Remote Sensing Letters Characterizing tree canopy loss using multi-source GIS data in Central Massachusetts, USA PLEASE SCROLL DOWN FOR ARTICLE

This article was downloaded by: [Clark University]On: 22 November 2013, At: 09:30Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Remote Sensing LettersPublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/trsl20

Characterizing tree canopy loss usingmulti-source GIS data in CentralMassachusetts, USAAndrew E. Hostetlera, John Rogana, Deborah Martina, VernaDeLauerb & Jarlath O’Neil-Dunnec

a Graduate School of Geography, Clark University, Worcester01610, USAb George Perkins Marsh Institute, Clark University, Worcester01610, USAc Rubenstein School of Environment and Natural Resources,University of Vermont, Burlington 05405, USAPublished online: 20 Nov 2013.

To cite this article: Andrew E. Hostetler, John Rogan, Deborah Martin, Verna DeLauer & JarlathO’Neil-Dunne (2013) Characterizing tree canopy loss using multi-source GIS data in CentralMassachusetts, USA, Remote Sensing Letters, 4:12, 1137-1146

To link to this article: http://dx.doi.org/10.1080/2150704X.2013.852704

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the“Content”) contained in the publications on our platform. However, Taylor & Francis,our agents, and our licensors make no representations or warranties whatsoever as tothe accuracy, completeness, or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Contentshould not be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions, claims,proceedings, demands, costs, expenses, damages, and other liabilities whatsoever orhowsoever caused arising directly or indirectly in connection with, in relation to or arisingout of the use of the Content.

This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &

Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [C

lark

Uni

vers

ity] a

t 09:

30 2

2 N

ovem

ber 2

013

Characterizing tree canopy loss using multi-source GIS data in CentralMassachusetts, USA

ANDREW E. HOSTETLER†, JOHN ROGAN*†, DEBORAH MARTIN†,VERNA DELAUER‡ and JARLATH O’NEIL-DUNNE§

†Graduate School of Geography, Clark University, Worcester 01610, USA‡George Perkins Marsh Institute, Clark University, Worcester 01610, USA

§Rubenstein School of Environment and Natural Resources, University of Vermont,Burlington 05405, USA

(Received 17 May 2013; accepted 27 September 2013)

Despite numerous ecosystem services provided by urban trees, they are continu-ally threatened by combined natural disturbances, invasive species, developmentand negligent management practices. This research characterizes the amount andcause of tree loss in Worcester, Massachusetts, in the northeast United States, andneighbouring towns between 2008 and 2010 using multi-source remotely sensedimagery and historical land cover maps (1976–2009). Historical land-changeanalysis reveals that proportional forest cover loss in the Worcester Countystudy area exceeds that of the state by 0.26% per year, 67% of which was drivenby the expansion of low-density residential land use. Between 2008 and 2010, 2%of Worcester County’s tree canopy was lost to high- and low-density urbandevelopment (47% of the total loss), United States Department of Agriculture(USDA) tree removal for Asian longhorned beetle (ALB) eradication (25%),timber harvest (15%) and ice storm damage (6%). The use of multi-source geo-graphic information system (GIS) data to characterize tree canopy loss makes it aflexible and replicable method to monitor urban tree health.

1. Introduction

Urban forests consist of patches of woody vegetation and shrub–grassland locatedon both private and public land within municipalities, including green features suchas street trees, forest woodlots, urban parks and trees on residential property (McGeeet al. 2012). Trees are an important component of urban landscapes, providing landsurface temperature regulation (Hardin and Jensen 2007), air and water qualitymaintenance (Plieninger 2012), storm water retention (Pataki et al. 2011) andhome energy usage regulation (Nowak and Greenfield 2012b). Additionally, urbantree canopy has been linked to reduced crime levels (Troy et al. 2005) and increasedreal estate values, reflecting the societal value of benefits rendered from trees(Mansfield et al. 2005).

Despite the recognized importance of urban trees, their health is continuallythreatened by negligent management practices (i.e., lack of maintenance and regularplanting) (Gilbertson and Bradshaw 2012), air and soil pollution (McLaughlin 1998),natural disturbances (i.e., pest infestation, disease and severe weather events)

*Corresponding author. Email: [email protected]

Remote Sensing Letters, 2013Vol. 4, No. 12, 1137–1146, http://dx.doi.org/10.1080/2150704X.2013.852704

© 2013 Taylor & Francis

Dow

nloa

ded

by [C

lark

Uni

vers

ity] a

t 09:

30 2

2 N

ovem

ber 2

013

(Morgenroth and Armstrong 2012) and new urban development within and extend-ing from the urban core (Nowak and Walton 2005, Pokorny 2006). As a result,urban forests in major US cities are estimated to lose 4 million trees (0.9%) each year(Nowak and Greenfield 2012a). Urban tree canopy assessment is potentially a cost-effective monitoring approach, and many cities use it to gauge their progress towardsurban forestry goals (Kenney et al. 2011).

To map urban forests and assess their health, previous research has predominantlyemployed fine spatial resolution (<4 m) digital aerial imagery and satellite imageryfor their ability to distinguish small features and characterize private and inaccessiblelocations (Rogan and Chen 2004, Troy et al. 2005, Li et al. 2012, Nowak andGreenfield 2012a). Additionally, aerial imagery such as US National AgricultureImagery Program (NAIP) imagery (<1 m) is often used in research related to theconterminous US due to its flexibility in data capture times, multitemporal coverageof the continental US and its availability in Google Earth™ (Al-Kofahi et al. 2012,Google Inc. 2013).

Although tree canopy monitoring has been pursued previously using multitem-poral land-cover classifications of aerial and satellite data, identifying causes ofobserved tree removal or canopy loss change can be very challenging. Ascribing acause to tree loss is made difficult by (1) the non-linear relationship between causesof loss and spectral reflectance (Singh 1989) and (2) a paucity or inaccessibility ofsupporting maps or documentation describing these causes of tree canopy loss(Rogan and Chen 2004). One common method for identifying causes of tree canopyloss is immediate assessment of tree damage or loss following events such as windstorms, ice storms and wildfires (King et al. 2005, Kokaly et al. 2007, Wang et al.2010). However, these methods are only applied to single-cause tree loss events andare not aggregated to comprehensively describe tree loss (Rogan et al. 2010). Kyleand Duncan (2012) addressed this problem in their study of historical tree covertrends in Victoria, Australia, by incorporating stakeholder and expert knowledge oftree cover to identify drivers of change such as timber harvest for industrial andprivate use and private and governmental land management. This approach requiredthree workshops involving twenty-nine stakeholders and experts to identify causes oftree canopy change, which, although informative, is a time-consuming process.

This research uses a combination of pre-existing data and maps to develop acomprehensive understanding of tree canopy loss in Worcester, Massachusetts. Thespecific objectives of this research are to (1) identify historical land-cover/use trendsin Central Massachusetts (1976–2009) and (2) determine the spatial distribution ofcauses of tree canopy loss using multiple types of geographic information system(GIS) data within the study area (2008–2010). Anthropogenic (i.e., negligent man-agement, urban expansion, timber harvesting and invasive pest and plant introduc-tion) and natural (i.e., disease and severe weather) forest disturbance have playedimportant historical roles in forest abundance, composition and condition in CentralMassachusetts (Herwitz and Nash 2001). Following these events during the twentiethcentury, maple trees (primarily Norway maple [Acer platanoides], also A. rubrum,A. saccharum and A. saccharinum) were planted to replace downed or removed trees,resulting in a maple tree monoculture. The recent ALB infestation in Worcester (firstreported in 2008) is exacerbated by the high concentration of maple species, whichare the ALB’s preferred hosts. Between 2008 and 2010, approximately 30,000 treeswere removed by the United States Department of Agriculture (USDA) andDepartment of Conservation and Recreation (DCR) from inside the USDA

1138 A.E. Hostetler et al.

Dow

nloa

ded

by [C

lark

Uni

vers

ity] a

t 09:

30 2

2 N

ovem

ber 2

013

established quarantine zone. Also occurring in this same time period (December2008) was an ice storm that was reported to have downed and damaged many treesin Worcester and surrounding towns (Ryan et al. 2008). Timber harvest is alsoknown to play an important role in forest disturbance on both state and privatelyowned land within and surrounding the Worcester study area (Kittredge et al. 2003,Lippit et al. 2008). Additionally, a 100–200% increase in population in Shrewsburyand Boylston and 50–100% population increase in Worcester, Holden and WestBoylston since 1970 and accompanying urban sprawl have contributed to ongoingforest cover loss (Nowak and Walton 2005).

2. Study area

The study area encompasses 337 km2 with elevation ranging from 51 to 422 m and iscomprised of five neighbouring towns in Central Massachusetts: Worcester, Holden,West Boylston, Boylston and Shrewsbury (figure 1). Summer temperatures inCentral Massachusetts peak in July and range from 13°C to 27°C, while wintertemperatures range from !9°C to 4°C and reach their lowest point in January(Dodds and Orwig 2011). Worcester supports a population of 1800 people per km2

and is the second largest city in the New England region after Boston (US Census2010). Although Worcester and Shrewsbury are predominantly urban, the study areais well forested (58%), accounting for expansive forests in Holden and Boylston,contiguous distributed patches of forest, maintained city parks and strips of treeslining streets and surrounding homes throughout Worcester, Shrewsbury and WestBoylston (Rogan et al. 2010). Forest patches are dominated by hardwood species(51% hardwood, 11% conifer and 38% mixed conifer-hardwood), predominantlymaple species (Rogan et al. 2010).

Land cover replacing forest

Study area

Water

0 2 4 8 12km

42° 11! N, 71° 39! W

42° 27! N, 72° 6! W

Interstate highwayALB quarantine zoneTowns

Figure 1. The study area located in Central Massachusetts including the USDA Asianlonghorned beetle quarantine zone boundary (2012).

Mapping amount and cause of urban tree loss 1139

Dow

nloa

ded

by [C

lark

Uni

vers

ity] a

t 09:

30 2

2 N

ovem

ber 2

013

3. Data

Land-cover/use maps produced from 30 m Landsat-5 Thematic Mapper (TM)imagery by the Human Environment Research Observatory (HERO)Massachusetts Forest Monitoring Program (MaFoMP) for the years 1976, 1984,1990, 2000 and 2009 were used to establish a baseline trend of forest cover loss forthe study area (Rogan et al. 2010). The map of land cover/use in 2009 was alsoused to help identify causes of tree canopy loss. Tree canopy data of the study areafor July 2008 and 2010 were provided by the University of Vermont’s SpatialAnalysis Lab. These maps were developed by applying an object-based classifica-tion scheme to 1 m NAIP imagery. A 30 m forest damage map of the December2008 ice storm, developed by using Landsat-5 TM imagery captured in June 2007and 2009, was used to identify tree canopy loss resulting from ice storm damage(MaFoMP 2012). Tree canopy removed for timber harvest was verified using expertopinion from the City of Worcester Parks and Recreation Department. Treecanopy removed by the USDA for ALB eradication was identified using a mapof points indicating USDA identified ALB infested trees, thus removed, in theWorcester County quarantine zone. Publicly available NAIP imagery fromGoogle Earth™ captured in summer 2008 and 2010 was also used to visuallyidentify and verify causes of tree canopy loss.

4. Methods

Identifying tree canopy loss and its causes required the following three steps: (1)analysing land change from 1976 to 2009, (2) quantifying tree canopy loss between2008 and 2010 and (3) ascribing cause to tree canopy loss (2008–2010).

Step 1: Land change analysis was performed using land-cover/use maps for 1976,1984, 1990, 2000 and 2009 (Rogan et al. 2010) to establish a historical baseline offorest cover loss to contextualize recent tree canopy loss (2008–2010). Forest coverloss to non-forest land-cover/use categories was assessed at state, study area andwithin-study area towns (Worcester, Holden, West Boylston, Boylston andShrewsbury). Step 2: Map comparison procedures were employed to reveal canopylosses between 2008 and 2010. Step 3: The ice storm damage map, USDA map ofALB infestations, 2009 MaFoMP land-cover/use map and Google Earth™ imagerywere employed to link causes of tree canopy loss (2008–2010) to six categories: icestorm damage, USDA tree removal, urban sprawl, timber harvest, unknown and maperror (table 1). The 2008 ice storm map was overlaid with the tree canopy loss map toassign values for that cause of tree loss (MaFoMP 2012). Tree canopy loss due toALB eradication was identified as any canopy that intersected points of ALBinfestation between 2008 and 2010.

Tree canopy loss that had not been identified as either ice storm or USDA treeremoval was overlaid and assigned cause using the 2009 MaFoMP land-cover map,which had been reclassified as urban sprawl, unknown or post-2009 tree canopy loss.The polygon-labelling procedure is presented in figure 2. The unknown class repre-sents the land-cover classes that could not be confidently associated with specificcauses of tree canopy loss. The post-2009 tree canopy loss class represents loss wherea forest category still existed in 2009, indicating that the loss occurred between 2009and 2010. The post-2009 tree canopy loss was manually assigned causes of loss usingGoogle Earth™ imagery based on a set of predetermined and progressively devel-oped guidelines (table 1).

1140 A.E. Hostetler et al.

Dow

nloa

ded

by [C

lark

Uni

vers

ity] a

t 09:

30 2

2 N

ovem

ber 2

013

5. Results

5.1 Historic land change 1976–2009

Analysis of historic forest cover loss in Massachusetts using Landsat maps revealedthat 7% (84,464 ha) of existing forest cover in 1976 was lost by 2009. By contrast,16% (2589 ha) of forest cover was lost in the Worcester County study area over thissame time period. Between 2000 and 2009, the study area lost the greatest proportionof its forest cover (7%) out of all other observed time periods between 1976 and 2009.Low-density residential land cover accounted for 67% of forest cover loss since 1976in the entire study area, 71% of forest cover loss in Worcester, 66% in Shrewsbury,68% each in Boylston and West Boylston and 63% in Holden (figure 3).

Legend: Unknown

(c)(b)(a)

Tree canopy polygonPost-2009 tree canopy loss

Tree canopy pointUrban sprawl

42° 19! N, 71° 26! W

Figure 2. Methodology used to ascribe cause of tree canopy loss using the 2009 land-cover/use map, demonstrated using a sample tree canopy loss polygon: (a) The tree canopy losspolygons (outlined in black above) were converted to point centroids (indicated by an astrixabove), (b) cause of tree canopy loss (depicted in green, red and grey above) was extracted ateach point and (c) the cause of tree canopy loss was assigned to the polygon using itsidentification number.

Table 1. Guidelines for the causes of tree canopy loss* and supporting data used in this study.

Class(thematic error) Identification guidelines

Data used, spatial resolution (m),error (±m)

USDA treeremoval (n/a)

Tree canopy loss overlaps points ofALB infestation.

Sites of ALB infestation (USDA),1 m, ±5 m

Ice stormdamage (8%)

Index damage values indicate severedamage.

2008 Ice storm damage map(MAFoMP 2012), 30 m, ±12 m

Urban sprawl(12%)

On, near or clearly connected tourban land cover/use; replaced byurban land cover/use.

MaFoMP land-cover/use maps(Rogan et al. 2010), 30 m andGoogle Earth™ imagery (NAIP),1 m, ±14 m

Timber harvest(n/a)

Contiguous patches of complete orpartial removal; visible access toa road or property.

Google Earth™ imagery (NAIP),1 m, ±4 m

Unknown (n/a) Downed trees are visible on ground;no sign of access to a main roador property; land cover/use doesnot indicate cause.

Google Earth™ imagery (NAIP),1 m, ±4 m

Map error (n/a) No clear sign of tree canopy loss;shading effect; loss polygonmatches cloud cover.

Google Earth™ imagery (NAIP),1 m, ±4 m

*Canopy loss polygons estimated to have ±2 m positional error based on NAIP photos.

Mapping amount and cause of urban tree loss 1141

Dow

nloa

ded

by [C

lark

Uni

vers

ity] a

t 09:

30 2

2 N

ovem

ber 2

013

5.2 Tree canopy loss 2008–2010

Assessment of the NAIP maps reveals that a total of 2% (395 ha) of tree canopy waslost between 2008 and 2010 in the Worcester County study area (figure 4). Worcesterlost the greatest proportion of its 2008 tree canopy (3.6%, 145 ha), while Shrewsburyand Holden both lost 1.9% (57 ha and 131 ha, respectively), West Boylston lost 1.7%(36 ha) and Boylston lost 0.8% (26 ha).

Of the 395 ha of tree canopy lost in the study area between 2008 and 2010, 47%was caused by conversion to developed land types (i.e., bare soil and residential),25% by USDA tree removal for ALB eradication, 15% was due to timber harvestand 6% was caused by damage sustained during the December 2008 ice storm (seefigures 4 and 5). Worcester City experienced the greatest proportion of tree canopyloss due to the USDA tree removal (60%) followed by West Boylston (35%).Shrewsbury experienced the largest proportion of tree canopy loss due to urbansprawl (82%). Holden experienced the highest proportion of tree canopy loss totimber harvesting (33%) and the 2008 ice storm (14%), most likely due to its higherelevation, resulting in high ice accumulation and strong winds that would haveintensified the ice storm damage.

For each data set used in the analysis, horizontal error (m) was calculated as thesum of data error presented in table 1, as is estimated categorical error. Therefore,the total positional error for each canopy loss category is as follows: (1) total errorfor USDA tree removal is ±5 m, derived from NAIP error (±2 m) + GPS error(±3 m); (2) total error for ice storm damage is ±12 or NAIP error (±2 m) + Landsatgeometric RMSE (±10 m); (3) total error for urban sprawl is ±17 m or NAIP error

Land cover replacing forestBare soil

High–density residential

Water

0 2 4 8km

12

Interstate highway

ALB quarantine zone

TownsStudy area

Commercial

Low–density residential

42° 27! N, 72° 6! W

42° 11! N, 71° 39! W

Figure 3. Land-cover/use classes replacing forest cover in the study area between 1976 and2009.

1142 A.E. Hostetler et al.

Dow

nloa

ded

by [C

lark

Uni

vers

ity] a

t 09:

30 2

2 N

ovem

ber 2

013

(±2 m) + Landsat geometric RMSE (±15 m) and (d) total error for timber harvest is±4 m or NAIP error (±2 m) + timber harvest records error (±2 m). The overallpositional error of the canopy loss map is estimated to range between 4 and 17 m,depending on the specific cause of tree loss.

The combination of positional and thematic error in this case study points to theimportance of holistic map error estimation, which obviously cascade and influencehow the end product could be interpreted/used. Clearly, the integration of multi-source GIS data with different magnitudes of positional error, as well as different

100

75

50

25

0

Urban sprawl

USDA tree removal

Error and unknown

Ice storm damage

Timber harvest

Study area

Pro

port

ion

of tr

ee lo

ss (%

)

Study a

rea

Worc

ester

Holden

W. Boy

lston

Boylst

on

Shrewsb

ury

Figure 5. Causes of tree canopy loss in the study area and study area towns 2008–2010.

Land cover replacing forest

Cause of loss

Pro

port

ion

of to

tal t

ree

loss

(%)

Urbansprawl

Are

a (h

a)

0

50

100

150

200

Ice stromTimberharvest

USDA

50

40

30

20

10

0

USDA tree removal

Timber harvest

Water

0 2 4 8km

12

Interstate highwayALB quarantine zoneTownsStudy areaUrban sprawl

Ice storm damage

42° 27! N, 72° 6! W

42° 11! N, 71° 39! W

Figure 4. Tree canopy lost between 2008 and 2010 to urban sprawl, USDA tree removal forAsian longhorned beetle eradication, timber harvest and the December 2008 ice storm.

Mapping amount and cause of urban tree loss 1143

Dow

nloa

ded

by [C

lark

Uni

vers

ity] a

t 09:

30 2

2 N

ovem

ber 2

013

intrinsic spatial resolutions, introduces uncertainty to the various areal estimatesprovided in this letter. Nonetheless, this work is justified because it is valuable foreducating the public as well as decision-makers about the consequences of tree loss aswell as the historical and current policies of land-use planning and urban forestmanagement in Central Massachusetts.

6. Discussion and conclusion

The purpose of this research was to characterize tree canopy loss between 2008 and2010 using multi-source GIS and remotely sensed data. The results provide a morecomprehensive understanding of tree loss in the study area and can be used inactivities such as forest conservation advocacy and urban tree management. Theuse of extant, multi-source GIS data allows this type of research to be performedrelatively quickly at a lower cost when compared to field survey and inventory,which may be important to small conservation organizations and local governments.Future use of this mapping approach in other locations could benefit from catalo-guing imagery and spatial statistics of tree canopy loss events such as severe weather,land clearing activities and logging.

Historical tree canopy loss in the study area, compared to the state, reveals apattern of chronic forest loss at both scales. Proportionally, the Worcester Countystudy area lost more than twice as much forest (16%) as the state of Massachusetts(7%) during the 33-year period. Low-density residential land cover, indicative ofurban sprawl, was the primary cause of forest loss (67%) throughout the study areaduring this historical period.

In recent years, tree canopy in the study area has been lost predominately todeveloped land types (47%), although USDA tree removal for ALB eradication(25%), harvesting timber (15%) and damage from the 2008 ice storm (6%) alsoresulted in recognizable tree canopy loss during this time. Reconciling the data setsused and their variable resolution was made possible by assuming mutual exclusivitybetween causes of change and by independently assessing the effects of each one ontree canopy.

Between 2008 and 2010, the December 2008 ice storm and the USDA removal oftrees for ALB eradication were given a great deal of state media attention. TheBoston Globe newspaper reported that in Holden and Worcester, the ice storm madea number of roads impassable due to downed trees and put residents at risk of fallingtree limbs (Ryan et al. 2008), while the ALB infestation has been presented as aserious threat to Massachusetts’ forest health in USDA news releases since the initialdiscovery of the infestation in 2008 (USDA-APHIS 2013). Although each of thesetree loss events reportedly adversely impacted the towns with power outages anddramatically changed neighbourhood landscapes, the amount of tree loss resultingfrom these events is considerably less than that caused by urban sprawl. Abrupt lossof trees to storm damage and widespread removal of ALB hosts along streets isdramatic and eye-catching compared to the long-term tree loss resulting from urbansprawl and timber harvest. Urban sprawl and timber harvest are also associated witheconomic benefits and are therefore seen in less of a negative light when resulting inlarge-scale tree removal. Additionally, timber harvest goes unnoticed likely becauseit takes place in core forest or on private land out of public view (Kittredge et al.2003). Urban sprawl and timber harvest are chronic causes of tree loss in the studyarea and do not receive attention proportional to their long-term impact due to the

1144 A.E. Hostetler et al.

Dow

nloa

ded

by [C

lark

Uni

vers

ity] a

t 09:

30 2

2 N

ovem

ber 2

013

public’s perception of these events and where they take place. Therefore, theapproach presented in this letter presents a novel way for stakeholders to understandthe causes and potential consequences of urban forest loss in an efficient and morecomprehensive manner.

Acknowledgements

A special thanks to A.J. Shatz and Crystal Fam for their assistance throughout thisresearch process.

Funding

This study was supported by the National Science Foundation (NSF) under grantno. SES-0849985 (REU Site) and the Clark University O’Connor ’78 Endowment.All findings expressed in this material are those of the authors and do not necessarilyreflect the views of the funders.

ReferencesAL-KOFAHI, S., STEELE, C., VANLEEUWEN, D. and ST. HILAIRE, R., 2012, Mapping land

cover in urban residential landscapes using very high spatial resolution aerial photo-graphs. Urban Forestry and Urban Greening, 11, pp. 291–301.

DODDS, K.J. and ORWIG, D.A., 2011, An invasive urban forest pest invades naturalenvironments – Asian longhorned beetle in Northeastern US hardwood forests.Canadian Journal of Forest Research, 41, pp. 1729–1742.

GILBERTSON, P. and BRADSHAW, A.D., 2012, Tree survival in cities: the extent and nature ofthe problem. Arboricultural Journal, 9, p. 1985.

GOOGLE INC., 2013, Google Earth™ 7.0.3.8542. Available online at: http://www.google.com/earth/ (accessed June 2012).

HARDIN, P.J. and JENSEN, R.R., 2007, The effect of urban leaf area on summertime urbansurface kinetic temperatures: a Terre Haute case study. Urban for Urban Green, 6,pp. 63–72.

HERWITZ, E. and NASH, R., 2001, Trees at Risk: Reclaiming an Urban Forest (Worcester,MA: Chandler House Press).

KENNEY, W.A., VAN WASSENAER, P.J.E. and ALEXANDER, L.S., 2011, Criteria andindicators for strategic urban forest planning and management. Arboriculture &Urban Forestry, 37, pp. 108–117.

KING, D.J., OLTHOF, I., PELLIKIA, P.K.E., SEED, E.D. and BUTSON, C., 2005, Modelingand mapping damage to forests from an ice storm using remote sensing and environ-mental data. Natural Hazards, 35, pp. 321–342.

KITTREDGE, D.B. JR, FINLEY, A.O. and FOSTER, D.R., 2003, Timber harvesting as anongoing disturbance in a landscape of diverse ownership. Forest Ecology andManagement, 180, pp. 425–442.

KOKALY, R.F., ROCKWELL, B.W., HAIRE, S.L. and KING, T.V.V., 2007, Characterizationof post-fire surface cover, soils, and burn severity at the Cerro Grande Fire, NewMexico, using hyper-spectral and multispectral remote sensing. Remote Sensing ofEnvironment, 106, pp. 305–325.

KYLE, G. and DUNCAN, D.H., 2012, Arresting the rate of land clearing: change in woodynative vegetation cover in a changing agricultural landscape. Landscape and UrbanPlanning, 106, pp. 165–173.

LI, W., RADKE, J., LIU, D. and GONG, P., 2012, Measuring detailed urban vegetation withmultisource high-resolution remote sensing imagery for environmental design andplanning. Environment and Planning B: Planning and Design, 39, pp. 566–585.

LIPPIT, C.D., ROGAN, J., LI, Z., EASTMAN, J.R. and JONES, T.G., 2008, Mapping selectivelogging in mixed deciduous forest: a comparison of machine learning algorithms.Photogrammetric Engineering and Remote Sensing, 74, pp. 1201–1211.

Mapping amount and cause of urban tree loss 1145

Dow

nloa

ded

by [C

lark

Uni

vers

ity] a

t 09:

30 2

2 N

ovem

ber 2

013

MANSFIELD, C., PATTANAYAK, S.K., MCDOW, W., MCDONALD, R. and PATRICK, H.,2005, Shades of green: measuring the value of urban forests in the housing market.Journal of Forest Economics, 11, pp. 1–23.

MASSACHUSETTS FOREST MONITORING PROGRAM (MaFoMP), 2012, Available online at: http://www.clarku.edu/departments/hero/researcharea/forestchange.cfm (accessed 5 February2013).

MCGEE III, J.A., DAY, S.D., WYNNE, R.H. and WHITE, M.B., 2012, Using geospatialtools to assess the urban tree canopy: decision support for local governments. Journalof Forestry, 110, pp. 275–286.

MCLAUGHLIN, D., 1998, A decade of forest tree monitoring in Canada: evidence of airpollution effects. Environmental Reviews, 6, pp. 151–171.

MORGENROTH, J. and ARMSTRONG, T., 2012, The impact of significant earthquakes onChristchurch, New Zealand’s urban forest. Urban Forestry and Urban Greening, 11,pp. 383–389.

NOWAK, D.J. and GREENFIELD, E.J., 2012a, Tree and impervious cover change in US cities.Journal of Urban Forestry and Urban Greening, 11, pp. 21–30.

NOWAK, D.J. and GREENFIELD, E.J., 2012b, Tree and impervious cover in the UnitedStates. Landscape and Urban Planning, 107, pp. 21–30.

NOWAK, D.J. and WALTON, J.T., 2005, Projected urban growth (2000–2050) and its esti-mated impact on the US forest resource. Journal of Forestry, 103, pp. 383–389.

PATAKI, D.E., CARREIRO, M.M., CHERRIER, J., GRULKE, N.E., JENNINGS, V., PINCETL,S., POUYAT, R.V., WHITLOW, T.H. and ZIPPERER, W.C., 2011, Coupling biogeo-chemical cycles in urban environments: ecosystem services, green solutions, and mis-conceptions. Frontiers in Ecology and the Environment, 9, pp. 27–36.

PLIENINGER, T., 2012, Monitoring directions and rates of change in trees outside foreststhrough multitemporal analysis of map sequences. Applied Geography, 32, pp.566–576.

POKORNY, J.D., 2006, Urban Tree Risk Management: A Community Guide to Program Designand Implementation, pp. 117–123 (St. Paul, MN: United States Department ofAgriculture).

ROGAN, J., BUMBARGER, N., KULAKOWSKI, D., CHRISTMAN, Z.J., RUNFOLA, D.M. andBLANCHARD, S.D., 2010, Improving forest type discrimination with mixed life-formclasses using fuzzy classification thresholds informed by field observations. CanadianJournal of Remote Sensing, 36, pp. 699–708.

ROGAN, J. and CHEN, D., 2004, Remote sensing technology for mapping and monitoringland-cover and land-use change. Progress in Planning, 61, pp. 301–325.

RYAN, A., BALLOU, B.R., ABEL, D., ELLEMENT, J.R. and BAKER, A., 2008, Nearly 1 millionwithout power after ice storm. The Boston Globe.Available online at: http://www.boston.com/news/local/breaking_news/2008/12/ice_storm_leave.html (accessed 23 March 2013).

SINGH, A., 1989, Digital change techniques using remotely-sensed data. International Journalof Remote Sensing, 10, pp. 989–1003.

TROY, A., GROVE, J.M. and O’NEIL-DUNNE, J., 2005, The relationship between treecanopy and crime rates along an urban-rural gradient in the greater Baltimore region.Landscape and Urban Planning, 106, pp. 262–270.

UNITED STATES CENSUS BUREAU, 2010, State & county quickfacts: Massachusetts. Availableonline at: http://quickfacts.census.gov/qfd/states/25000.html (accessed 13October 2012).

UNITEDSTATESDEPARTMENT OFAGRICULTURE (USDA):Animal andPlantHealth Inspection Service(APHIS), 2013, News room: Asian longhorned beetle. Available online at: http://www.aphis.usda.gov/newsroom/hot_issues/alb/alb_news.shtml (accessed 20 January 2013).

WANG, W., QU, J.J., HAO, X., LIU, Y. and STANTURF, J.A., 2010, Post-hurricane forestdamage assessment using satellite remote sensing. Agricultural and ForestMeteorology, 105, pp. 122–132.

1146 A.E. Hostetler et al.

Dow

nloa

ded

by [C

lark

Uni

vers

ity] a

t 09:

30 2

2 N

ovem

ber 2

013