R

Fp

PE

h

•••••

a

ARRAA

KLPASAU

1

eobrsor

M

0h

Landscape and Urban Planning 121 (2014) 55–64

Contents lists available at ScienceDirect

Landscape and Urban Planning

journa l homepage: www.e lsev ier .com/ locate / landurbplan

esearch Paper

actors driving land use change and forest distribution on the coastallain of Mississippi, USA

eter E. Schweizer, Glenn R. Matlack ∗

nvironmental and Plant Biology, Porter Hall 405, Ohio University, Athens, OH 45701, United States

i g h l i g h t s

Historical aerial photos allow reconstruction of land use history at a Gulf Coast site.Land cover change has been synchronized by the lumber boom of the early 20th century.A cycle of short-rotation forestry is imposed on the longer cycle of regrowth.A wave-front of urban expansion has been replaced by a pattern of nucleation.Modern land cover is dictated by proximity to an urban area rather than by regional trends.

r t i c l e i n f o

rticle history:eceived 12 June 2013eceived in revised form 8 September 2013ccepted 9 September 2013vailable online 8 October 2013

eywords:and use historyinus palustrisbandonmentuccessiongriculture

a b s t r a c t

Forest distribution is controlled by broad regional trends in land use and by the specific natural andanthropogenic features of a particular site. To separate these influences in landscapes of the Southeast-ern coastal plain we describe land cover history outside the small city of Hattiesburg, Mississippi, USA, arural landscape originally occupied by pine savanna and mixed forests. Land cover was recorded at 296point locations regularly spaced on a 1 km grid. Aerial photographs from 1938, 1958, 1970, 1982, 2000,and 2010 illustrated a progression from open land to pine savanna, Southern Mixed Hardwood Forest(SMF), and built land cover, with low-density residential development encroaching after 1980 – a pat-tern reflecting broad regional trends in the mid and late 20th century. Examination of point-transitionsshowed frequent conversion between recent clearcuts and SMF, indicating rapid cycling of small parcelsin short rotation forestry, and long-term conversion of abandoned agricultural land to SMF, reflectingregional regrowth following the lumber boom of the early 20th century. Pine savanna declined by intro-

rban expansion gression of hardwood species rather than by cutting. Logistic regression identified land on floodplainsand distant from developed areas as most likely to regenerate as SMF. After 1980 urban expansion wasmost likely to occur close to existing buildings and arterial roads, suggesting nucleation outside the his-torical urban core. Thus, modern forest distribution has been decoupled from the natural environmentaltemplate. Recent land use changes appear to be driven by proximity to the expanding city rather than

.

regional economic trends

. Introduction

Human-generated pattern dominates the landscape matrix ofastern North America, clearly evident in the fine-scale distributionf forest (Forman, 1995). In rural areas land use may be channeledy land form, soils, and land ownership, with land cover ofteneflecting historical land use rather than current activity. Most

tudies of land cover change in eastern North America have focusedn geologically old and stable regions which are characterized byelatively fertile soils, well-defined topography, deciduous tree

∗ Corresponding author. Tel.: +1 740 593 1131; fax: +1 740 593 1130.E-mail addresses: Schweizer.peter@gmail.com (P.E. Schweizer),

atlack@ohio.edu (G.R. Matlack).

169-2046/$ – see front matter © 2013 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.landurbplan.2013.09.003

© 2013 Elsevier B.V. All rights reserved.

species, and long human occupation (e.g. Duram, Bathgate, & Ray,2004; Foster, 1992; Hall, Motzkin, Siefert, & Burk, 2002; Matlack,1997a,b). The Atlantic and Gulf coastal plains are comparativelyless studied. In the last decades of the twentieth century the South-eastern coastal plain has experienced rapid population growth withattendant changes in land use (Wear & Greis, 2002). Because suchchanges have the potential to affect natural processes over largeareas, there is a need to document land cover change in the regionand to understand the factors controlling its spatial distribution.We describe land cover change in an urbanizing area on the coastalplain of southern Mississippi. The region displays a distinctive

combination of low relief, rapidly draining soils of poor agriculturalquality, fire-shaped forest communities, and a history of resourceexploitation. We ask how these features and events have shapedthe modern landscape mosaic, and whether principles of land use

5 ape an

ap

s(mtfssrlti2KccchmPbssi(itB

aicIpmtitths

2

2

MeMbbastalbu1sST

6 P.E. Schweizer, G.R. Matlack / Landsc

nd land cover change observed elsewhere apply to the coastallain.

Natural communities of the Southeastern coastal plain show atrong upland-lowland contrast caused by rapidly draining soilsGemborys & Hodgkins, 1971; Titus, 1990). Notwithstanding the

odest variation in elevation, the upland-lowland contrast leadso a dendritic distribution of forest types with mixed hardwoodorest closely following stream courses in a matrix of upland pineavanna. Widespread cutting in the early 20th century and sub-equent plantation forestry have obscured the natural pattern,eplacing it with large blocks of managed forest independent ofand form (Perkins & Matlack, 2002). In the last three decadeshere has been a significant increase in the amount of urban-zed land in the region (e.g. Gallant, Loveland, Sohl, & Napton,004; Griffith, Stehman, & Loveland, 2003; Price, Dorcas, Gallant,laver, & Willson, 2006). The biological effects of these landusehanges are not well documented, but observations of naturalommunities suggest substantial alteration of landscape-scale pro-esses. Landscape fragmentation and widespread fire suppressionave reduced fire frequency, causing a shift from pine savanna toixed pine-hardwood forest over large areas (Batista & Platt, 1997).

lantation forestry appears to influence the spread of pathogensy increasing spatial connectedness and homogenizing landscapetructure (Perkins & Matlack, 2002). Changes in land cover haveeverely affected rain water runoff and erosion, in turn impair-ng water quality and stressing biological stream communitiesSchweizer & Matlack, 2005). Land use change is of particularnterest because the region preserves isolated examples of thehreatened longleaf pine savanna ecosystem (Landers, van Lear, &oyer, 1995).

It is important to understand the factors driving these trends atfine scale because management decisions are made at the level of

ndividual ownership parcels. We document point-changes in landover using a time series of aerial photographs from 1938 to 2010.ndividual point transitions were compared with aspects of thehysical, vegetational, and anthropogenic environment to deter-ine the drivers of land-use conversions at various scales. We test

he null hypothesis that land use history in a particular study areas shaped by broad regional tends irrespective of local environmen-al and cultural variation. Our goal is to determine to what extenthe peculiar natural and cultural circumstances of the coastal plainave guided land cover change and to identify the scale at whichuch factors work.

. Methods

.1. Study area

The study area was located in Forrest and Lamar Counties,ississippi (31.31◦ N, 89.31◦ W), on the coastal plain of the south-

astern United States, ca. 100 kilometers north of the Gulf ofexico (Fig. 1). The area was well suited to testing our hypothesis

ecause it includes all the principal landuse types of the region, andecause extensive historical and photo documentation was avail-ble. The climate is subtropical with mild winters and warm, humidummers (National Climate Data Center, 2007). Mean monthlyemperatures range from 8.8 ◦C in January to 27.6 ◦C in July. Therea receives an average annual precipitation of 158 cm from pro-onged low-intensity rainfall between late fall and early spring andrief thunderstorms during summer months. Soils are fast-drainingltisols in Pleistocene shoreline terraces (USDA, 1993; Walker,

994) characterized by sand and sandy loams with a high shrink-well potential and low agricultural value (USDA Soil Conservationervice, 1993). The area falls within EPA Ecoregion IX; Southeasternemperate Forested Plains and Hills (Omernik, 1987; Sohl, 2013)

d Urban Planning 121 (2014) 55–64

close to its southern boundary with the Southern Coastal PlainEcoregion (Drummond, 2013).

The study area covers ca. 340 km2, defined by 17 small water-sheds draining into the nearby Leaf and Bouie Rivers. Landformsinclude the flood plains of the Leaf and Bouie Rivers and tributarycreeks (ca. 45 m a.s.l.) flanked by gently rolling hills rising to 120 min the southwestern part of the study area. Natural vegetation hashistorically been an open, fire-maintained savanna with tree coverconsisting almost exclusively of longleaf pine (Pinus palustris) inupland sections. The dominant vegetation today is a closed-canopypine-hardwood forest roughly corresponding to Quarterman andKeever’s (1962) Southern Mixed Hardwood Forest, with smallamounts of managed pine plantation (largely fast-growing Pinustaeda and P. elliottii). Human activity has produced a mosaic offorests, agricultural grasslands, lawn, impervious surface, wetlands,and small lakes.

2.2. Local history

Until the late 19th century the study area was sparsely pop-ulated, dominated by old-growth pine savanna and bottomlandhardwood forest. Between ca. 1890 and 1930, the area experiencedrapid deforestation in a period of intensive timber exploitationreferred to as the “Southern Lumber Boom” (Williams, 1989). Thesmall city of Hattiesburg (pop. 46,626; US Census, 2011), aroseat the east edge of the study area as a transportation hub andmilling center for the lumber industry. Timber extraction peakedca. 1925 and declined rapidly in the economic depression of the1930s (Williams, 1989). After exhaustion of the timber resource,modest population growth was supported by establishment of aregional hospital, educational facilities including the University ofSouthern Mississippi, and a military training facility. Lacking therecreational advantages of the Gulf Coast and the established indus-try of the adjacent Piedmont zone, the Southern Plains lagged inpopulation growth (Napton, Auch, Headley, & Taylor, 2010). Whileother parts of the country experienced rapid suburban growth inthe years following World War II (Garreau, 1991; Platt Boustan,2007), the study area remained in a state of economic stagnationuntil the 1960s, when an interstate highway connected the regionwith urban centers along the Gulf Coast. In the 1970s and 1980s, thelocal economy shifted from dependence on marginal agriculture toan economy supported mainly by service industries in the smallurban area (Hattiesburg Area Chamber of Commerce, 2002). Sub-urban development accelerated with the construction of arterialroads westward from the town center. In the 1980s and 1990s aninflux of senior citizens attracted by the mild climate and affordableliving contributed to population increase. On August 29, 2005, theeye of Hurricane Katrina, a strong Category 3 storm, passed directlyover Hattiesburg causing extensive destruction to buildings andforests (Stanturf, Goodrick, & Outcalt, 2007).

2.3. Land cover data

Land cover was recorded at 296 sample points on a square gridwith a 1000 meter interval. The interval was selected to providehigh resolution while exceeding the scale of most managementunits, thus ensuring spatial independence of point data. Individ-ual points were relocated and examined on aerial photos taken in1937–1938, 1958, 1970, 1982, 2000, and 2010 covering a periodof ca. seventy years. The earliest photos were taken at a scale of1:18,000 for purposes of petroleum exploration (Tobin Aerial Sur-veys, San Antonio, TX); photos from 1958 onward were taken in

the federal NHAP, NAPP, and Farm Service Agency survey pro-grams at a scale of ca. 1:8000. At all dates individual trees wereclearly distinguishable, although grass- and scrubland texture wasoccasionally problematic in the 1937–1938 photos. At each date,

P.E. Schweizer, G.R. Matlack / Landscape and Urban Planning 121 (2014) 55–64 57

d proliferation of built land along primary roads. Sample points are indicated by “×”.

pAgalaSsamsCwsev

eVaSsai(tepdet

Table 1Land cover classification.a

Southern mixed forest (SMF)Mixed pine/deciduous forest with closed canopy occurring in uplands and

along streams, excluding orchards and urban tree cover with understory oflandscaped vegetation.

Managed Pine Forest (MPF)Managed patches ranging from multi-hectare industrial holdings to patches

<1 ha. MPF were identified by spectral signature, heads-up digitizing, andground truthing.

Forest clearcut and regrowthAreas of recent clearcut with bare soils visible or forest regrowth with woody

vegetation and shrubs but canopy closure <50%.

Pine savannaNative grassland ecosystem with open-canopy longleaf pine, dependent on

previously naturally occurring burning cycles with low-intensity fires.

OpenAll non-forest rural land including agriculture, pasture and hay production.May include isolated trees

Built areaUrban and residential development, including impervious surfaces,

recreational and residential lawns, and golf courses.

Transient land coverCharacterized by removed vegetation and exposed soil surfaces, including

riparian sand bars, active construction sites, sand and gravel mining, andpreparation for future development.

Fig. 1. Study area surrounding Hattiesburg, Mississippi, showing urban area an

oints were assigned to one of eight land cover types modified fromnderson, Hardy, Roach, and Witmer (1976) (Table 1). Managedreen areas (lawns, golf courses, athletic fields, and road margins)nd impervious surfaces (roads, parking lots, and roofs) were col-ectively referred to as Built Area, characterized by compacted soilsnd sealed surfaces with rapid runoff. We distinguished betweenouthern Mixed Forest (SMF), a mixture of Pinus and hardwoodpecies regenerating naturally (Batista & Platt, 1997), and Man-ged Pine Forest (MPF) which was easily recognized as a pineonoculture with trees arranged in rows. Cleared land showing

ome degree of forest regrowth was assigned a separate category:learcut/Regrowth. Cultivated land, rural grasslands, and pasturesere combined in the Open category. Pine Savanna, Transient (with

and or exposed soil), and Water Bodies contributed additional cat-gories. Assignment of land cover categories in 2000 and 2010 waserified for a subset of points by field-checking at ground level.

Several environmental variables were noted which might bexpected to influence land-use transitions (Erickson, 1995; Flinn,ellend, & Marks, 2005; Foster, 1992; Matlack, 1997b). Elevation,spect, and slope at each sample point were extracted from ahuttle Radar Topography Mission digital elevation model at 28 mpatial resolution (Taverna, Urban, & MacDonald, 2004). Total patchrea, distance to the forest edge, and distance to the nearest build-ng and paved road were measured for each forested sample pointSMF or MPF). Four-lane arterial roads were distinguished fromwo-lane streets and rural roads. In the 2000 and 2010 photosdge, building, and road distances were measured for all sample

oints regardless of cover type. Data were available for 228–254ata points in 1938, 1958, and 1970, omitting a small section ofastern Lamar County. Data were available for the whole study areahereafter, including 296 sample points.

Water bodiesLakes, ponds, and reservoirs.

a Modified from Anderson et al. (1976).

58 P.E. Schweizer, G.R. Matlack / Landscape and Urban Planning 121 (2014) 55–64

Fl

isgAoetcbo

2

uaclmtdfielmDaalort

3

3

rdr1e7mPdd

Edge Distance

1938 1958 1970 1982 2000 2010

Dis

tanc

e (m

eter

s)

0

20

40

60

80

100

120

140

Building Distance

Year

1938 1958 1970 1982 2000 2010

Dis

tanc

e (m

eter

s)

0

100

200

300

400

500

600

700

Forest Patch Area

1938 1958 1970 1982 2000 2010

Are

a (h

ecta

res)

0

20

40

60

80

MedianMean

Fig. 3. Properties of forested sample points. Data include both Southern Mixed For-

ig. 2. Land cover at sample points near Hattiesburg, Mississippi; 1938–2010. Eachand cover is expressed as a percentage of the total sample area.

Agricultural land use was difficult to determine in the aerialmages, particularly at earlier dates, so all forms of agriculture areubsumed into the Open category. However, insight can also beained from agricultural statistics published in the US Census ofgriculture. Although the study area covers only the upper ca. 1/3f Forrest County, county-wide data provide an indicator of localconomic and land use trends. Farming activity was assessed inerms of farm number and total area. Corn, the most common rowrop, provided an indicator of cultivation intensity, and the num-er of cattle was used as an index of grazing activity (U.S. Censusf Agriculture, 1910–2007).

.4. Analysis

The frequency of points in each land cover category was tab-lated at each date. Summary statistics were calculated for patchrea, edge distance, and building distance in both the reduced andomplete data sets. For each sampling interval, transitions betweenand cover types were tabulated and summarized as transition

atrices. Forest rotation time was estimated with the assump-ion that stands were cut and regenerated midway between sampleates. On the basis of changes in the dominant land cover categoriesve critical transitions were selected for closer examination. Inach case, potential controlling factors were assessed by stepwiseogistic regression of a binomial transition variable on environ-

ental and landscape factors using the GLM procedure in R (Revelopment Core Team, 2011). An inclusive model was derived,nd nonsignificant variables were dropped individually in order ofscending Z value until only significant variables remained. Corre-ated variables were tested alternately in the model, and selectedn the basis of their ability to lower the Akaike Information Crite-ion. Goodness-of-fit of the final model was tested by comparinghe residual deviance with the appropriate chi-square distribution.

. Results

.1. Land cover

“Open” was the dominant land cover between 1938 and 1970,eaching a maximum of 58.3% of the landscape in 1958, buteclining to only 12.3% by 2010 (Fig. 2). Southern Mixed Forestemained stable at 21–26% of the landscape between 1938 and970, but increased rapidly thereafter, reaching a maximum cov-rage of 44.2% in 2000. Built land cover increased steadily from.5% in 1938 to 25.3% in 2010. Other land covers were less com-

on but remain important as indicators of landscape processes.

ine Savanna, the historically dominant land cover in upland sites,ecreased to less than 2% by the year 2000 showing the sharpestecline between 1970 and 1982. Clearcut/Regrowth has been a

est and Managed Pine Forest. (a) Mean and median forest patch area. (b) Distanceof forested sample points from the nearest forest edge and (c) from the nearestbuilding. Distance data are not available for 1982. Bars indicate one standard error.

small but significant component in the landscape matrix, serv-ing as an indicator of rate of forest clearance. Undetectable from1958–1970, Clearcut/Regrowth increased rapidly to 14% in 2010.Managed pine forests were not observed in the study area until1982 and have remained only a small portion of the landscape.Water surface first appeared in 1982 and increased to 2.7% by 2010due to construction of artificial ponds. Transient land cover was aminor presence throughout the period.

Forest patch area increased from 1958 to 2010 (Fig. 3a), mostclearly seen as an increase in median patch area from 4.7 to 21.0 ha.

Mean values were higher due to the presence of a few very largestands in the southern and northwestern sections of the study areawhich strongly affected variance after 1970. Although the medianpatch size was 12.7 ha in 1982, for example, the four largest patches

P.E. Schweizer, G.R. Matlack / Landscape and Urban Planning 121 (2014) 55–64 59

Farm number

Year

1900 1920 1940 1960 1980 2000 2020

Num

ber o

f far

ms

0

200

400

600

800

1000

1200

1400

1600

Land in farms

Year

1900 1920 1940 1960 1980 2000 2020

Tota

l are

a (th

ousa

nds

of a

cres

)

0

20

40

60

80

100

120

140

Cattle

Year1900 1920 1940 1960 1980 2000 2020

Num

ber o

f cat

tle

0

5000

10000

15000

20000

25000

30000

Land in corn

Year1900 1920 1940 1960 1980 2000 2020

Tota

l are

a (a

cres

)

0

2000

4000

6000

8000

10000

12000

14000

16000

18000

1954 1954

19401974

F (US Cei

cfhmhb2se

1t(bCrtb

3

obaiOtmOfim

ig. 4. Agricultural activity in Forrest County, Mississippi, between 1910 and 2007n corn, and (d) number of cattle.

overed 125, 135, 597, and 1062 ha, respectively. Distance fromorested sample points to the nearest forest edge declined from aigh value in 1958 (median = 79 m) to a low in 2010 (27 m) althoughany points were considerably farther from the edge, leading to

igh mean values (Fig. 3b). Distance to the nearest residentialuilding (Fig. 3c) similarly declined from 1958 (median = 331 m) to010 (150 m). Both the core area (N = 228 points) and the extendedample (N = 296) showed similar patterns of change in these param-ters.

Agricultural activity in Forrest County increased between ca.910 and 1950, and then decreased sharply in the second half ofhe century (Fig. 4) with the most abrupt decline in farm number−46%) between 1954 – 1969 and a – 68.3% decline in total farm areaetween 1954 and 1992 (U.S. Census of Agriculture, 1910–2007).ultivation of corn increased from 1910 to 1940, and then declinedapidly to the present; by 2007 the crop was no longer recorded inhe County. Number of cattle increased from 1910 to a high periodetween 1950 and 1974, after which it declined sharply.

.2. Transitions between land cover types

The overall frequency of land cover transition ranged from 40.3%f sample points which changed to another land use in the most sta-le interval (1958–1970) to 50.2% which changed between 1982nd 2000. Most land cover types were relatively stable over mostntervals, with large portions of the landscape remaining in the SMF,pen, and Built categories for extended periods (Table 2). Never-

heless, gradual trends in land conversion cumulatively produced

ajor changes over the whole 70-year period. Most noticeably,pen land was replaced by SMF as the dominant land cover. In therst interval (1938–1958) removal of 67% of existing SMF cover, pri-arily due to transitions to Open land cover, exceeded the potential

nsus of Agriculture): (a) farm number, (b) land area in farms, (c) land area planted

40% increase caused by ecological succession from Pine Savannaand Open. Built area showed a modest increase (7.5–9.3% of thelandscape) mainly due to conversion of Open land. Pine savannasuffered a major contraction due to ecological succession to SMF(32.1% of existing savanna lost) and to Open (39% lost).

Between 1958 and 1970 land cover was relatively stable(Table 2b). Forest cover increased slightly as a result of frequenttransitions from SMF to Open and the reverse (6.6% of observedtransitions in both directions). Notably, 10.9% of the landscapechanged from other land uses to Pine Savanna, primarily due torecruitment of pines in Open sites (9.6% of transitions). This gainmore than compensated for the loss of savanna to SMF (3.9% oftransitions) and Open (4.4%), leading to a modest overall increasein Savanna.

Between 1970 and 1982, the frequency with which Open landand Pine Savanna succeeded to Southern Mixed Forest greatlyincreased (16% and 7.6% of transitions, respectively), far outstrip-ping the frequency of forest clearance (7.9%), which had maintainedOpen land in previous intervals (Table 2c). Built area continued itssteady increase, largely by transition of points from Open land. Thefrequency of Pine Savanna dropped from 13.3 to 2.7% of the totallandscape due to conversion to SMF (7.6% of transitions) and Open(4.0% of transitions).

From 1982 to 2000 Open land continued to shrink (28.4% oftransitions), again primarily due to conversion to SMF, but loss toClearcut/Regrowth, MPF, and Built area were collectively as impor-tant as the transition to SMF (Table 2d). Built area increased 24% dueto transitions from Open area and (to a lesser extent) SMF. Among

less-frequent land covers, MPF showed a modest increase almostexclusively due to planting on Open land. Clearcut/Regrowth for-est increased at points that were previously SMF as well asOpen land.

60 P.E. Schweizer, G.R. Matlack / Landscape and Urban Planning 121 (2014) 55–64

Table 2Transition matrices showing land cover conversions (% of total area) over five intervals from 1938 until 2010 near Hattiesburg, Mississippi, USA.

Land cover 1958 1937/38Land cover

Southern mixedforest (SMF)

Forest regrowth/clear cuts

Managed pineforest (MPF)

Pinesavanna

Open/agriculture

Builtarea

Transient/barren

Water

(a)Southern mixed forest (SMF) 9.8 – – 4.4 7.3 – –Forest regrowth/clear cuts – – – – – – –Managed pine forest (MPF) – – – – – – – –Pine savanna 2.0 – – 3.4 4.4 0.5 – –Open/agriculture 16.6 – – 5.4 31.2 0.5 – –Built area 1.0 – – 0.5 2.9 7.3 – –Transient/barren – – – – – – 0.5 –Water – – – – – – – –

Land cover 1970 1958 Landcover

Southern mixedforest (SMF)

Forest regrowth/clear cuts

Managed pineforest (MPF)

Pinesavanna

Open/agriculture

Builtarea

Transient/barren

Water

(b)Southern mixed forest (SMF) 11.0 – – 3.9 6.6 0.9 0.4 –Forest regrowth/clear cuts 1.3 – – – 0.4 – – –Managed pine forest (MPF) – – – – – – – –Pine savanna 1.3 – – 2.2 9.6 – – –Open/agriculture 6.6 – – 4.4 37.3 – – –Built area – – – 0.4 2.6 8.8 – –Transient/barren 0.9 – – – 0.4 – 0.4 –Water – – – 0.4 – – – –

Land cover 1982 1970 Landcover

Southern mixedforest (SMF)

Forest regrowth/clear cuts

Managed pineforest (MPF)

Pinesavanna

Open/agriculture

Builtarea

Transient/barren

Water

(c)Southern mixed forest (SMF) 13.3 0.4 7.6 16 – – –Forest regrowth/clear cuts 1.3 – – – 0.9 – – –Managed pine forest (MPF) – 0.4 – – – – – –Pine savanna 1.3 – – 0.4 1.3 – – –Open/agriculture 4.0 0.9 – 4.0 26.2 0.4 – –Built area 1.3 – – 1.3 4.9 11.6 – –Transient/barren – – – – – – 1.8 –Water – – – – – – – 0.4

Land cover 2000 1982 Landcover

Southern mixedforest (SMF)

Forest regrowth/clear cuts

Managed pineforest (MPF)

Pinesavanna

Open/agriculture

Builtarea

Transient/barren

Water

(d)Southern mixed forest (SMF) 24.4 1.0 – 4.7 13.2 2.7 – –Forest regrowth/clear cuts 3.1 0.7 – – 4.4 – – –Managed pine forest (MPF) 1.7 0.7 – – 3.4 – – –Pine savanna 0.3 – – – – – – –Open/agriculture 1.4 – – – 11.2 0.7 0.7 –Built area 2.0 0.3 – – 6.1 12.9 0.3 –Transient/barren 1.0 – – – 1.0 0.3 0.3 –Water 0.3 – – – 0.3 0.3 – 0.3

Land cover 2010 2000 Landcover

Southern mixedforest (SMF)

Forest regrowth/clear cuts

Managed pineforest (MPF)

Pinesavanna

Open/agriculture

Builtarea

Transient/barren

Water

(e)Southern mixed forest (SMF) 24.0 3.7 1.3 0.7 3.0 2.7 0.7 –Forest regrowth/clear cuts 8.7 1.3 1.7 – 2.0 0.3 – –Managed pine forest (MPF) 0.7 0.7 2.3 – – – – –Pine savanna 1.3 – – – – – 0.3 –Open/agriculture 3.0 1.0 0.3 – 4.7 3.0 0.7 –

sC(cao

atr

Built area 6.3 1.0Transient/barren 0.7 –Water 0.7 –

Over the last interval (2000–2010) the four-decade expan-ion of SMF ended with a net loss of 14.7%, largely converted tolearcut/Regrowth (8.7% of transitions) and conversion to Built6.3%) (Table 2e). Removal of SMF increased the Clearcut/Regrowthategory to 14.2% of the landscape, the largest proportion observedt any date. Similarly, Built land continued to increase reaching 27%f the landscape, primarily at the expense of SMF.

SMF stands persisted for relatively short periods. Stands firstppearing in 1958, 1970, 1982, and 2000 showed median longevi-ies of 28, 27, 29, and 13 years, respectively. Modal longevitiesanged from 12 to 16 years except in the 1982 cohort, most of

0.3 – 3.0 16.0 0.3 –– – 0.3 0.7 0.7 0.3– – 0.7 0.3 – 1.0

which were still standing in 2010 (implying a rotation time >34years). Collectively rates of SMF removal were low between 1958and 2000 while rates of regeneration were high, indicating a pulseof recruitment following the lumber boom (Fig. 5).

3.3. Factors influencing land cover

Environmental factors controlling land cover change variedbetween intervals and land cover type. Conversion to Open habitatin the first interval was influenced by a group of correlated variablesreflecting landscape position and location within the sampling

P.E. Schweizer, G.R. Matlack / Landscape an

Southern Mixed Forest

Interval

1938-1958 1958-1970 1970-1982 1982-2000 2000-2010

Cha

nge

in c

over

(% o

f lan

dsca

pe)

5

10

15

20

25

30

To SMF From SMF

Fu

alssr

TEcoii

ig. 5. Rates of loss and gain of Southern Mixed Forest as a proportion of total landse transitions near Hattiesburg, Mississippi; 1938–2010.

rea (Table 3a). Open habitat was more likely to be formed at

ow values of the dummy variable “row”, indicating the northernection of the study area, and at points of high elevation and highlope. Human-generated landscape structure, e.g. proximity tooads and buildings, was not significant. Residual deviance was not

able 3nvironmental factors influencing five critical land-cover transitions on the outeroastal plain of Mississippi, USA. Logistic regressions of transition/no transitionn geographic position, physical environment, and human land use. Final modelsnclude only variables which had a significant effect (p < 0.05). “Row” and “column”ndicate position in the sampling grid.

Estimate

Conversion to “Open” 1938–1958Intercept −1.208NS

Row −0.173***

Elevation 0.035*

Slope 0.163*

Residual deviance 130.14NS

107 df

Conversion to southern mixed forest 1970–1982Intercept −1.04**

Building distance 2000 0.0024***

Stream distance −0.0017*

Residual deviance 192.79NS

175 df

Conversion to “Built” 1982–2000Intercept −0.346NS

Building distance 2000 −0.0057****

Residual deviance 32.4NS

240 df

Conversion to Managed Pine Forest 1982–2000Intercept −2.77****

Row −0.115*

Road 4 lanes 0.0005***

Residual deviance 114.85NS

292 df

Loss of forest (southern mixed + managed) 2000–2010Intercept 0.250****

Edge distance 2000 −0.0056**

Residual deviance 199.47**

150 df

* <0.05.** <0.01.

*** <0.001.**** <0.0001.NS >0.05.

d Urban Planning 121 (2014) 55–64 61

significantly different from the appropriate value of Chi-square(p ∼ 1.0), indicating that the model was a satisfactory fit to the data.

Conversion to SMF between 1970 and 1982 was most likelyclose to streams and at greater distance from buildings (Table 3b).Variables indicating landscape position and location within thesample grid were not significant, implying that regeneration couldpotentially occur throughout the region irrespective of topography.Residual deviance was not significantly different from the appro-priate value of Chi-square (p > 0.05) indicating that this model wasa reasonable fit to the data. Between 1982 and 2000 conversionof all land covers to Built was negatively influenced by buildingdistance, suggesting clustering of building activity (Table 3c). Land-scape position was not important, judged by nonsignificance ofelevation, slope, stream distance, and aspect. Residual deviance wasnot significantly different from the appropriate value of Chi-square(p ∼ 1.0).

Between 1982 and 2000, conversion of all land uses to ManagedForest was significantly related to the locational variable “row’ andfour-lane road distance (Table 3d), indicating formation of plan-tations in the northern portion of the study area and at greaterdistances from major roads. The residual deviance was not sig-nificant (p > 0.05). Between 2000 and 2010 the modeling processsuggests that sample points close to a forest edge were signifi-cantly more likely to be cleared, consistent with erosion of foreststands at the periphery (Table 3e). However, residual deviance wassignificantly different from the appropriate value of Chi-square(p = 0.0043) so this model must be considered suspect.

4. Discussions

Land cover change in the study area mirrored trends across theSoutheastern Plains and Hills Ecoregion (Bragg & Shelton, 2011;Shirley & Battaglia, 2006): A generally open landscape in the mid-twentieth century developed into one dominated by forest andhuman-constructed land covers in the early twenty-first century.Land cover change was driven by several distinct phases of land useincluding marginal agriculture, abandonment and reforestation,and urban/suburban expansion, a sequence observed throughouteastern North America (Flinn et al., 2005; Foster, 1992; Hall et al.,2002; Hart, 1978; Matlack, 1997a, 1997b), although it occurred rel-atively later in the study area. Longleaf pine savanna, the naturalupland land cover, declined to a small fraction of the landscape asit has throughout the Gulf South (Landers et al., 1995).

In other respects the pattern of land cover change has been dis-tinctive to the study area. The area had a higher rate of change thanthe Ecoregion as a whole, probably because it includes a growingurban area (Napton et al., 2010; Sohl, 2013). Forrest County hada relatively low proportion of land in agriculture through mostof the twentieth century reflecting low soil quality (equivalentto ca. half of values seen in the rest of the state; U.S. Census ofAgriculture, 1910–2007). The schedule of cutting and regrowthwas distinctive, as well. Forest cover expanded between 1958 and2000 in a period when forests of nearby northern Florida wereexperiencing severe cutting (Drummond, 2013). The decline inforest observed in the last interval was partially caused by Hur-ricane Katrina, a localized disturbance which caused substantialdestruction to forests in the southern-most counties of Mississippi(Stanturf et al., 2007). Plantation forestry, a dominant land coverin many parts of the Ecoregion since the 1950s, only appeared inthe study area in the 1980s and remains a minor landscape fea-ture today. Water surface, which is generally rare in the Ecoregion,

continues to increase due to construction of artificial ponds foresthetic and runoff-detention purposes. Constructed ponds shouldbe considered an aspect of the built landscape, but they havebiological significance in substantially increasing the availability

6 ape an

oeiap

4

vbsawdattocbgIfntsllatsowm

ll(saobioapiflff

csho(fiu&tpsodK

2 P.E. Schweizer, G.R. Matlack / Landsc

f aquatic habitat. Such local deviation suggests the cultural andnvironmental factors that control forest distribution. Examiningndividual point transitions, it becomes apparent that the studyrea is not merely a late-developing variation on the Ecoregionattern.

.1. Forest regeneration

Land use change through most of the twentieth century can beiewed as an extended response to forest clearance during the lum-er boom. The growth in farm number and area from 1910 to 1950uggests that much of the cleared land was initially converted togriculture, a subset of the Open category. The agricultural phaseas short-lived, however, declining rapidly after 1954, probablyue to the marginal nature of agriculture on nutrient-poor soilsnd emergence of economic opportunity outside the region afterhe Second World War (Cowdrey, 1996). Much abandoned agricul-ural land appears to have regenerated as forest. Transitions to SMFccurred predominantly on Open land in all intervals up to 2000,onsistent with this interpretation. Regeneration appears to haveeen delayed, however, with the strongest increase in SMF lag-ing 20–30 years behind agricultural abandonment in the 1950s.t is possible that grazing by livestock prevented regeneration oforest between 1950–1980. However, even at the peak of cattleumbers in 1974 farmland occupied only 24% of the land area ofhe county, substantially below the 49% Open observed in 1970,o the effect of grazing should not be overemphasized. It seemsikely that much of the Open land described here was unmanagedand in early stages of regeneration rather than actively engaged ingriculture. SMF typically regenerates in 3–5 years following dis-urbance in sites with adequate seed, resprouting, and/or a pool ofuppressed seedlings (Batista & Platt, 1997; Clewell, 2011). Largepen areas lacking advanced regeneration (for example a cutoverhich experienced a slash fire) would be expected to regenerateuch more slowly, perhaps accounting for the delay noted here.Forest has long been the default condition of the coastal plain

andscape, regenerating when the commercial value of all otherand uses declines (Napton et al., 2010). Hart (1978) and Healy1985) identify soil quality and slope as factors controlling rever-ion of agricultural land to forest in the 1940s - 1970s, with steepernd less fertile sites more quickly abandoned to forest. Slope wasnly occasionally an issue in the low relief of southern Mississippi,ut comparison of crop production data suggests that soil qual-

ty was important in causing abandonment, at least at the scalef counties (U.S. Census of Agriculture, 1910–2007). Reforestationlso reflected the relative suitability of sites for construction. In theeriod of greatest forest expansion (1970–1982) forest was signif-

cantly more likely to establish close to streams, corresponding toood plains of the Leaf and Bouie Rivers, and at greater distance

rom existing buildings, areas which were considered undesirableor construction by virtue of flooding and/or isolation.

Eleven percent of the landscape transitioned to the Savannaategory between 1958 and 1970 suggesting that longleaf pineavanna has the capacity to regrow after cutting. In 1970–1982,owever, point-transitions showed a striking decline in Savannaf which 55% was due to introgression of hardwood speciesi.e. conversion to the SMF category). In the absence of naturalre fire-intolerant hardwood species were able to recruit inpland areas, removing the upland/lowland distinction (PerkinsMatlack, 2002). The Savanna category appears to have been a

ransitional phase leading to SMF rather than a sustainable exam-le of the original upland community. Long-term observations in

outheastern Arkansas and central Florida suggest that conversionf pine savanna to SMF may take 35–80 years in the absence ofisturbance by fire or wind (Bragg & Shelton, 2011; Hartnett &rofta, 1989) – a range consistent with the rate of decline observed

d Urban Planning 121 (2014) 55–64

here. Presumably the delay in establishment of SMF reflects thetime required for hardwood species to colonize from off-site seedsources. Clearly longleaf pine savanna requires active manage-ment; the few examples of low-density pine forest remainingin the study area may be carrying a heavy extinction debt assuccession propels them toward a mixed hardwood community.

4.2. Forest removal

In the last two intervals (1982–2010) transitions into and out ofSMF intensified, dominating land cover change in the study area.At the level of individual stands this can be interpreted as short-rotation management reflecting cutting decisions at the scale of15–20 ha management units. Although the long interval betweenphotos does not allow precise documentation of cutting, it is clearthat the cutting interval is fairly brief (estimated median inter-vals of 27–29 years), suggesting a pressing need for short-termreturns on investment (J. Kemp, pers. com.). Individual land ownerstypically use naturally regenerated SMF to supplement limitedfinancial resources while speculating on an increase in propertyvalue (Walker & Oswald, 2000). Small-parcel ownership and theperception of rising land prices near an expanding city also mili-tate against plantation forestry – a land use which is common in theregion at a greater distance from towns. The increased likelihood ofMPF in the remote northwestern section is consistent with the ideathat plantation forestry was discouraged by high land prices nearerto the expanding city (Mitchell & Duncan, 2009; Schultz, 1999).

Short-rotation forestry in the Southeastern Plains and HillsEcoregion generally produces a cyclical change which may result infrequent shifts in land cover but does not lead to long-term changesin land cover proportions (Napton et al., 2010). In the study area,however, short term parcel-scale forestry has been imposed upona much longer historical trend with a periodicity determined bythe lumber boom and agricultural abandonment. Cutting of SMFwas minor between 1958 and 2000 while SMF regenerated, butincreased between 2000 and 2010 as the pulse of forest estab-lished in 1970–1982 reached commercial maturity and high lumberprices encouraged harvesting (J. Kemp, pers. com.). The net resultwas a substantial (120%) increase between 1958 and 2000 mak-ing SMF the dominant land cover. Loss of forest between 2000and 2010 was to a lesser extent due to Hurricane Katrina, whichcaused substantial damage south of the study area (Stanturf et al.,2007) and must have caused financial hardship to individual for-est owners. A similar effect was observed in western New Englandas a large cohort of white pine which recruited on marginal landin the 1890s reached commercial maturity in the 1930s, only tobe destroyed in the 1938 hurricane (Foster, 1992). The historicalresult in Massachusetts was a reduction in lumber activity for sev-eral decades and further depopulation of the rural landscape. Whileforestry may also decline following the hurricane in our studyarea, the human population seems likely to increase as the Builtlandscape expands.

4.3. The built landscape

Transitions from Open to Built land cover increased graduallybetween 1958 and 2010, becoming second only to turnover of SMFas the dominant landscape dynamic. Conversion to Built reflects thenational growth of sun-belt communities as retirement and recre-ational destinations, a broad economic driver operating above theEcoregion scale (Napton et al., 2010; Drummond, 2013). In ruralareas the trend is commonly expressed as low-density exurban

development (Benfield, Raimi, & Chen, 1999), a pattern also seen inour study area.

Before the 1980s Built land cover expanded in small steps at theedge of the existing urban area, resulting in concentric rings of high

pe an

dpombpSscwrunIaabsTdo

5

rupiutoiUttpdaoedutoau2mdpo

A

CppppI(Cw

P.E. Schweizer, G.R. Matlack / Landsca

ensity development. After ca. 1980, logistic regression suggests aattern of low-density nucleation, in which local clusters appearedutside the urbanized area around preexisting seeds of develop-ent. Commercial building was concentrated on land near existing

uildings and adjacent to multi-lane arterial roads, consistent withatterns of urban decentralization throughout the eastern Unitedtates (Diamond & Noonan, 1996). A glance at the map (Fig. 1)hows several major roads radiating from the urban center, carryingommercial development into formerly rural areas. In a landscapeith low population density and without strong relief or major

ivers, construction of these roads was largely independent of nat-ral features. Large parking lots appeared in the road-side strip – aovel form of land cover in the region (Albanese & Matlack, 1999).

n a second form of nucleation, residential development appeareds small clusters of houses in the outlying rural communities southnd west of Hattiesburg. Declines in median distance to the nearestuilding and forest edge suggest a process of invagination, in whichmall residential projects expand within large forested sections.he cumulative effect is an anthropogenic, radial pattern of landevelopment, in which building occurs in satellite communitiesutside the urban core.

. Conclusions

Early in the study period land use changes can be interpreted asegional-scale responses to national economic trends, with individ-al plots gravitating to their ‘highest and best use’ as land use theoryredicts (Barlowe, 1986). Later in the period, land use was more

diosyncratic, apparently influenced by proximity to the growingrban area rather than representing the Ecoregion as a whole. Fac-ors influencing land use decisions have evolved from dependencen natural features such as slope, elevation, and soil quality earlyn the study period, to a purely anthropogenic pattern at the end.rbanization has effects on forest management out of proportion

o the area actually developed. The scarcity of managed pine plan-ation, for example, appears to be a byproduct of increasing landrices adjacent to the expanding city. Interdigitation of residentialevelopment into existing forest prevents reintroduction of fire,nd landscape fragmentation restricts the spread of fire when itccurs naturally. The net effect is a homogenized landscape cov-red with small-stature mixed forest in which spatial structure isecoupled from natural environmental gradients. Land currentlysed for small-plot silviculture on the edge of suburbia is likely toransition to residential properties, and low-density exurban devel-pment will increasingly fragment the forest matrix. These changesre irreversible; sample points rarely transition out of Built landses. Urban growth will eventually pass a threshold (Wear & Greis,002), creating a fine-grained landscape mosaic in which manage-ent for ecological or silvicultural purposes is no longer possible

espite a substantial forest area remaining. Although atypical atresent, this urban-edge forest dynamic may represent the futuref the Southeastern Plains and Hills Ecoregion as a whole.

cknowledgments

We are grateful to the Hattiesburg office of the Natural Resourceonservation Service for allowing access to their historical aerialhotos. Tobin Aerial Surveys, San Antonio, kindly made the 1937hotos available from their archive. Jake Kemp and Mike Leerovided insight in regional forestry economics. This work was sup-orted by grants from the Mississippi Water Resources Research

nstitute and NASA’s Scientific Data Purchase program. One of usPES) was supported by a fellowship from the Mississippi Spaceommerce Initiative. Two anonymous reviewers made suggestionshich greatly improved the manuscript.

d Urban Planning 121 (2014) 55–64 63

References

Albanese, B., & Matlack, G. R. (1999). Utilization of parking lots in Hattiesburg, Mis-sissippi and impacts on local streams. Environmental Management, 24, 265–271.

Anderson, J. R., Hardy, E. E., Roach, J. T., & Witmer, R. E. (1976). A land-use and land-cover classification system for use with remote sensor data. Washington, DC, US:Government Printing Office.

Barlowe, R. (1986). Land resource economics (4th ed.). Upper Saddle River: PrenticeHall.

Batista, W. B., & Platt, W. J. (1997). An old-growth definition for southern mixed hard-wood forests. (General Technical Report SRS-9). Ashville, NC: US Department ofAgriculture Forest Service, Southern Research Station.

Benfield, F. K., Raimi, M., & Chen, D. D. (1999). Once there were Greenfields, how urbansprawl is undermining America’s environment, economy, and social fabric. NewYork: Natural Resource Defense Council.

Bragg, D. C., & Shelton, M. G. (2011). Lessons from 72 years of monitoring a once-cutpine-hardwood stand on the Crossett Experimental Forest, Arkansas, USA. ForestEcology and Management, 262, 911–922.

Clewell, A. (2011). Forest succession after 43 years without disturbance on ex-arableland, northern Florida. Castanea, 76, 386–394.

Cowdrey, A. E. (1996). This land, this south, an environmental history (revised ed.).Lexington, Kentucky: The University Press of Kentucky.

Diamond, H. L., & Noonan, P. F. (1996). Land use in America. Washington, DC: IslandPress.

Drummond, M. A. (2013). Southern Coastal Plain; USGS Land Cover TrendsProject [Online]. Retrieved from http://landcovertrends.usgs.gov/east/eco75Report.html

Duram, L., Bathgate, J., & Ray, C. (2004). A local example of land-use change, SouthernIllinois—1807, 1938, and (1993). Professional Geographer, 56, 127–140.

Erickson, D. L. (1995). Rural land cover and land cover change, implications for localplanning in the River Raisin watershed. Land Cover Policy, 12, 223–236.

Flinn, K. M., Vellend, M., & Marks, P. L. (2005). Environmental causes and conse-quences of forest clearance and agricultural abandonment in central New York,USA. Journal of Biogeography, 32, 439–452.

Forman, R. T. T. (1995). Land mosaics, the ecology of landscapes and regions. New York,USA: Cambridge University Press.

Foster, D. R. (1992). Land cover history (1730–(1990)) and vegetation dynamics inCentral New England USA. Journal of Ecology, 80, 753–771.

Gallant, A. L., Loveland, T. R., Sohl, T. L., & Napton, D. E. (2004). Using an Ecore-gion framework to analyze land-cover and land-use dynamics. EnvironmentalManagement, 34, S89–S110.

Garreau, J. (1991). Edge city, life on the new frontier. New York, USA: Doubleday,Bantam Doubleday Dell Publishing Group, Inc.

Gemborys, S. R., & Hodgkins, E. J. (1971). Forests of small stream bottoms in thecoastal plain of Southwestern Alabama. Ecology, 52, 70–84.

Griffith, J. A., Stehman, S. V., & Loveland, T. R. (2003). Landscape trends in Mid-Atlantic and Southeastern United States Ecoregions. Environmental Management,32, 572–588.

Hall, B., Motzkin, G., Foster, D. R., Syfert, M., & Burk, J. (2002). Three hundred years offorest and land-use change in Massachusetts, USA. Journal of Biogeography, 29,1319–1335.

Hart, J. F. (1978). Cropland concentrations in the South. Annals of the Association ofAmerican Geographers, 68, 505–517.

Hartnett, D. C., & Krofta, D. M. (1989). Fifty-five years of post-fire succession ina southern mixed hardwood forest. Bulletin of the Torrey Botanical Club, 116,107–113.

Hattiesburg Area Chamber of Commerce. (2002). Chamber of commerce,economic development, area development partnership. Available fromhttp://www.theadp.com/. Accessed 02.01.08

Healy, R. G. (1985). Competition for land in The American South, Agriculture, HumanSettlement and The Environment. Washington, DC: The Conservation Foundation.

Landers, J. L., van Lear, D. H., & Boyer, W. D. (1995). The longleaf pine forests of thesoutheast, requiem or renaissance? Journal of Forestry, 93(11), 39–44.

Matlack, G. R. (1997a). Four centuries of forest clearance and regeneration in thehinterland of a large city. Journal of Biogeography, 24, 281–295.

Matlack, G. R. (1997b). Land cover and forest habitat distribution in the hinterlandof a large city. Journal of Biogeography, 24, 297–307.

Mitchell, R. J., & Duncan, S. L. (2009). Range of variability in South-ern coastal plain forests, Its historical, contemporary, and future rolein sustaining biodiversity. Ecology and Society, 14, 17. Retrieved fromhttp://www.ecologyandsociety.org/vol14/iss1/art17/

Napton, D. E., Auch, R. F., Headley, R., & Taylor, J. L. (2010). Land changes and theirdriving forces in the Southeastern United States. Regional Environmental Change,10, 37–53.

National Climate Data Center (NCDC). (2007). Historic climate data for Hat-tiesburg Chain Municipal Airport from August 1942 to present. Retrieved from:http://www4.ncdc.noaa.gov/cgi-win/wwcgi.dll?wwDI∼StnSrch∼StnID∼(2001)1170

Omernik, J. M. (1987). Ecoregions of the conterminous United States. Annuals of theAssociation of American Geographers, 77, 118–125 (and map supplement).

Perkins, T. E., & Matlack, G. R. (2002). Human-generated pattern in commercialforests of southern Mississippi and consequences for the spread of pests and

pathogens. Forest Ecology and Management, 157, 143–154.

Platt Boustan, L. (2007). Was postwar suburbanization “white flight”? Evidence fromthe black migration. Working paper 13543 National Bureau of Economic Research.Available from http://www.nber.org/papers/w13543. Accessed 01.01.08

6 ape an

P

Q

R

S

S

S

S

S

T

Williams, M. (1989). Americans and their forests, a historical geography. New York,USA: Cambridge University Press.

Wear, D. N., & Greis, J. G. (2002). Southern forest resource assessment (General Tech-

4 P.E. Schweizer, G.R. Matlack / Landsc

rice, S. J., Dorcas, M. E., Gallant, A. L., Klaver, R. W., & Willson, J. D. (2006). Threedecades of urbanization, estimating the impact of land-cover change on streamsalamander populations. Biological Conservation, 133, 436–441.

uarterman, E., & Keever, C. (1962). Southern mixed hardwood forest, climax in thesoutheastern coastal plain. Ecological Monographs, 32, 167–185.

Development Core Team. (2011). R, A language and environment for statisticalcomputing. Vienna, Austria: R Foundation for Statistical Computing., ISBN 3-900051-07-0. URL http://www.R-project.org/

chultz, R. P. (1999). Loblolly – The pine for the twenty-first century. New Forests,17, 71–88.

chweizer, P. E., & Matlack, G. R. (2005). Annual variation in fish assemblages ofwatersheds with stable and changing land use. American Midland Naturalist, 153,293–308.

hirley, L. J., & Battaglia, L. L. (2006). Assessing vegetation change in coastal land-scapes of the northern Gulf of Mexico. Wetlands, 26, 1057–1070.

ohl, T. L. (2013). Southeastern Plains. USGS Land Cover Trends Project. Retrieved fromhttp://landcovertrends.usgs.gov/east/eco65Report.html

tanturf, J. A., Goodrick, S. L., & Outcalt, K. W. (2007). Disturbance and coastal forests,

a strategic approach to forest management in hurricane impact zones. ForestEcology and Management, 250, 119–135.

averna, K., Urban, D. L., & McDonald, R. I. (2004). Modeling landscape vegetationpattern in response to historic land cover, a hypothesis-driven approach for theNorth Carolina Piedmont, USA. Landscape Ecology, 20, 689–702.

d Urban Planning 121 (2014) 55–64

Titus, J. H. (1990). Microtopography and woody plant regeneration in a hard-wood floodplain swamp in Florida. Bulletin of the Torrey Botanical Club, 117,429–437.

U.S. Bureau of the Census. (1910–2007). Economic areas, Census of Agriculture,(1910–2007). Mississippi Counties and State. Washington, DC: U.S. GovernmentPrinting Office.

U.S. Census Bureau, State and County Quick Facts. (2011). Hatties-burg, Mississippi, Consolidated Federal Funds Report. Retrieved fromhttp://quickfacts.census.gov/qfd/states/28/2831020.html

USDA Soil Conservation Service. (1993). Soil survey of forest county, Mississippi. Wash-ington, DC: Government Printing Office.

Walker, L. C. (1994). The southern pine region. In J. W. Barrett (Ed.), Regional silvi-culture of the United States (3rd ed., pp. 271–333). New York: Wiley.

Walker, L. C., & Oswald, B. P. (2002). The southern forest, geography, ecology, andsilviculture. New York: CRC Press, LLC., 332 p.

nical Report SRS-53). Ashville, NC: US Department of Agriculture Forest Service,Southern Research Station.