LAKELINE ■ Spring 2005 17

Satellite Tabulation

Marvin Bauer and Bruce Wilson

of Impervious Surface Areas

The past decade has been one ofsteady growth across many areasof North America including

Minnesota, and particularly in our lakeregions and associated communities asexemplified in Figure 1, which is anaerial photo of Spicer, Minnesota (withGreen Lake in the background). Thechange in imperviousness from thecensus time periods of 1990 to 2000,shown in Figure 2, is about 36% (withthe red and yellow colors depicting thehighest areas of imperviousness – keepreading for details) With future growthprojections as much as 100+ percent insome areas over the next 30 years, thereis concern about the impacts to ourwater resources as new developments ofhomes and associated commercial/industrial zones create new stormwaterdischarges to lakes and streams. A majorfactor in determining the potency of thestormwater relates to the amount ofimperviousness – highways and streets,parking lots, and rooftops within thedrainage areas.

Many studies have shown that theamount of impervious surface area in awatershed is related to the alteration ofstreams and loss of fish habitat as wellas degradation of water quality andwetlands (Schueler and Holland 2000).In the case of stream impacts, tenpercent or less of drainage basinimperviousness can begin alteringstreambeds, negatively impactingfisheries and critters (macroinverte-brates), and decreasing importantinfiltration of ground waters.As the amounts of imperviousnessincrease, the runoff velocities andvolumes increase, ripping away thesides of the streams to carry eroded soilsdownstream to sensitive lakes andstreams.

Figure 1. View of impervious surfaces in Spicer, Minnesota with Green Lake in background(MDNR photo provided by Skip Wright).

Until relatively recently, the linkagebetween impervious surfaces (or mostly“car habitat” in the words of TomSchueler of the Center for WatershedProtection) and impacts to receivingwaters was not fully realized. And, evennow, these impacts are only beginning tobe appreciated by developers, citymanagers, councils, and industries asthey implement the U.S. EnvironmentalProtection Agency’s Phase II municipalseparate storm sewer system (MS4s)permits. While managing to prevent/avoid/minimize imperviousness ispreferable, most MS4 communities havehad little time to understand andappreciate the ramifications of increasesin impervious surfaces to receiving

waters. MS4 communities are beingasked to improve stormwatermanagement via stormwater pollutionprevention plans (SWPPP). Eachregulated party determines theappropriate pollution preventionpractices or “best managementpractices” to minimize pollution fortheir specific site. Recommended firststeps in BMP siting include:

1. advanced planning and mapping ofimpervious areas within drainagebasins of sensitive streams and lakes;

2. locating existing runoff managementpractices, potential buffer, andconservation areas that would limitinterconnectedness of impervioussurfaces; and

18 Spring 2005 ■ LAKELINE

3. linking of several BMPs to providesequential treatment of stormwaterrunoff.

The effectiveness of these efforts isimproved by having numerical/narrativeresource management goals for thereceiving water bodies for typicalpollutants. Advanced simulation modelssuch as P8 (Walker 1990) can offerconsiderable assistance in estimating theeffectiveness of BMP treatment systems.

Figure 2. Spicer, Minnesota’s impervious surface percentages in 1990 and 2000 bygeneral categories (0–19%, 20–39%, and 40–100% cover) – keep reading for details.

Until satellite technologies wereavailable, impervious surface wascommonly calculated by using averagepercent impervious values by land useclasses developed from parcel maps orinterpretation of aerial photography. Forexample, single family residential is35% impervious, commercial 85%impervious, industrial 75% impervious,and parks 15% impervious. However,the estimated values may not accuratelyrepresent any particular location due to

building and design variations, and it isexpensive to obtain the land use data.And, if results are needed over largergeographic regions, including multiplecities, there will likely be differences inmethods and dates of data collection thatlead to inconsistencies in results.

Satellite Mapping of ImperviousnessRecently a number of researchers,

including the University of Minnesota’sRemote Sensing and GeospatialAnalysis Laboratory, have investigatedthe potential of satellite remote sensingas a means to accurately andeconomically map impervious surfacearea. The approach developed and testedby the University of Minnesota hasshown that digital multispectral satelliteimagery can be used to accuratelydetermine and map the amount ofimpervious surface area. The methodsdepict the degree of imperviousnessfrom 0 to 100% at the pixel level – 30meters on a side or about a quarter-acrefor Landsat Thematic Mapper (TM)data.

The method is based on a strongrelationship of the “greenness”component of the “tasseled cap”transformation of Landsat TM data tothe amount of green vegetation, andtherefore to the lack of green vegetation,or impervious surfaces, in a pixel (orquarter-acre cells). Although themajority of Landsat pixels in an urbanarea are mixtures of two or more covertypes, this method provides a means toestimate the fraction of each pixel that isimpervious. Following conversion of thesix reflective spectral bands of theLandsat data to greenness, a polynomialregression model of the relationship ofgreenness to percent impervious area isdeveloped. The relationship isdetermined by carefully measuring theamount of impervious surfaces inselected sample areas from high-resolution aerial photographs that areregistered to the Landsat imagery. Anexample of the relationship of greennessto impervious area is shown in Figure 3.Once the regression model is developedfrom a sample of approximately 50 sitesrepresenting different kinds and amountsof impervious areas, it is applied to allof the pixels in the imagery that havebeen classified as urban or developed. Acomparison of measurements from a

LAKELINE ■ Spring 2005 19

Figure 5. Landsat classification of imperviousness for theseven-county Twin Cities Metropolitan area for 2002.

0

20

40

60

80

100P

erc

en

t Im

perv

iou

s A

rea

-60 -40 -20 0 20 40 60

Landsat Greenness

R2 = 0.91

Std. Error = 10.6

Y = 48.4 - 1.16 G – 0.00055 G2

Figure 3. Relationship of Landsat “greenness” toimperviousness for 2002.

R2 = 0.85

Std. Error = 10.7

0

20

40

60

80

100

La

nd

sa

t E

sti

ma

ted

Im

perv

iou

s A

rea (

%)

0 20 40 60 80 100

Measured Impervious Area (%)

Figure 4. Comparison of Landsat estimates and measurementsof imperviousness.

second, independent set of samples tothe Landsat estimates of the amount ofimpervious area is shown in Figure 4.The accuracy or agreement between theLandsat estimates and measurementsfrom the aerial photography has beenquite good – with average accuracy of90% (as measured by R2) and standarderrors of 8–10%.)

The Landsat classification maps canbe transcribed into geographicinformation system (GIS) data layers foruse by common planning and analysis

packages and used to generate maps andsummary statistics as well as inputs fora variety of runoff prediction softwarepackages (from the “simple method” tocomplicated hydrologic runoff models).The satellite products can be developedfor different scale needs such as:• Large county and regional

assessments such as the seven-countyTwin Cities Metropolitan Area(TCMA). An example map of theTCMA for 2002 is shown in Figure 5.For the entire seven-county area, theamount of impervious area increased

from 171,700 acres (9% of the totalarea) in 1986 to 247,000 acres(12.9%) in 2002. The trends inpercent imperviousness for each of theTCMA counties, by ~ five yearintervals, are depicted in Figure 6.

• Smaller scale tracking of changes inimperviousness over ~5 year intervalsfor all 200 municipalities within theTwin Cities Metropolitan Area. Thechanges in impervious area due todevelopment from 1986 to 2002 forWoodbury, a rapidly growing suburbeast of St. Paul, are readily apparentin Figure 7.

0%

5%

10%

15%

20%

25%

30%

35%

Anoka

Carve

r

Dakot

a

Hennip

in

Ramse

ySco

tt

Was

hingt

on

Impe

rvio

us A

rea

(%)

1986199119982002

Figure 6. Changes in percent impervious surface area for the TCMAcounties, 1986 – 2002.

20 Spring 2005 ■ LAKELINE

• Statewide changes in imperviousness– as are being completed forMinnesota at this writing, for thecensus years 1990 and 2000. Work isalso underway to define changes inimpervious cover from 1990 to 2000census periods, for about 20 lakewatersheds in the heart of theMinnesota Lake regions. We’ll reportback to you when that work iscompleted.

ConclusionWe believe the satellite work is an

invaluable new tool for showingcumulative impacts of urbanization. TheGIS data layers and associated maps canbe used to quantify imperviousness bylake and stream drainage areas, modelrunoff characteristics, and begin todevise Best Management Practices(BMPs) for rehabilitation goals orprotective purposes. Landsat satellite-based impervious classification andmapping provides a cost-effective wayto accurately assess the degree ofimpervious cover over large geographicareas and by municipality (e.g.,probably less than $1000 per

Figure 7. Landsat classifications of imperviousness for the City of Woodbury, 1986–2002.

municipality for defining trends overseveral years). One of our greatestchallenges in natural resourcemanagement has been to show how tobalance the competing needs of growthwhile protecting/restoring our watersupport system of lakes, streams, andgroundwater. The satellite remotesensing is helping us in this criticallyimportant task.

ReferencesSchueler, T. and H.K. Holland (Eds.). 2000.

The Importance of Imperviousness.Center for Watershed Protection,Watershed Protection Techniques 1(3):100-111.

Walker, W.W. Jr. 1990. P8 Urban CatchmentModel Program. Prepared for IEP, Inc.Northborough, MA and Narragansett BayProject, Providence, RI.

Marvin Bauer is professor of remote sensingat the University of Minnesota in the Collegeof Natural Resources. He has been involvedwith research and development of satelliteremote sensing for monitoring naturalresources since 1970. He is director of theRemote Sensing and Geospatial AnalysisLaboratory (http://rsl.gis.umn.edu/) andteaches remote sensing classes. He is afellow of the American Society of RemoteSensing and has been editor-in-chief ofRemote Sensing of Environment since 1980.

Bruce Wilson, pastNALMS president, hasbeen a biologist andresearch scientist forthe Minnesota PollutionControl Agency since1979. He has ledseveral water qualityassessment efforts ofmajor lake and river

systems in Minnesota and has providedtechnical assistance to over 30 locally ledwatershed management projects. Presentwork includes satellite remote sensing oflake and river water quality, imperviousnessand stormwater Best Management Practiceeffectiveness monitoring. You can reachBruce at bruce.wilson@pca.state.mn.us x