− ln (KsD) (2)

here Ks is soil hydraulic conductivity in meters per day and Ds the soil depth to restrictive layer in centimeters, up to a maxi-

um of 2 m (a characteristic of the SSURGO dataset for this region).dmittedly, the altered hydraulic properties of urban soils are oftenifficult to map (i.e., compaction, irrigation, etc.) due to sparsebservation points. However, we chose to use SSURGO soil dataecause it represents the best-available option. To ensure continu-us coverage of the soil water storage index across the study area,ncluding land types for which no value for Ks existed in the soilatabase, for grid cells with a hydraulic conductivity of zero theoil water storage index was reassigned to −3 (sensu Qiu, 2009)his is lower than the highest calculated soil water storage index inhe watershed and had no significant effect on topographic indexalues, since most grid cells with a hydraulic conductivity of zero inhe Lake Thunderbird Watershed are water and had high wetnessndex values.

To expand the ability of the topographic index to model sur-ace runoff in urbanized areas, we altered the depth-to-restrictiveayer raster to account for impervious surface area (ISA) of each gridell. This alteration modifies the proportion of soil depth that doesot have a water storage capacity during rain fall events, making itossible to simulate runoff from impervious surfaces onto perviousurfaces. The altered D value is represented by DISA and is defineds:

ISA = D − (D × ISA) (3)

here D is the original soil depth layer and ISA is the proportionf the pixel (0–1) covered by impervious surfaces. The proportionf soil depth with no water storage capacity was then subtractedrom the original soil depth raster. The modified topographic index* is:

∗ = ln(

˛

tan ˇ

)− ln (KsDISA) (4)

High topographic index values correspond to pixels most likelyo become saturated (relative to the rest of the pixels within theatershed) and act as a source of overland runoff (Walter et al.,

000). In this study, clusters of pixels with high topographic indexalues are defined as HSAs. Because the topographic index is heavilynfluenced by flow accumulation, selecting the highest topographicndex values resulted in identification of the pixels furthest down-tream, whereas selecting a very broad range of topographic indexalues resulted in an overabundance of potential sites. To iden-

ify a manageable number of LID sites, this study targets 6% ofhe watershed area with the highest topographic index values (1.5tandard deviations above the mean) as HSAs (Fig. 5), using sim-lar practical methods to Qiu (2009). HSAs were ranked in order

rban Planning 140 (2015) 29–41 33

of sensitivity: 20 sites were prioritized for local-scale LID, 50 sitesthe intermediate-scale, 50 sites for the catchment-scale LID, and20 sites for reach-scale LID, for a total of 140 prioritized LID sitesacross the Lake Thunderbird Watershed.

2.5. Combining HSAs with land use and land cover data toprioritize locations for LID

Land use and land cover were used in conjunction with HSAs toprioritize locations where LID will be most effective. We selectedthe eight most described and mapped LID techniques from theInternational Stormwater BMP Database (Geosyntec Consultants& Wright Water Engineers, 2012) grouped according to scale andgeneral site requirements (Table 2): (1) local-scale LID techniquesfor highly developed source areas, (2) intermediate and catchment-scale LID techniques that capture runoff from multiple source areas,and (3) reach-scale riparian buffers for land adjacent to streams. Aland use and land cover sub-model was developed to prioritize loca-tions where each group of LID techniques would be most effective(Table 3).

2.5.1. Local-scale LID selectionTo prioritize locations for local-scale LID (i.e., rain barrels, green

roofs and porous pavement), the contributing drainage was derivedfor each HSA that intersected with a headwater stream (Fig. 3a).Each impervious area within the contributing drainage to the HSA,when fitted with LID, has the potential to reduce runoff, and there-fore stress, to the down-slope HSA. The contributing drainage areaswere ranked in order of most likely to least likely to generaterunoff based on the mean topographic index value of each HSA.We mapped the contributing drainage areas for the twenty most-sensitive HSAs.

2.5.2. Intermediate-scale and catchment-scale LID selectionOptimal locations for intermediate (i.e., rain gardens and

grassed swales) and catchment-scale (i.e., detention and reten-tion ponds) LID were identified by masking unsuitable land coverand land use types from the HSA raster. Land cover and land usesunsuitable for LID are watershed-dependent, based on resourceavailability and existing land cover (i.e., forests) and land uses (i.e.,existing structures). In the Lake Thunderbird watershed, forestedland, floodplains, building footprints, roads and all land within 30-m of the stream network were masked from the HSA raster. The30-m buffer was used for reach-scale LID, described in the nextsection. A pond vector layer available from the National Hydrog-raphy Dataset (NHD) was incorporated into the land cover datalayer to facilitate possible retrofitting of pre-existing constructedponds (e.g., for livestock) into detention ponds or retention ponds(Table 3). HSAs were divided into two groups based on contributingdrainage area in accordance with general EPA LID sizing guide-lines that suggest a range of contributing drainage area for eachLID type (USEPA, 2014); HSAs that drained <4.5 ha were priori-tized for intermediate-scale LID (Fig. 3b), and HSAs draining >4.5 hawere prioritized for catchment-scale LID (Fig. 3c). HSAs within eachgroup were prioritized by the mean topographic index value of thepixels within each HSA. For the Lake Thunderbird Watershed, the100 sites prioritized for intermediate and catchment-scale LID weremapped for the watershed (50 for each scale).

2.5.3. Reach-scale LID selectionTo prioritize where reach-scale LID (i.e., riparian buffers along

segments of the channel) would be most effective, the topographicindex was calculated for 30-m on either side of the stream network.In this sub-model, building footprints and roads were masked asunsuitable land uses from the HSA raster (Table 3). The result was

34 C.J. Martin-Mikle et al. / Landscape and Urban Planning 140 (2015) 29–41

Fig. 3. Examples of the results from each sub-model demonstrating the functionality of the LID siting approach: (a) estimated ideal location for rain barrels, a green roof, orporous pavement, (b) estimated ideal location for rain gardens or vegetated swales, (c) site where a detention pond or retention pond is estimated to be most effective and(d) an estimated ideal stream segment for riparian buffer implementation identified by the reach-scale sub-model. Note: example b was located in the path of the May 2013Moore Tornado. This example was included to illustrate that LID could be easily implemented into site reconstruction. (For interpretation of the references to color in thisfigure legend, the reader is referred to the web version of this article.)

C.J. Martin-Mikle et al. / Landscape and Urban Planning 140 (2015) 29–41 35

Table 3Land use and land cover data used in each sub-model to determine suitable LID sites.

Scale of LID Roads Building footprints Flood plain Stream buffer zone Existing ponds Land cover

Local Yes Yes Yes Yes No All categories of land that include someimpervious surface

Intermediate No No No No Yes Water (ponds only), partial imperviousareas, soil/barren,grass/herbaceous/agriculture

Catchment No No No No Yes Water (ponds only), partial imperviousareas, soil/barren,grass/herbaceous/agriculture

Reach No No Yes Yes No Partial impervious areas, soil/barren,grass/herbaceous/agriculture, trees/forest

YN

aa

2

icIF2iulfp(mEt2

2

aic1ifvsd5pooqaelooltaecaa

from the OCC (2008), which were derived from the Soil & WaterAssessment Tool (SWAT) and are as follows (g/m2/yr): 0.041–0.063TP, 0.050–0.075 TN, and 11.9–16.3 TSS. The nutrients/sedimentsremoved at each site were summed and divided by the total

Table 4Estimated percent removal efficiency of LID techniques1.

LID Type TP2 TN TSS

Rain barrel NI NI NIGreen roof 0 0–91%3 0–93%Porous pavement 25–50% 0–42%3 68–86%Bioretention/rain garden 0–42% 0–58%3 69–89%Vegetated swale 0 0–32% 6–55%Detention pond 4–37% 0 50–75%Retention pond 48–61% 15–40% 75–85%Riparian buffer 41–93%4 56–87%5 58–100%6

1 All % removal ranges calculated based on the International Stormwater BestManagement Practice (BMP) Database (Geosyntec & WWE, 2012), unless otherwisenoted.

2 TP = total phosphorus; TN = total nitrogen; TSS = total suspended solids.3 Collins et al., 2010; concentration-based removal efficiencies.4 Hoffmann, Kjaergaard, Uusi-Kämppä, Hansen, & Kronvang (2009); ranges cal-

culated across several buffer widths, expressed as percentage yearly retention as

es—data layer included in sub-model.o—data layer excluded in sub-model.

map prioritizing the 20 most hydrologically sensitive locationsdjacent to streams for buffer placement (Fig. 3d).

.6. Development of 1 m land cover data

To determine whether high resolution land cover data wouldmprove ability to prioritize sites for LID, we developed a 1 m landover map. Forty-six image tiles from the National Agriculturalmagery Program (NAIP), obtained from http://www.fsa.usda.gov/SA/apfoapp?area=home&subject=prog&topic=nai, taken in May010 were combined with a normalized difference vegetation

ndex (NDVI) layer and utilized to develop a 1 m land cover mapsing a maximum likelihood classification procedure. Maximum

ikelihood classification, based on the assumption that the statisticsor each class in each band are normally distributed, assigns eachixel to the class to which it has the highest probability of belongingRichards & Jia, 1999). The classification was created using approxi-

ately 2000 training samples identified and classified using Googlearth imagery. The classes used to classify the image aligned withhose in the level 1 NLCD land cover classification scheme (Fry et al.,011).

.7. Validation

To enable quantification of potential nutrient/sediment removalnd to facilitate on-ground evaluation, two sub-catchments thatncluded sites from each sub-model and were in urban areas werehosen for validation. Combined, the sub-catchments contained7 of the 140 priority sites without regard for mean topographic

ndex value. We were aiming to identify a true proportion of 0.95or the correct identification of LIDs. We set our confidence inter-al at 0.90 and the desired precision at 0.1. With a populationize of 140 sites, we can calculate that the minimum number ofesired samples is 12. To be on the safe site we decided to select

more sites than the minimum. All 17 sites were validated on aass/fail basis through field visits and suitability determined basedn whether construction of LID was feasible, the required amountf open space existed for LID implementation, and topographicualities (i.e., slope and stream elevation relative to surroundingreas) were conducive to supporting effective LID. For example,ach site was evaluated to determine whether existing land use andand cover permitted LID implementation (i.e., changes may haveccurred since the time of data acquisition such as construction,r a site may not have been large enough due to coarse resolutionand cover data) and to determine whether water could be routedo the location (e.g., via grassed swale) from surrounding areas. Thebility to remotely prioritize potential LID sites is one advantage of

mploying a GIS-based approach for siting LID, as opposed to theostly and time-consuming practice of visiting all potential sitescross a watershed. Additionally, the ability to prioritize LID sitest the watershed scale ensures cost-effective implementation. The

error rate in percent (E) associated with the ability of this approachto effectively eliminate unsuitable land use and land cover wascalculated using the following equation:

E =(

X

Y

)× 100 (5)

where (X) is the number of validation sites where unsuitable landuse and land cover had not been effectively eliminated by theapproach, and (Y) is the total number of validated sites. Accordingto basic statistics, we can calculate the error rate of this proportionas follows:√(

((1 − E) × E)n

)(6)

where E is calculated in Eq. (5) and n is 17, the number of samples.

2.8. Nutrient and sediment removal efficiency of prioritized LIDsites

Nutrient and sediment load estimates were used in conjunc-tion with removal efficiency ranges from the literature (Table 4) toapproximate potential nutrient/sediment load removal at each pri-oritized LID site in validation sub-catchment 1 (Fig. 5). TP, TN andTSS load estimates for validation sub-catchment 1 were obtained

compared with TP load.5 Mayer et al. (2007); for buffer widths between 26 and 50 m.6 Zhang, Liu, Zhang, Dahlgren, and Eitzel (2010); range represents 95% confi-

dence interval based on the median Note: all values < 0 are expressed as 0. NI = noinformation.

36 C.J. Martin-Mikle et al. / Landscape and Urban Planning 140 (2015) 29–41

0

20

40

60

80

100

120

140

160

1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33

Num

ber o

f pix

els (

mill

ions

)

phi

bird W

sp

3

3

ae

Ft

Topogra

Fig. 4. Distribution of topographic index values across the Lake Thunder

ub-catchment loads to determine the removal potential (%) ofrioritized LID sites in validation sub-catchment 1.

. Results

.1. Prioritized HSAs

Across the 666 million grid cells (N) topographic index valuescross the Lake Thunderbird Watershed ranged from 1–33 andxhibited a near normal distribution with a slight right skew, a

ig. 5. Spatial distribution of hydrologically sensitive areas (HSAs) in the Lake Thunderbihan 1.5 standard deviations above the mean and cover approximately 6% of the total wa

c Index Values

atershed. Values above 9.5 are 1.5 standard deviations above the mean.

mean of 6.67 and a standard deviation of 2.12 (Fig. 4). HSAs corre-sponded to index values greater than 1.5 standard deviations abovethe watershed mean (8.79) (i.e., approximately 6% of the Lake Thun-derbird Watershed) (Fig. 5). A total of 140 HSAs were prioritizedacross the Lake Thunderbird Watershed for LID placement based onthe topographic index (�*) (Eq. (4)) derived from drainage area per

unit contour length (˛), slope (ˇ), soil hydraulic conductivity (Ks),soil depth to restrictive layer (D) and impervious surface area (ISA).For the 140 LID sites, these relative values had the range of 10–33(�*), 0–0.7 km2/m (˛), 0–7 degrees (ˇ), 2–38 m−1 (Ks), 53–200 cm

rd Watershed. HSAs were defined as regions with topographic index values greatertershed area. The two validation sub-catchments are outlined in black.

C.J. Martin-Mikle et al. / Landscape and Urban Planning 140 (2015) 29–41 37

F rbird

v

(sdc

fitdbrc

ig. 6. Sub-catchment 1 used to verify prioritized LID sites located in Lake Thundeerified as suitable for LID through on-the-ground visits.

D) and the proportion of the pixel (0–1) covered by imperviousurfaces (ISA). A total of 17 sites in two sub-catchments were vali-ated (Figs. 5 and 6); one local-scale, five intermediate-scale, nineatchment-scale, and two reach-scale sites.

The suitability of each validation site was verified througheld investigation (Fig. 6). Placement of LID was as intended; six-een of the seventeen sites were suitable for LID (94 ± 5.7%). The

rainage area identified for local-scale LID placement containeduildings and roads where rain barrels and porous pavement couldeduce runoff volume and nutrient/sediment loads. All prioritizedatchment-scale LID sites were located in low-lying areas that could

Watershed in Moore, Oklahoma. Eleven potential LID sites were located and each

facilitate ponding and to which it would be possible to route runofffrom surrounding areas. Additionally, each site was large enough toaccommodate detention or retention ponds. Both sites prioritizedfor reach-scale LID were ideal for riparian buffer placement; exist-ing land use and land cover within 30-m on either side of the streamnetwork was conducive to planting vegetation to slow and filterrunoff. One intermediate-scale LID location was ruled out because

the site was partially covered by a parking lot, an instance in whichland cover was likely misrepresented as a result of the relativelycoarse resolution of the 10-m impervious surface data. The par-king lot was not masked from this analysis as unsuitable land use

38 C.J. Martin-Mikle et al. / Landscape and Urban Planning 140 (2015) 29–41

F resol( ricultc

bfi1mewo

3

aIswNdiFtTgauw

3t

tTr(2cst(t(

ig. 7. Land cover data comparison. To compare sites prioritized for LID using a highFry et al., 2011), a 1 m resolution land cover map was developed from National Agity-level land use lend a similar level of detail to the 1 m land cover map.

ecause parking lots are not included in either the roads or buildingootprint land use data layers. Nevertheless, there was still room tomplement LID. However, because the available space for 1 of the7 LID sites was smaller on-the-ground than as denoted by the sub-odel, the error rate associated with the ability of this approach to

ffectively eliminate unsuitable land use and land cover (Eq. (5))as approximately 5.7% (i.e., this approach correctly sites LID 94%

f the time).

.2. Accuracy of 1 m land cover map

Accuracy of the 1 m land cover map (Fig. 7) was assessed usingpproximately 1000 ground observations based on Google Earthmagery. An overall accuracy ((the number of pixels correctly clas-ified/the total number of pixels) × 100) of 89.9% was achieved,hich is higher than the 78% overall accuracy achieved for the 30 mational Land Cover Database which is used as the 30 m land coverataset in this study (Wickham et al., 2013), and higher than what

s generally considered an accurate land cover classification (85%,oody, 2002). The overall accuracy represents the percent of pixelshat were classified into the correct ground truth land cover class.here was some confusion between impervious surfaces and barrenround, and between vegetation shadows and building shadows,nd water. However, the 1 m land cover map effectively depictedrban land use features (e.g., buildings, driveways and roads), asell as land cover (e.g., individual trees, vegetation type).

.3. Nutrient removal efficiencies associated with prioritized LIDechniques

Assuming the lower limits of nutrient/sediment input (Table 5)o the 12.5 km2 validation sub-catchment 1 (kg/yr), 16–84P, 22–94 TN, and 16,323–25,692 TSS may potentially beemoved. Assuming the upper limits of nutrient/sediment inputTable 5) to the sub-catchment (kg/yr): 25–129 P, 33–141 N, and2,358–35,192 TSS may potentially be removed. These values indi-ate that by implementing LID at the 11 validated sites acrossub-catchment 1 (which address runoff from approximately 25% of

he sub-catchment), between 3–16% (TP), 3–15% (TN) and 11–17%TSS), with respect to the upper and lower limits of input, of theotal sub-catchment concentrations could be removed annuallyTable 5).

ution land cover to sites identified using readily-available 30-meter land cover dataural Imagery Program photographs. The combination of 30-m land cover data and

4. Discussions

The LID-siting approach presented here prioritizes HSAs asthe foundation for effective LID implementation across mixed-usewatersheds. The advantages of this approach include a systematicand reproducible methodology based on publically-available datafor siting LID at the watershed scale. As LID gains popularity forits ability to mitigate the effects of altered land use on watershedprocesses, a spatially-explicit, mechanistic approach for prioriti-zing LID sites in urban areas will be useful for watershed managers,landscape architects, and urban planners (Palmer, 2009; Walshet al., 2005). Most importantly, knowledge of where LID would bemost effective can inform cost-effective land-use planning deci-sions (Montalto et al., 2007; Strauss et al., 2007).

Our approach expands upon HSA-based techniques previouslyshown to effectively reduce land use impacts to water quality andstream health (Agnew et al., 2006; Ambroise, Beven, & Freer, 1996;Qiu, 2009; Walter et al., 2000) by identifying HSAs in the complexurban environment. Our study demonstrated that using publically-available data, we were able to depict areas likely to generate runoffin mixed-use watersheds through a slight, yet important modifica-tion of the topographic index that accounts for decreased soil waterstorage capacity as a result of impervious surfaces. The inclusion ofhigher resolution data (i.e., soil, DEM, impervious surface) wouldfurther improve the ability of this approach to prioritize LID at thewatershed scale. Topographic index values across the study areawere consistent with Qiu (2009), who obtained values between 1and 28. The Lake Thunderbird Watershed exhibited values between1 and 33, with most values between 5 and 8. We prioritized HSAsdefined by topographic index values greater than 1.5 standard devi-ations above the mean as ideal LID sites.

4.1. Distribution of HSAs

In the Lake Thunderbird watershed, HSAs were more promi-nent in the western third of the watershed (Fig. 5) due primarilyto the low infiltration capacity that results from the clayey soilsand large impervious surface areas of this region (9.58% ISA); bothwhich lead to high volumes of runoff during intense rainfall events(Figs. 1 and 2). The eastern two-thirds of the watershed had more

sandy soils and relatively little urban development (1.28% ISA).These findings are consistent with the OCC (2008), who found thatdeveloped areas in the western third of the watershed contributedthe most water and nutrient runoff to Lake Thunderbird.

C.J. Martin-Mikle et al. / Landscape and Urban Planning 140 (2015) 29–41 39

Table 5Nutrient removal estimates for range of nutrient input and range of LID efficiency removal.

Removal (kg/yr) TP input TN input TSS input

Lower limit Upper limit Lower limit Upper limit Lower limit Upper limit

Lower limit 16 25 22 33 16,323 22,358Upper limit 84 129 94 141 25,692 35,192

3

N

4s

vstnees2adiu(piwaLgeac1nspgsb

F3w

Sub-catchment % reduction 3–16%

ote: TP = total phosphorus; TN = total nitrogen; TSS = total suspended solids.

.2. Nutrient and sediment removal capabilities of prioritized LIDites

The nutrient and sediment removal estimates calculated foralidation of sub-catchment 1 indicate that individual LID canignificantly reduce TP, TN and TSS at both the site level and athe catchment scale. The necessity for large-scale reductions inutrient input to the Lake Thunderbird reservoir has driven sev-ral ongoing LID pilot-projects in the Norman area; all of whichmphasize intensive LID implementation at the neightborhoodcale (R. Coffman, personal communication, April 8, 2013; OCC,008). Additionally, recent studies have examined the effect ofpplying management practices such as LID to the entire Lake Thun-erbird Watershed (i.e., not only HSAs). The OCC (2008) found that

mplementing management practices in all locations where landse and land cover permit would result in a 74% reduction in TPassuming the lower limits of nutrient input). While this exam-le illustrates the potential for LID to significantly reduce nutrient

nput to receiving water bodies, the feasibility of this scenarioould depend on policies and resources to implement LID practices

s there is usually a lmited amount of funding for implementingID (Muthukrishnan & Field, 2004; Palmer, 2009). Additionally, theeographic placement of various types of LID impacts both removalfficiency and cost. We demonstrate that by strategically selectingnd placing LID techniques that cover less than 1% of the sub-atchment (Fig. 6), 25% of runoff from validation sub-catchment

(12.5 km2) is addressed, leading to reductions in sediment andutrient loads of as great as 15%. In theory, numerous suitable LIDites can be identified. However, cost effectiveness declines as high-

riority sites (i.e., locations with high topographic index values)ive way to lower priority sites (Fig. 8). We focused on the priorityites for LID in order to demonstrate where, along a continuum ofenefits, the greatest return would be realized.

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

0 5 10

% w

ater

shed

are

a dr

aine

d by

pot

en�a

l LI

D si

te

Topo gr

ig. 8. Relationship between the topographic index of a potential LID site and percentage000 randomly selected cells. This figure illustrates that placing LID at sites with high topith comparatively few LID projects. Percent of drained watershed area is one of five dat

–15% 11–17%

4.3. Land cover data resolution

The 1 m land cover map provided more detail than the 30-m landcover map, however, after combining the 30-m land cover map withcity-level land use data, the difference between the approaches wasnegligible (Fig. 7). As a result, land cover resolution did not play asignificant role in effectively siting LID. Land cover did affect thesize of LID-sites identified by both the intermediate and catchment-scale LID-siting modules. For these LID sites, the more accurateimpervious surface information given by the 1 m land cover data(i.e., location of driveways and parking lots) improved LID site iden-tification. The validation process resulted in one of the 17 prioritysites being smaller than initially estimated due to coarser resolu-tion land cover data. It is important to note that the usefulnessof 30-m land cover data is dependent upon availability of city-level land use data. In the absence of a spatially-explicit land usedataset, 1 m land cover data (as opposed to 30-m data) would likelybe needed to properly map useful LID sites. The development of a1 m land cover map, however, requires additional expertise andresources. Our overall approach was intended to be widely usefulby employing publically-available data.

A higher-resolution hydrography dataset may have shiftedsome of the priority LID sites upstream, particularly at the localscale. The hydrography dataset was derived from the DEM basedon the 1:24,000 NHD dataset, which underestimates the actualdrainage network considerably, especially headwaters (Julian,Elmore, & Guinn, 2012). A higher resolution or more accuratestream network (e.g., Wilgruber et al., unpublished data) wouldbetter characterize HSA locations and thus, result in more effi-

cient LID prioritization (i.e., sites would likely be in same locations,but siting would be more precise). Regardless of the resolutionof the hydrography dataset, ground truthing would be neces-sary to finalize LID locations. Use of the “best available” NHD

15 20 25

aphic ind ex

of Lake Thunderbird Watershed that contributes runoff to each site/cell based onographic index values is cost-effective because large drainage areas are addressed

a layers used to calculate the topographic index.

4 and U

dm

4

jcegrfiair

tftBsPwgihcanlwawu

5

Suu2itmmnbY

ehrpic3RELcrw

d

0 C.J. Martin-Mikle et al. / Landscape

ataset here, however, facilitates widespread use of the LID-sitingethod.

.4. Applicability of approach in other regions

From a mechanistic perspective, our approach should performust as well in any region with similar runoff processes. We adviseaution when applying our method to watersheds with differ-nt hydrologic drivers. For example, watersheds dominated byroundwater hydrology have different runoff processes and flowegimes, which would need to be incorporated into the modi-ed topographic index. Along this rationale, precipitation intensitynd amount may need to be included in the modified topographicndex for watersheds that encompass vastly different precipitationegimes.

The systematic framework described here is also applicableo any region. Our approach was developed based exclusively onreely available data to ensure access by potential users. In addi-ion, all data processing was performed within ESRI ArcGIS Modeluilder, a function that enables models to be exported to pythoncript and run in open-source software programs such as GRASS,ython, and R. Some aspects of our approach may vary amongatersheds. For example, we define HSAs as regions with topo-

raphic index values > 1.5 standard deviations above the mean. Todentify the same land area of HSAs in a different watershed asydrologically sensitive, the range of topographic index values mayhange. In general, topographic index values > 1.5 standard devi-tions above the mean in watersheds with more dense streametworks will correspond to more stream network area and less

and area. Therefore, to identify the same land area as HSAs, aider range of topographic index values is necessary. Our system-

tic approach for prioritizing LID sites has the potential to facilitateidespread adoption of LID to mitigate the effects of urban landse on stream ecosystems across diverse landscapes.

. Future implications

The potential that LID holds to remediate the ‘Urban Streamyndrome’, the consistent degradation of stream ecosystems inrban catchments (Walsh et al., 2005), remains understated andnderstudied (Burns, Fletcher, Walsh, Ladson, & Hatt, 2012; Palmer,009). Future research is needed to investigate the potential for

mproving our prioritization approach by applying high resolu-ion ground surface topography derived from LiDAR to provide

ore accurate drainage network datasets in the urban environ-ent. However, the challenge of simulating underground drainage

etworks in urban areas remains, due in part to unregulated streamurial during development (Elmore & Kaushal, 2008; Stammler,ates, & Bailey, 2013).

Research is also needed to expand this approach so that LIDfforts that address water quality by facilitating infiltration, andigh runoff volume via water retention, can be constructed toesolve site-specific and watershed-specific problems. For exam-le, in addition to facilitating watershed-level initiatives aimed at

mproving stream quality, LID may assist local municipalities inomplying with federal regulatory considerations such as section03(d) of the Clean Water Act, section 6217 of the Coastal Zone Acteauthorization Amendments, and critical habitat areas under thendangered Species Act (Clar et al., 2002) effectively and affordably.ID also may be used to comply with consent decrees that manyities face from the U.S. Justice Department that obligate cities to

evamp their combined sewer overflow (CSO) systems to complyith water quality standards (Smith & Hutchinson, 2001).

At a time when urban land use is a leading cause of streamegradation (Paul & Meyer, 2001; OCC, 2008), it is critical to

rban Planning 140 (2015) 29–41

develop an effective approach for siting LID in urban areas. The LID-prioritization approach presented here spans municipal boundariesto assist in developing watershed-scale LID implementation plans.The benefits from watershed-scale LID implementation includeboth watershed coverage and cost-effectiveness. Additionally, ourstudy suggests that readily-available data can be sufficient for thepurpose of siting LID within the complex urban environment. Ourapproach for prioritizing LID in urban watersheds not only provideswater managers, landscape architects, and urban planners with avaluable tool, it also provides communities an affordable means toevaluate their most important resources.

This research was supported by the University of OklahomaFoundation Fellowship and U.S. Environmental Protection AgencyNNEMS Award [2012–301]. The information in this document hasbeen subjected to U.S. Environmental Protection Agency peer andadministrative review, and it has been approved for publication as aU.S. Environmental Protection Agency document. Mention of tradenames or commercial products does not constitute endorsementor recommendation for use. We thank Robert Nairn for improvingearlier drafts of this manuscript.

Acknowledgments

This research was supported by the University of OklahomaFoundation Fellowship and U.S. Environmental Protection AgencyNNEMS Award [2012-301]. The information in this document hasbeen subjected to U.S. Environmental Protection Agency peer andadministrative review, and it has been approved for publication as aU.S. Environmental Protection Agency document. Mention of tradenames or commercial products does not constitute endorsementor recommendation for use. We thank Robert Nairn for improvingearlier drafts of this manuscript.

References

Agnew, L. J., Lyon, S., Gérard-Marchant, P., Collins, V. B., Lembo, A. J., Steenhuis,T. S., et al. (2006). Identifying hydrologically sensitive areas: Bridging the gapbetween science and application. Journal of Environmental Management, 78(1),63–76. http://dx.doi.org/10.1016/j.jenvman.2005.04.021

Ambroise, B., Beven, K., & Freer, J. (1996). Toward a generalization of the TOPMODELconcepts: Topographic indices of hydrological similarity. Water ResourcesResearch, 32(7), 2135–2145. http://dx.doi.org/10.1029/95wr03716

Berry, J. K., Delgado, J. A., Khosla, R., & Pierce, F. (2003). Precision conservationfor environmental sustainability. Journal of Soil and Water Conservation, 58(6),332–339.

Beven, K., & Kirkby, M. J. (1979). A physically based, variable contributing area modelof basin hydrology/Un modèle à base physique de zone d’appel variable del’hydrologie du bassin versant. Hydrological Sciences Journal, 24(1), 43–69.

Burns, M. J., Fletcher, T. D., Walsh, C. J., Ladson, A. R., & Hatt, B. E. (2012). Hydrologicshortcomings of conventional urban stormwater management and opportuni-ties for reform. Landscape and Urban Planning, 105(3), 230–240. http://dx.doi.org/10.1016/j.landurbplan.2011.12.012

Clar, M., Barfield, B., & Yu, S. (2002). Considerations in the design of treatmentbest management practices to improve water quality, US EPA. In EPA/600/R-03.Cincinnati, OH: Office of Research and Development (ORD).

Collins, K. A., Lawrence, T. J., Stander, E. K., Jontos, R. J., Kaushal, S. S., Newcomer, T.A., et al. (2010). Opportunities and challenges for managing nitrogen in urbanstormwater: A review and synthesis. Ecological Engineering, 36(11), 1507–1519.http://dx.doi.org/10.1016/j.ecoleng.2010.03.015

Craig, L. S., Palmer, M. A., Richardson, D. C., Filoso, S., Bernhardt, E. S., Bledsoe, B.P., et al. (2008). Stream restoration strategies for reducing river nitrogen loads.Frontiers in Ecology and the Environment, 6(10), 529–538. http://dx.doi.org/10.1890/070080

Delgado, J. A., & Berry, J. K. (2008). Advances in precision conservation. In L. S. Donald(Ed.), Advances in Agronoour (98) (pp. 1–44). http://dx.doi.org/10.1016/S0065-2113(08)00201-0

Dietz, M. (2007). Low impact development practices: A review of current researchand recommendations for future directions. Water, Air, and Soil Pollution,186(1–4), 351–363. http://dx.doi.org/10.1007/s11270-007-9484-z

Elmore, A. J., & Kaushal, S. S. (2008). Disappearing headwaters: Patterns of stream

burial due to urbanization. Frontiers in Ecology and the Environment, 6(6),308–312. http://dx.doi.org/10.1890/070101

Foody, G. M. (2002). Status of land cover classification accuracy assessment.Remote Sensing of Environment, 80, 185–201. http://dx.doi.org/10.1016/S0034-4257(01)00295-4

and U

F

F

G

G

G

G

H

H

H

J

J

K

M

M

M

M

M

M

N

OOO

O

O

P

P

C.J. Martin-Mikle et al. / Landscape

ry, J. A., Xian, G., Jin, S., Dewitz, J. A., Homer, C. G., Limin, Y., et al. (2011). Completionof the 2006 national land cover database for the conterminous United States.Photogrammetric Engineering and Remote Sensing, 77(9), 858–864.

uhrer, G. J. (1999). The quality of our nation’s waters: nutrients and pesticides: USGeological Survey Circular.

burek, W. J., Drungil, C. C., Srinivasan, M. S., Needelman, B. A., & Woodward, D.E. (2002). Variable-source-area controls on phosphorus transport: Bridging thegap between research and design. Journal of Soil and Water Conservation, 57(6),534–543.

eosyntec Consultants and Wright Water Engineers. (2012). Pollutant cat-egory summary statistical addendum: TSS, bacteria, nutrients and metals,international stormwater BMP database (1996–2012). Retrieved from〈www.bmpdatabase.org〉.

esch, D. B. (2007). The national elevation dataset. In D. Maune (Ed.), Digital ele-vation model technologies and applications: The DEM users manual (2nd ed., 98,pp. 99–118). Bethesda, Maryland: American Society for Photogrammetry andRemote Sensing.

old, A. J., Groffman, P. M., Addy, K., Kellogg, D. Q., Stolt, M., & Rosenblatt, A. E.(2001). Landscape attributes as controls on groithd water nitrate removal capac-ity of riparian zones1. JAWRA Journal of the American Water Resources Association,37(6), 1457–1464. http://dx.doi.org/10.1111/j.1752-1688.2001.tb03652.x

ewlett, J. D., & Hibbert, A. R. (1967). Factors affecting the response of small water-sheds to precipitation in humid areas. In W. E. Sopper, & H. W. Lull (Eds.),Forest hydrology (pp. 275–290). Oxford, UK: Symposium Publication Division,Pergamon Press.

offman, R. S., Capel, P. D., & Larson, S. J. (2000). Comparison of pesticides in eightU.S. urban streams. Environmental Toxicology and Chemistry, 19(9), 2249–2258.http://dx.doi.org/10.1002/etc.5620190915

offmann, C. C., Kjaergaard, C., Uusi-Kämppä, J., Hansen, H. C. B., & Kronvang, B.(2009). Phosphorus retention in riparian buffers: Review of their efficiency.Journal of Environmental Quality, 38(5), 1942–1955. http://dx.doi.org/10.2134/jeq2008.0087

ulian, J. P., Elmore, A. J., & Guinn, S. M. (2012). Channel head locations in forestedwatersheds across the mid-Atlantic United States: A physiographic analysis.Geomorphology, 177, 194–203. http://dx.doi.org/10.1016/j.geomorph.2012.07.029

ulian, J. P., & Torres, R. (2006). Hydraulic erosion of cohesive riverbanks. Geomor-phology, 76(1), 193–206. http://dx.doi.org/10.1016/j.geomorph.2005.11.003

aushal, S., & Belt, K. (2012). The urban watershed continuum: Evolving spatial andtemporal dimensions. Urban Ecosystems, 15(2), 409–435. http://dx.doi.org/10.1007/s11252-012-0226-7

ashburn, S. L., Ryter, D. W., Neel, C. R., Smith, S. J., & Magers, J. S. (2013).Hydrogeology and simulation of groundwater flow in the Central Oklahoma(Garber-Wellington) Aquifer. In Oklahoma, 1987–2009, and simulation of avail-able water storage, 2010–2059. U.S. Geological Survey Scientific InvestigationsReport 2013-5219. http://dx.doi.org/10.3133/sir20135219

ayer, P. M. (2005). Riparian buffer width, vegetative cover, and nitrogen removaleffectiveness: A review of current science and regulations. (October 2005,EPA/600/R-05/).

ayer, P. M., Reynolds, S. K., McCutchen, M. D., & Canfield, T. J. (2007). Meta-analysisof nitrogen removal in riparian buffers. Journal of Environmental Quality, 36(4),1172–1180. http://dx.doi.org/10.2134/jeq2006.0462

ontalto, F., Behr, C., Alfredo, K., Wolf, M., Arye, M., & Walsh, M. (2007). Rapidassessment of the cost-effectiveness of low impact development for CSO con-trol. Landscape and Urban Planning, 82(3), 117–131. http://dx.doi.org/10.1016/j.landurbplan.2007.02.004

oore, I. D., Grayson, R., & Ladson, A. (1991). Digital terrain modelling: A review ofhydrological, geomorphological, and biological applications. Hydrological Pro-cesses, 5(1), 3–30. http://dx.doi.org/10.1002/hyp.3360050103

uthukrishnan, S., & Field, R. (2004). The use of best management practices (BMPs)in urban watersheds. In EPA/600/R-04. National Risk Management ResearchLaboratory, Office of Research and Development, US Environmental ProtectionAgency.

RCS. (2012). Soil survey geographic (SSURGO) database. Retrieved November 18,2012, from 〈http://websoilsurvey.sc.egov.usda.gov/〉.

DEQ. (2008). 2008 Integrated report: water quality in Oklahoma.DEQ. (2010). 2010 Integrated report: water quality in Oklahoma.klahoma Conservation Commission, Water Quality Division. (2008). Lake Thunder-

bird watershed analysis and water quality evaluation report 15976. Retrieved from〈http://digitalprairie.ok.gov/cdm/ref/collection/stgovpub/id/15976〉 (15 p.).

mernik, J. M. (1976). The influence of land use on stream nutrient levels. US Envi-ronmental Protection Agency, Office of Research and Development, CorvallisEnvironmental Research Laboratory, Eutrophication Survey Branch (76).

sborne, L. L., & Kovacic, D. A. (1993). Riparian vegetated buffer strips in water-quality restoration and stream management. Freshwater Biology, 29(2), 243–258.http://dx.doi.org/10.1111/j.1365-2427.1993.tb00761.x

almer, M. (2009). Reforming watershed restoration: Science in need of applicationand applications in need of science. Estuaries and Coasts, 32(1), 1–17. http://dx.doi.org/10.1007/s12237-008-9129-5

asseport, E., Vidon, P., Forshay, K. J., Harris, L., Kaushal, S. S., Kellogg, D. Q., et al.(2013). Ecological engineering practices for the reduction of excess N in human

rban Planning 140 (2015) 29–41 41

influenced landscapes: A guide for watershed managers. Environmental Man-agement, 51, 392–413. http://dx.doi.org/10.1007/s00267-012-9970-y

Paul, M. J., & Meyer, J. L. (2001). Streams in the urban landscape. Annual Review of Ecol-ogy and Systematics, 32, 333–365. http://dx.doi.org/10.1146/annurev.ecolsys.32.081501.114040

Pyke, C., Warren, M. P., Johnson, T., LaGro, J., Jr., Scharfenberg, J., Groth, P., et al.(2011). Assessment of low impact development for managing stormwater withchanging precipitation due to climate change. Landscape and Urban Planning,103(2), 166–173. http://dx.doi.org/10.1016/j.landurbplan.2011.07.006

Qiu, Z. (2009). Assessing critical source areas in watersheds for conservation bufferplanning and riparian restoration. Environmental Management, 44(5), 968–980.http://dx.doi.org/10.1007/s00267-009-9380

Richards, J. A., & Jia, X. (1999). Remote sensing digital image analysis (3). Berlin:Springer.

Roy, A., Wenger, S., Fletcher, T., Walsh, C., Ladson, A., Shuster, W., et al. (2008).Impediments and solutions to sustainable, Watershed-Scale Urban Stormwa-ter Management: Lessons from Australia and the United States. EnvironmentalManagement, 42(2), 344–359. http://dx.doi.org/10.1007/s00267-008-9119-1

Smith, D. C., & Hutchinson, R. (2001). Case study: City of Atlanta CSO program uti-lization of GIS and SSES investigations to develop a watershed approach tosewer separation planning storm water management. Proceedings of the WaterEnvironment Federation, 2001(9), 361–383.

Stammler, K. L., Yates, A. G., & Bailey, R. C. (2013). Buried streams: Uncoveringa potential threat to aquatic ecosystems. Landscape and Urban Planning, 114,37–41. http://dx.doi.org/10.1016/j.landurbplan.2013.02.008

Strahler, A. N. (1957). Quantitative analysis of watershed geomorphology. Civil Engi-neering, 101, 1258–1262.

Strauss, P., Leone, A., Ripa, M., Turpin, N., Lescot, J. M., & Laplana, R. (2007). Usingcritical source areas for targeting cost-effective best management practices tomitigate phosphorus and sediment transfer at the watershed scale. Soil Useand Management, 23(s1), 144–153. http://dx.doi.org/10.1111/j.1475-2743.2007.00118.x

Urbonas, B. (2000). Assessment of stormwater best management practice effective-ness. In Innovative urban wet-weather flow management systems.

Urbonas, B., & Stahre, P. (1993). Stormwater: best management practices and detentionfor water quality, drainage, and CSO management.

U.S. Census Bureau. (2012). Data file from the Geography Division based on theTIGER/Geographic Identification Code Scheme (TIGER/GICS) computer file. RetrievedJanuary 3, 2014, from 〈http://www.census.gov/geo/www/tiger/index.html〉.

USCB. (2013). Topologically integrated geographic encoding and referencing(TIGER) products. Retrieved from 〈http://www.census.gov/geo/maps-data/data/tiger.html〉.

USDA. (2011). Major uses of land in the United States. Retrieved February 12, 2014,from 〈http://www.ers.usda.gov/data-products/major-land-uses.aspx#25984〉.

USEPA. (2013). Green infrastructure modeling tools. Retrieved January 15, 2014, from〈http://water.epa.gov/infrastructure/greeninfrastructure/gi modelingtools.cfm#simpleModels〉.

USEPA. (2014). National menu of stormwater best management practices. RetrievedJanuary 7, 2014, from 〈http://water.epa.gov/polwaste/npdes/swbmp/〉.

USGS. (2012). National hydrography dataset—based on 1:24,000-scale USGStopographic map content. 〈http://viewer.nationalmap.gov/viewer/〉 (accessedOctober 19 2012).

Van Roon, M. (2005). Emerging approaches to urban ecosystem management: Thepotential of low impact urban design and development principles. Journal ofEnvironmental Assessment Policy and Management, 07(01), 125–148. http://dx.doi.org/10.1142/s1464333205001943

Walsh, C. J., Fletcher, T. D., & Ladson, A. R. (2005). Stream restoration in urbancatchments through redesigning stormwater systems: Looking to the catch-ment to save the stream. Journal of the North American Benthological Society,24(3), 690–705. http://dx.doi.org/10.1899/04-020.1

Walsh, C. J., & Kunapo, J. (2009). The importance of upland flow paths in determiningurban effects on stream ecosystems. Journal of the North American BenthologicalSociety, 28(4), 977–990. http://dx.doi.org/10.1899/08-161.1

Walter, M. T., Steenhuis, T. S., Mehta, V. K., Thongs, D., Zion, M., & Schneiderman, E.(2002). Refined conceptualization of TOPMODEL for shallow subsurface flows.Hydrological Processes, 16(10), 2041–2046. http://dx.doi.org/10.1002/hyp.5030

Walter, M. T., Walter, M. F., Brooks, E. S., Steenhuis, T. S., Boll, J., & Weiler, K. (2000).Hydrologically sensitive areas: Variable source area hydrology implications forwater quality risk assessment. Journal of Soil and Water Conservation, 55(3),277–284.

Wickham, J. D., Stehman, S. V., Gass, L., Dewitz, L., Fry, J. A., & Wade, T. G. (2013).Acuracy assessment of NLCD 2006 land cover and impervius surface. RemoteSensing of Environment, 130, 294–304. http://dx.doi.org/10.1016/j.rse.2012.12.001

Wolman, M. G., & Schick, A. P. (1967). Effects of construction on fluvial sediment,urban and suburban areas of Maryland. Water Resources Research, 3(2), 451–464.

http://dx.doi.org/10.1029/WR003i002p00451

Zhang, X., Liu, X., Zhang, M., Dahlgren, R. A., & Eitzel, M. (2010). A review of vegetatedbuffers and a meta-analysis of their mitigation efficacy in reducing nonpointsource pollution. Journal of Environmental Quality, 39(1), 76–84. http://dx.doi.org/10.2134/jeq2008.0496