In-Young Yeo & Chengquan Huang

thumbsshoesSoftware and s/w Development

Dec 11, 2013 (3 years and 8 months ago)

114 views

1 23
Regional Environmental Change

ISSN 1436-3798

Reg Environ Change
DOI 10.1007/s10113-012-0340-3
Forest dynamics in Mississippi, USA: a
hybrid statistical and geospatial analysis
In-Young Yeo & Chengquan Huang
1 23
Your article is protected by copyright and
all rights are held exclusively by Springer-
Verlag. This e-offprint is for personal use only
and shall not be self-archived in electronic
repositories. If you wish to self-archive your
work, please use the accepted authors
version for posting to your own website or
your institutions repository. You may further
deposit the accepted authors version on a
funders repository at a funders request,
provided it is not made publicly available until
12 months after publication.
ORIGINAL ARTICLE
Forest dynamics in Mississippi,USA:a hybrid statistical
and geospatial analysis
In-Young Yeo

Chengquan Huang
Received:10 October 2011/Accepted:27 July 2012
￿
Springer-Verlag 2012
Abstract
Understanding forest changes and its trajectory
is important to develop policy options and future scenarios
for climate analysis.This research is conducted to gain
insights on secondary forests change using Mississippi,
USA,as a case study.We investigate the spatial patterns
and temporal dynamics of secondary forests at high reso-
lution and examine the forces driving their changes.
An extensive literature review is conducted to refine the
conceptual framework of forest changes and identify the
underlying key factors.Forest changes are quantified at
high spatial (30-m) and temporal (biennial) resolutions,
using time series remotely sensed data between 1984 and
2007.A number of geospatial and socioeconomic data
were compiled to analyze the spatial variations of forest
disturbances and their linkages to various socioeconomic,
political,and biogeophysical factors.The results show that
the secondary forests are highly dynamic and variable.
Disturbances and regeneration occur continuously every-
where in a systematic and coordinated fashion.This pattern
prevents an extensive disturbance and increases total forest
cover.Market conditions (i.e.,timber price) are the key
predictor of the level and overall trend of forest distur-
bances.However,spatial patterns of forest dynamics
cannot be explained by location-specific biophysical,
socioeconomic,and policy factors identified in the litera-
ture.They can best be described by the ecological char-
acteristics of the forests (i.e.,the forest type and age
distribution),which have a clear economic linkage.The
research shows that regenerated forests frequently experi-
ence loss and gain of their extent,and their ecological
characteristics change drastically on a short-term basis.
These results point out challenges and opportunities in
forest management and policy with regard to reforestation.
Keywords
Secondary forests

Forest management

Geospatial analysis

Forest transition theory

Southern USA
Introduction
Land-use change depends on choices made to balance
societal needs and ecological services and functions
(DeFries et al.
2004
).Population growth and economic
development pressures have resulted in deforestation on a
global scale.The elevated rates of deforestation observed
during the 1990s,combined with global climate change,
were alarming and underlined the urgency for scientific and
policy solutions.Since then,reforestation has gained pop-
ularity as an important sustainability policy.Reforestation is
the process to reestablish tree cover in previously cleared
areas.It can happen naturally or artificially,and may take
many different forms,depending on the land management
objectives.Cleared forests recover to the original forest
types through natural processes,or can be replaced by timber
plantations with fast-growing trees (Nagendra and South-
worth
2010
;Le et al.
2012
).Governments and international
aid agencies commit substantial resources to restore and
protect forests,and tree farming has been effective in
expanding forest covers (Rudel
2009
;Mather
2007
;Perz and
Skole
2003
;Nagendra and Southworth
2010
).Understand-
ing the changes and dynamics of secondary forests and the
driving mechanismis critical to develop policy options and
future scenarios for climate analysis (Chazdon
2008
).
A number of previous studies have already investigated
forest change patterns and processes,but there is a great
I.-Y.Yeo (
&
)

C.Huang
University of Maryland,College Park,MD,USA
e-mail:iyeo@umd.edu
123
Reg Environ Change
DOI 10.1007/s10113-012-0340-3
need to improve understanding on the change dynamics of
secondary forests,and their implications for ecosystems
over the mid-term scale (several years to several decades).
For example,econometric-based studies have examined the
impacts of key physical and social variables on forest
management (Karppinen 1998;Joshi and Arano 2009;
Hyberg and Holthanusen 1989;Beach et al.2005).Com-
munity planning approaches provide key insights to assess
the impacts of public attitudes toward forest management
(Best and Wayburn 2001;Schaaf and Broussard 2006).
Land-use science has emerged to explain the mechanisms of
deforestation and feedback to the environment (Lambin
1997;Liverman et al.1998;Nagendra and Southworth
2010).Land-use science also offers forest transition theory
(FTT) to describe the long-term trajectory of forest change
([100 years),in relation to the development stages of
human society and the scarcity of ecosystem services
(Mather 1992;Rudel et al.2002,2005;Perz 2007;Grainger
1995;Lambin et al.2001;Perz and Skole 2003).However,
these three approaches emphasize either patterns or pro-
cesses over a short-termperiod,or the long-run trajectory of
forest cover,and they tend to focus on mature forests.
Therefore,there is a great need for studies on the spatio-
temporal dynamics of secondary (especially managed)
forests over the medium term (several years to decades).
These studies can complement prior work that focuses on
the short-term or cross-sectional studies or a longer term
([100 year).It also offers an important empirical basis to
refine FTT.It can be used to develop ideas on macroscopic
changes in the forest sector in the later stages of forest
transition and how this would vary from place to place.
The objectives of this research are to:(1) analyze
changes in secondary forests across multiple spatial and
temporal scales and (2) explore the driving forces for forest
changes.This research addresses not only areal changes in
forest cover,but also the conditions,dynamics,spatial
patterns,and transitional pathways of forest changes that
previous studies failed to address.Forest changes are
matched with various social and physical variables to
explore key driving factors,identified by integrating three
research frameworks of forest management discussed ear-
lier.We quantified forest dynamics at high spatial (30-m)
and temporal (biennial) resolution,using time series of
remotely sensed data from 1984 to 2007 and an innovative
land-change detection method (Huang et al.2010).This
geospatial data set shows the trajectory of forest changes
and dynamics at a high spatial and temporal resolution.We
chose Mississippi (MS) in United States,as a case study
(see Fig.1),because MS provides a unique opportunity to
study changes in secondary forests in the subtropics.MS,in
the southeastern USA,is well known for abundant forest
resources and a rich history of forest management;and
forestry has been very important to the local economy (Fox
et al.2007;Henderson et al.2008).In MS,forest land-
scapes experienced dramatic changes from major clearings
to full recovery (Fickle 2001;Yeo and Huang 2012).
Reforestation efforts have been initiated as a response to
the severity of forest clearing,and the scarcity of its
resources.Forest landscapes have been intensively man-
aged by the public,government,and industry,and have
fully recovered,largely through plantation.The humid
subtropical climate and other physical conditions provide a
Fig.1 Location of the study
area
I.-Y.Yeo,C.Huang
123
highly productive environment that favors the fast growth
of secondary forests.
Theoretical background:drivers of forest change
We reviewed three research streams that highlight different
aspects of forest management and change,in relation to
human society and the physical environment:econometric-
based analyses,land-use science,and community-based
planning.The three frameworks offer different conceptual
perspectives to understand decision-making processes on
forest resources,but they are highly interconnected and
complementary.In terms of scale,econometric-based
approaches aregenerallyappliedat the micro-level andtendto
be ‘‘cross-sectional’’ or ‘‘case-specific’’ over a short-term
period (Zhang and Mehmood 2001;Karppinen 1998;Joshi
and Arano 2009;Arano and Munn 2006).Land-use science is
applied at the regional or macro-level.Community-based
planning bridges both by considering the impacts of social
networks and public views.Based on this literature and data
availability,we have developed a list of key factors that are
relevant to the spatial and temporal variations in forest land-
scapes.We first briefly discuss the three general approaches
and then the key factors that affect forest management.
Review on relevant research streams
Econometric approaches
In general,econometric-based approaches use micro-econo-
metric theory as a conceptual framework for understanding
decision-making processes.They assume individuals make
decisions based on their values and long-term objectives
(Karppinen 1998;Joshi and Arano 2009).Two basic micro-
economic concepts are used to investigate the link between
the socioeconomic systems and forest condition:profit max-
imizationor utilitymaximization.The first assumes that forest
owners maximize their profits over time,without considering
the benefits associated with non-market goods.The latter,
however,recognizes that landowners may have multiple
objectives for forest management (e.g.,esthetics,recreation,
and wildlife habitat,in addition to timber value).These land
owners mayharvest less intensivelybut invest moreheavilyin
reforestation (Hyberg and Holthanusen 1989).Recent studies
indicate that nonindustrial private forest (NIPF) landowner
behavior is better understood by factors other than profits.
Beach et al.(2005) used a simple conceptual framework to
explain these behaviors and provided an organizing structure
based on synthesis of the empirical micro-economic litera-
ture.Assuming that landowners derive utility from non-
market timber amenities and all other goods,the utility
function of the individual owner can be expressed as:
U ¼ UðY;NÞ
where Y is the present value of all future income and N is
non-market timber amenities.The total income Y is the sum
of timber income (Y
tim
) and non-timber income (Y
nontim
).
Land-use science approaches
The second framework stems from land-use science,which
examines forest cover change in the context of regional or
national land-use dynamics.In general,two different
approaches are used.The first one is driven by remotely
sensed data.Land-use patterns observed from remotely
sensed data generate questions on changes and their driving
mechanisms (Lambin 1997;Liverman et al.1998;Nagendra
and Southworth 2010).Key geographical indicators (e.g.,
distance from a road) are identified empirically to explain
deforestation or its rate of change.Remotely sensed data can
be easily aggregated to a higher level (e.g.,the local com-
munity or larger) to examine key land changes and their
relationship to other social or physical variables.This type of
research is typically more spatially explicit,compared to the
econometric approach.Another approach in land-use sci-
ence,especially related to forest changes,is derived from
FTT,which offers a conceptual framework to explain long-
term forest trends in light of regional development stages.It
describes how changes in population,industry and technol-
ogy,economic market,and public perception and values
affect regional or national forests over a long time span
([100 years).These studies typically use historical obser-
vations and data from the Organization for Economic
Co-Operation and Development (OECD) countries.Histori-
cal records date back to the nineteenth and twentieth centu-
ries,and show a slow transition in forest cover from
shrinkage to expansion,after changes in industrial activities
and technology brought a dramatic shift in population and
land-use demand in rural and urban areas.At the last stage of
development,an increase in forest cover occurs naturally or
artificially,through succession to secondary forests or plan-
tations on abandoned agricultural fields (Mather 1992;Rudel
et al.2005;Perz 2007;Grainger 1995).Two general path-
ways are assumed to describe the mechanisms that trigger
FT:‘‘economic development path’’ and ‘‘forest scarcity path’’
(Mather 2007;Rudel et al.2005).The first mechanism
assumes that increases in forestlands are achieved by
recovering forests fromagricultural fields,after migration of
rural populations to urban areas.The second assumes that FT
occurs as response to shortage in forest products and other
ecosystem services.There has been a growing interest in
explaining the drivers and processes of forest change within
the FTT framework,using remote sensing,especially for
developing countries in tropical regions (Rudel et al.2002;
Lambin et al.2001;Perz and Skole 2003).
Forest dynamics in Mississippi
123
Community planning approaches
The third research area is largely influenced by ideas of
community-based planning.This research seeks to improve
our understanding of the public’s attitudes toward timber
harvesting and forest management.It focuses on the fact
that the public receives many tangible and intangible
benefits not only from public forests,but also from private
forests.Current development trends point to amenity-dri-
ven relocation (e.g.,exurbanization) and an increase in the
recreational uses of forests.Exurban development in forest-
dominated landscapes causes more forest parcelization and
diversifies its ownership across different demographic
groups.The public has become more cognizant of the value
of private forests and is willing to invest in them (Best and
Wayburn 2001).Public awareness may lead to more suc-
cessful forest conservation strategies and policies through
citizen engagement in policy-making processes.Key issues
here lie in understanding the public preferences and views
on private forest policy and resources management,and
identifying key (proxy) variables fromdemographic factors
that can be used to gauge public support for forest man-
agement (Schaaf and Broussard 2006).
Key factors affecting forests
Because the econometric approach offers a general orga-
nizing structure,we have adapted its framework as shown in
Beach et al.(2005),with additional variables fromland-use
sciences and community planning.These variables are
related to the geographical characteristics of the physical and
social environment,mayindicate long-runchanges inhuman
society,or informthe level of public awareness.Beach et al.
(2005) present four primarycategories of forest management
factors that produce Y:market drivers (MD),land owner
characteristics (OC),policy (PV),and plot/resource condi-
tion (PR) factors.These factors are identified from results
presented in the literature through meta-analysis and vote
counting.Previous empirical studies have produced various
model forms and variables to specify Y = f(MD,OC,PV,
PR),as summarized by Beach et al.(2005).We explain the
four primary factors and identify specific variables under
each category by integrating the three theoretical approaches
(see Table 1 for a summary of variables).
Market drivers
Market drivers (MD) are often considered the most
important variables.They include factors that explicitly
represent costs or return of forest and related products.The
price of timber is the most commonly used market variable.
In general,an increase in timber profits (e.g.,decrease in
planting costs,increase in the return to forestry relative to
croplands and other land uses) tends to increase invest-
ments in forests.Land-use science provides additional MD.
For example,distance to roads or processing mills are
important factors related to accessibility to/from forests,
and key determinants of production costs.Land price or
population change,key variables driving land-use pattern,
may be used as indicators of the demand for competing
land uses.For example,a growing population increases the
demand for urban land and results in more forest clearing.
The long-term population trend may reflect the develop-
ment stage of the region,which is closely related to forest
transition in rural areas (Mather 1992).
Land owner characteristics
Land owner characteristics (OC) are used to measure the
preference,resources,perception,and values of landowners
and the characteristics of the local communities.Previous
econometric studies have often been based on surveys of
individual owners to determine the likelihood and extent of
different silvicultural practices.Age and education are used to
approximate preferences and attitude toward forests,and
incomeis usedtomeasureavailableresources for investment in
forestry or preference for the amenities that forests provide.
Some studies have reported the importance of ownership
characteristics (e.g.,property size,length of tenure) as key
determinants of private forest management decisions (Joshi
and Arano 2009).However,individual surveys are of limited
value,as they are specific to a location or period,and are rarely
available publicly in a spatially explicit format.Schaaf et al.
(2006) show little difference in timber harvesting attitude
between those who own forestlands and those who do not,and
also little difference across ownership characteristics (e.g.,
landholding size,existence of a management plan).Schaaf
et al.(2006) argued indifference toward forest management is
mostly due to shifting demographics.An aging forest owner
population,an increasing number of forest owners in general,
and a growing number of previously urban residents owning
forests,all makeownershipnolonger unique totraditional rural
residents (Birch 1996;Schaaf et al.2006).It may be useful to
study whether there is any difference in forest changes due to
the dissimilarity in socioeconomic characteristics and land
demand (or value) across rural and urban communities (Munn
et al.2002).Overall,earlier findings support the importance of
public records,such as the census,to provide insights on the
value and motivation of individual owners and communities.
Policy variables
Policy variables (PV) are generally related to forest man-
agement practices,and include federal,state,or local
programs designed to change the allocation of forest lands
or resources,or to reduce forest management costs (e.g.,
I.-Y.Yeo,C.Huang
123
tax incentives,cost sharing,and technical assistance),which
would increase forestry activities.Policy effectiveness and
receptiveness are alsoimportant factors togauge their impacts
(Sun et al.2009).However,the effects of PV are extremely
difficult to measure,as information on forestry policy and
programs (e.g.,allocation of financial resources to policies,
distributionamongindividual landowners,andthe impacts on
forest investment andmanagement) is not available.Instead,it
may be possible to indirectly evaluate their effectiveness and
receptiveness through local communities and individual
landowners’ characteristics.Previous studies have shown that
the effectiveness of policy variables is related to the knowl-
edge and available resources of landowners (Sun et al.2009).
On the other hand,land-use science takes a geographically
explicit approach to assess the impacts of conservation man-
agement on forest disturbances.For example,it asks whether
forest changediffers bylandconservationstatus or bydistance
to natural features that provide various ecosystem services
(e.g.,biodiversity,water quality).These variables can be
derived in a spatially explicit way.
Plot/resource conditions
Plot/resource conditions (PR) are used to assess the phys-
ical condition and quality of forest production.It is
assumed that there is a higher incentive to engage in forest
Table 1 Key variables categorized by the four factors of forest change
Category Factors Variables Temporal
scale
Spatial
scale
Data source
Market variables (MD) Output price for wood products Timber price 1984–2006
(annual)
State-level Mississippi
extension
Distance to roads,
distance to mills
1990–2000
(decadal)
30-m Mississippi
extension
NLCD
US census
Competing land use Population change 1970–2000 County US census
Regional land-use
change/dynamics
1990–2000
(decadal)
30-m NLCD
US Census
Characteristics of owners and
local communities (OC)
Value/perception/preference Age,education,
employment,race
1990–2000
(decadal)
County or
block
group
US census
Urban and rural land uses 1990–2000 30-m NLCD
MS-GAP
US census
Population density Proxy for the plot size
(group size per unit
area)
1990–2000 County or
block
group
US census
Available resources for investment in
forestry or preference for amenities
Income,Income
distribution
1990–2000
(decadal)
County or
block
group
US census
Long-term objective Land ownership
categories
1992–1995 30-m MS-GAP
Policy variables (PV) Management (conservation) level and
status
Land ownership
categories
1992–1995 30-m MS-GAP
Water policy Distance from waterbody 1980–2006 30-m NLCD
USGS
Plot/resource conditions (PR) Physical production processes and
condition
Regional land-use
change/dynamics
1990–2000
(decadal)
30-m NLCD MS-
GAP
Site conditions—soil quality Geologic soil map 1969 – MS
geologic
maps
Site conditions—slope Slope 2001 30-m DEM
Forest species Forest species 1977–2006 30-m NLCD
MS
geologic
maps
MS-GAP
Forest dynamics in Mississippi
123
production with better PR conditions.Plot size,slope of
land,and/or soil quality is often used as plot/resource
conditions.The geographical approach of land-use science
offers a number of physical variables and provides useful
tools.Certain physical variables,such as forest species or
biodiversity,are used to measure amenity and recreational
opportunities,which may indicate a need for conservation.
Description of the study area
MS is located in the eastern south-central US,with a land
area of 123,514 km
2
and a 2002 population of 2.8 million.
It is one of the most rural states of the United States.While
its urban population has increased over the years,only
13 % of state residents live in cities.Overall,the MS
economy remains poor and stagnant.Its gross state product
was ranked 35th among US states in 2001,with the lowest
per capita personal income (PCPI),at 71 % of the national
average.Over 1999–2001,the average poverty rate was
16.8 %,ranking MS as 49th among the 50 states.
MS lies entirely within two lowland plains,the Missis-
sippi Alluvial Plain (also known as the Delta) and the Gulf
Coastal Plain,and its elevation varies from0 to 246 m.MS
belongs to the humid subtropical climate zone,with
monthly average temperatures ranging from1.6 to 33.6 ￿C.
This climate supports abundant forest resources.Changes
in forest landscapes are largely due to timber harvesting
and regeneration practices designed to ensure the supply of
sawtimber and pulpwood to the processing industry.
Dramatic land-use changes have occurred throughout the
nineteenth and twentieth centuries.Major forest clearing
was started by logging companies in the late nineteenth
century and left much land abandoned or converted to
agriculture.It was not until the 1930s that the federal
government initiated a major reforestation effort,with land
acquisition and the establishment of National Forests.Most
abandoned and disturbed lands were surveyed and reha-
bilitated during the Great Depression (Fickle 2001).Other
rehabilitation efforts were undertaken to improve water
quality,erosion problems,and the health of ecological
communities and biodiversity,including planting native
species in specific ecoregions (e.g.,bottomland hardwoods)
(Duffy and Ursic 1991).In the private sector,the rapid
growth of the timber industry began during the 1930s,with
extensive tree-farming practices beginning in the 1940s and
major scientific efforts to improve timber production.This
led to the intensification of plantation forests (Fickle 2001).
Inthe 1970s,the Mississippi ForestryCommission(MFC)
started an effort to promote tree planting by NIPF owners.In
the 1980s,the MFC cooperated with the US Forest Service
(USFS) to develop management plans,and provided guid-
ance to assist NIPFs in regenerating their timberlands.
Financial and technical assistance [e.g.,the Mississippi
Forest Resource Development Program (FRDP),the Forest
Incentive Program (FIP),the Mississippi Reforestation
Tax Credit (RTC),and the Conservation Reserve Program
(CRP)] was initiated to encourage forestation,particularly
targeting NIPF owners,who own 78 % of forestlands in
MS (MFC 2010).These programs provided cost sharing
for forest planting and timber stand improvement,
encouraged compliance with management plans by MFC,
promoted the conversion of agricultural lands to forest-
lands,and provided credits for state income tax (Sun et al.
2009).These programs,especially the CRP program of
1984,have encouraged plantation,such as a short-rotation
(15- to 30-year) loblolly pine plantation (Wear and Greis
2002).
By the late 1990s,MS forests had fully recovered from
the major clearing of the early twentieth century.Forest
cover has increased in both public and private sectors.
Changes in government policy to support multiple use of
forests and ecological management,combined with
increasing public awareness and appreciation for forests,
have resulted in a significant reduction in harvesting from
public forests,and better forest management and tree-
farming practices from NIPFs and industry.Details on the
MS forest history are available in Fickle (2001).
Data and methods
Data sources
Key geospatial data characterizing forest dynamics were
obtained from a time series of remote-sensing data and an
innovative image processing technique that tracks changes
in forest growth stages (Huang et al.2010).This forest
mapping product was derived using spectral and temporal
information from the historical records of Landsat The-
matic Mapper (TM) and Landsat Enhanced Thematic
Mapper Plus (ETM?).The time series information was
used to develop forest likelihood functions (or profiles),by
detecting changes in vegetation throughout the observation
period.This allowed the classifying of forests as persisting
forest,disturbed forests,and regenerating (or immature)
forests,and the characterizing of the short-term dynamics
of forests in each 30-m cell.Disturbed forests include
any type of disturbing activities (e.g.,harvesting,thinning,
fire),and classified as single or multiple disturbances.
Single disturbances are usually related to ‘‘clear-cut’’ tim-
ber harvesting,multiple disturbances to other natural dis-
turbances or forest management actions (Huang et al.
2010).The data show changes in forest cover categories
every 2 years up to 2007,based on forest cover information
between 1984 and 1986.The data were generated from132
I.-Y.Yeo,C.Huang
123
Landsat scenes over the period of 1984–2007,and vali-
dated with detailed field studies (Huang et al.2009).
Other spatial information was derived from the National
Land Cover Databases (NLCD),the Mississippi Gap
Analysis Project (MS-GAP),and Digital Elevation Models
(DEM) from the US Geologic Survey (USGS).DEM data
were used to calculate slopes using ArcGIS.The NLCD of
1992 and 2001 were used to understand the relationship
between forest cover changes and regional land-use
dynamics.The MS-GAP data provided spatial information
on vegetation cover and more detailed land-use patterns
than the NLCD,at a 30-m resolution.MS-GAP has built
upon the NLCD of 1992 with 18 additional Landsat TM
data obtained over 1991–1993 and from information
derived from 273 color infrared aerial photographs.Highly
cooperative and iterative efforts among government agen-
cies,universities,conservation groups,and individual land
owners helped to improve the accuracy of vegetation
mapping and land-use information through extensive
ground studies (MS-GAP 2012).The data set includes
valuable information on the ecological characteristics of
forests (e.g.,vegetation types,their succession stages),
their spatial distribution,and the conservation status of
biodiversity under existing land ownership and manage-
ment regimes.
Historical records on agricultural statistics,forest cen-
sus,and demographic variables were also collected and
analyzed.The forest census data were obtained from the
Forest Inventory Analysis (FIA) Program of the US Forest
Service (USFS).The FIA database provides information on
forest areas and locations,including species,size,health,
growth,mortality,and removals of forests.This data set
was collected from sampled plots,regardless of ownership,
management status,or political boundaries,and aggregated
to provide forest information at the county or state level.
The survey cycle was changed recently from a 10–12-year
cycle (at its inception) to a 5-year cycle in all states since
1998.Details on the FIA data set are provided by Smith
(2002) or are available online (USFS 2010).The study area
has detailed information on forest species and forestlands
at the county level in 1977,1987,1994,and 2006.The first
three data sets (1977,1987,and 1994) include county-level
statistics on broad categories of forest species and regen-
eration methods (artificial vs.natural) by ownership cate-
gories (i.e.,government,industry,and NIPF private sector).
The 2006 FIA data were processed to generate consistent
information,by re-grouping forest species into broad cat-
egories.However,the 2006 data could not be analyzed by
ownership category at the county level,due to the
unavailability of the data.Agricultural statistics on farm
lands,harvested farm lands,and forested areas were
obtained from the US Department of Agriculture (USDA
Economic Research 2010;USDA-National Agricultural
Statistics Service (NASS) 2010).Demographic,social,and
economic data were obtained from the US Census Bureau
at various community scales (i.e.,block,county,and state).
Individual owner information was not available publicly,
and the analysis was limited to analyze the impacts of local
community characteristics on forest cover changes.Table 1
provides a summary of the compiled variables categorized
by the four factors of forest change.
Data processing and method
Geospatial analysis of forest change
A number of statistical and geospatial methods were
applied to quantify forest dynamics.Forest mapping data
and land-use maps of the NLCD and MS-GAP were geo-
referenced and processed using a Geographic Information
System (GIS).They supported an explicit geospatial
analysis,and were integrated at various spatial scales by
location.As there were differences in data sources (e.g.,
timing of data acquisition,image seasonality,data quality
issues due to differences in remote-sensing sensors),geo-
registration,and classification methods,a direct compari-
son of the various geospatial data was not possible.Spatial
data from different sources were reviewed carefully to
understand variable definitions and to generate consistent
and compatible land-use and vegetation categories.
Therefore,the spatial data were cross-compared and tested
at various spatial scales by checking the rate,extent,and
transitional pathways of land-use changes.Forest changes
were characterized by computing forest gain,loss,net
change,the frequency of disturbances,and cumulative
disturbances.Special attention was paid to the timing of
data acquisition and the processing algorithms used in
various geospatial data to better understand the nature of
the vegetation covers,their ecological characteristics,and
their growth stage.
Assessment of key drivers:a hybrid approach of statistical
and geospatial methods
Various exploratory statistical methods,such as descriptive
statistics,t tests,analysis of variance (ANOVA),and
regression analysis,were used to investigate the impor-
tance of key factors.Overall,the statistical analysis con-
sisted in a series of bivariate tests to examine the
importance of linkage between each variable and forest
change over a range of temporal and spatial scales.Most
physical and social variables were available at a coarser
resolution than forest mapping data.Therefore,forest
mapping data were spatially aggregated or temporally
integrated to match the scales of the other variables.Since
the scope of this study was to explore linkages,the analysis
Forest dynamics in Mississippi
123
remained exploratory but could be further expanded to a
multivariable framework,as presented in Table 1.Details
on the applied methods are briefly discussed.
We analyzed the effects of four MD factors:(1) timber
price,(2) distance to processing mills,(3) distance to roads,
and (4) population dynamics.The overall impacts of the
timber market on forest land were analyzed using the
standing timber prices observed over 1986–2007.Due to
limited data availability,the relationship between timber
price and forest changes was assessed qualitatively by
comparing the changes in timber price with the changes in
forest covers.Both disturbance and regeneration were
considered.The impacts of distances to processing mills
and roads were analyzed by correlating the relative
cumulative disturbances with the distance to mills or roads.
The relationship between population and forest distur-
bances was examined in a two-step procedure.First,we
studied the overall trend of county populations since 1930.
Among various indicators of population change (e.g.,
population density,total population,the number of housing
units),we identified a single variable that best represented
the population change over the time.We used the relative
population growth rate as a key variable and related it to
the cumulative forest disturbances by county,using cor-
relation analysis.ANOVA was used to study forest dis-
turbances by population groups.
Various social variables from the US Census Bureau
were correlated with forest disturbances in a three-step
procedure.First,the temporal variations within the vari-
ables were analyzed using historical US census data over
1930–2000,at the county level.Because the longitudinal
study showed little variation over time,we conducted two
cross-sectional analyses of forest disturbance observed
between 1999 and 2001,and of the cumulative disturbance
between 1984 and 2007,using 2000 demographic vari-
ables.Lastly,a similar analysis was done at the block
group level to assess the impact of spatial scale.In addi-
tion,the impacts of urban and rural areas on forest dis-
turbances were analyzed by quantifying the extent and
frequency using their geographical boundaries.
Three variables—management effects,land conserva-
tion,and water conservation—were considered under the
PV category.Forest losses and gains,cumulative distur-
bances,and frequency of disturbances were quantified in
terms of the conservation management,land tenure,and
distance to protected area and streams.Two PR variables
were considered:slope and forest species.Using the
DEMs,ArcGIS calculated the slopes and categorized them
into three groups.Forest disturbances were aggregated by
slope categories and compared.Finally,forest changes
were analyzed across broad categories of ecological com-
munities.Forest species information was derived from the
FIA and MS-GAP data.The overall trends in forest areas
by forest types were analyzed using historical FIA data at
the county level.Accordingly,forest disturbances were
estimated at the county level,by aggregating disturbances
fromforest mapping data,and were correlated with the FIA
data set to investigate forest disturbances by forest types.
The spatiotemporal analysis was expanded by cross-com-
paring MS-GAP,NCLD 2001,and forest mapping prod-
ucts.We derived time series of forest change data at a high
spatial resolution by broad categories of ecological com-
munities.MS-GAP provided the spatial distributions of
trees.Their age distribution was derived from forest
mapping data,by tracking how long an area has remained
forests since regrowth occurred (Yeo and Huang 2012).
This information was used to refine the land-use categories
of the 2001 NLCD for vegetation cover,based on spectral
characteristics.Time series of forest gain and loss were
calculated based on these new land-use categories.
This study has two limitations.Because of limited data
availability,this study could not explicitly assess the impacts
of individual owners’ characteristics on forest changes and
management practices.However,Schaaf et al.(2006)
showed that there is little difference in timber harvesting
attitudes across ownership characteristics,especially for
highly fragmented forest landscapes with shifting demo-
graphics.Therefore,a study of local community character-
istics may provide insight into the decisions of individual
owners,and our analysis was done using publicly available
demographic variables that reflect characteristics of local
communities.Policy variables (PV) generally influence for-
est investment decisions and management practices.For
example,earlier studies have included a variable for gov-
ernment cost sharing,to determine whether this has an effect
on overall reforestation (Beach et al.2005).In the study area,
a number of government programs have been available to
provide incentives and technical assistances.However,
detailed information on forestry policy and programs is not
available to directly assess changes in forest dynamics due to
forest policy.As these programs have been well perceived by
the local communities and individual land owners (Fickle
2001;MFC 2010;Sun et al.2009),we assume that the
effectiveness and receptiveness of these policies (assuming
their availability to all land owners) would vary by the
characteristics of local communities and individual land
owners (e.g.,their knowledge and available resources in
forest management and investment).Instead,we focus on
assessing the impacts of conservation management on forest
management patterns.
Results
Section ‘‘Forest change:spatial and temporal patterns’’
reports the finding on the extent of change,pattern,and
I.-Y.Yeo,C.Huang
123
dynamics of secondary forests at multiple spatial and
temporal scales.Section ‘‘Statistical assessment of key
determinants of forest disturbance’’ describes how these
changes are related to the four primary factors (MD,OC,
PV,and PR) and key variables,as presented in Table 1,
and sections ‘‘Key factors affecting forests’’ and ‘‘Data
processing and method’’.
Forest change:spatial and temporal patterns
Table 2 summarizes the changes in forest cover over
1984–2007.The forest mapping products show that 54 %
of MS was covered by forests in 2007.During the moni-
toring period 1987–2007,MS lost 2 % of forests on an
annual basis,but this loss was offset by forest regeneration.
As a result,only 2.2 %of disturbed forests were converted
to urban use or remained barren as in 1984.Despite a small
annual net loss,MS forests have experienced strong spatial
and temporal changes over the last two decades.First,a
considerable disturbance has taken place at the rate of
5–10 % of the total land surface statewide.Most distur-
bances (C 96 %) are ‘‘single’’ disturbance events,often
related to timber harvesting.Second,Table 2 shows that
forests include significant shares of disturbed (e.g.,
12–17 %) or regenerated (e.g.,21–26 %) forests.Distur-
bance and regeneration are inversely correlated (r =
-0.74,p value = 0.056,n = 7).The greater the distur-
bance,the less the regeneration.Third,Table 2 indicates
that regeneration does not immediately follow disturbance.
Rather,there is a time lag fromdisturbance to regeneration.
This was further analyzed by tracking the number of years
for disturbed areas to show signs of vegetation (i.e.,
regeneration) at the cell (30-m) level,using forest mapping
products.The results point to less than 30 % regeneration
within 3 years.It takes more than 10 years for 70 % of the
disturbed areas to become forested again (Yeo and Huang
2012).Fourth,forest disturbance is spatially extensive.
Figure 2 presents biennial disturbance patterns during the
monitoring period,aggregated at the county level.A sys-
tematic and coordinated pattern is evident.Disturbances
are widespread at low intensity (i.e.,\10 %of forest cover
in a county).Their spatial patterns vary highly and dis-
turbed locations rotate during the monitoring period.This
has incrementally affected the entire state,causing distur-
bance of 40 % of the total forest land.Mean cumulative
disturbance for each county is approximately 54 %,with a
standard deviation of 11 %.The highly dynamic distur-
bance and regrowth patterns indicate that MS forests are
intensively managed to maintain and increase forest covers,
with short-term fluctuation.The spatial and temporal
analysis with a time series remote-sensing data over the last
two decades revealed MS forests experienced very frequent
changes and disturbance,and these change patterns were
not apparent when considering the total forest area esti-
mates.This indicates the need for conducting detailed
spatiotemporal analysis to develop a comprehensive
understanding of the changes in highly human-dominated
secondary forests.
Statistical assessment of key determinants of forest
disturbance
Market-driven (MD) variables
Timber Price Figure 3 presents changes in standing
timber prices and changes in forest cover (%).Overall,
timber price has increased over the last two decades.Prices
of most timber products have more than doubled.During
the period of rising timber prices (prior to 1996–1998),the
total forest cover increased with increasing regeneration
but decreasing disturbance.However,after the maximum
timber price is reached and timber prices stabilize,the
forest cover decreases with increasing harvesting and less
reforestation.The temporal pattern shows that expectation
of high economic returns,due to rapidly increasing timber
prices,tends to increase regeneration,accumulating for-
estry capital and growing stocks as an investment oppor-
tunity.When the market stabilizes at a high timber price,
the trend is reversed,with more harvesting but less
regeneration.
Distance to processing mills and roads The results show
a gradual decrease in forest disturbances with increasing
distance to the mills.The highest cumulative relative dis-
turbance (i.e.,total disturbance over 1984–2007 normal-
ized by the forest area) occurs closest to the mills
Table 2 Changes in forest cover estimated from remotely sensed
data (in 1,000 km
2
) and forest land area estimated from FIA (in
1,000 km
2
),1987–2006
Years Total forest
cover
Disturbed forest
cover
Regenerated forest
cover
1987–1989 66.13 (53.6 %) 18.51 (15.0 %) 27.40 (22.20 %)
1990–1992 69.00 (55.9 %) 15.47 (12.5 %) 30.44 (24.70 %)
1993–1995 72.44 (58.7 %) 14.49 (11.7 %) 31.96 (25.90 %)
1996–1998 69.10 (56.0 %) 16.40 (13.3 %) 30.05 (24.30 %)
1999–2001 70.17 (56.8 %) 18.96 (15.4 %) 27.48 (22.30 %)
2002–2004 69.81 (56.5 %) 15.65 (12.7 %) 30.79 (24.90 %)
2005–2007 66.04 (53.5 %) 20.87 (16.9 %) 25.58 (20.70 %)
The percent change is relative forest cover to total state area
Total forest cover includes areas that are persistent forest cover,
disturbed forest area during the monitoring period,and regenerated
forest cover
Disturbed forest area includes areas disturbed during the monitoring
period and previously disturbed area but remained non-forest
Forest dynamics in Mississippi
123
(0–50 m),at a rate of 46 %.However,there was little
variation in total cumulative disturbances by distance (in
the range of 0–80 km) from the mills.The mean cumula-
tive disturbance over the distance range is approximately
42 %,with a standard deviation of 1.9 %.The correlation
coefficient (r) between the distance to the mills and dis-
turbance is -0.38 (p = 0.18),indicating that this distance
does not significantly affect the amount of disturbance.We
also analyzed the correlation between cumulative distur-
bance (%) and distance to roads (Fig.4).The results
indicate a negative relationship (r = -0.89,p = 0.0001):
the closer the roads,the higher the disturbance.This effect
is most significant within a 1-km radius from the road.
Approximately 50 % of the forest cover within 1 km is
disturbed.Although most disturbances ([90 %) occur only
once (related to timber harvesting),more frequent and mul-
tiple disturbances are observednear the roads.The cumulative
disturbance sharply drops at 10 kmfromthe roads.
Population dynamics This variable is used as a proxy to
assess the competing demands for and prices of land.It is
generally assumed that an increasing population and
urbanization increase the demand for land and lead to
increasing land values.The data indicate very little change
in population during the period 1930–1970,but an
increasing trend since 1970.Over the period 1970–2000,
MS population increased by 28 %,with the growth rate
being highly variable across counties (-36 to 198 %,with
a mean and standard deviation of 26 and 42 %).The
population growth rate is positively related to the 2000
Fig.2 Changes in forest disturbance pattern,1987–2006.Note:Disturbance (%of forest) indicates the share of disturbed forest,estimated from
the total forest cover in county
I.-Y.Yeo,C.Huang
123
population density (r = 0.44,p\0.0001),2000 popula-
tion (r = 0.87,p\0.0001),and 2000 total housing units
(r = 0.996.p\0.0001).Therefore,the relative population
growth rate is used to represent changes in population and
is correlated with cumulative forest disturbance (%)
(Fig.5).Although a weak but statistically significant
relationship exists (r = 0.35;p = 0.001),this relationship
appears to be strongly affected by outliers.If extreme
values are removed at the 95 % threshold (i.e.,data values
out of the 95 % range of population and disturbance data),
the relationship becomes random without much statistical
significance (r = 0.12;p = 0.33).Using the range of rel-
ative growth rates,the population was subdivided into 5
groups (ANOVA,F statistics = 134.5;p = 0) to study
forest disturbance by population group.The results show
that forest disturbances are highly variable within the
population groups,indicating population dynamics offer
little explanation on forest disturbance (ANOVA,F statis-
tics = 1.09,p = 0.37).
Fig.3 Timber prices and forest
cover change
0
2
4
6
8
10
12
14
36
38
40
42
44
46
48
50
52
Distance from roads, km
Cumulative forest disturbance (%)
0
5
10
10
20
30
40
50
60
70
80
90
Distance from roads, km
Cumulative forest disturbance (%)
Single Disturbance
Multiple Disturbance
Fig.4 Cumulative forest
disturbance with distance from
major roads
-50
0
50
100
150
200
10
20
30
40
50
60
70
Relative Population Changes (%), 1970-2000
Cumulative Forest Disturbance (%)
All data
Data without outliers
Fig.5 Relationship between population changes and cumulative
forest disturbance
Forest dynamics in Mississippi
123
Characteristics of owners and local communities (OC)
Social characteristics We tested various demographic
variables,such as income,education,age,race,industry
sector employment,and population density,to explore their
relationship to forest disturbances.The results indicate
weak or insignificant relationships between OC variables
and the 1999–2001 forest disturbances,in contrast to the
relationship with cumulative disturbances.Summary
results with OC variables are presented in Table 3,based
on 2000 US Census block group data,that provide the best
explanations for forest disturbance.Although the r value is
very small,the results generally suggest increasing forest
disturbance in areas with higher population density and an
increasing number of new housing units.More distur-
bances are observed in local communities with lower-
income groups and higher employment in agriculture,
including silviculture.Education variables provide mixed
results.Overall,the results indicate that most of the OC
variables have either no (i.e.,high p value,p[0.1) or little
(i.e.,low r value,r\0.1) association with the cumulative
forest disturbance.
Forest disturbance across urban and rural areas Because
urban areas are characterized by a higher population den-
sity and more new housing stock,we further analyzed
forest change and management across the urban/rural
boundary.This variable may capture the impacts of com-
peting land uses and differences in public attitude and
perception across rural and urban communities.The results
show that 96 % of disturbances occurred in rural areas,
where 98 % of forest lands are located.However,urban
forests also experienced large disturbances,given the share
of forests in the urban category.Urban areas have less than
2 % of forest cover,but more than half of urban forests
(54 %) experienced disturbances,a rate much higher than
rural forests (41 %disturbance).Urban forests experienced
more frequent multiple disturbances than rural forests.
Approximately 10 % of disturbances in urban forests are
multiple disturbances,in contrast to 5 % for rural forests.
Policy variables
Management effects We examine whether forest distur-
bances differ by land tenure and level of conservation.
Conservation management status is based on land owner-
ship and level of conservation afforded for the biological
diversity (MS-GAP 2012).Approximately 96 % of cumu-
lative forest disturbances over 1984–2007 have occurred in
unprotected private lands.Overall 20 % of the relative
cumulative disturbances (i.e.,the cumulative amount of
forest disturbance over 1984–2007,normalized by the total
forest cover) are observed in protected areas,a rate that is
less than a half the rate observed on private lands.Most
disturbances are ‘‘single’’ disturbances.In general,pro-
tected lands display fewer disturbances and higher net
forest gains.A significantly higher forest net gain on pri-
vate lands has occurred over 1990–1992 and 2002–2004,
when timber prices were highest.However,the overall
Table 3 Statistical summary of social variables and their relationships with cumulative forest disturbance
Social factors Selected variables Mean SD Min Max R value P value
Demographic Population density 4.3 6.3 0.0 60.0 0.11 0.00***
Racial Ratio white/black 68.8 270.4 0.0 2980.0 0.03 0.16
Age 18–65 (%) 60.0 6.7 0.0 104.3 -0.01 0.60
65 (%) 12.9 5.5 0.0 51.6 0.02 0.46
Income\$ 10 k (%) 26.1 9.6 0.0 94.1 0.05 0.03**
\$ 30 k (%) 72.8 12.4 0.0 100.0 -0.01 0.67
Poor group (%) 17.7 13.3 0.0 100.0 -0.02 0.45
Non poor group (%) 82.0 13.9 0.0 100.0 0.03 0.14
Education 0 grade (%) 1.7 2.6 0.0 67.6 -0.04 0.08**
10 grade (%) 12.2 7.4 0.0 50.0 -0.07 0.00***
High school (%) 29.2 8.9 0.0 61.3 -0.06 0.00***
Bachelor (%) 16.1 9.5 0.0 63.7 0.10 0.00***
Employment Unemployed (%) 8.0 6.2 0.0 60.3 -0.01 0.68
Employed in agriculture (%) 1.5 2.3 0.0 18.7 0.05 0.03**
Urbanization Number of houses built 16.8 29.0 0.0 392.0 0.08 0.00***
Selected variables are collected from 2000 US Census block group data.Most of variables are normalized by block group population
The poor group under income category is defined by US Census,according to US federal criteria for ‘‘poor’’
Employed in agriculture under employment category includes sivilcultural forestry
*** Indicates that the variable is statistically significant at\1 %,and ** at\5 %
I.-Y.Yeo,C.Huang
123
changes in forest disturbances and net changes from pro-
tected areas have trends similar to those on private lands
(Fig.6).
Land and water conservation policy Variation in forest
disturbances with distances from protected areas and water
features may indicate how forests have been managed to
support multiple ecosystem services,such as biodiversity
and water quality.Providing buffer areas from the edge of
the highly protected areas or water features is critical to
maintain and improve biodiversity or water quality.The
results point to a rapid increase in forest disturbances away
from the limits of protected areas.Within 50 m,forest dis-
turbances represent 30 %of the total forest areas,increasing
up to 40 %100 maway.This disturbance rate is very close
to the mean cumulative disturbance rate in MS,indicating
that there is little buffer between protected areas and private
lands.Most disturbances are concentrated within 3 km of
water features.Although riparian buffer areas within 50 m
fromthe water have the smallest amount of disturbances,the
disturbance rate is not insignificant,ranging up to 57 %.On
average,67 % of disturbances are within a 15 km radius
from water.Riparian buffer areas experience the smallest
net forest gain during reforestation periods (forest
loss\forest gain:1990–1992 and 2002–2004),but also the
smallest net forest loss during deforestation periods (forest
loss [forest gain:1999–2001).This suggests that forest
changes in buffer areas near water bodies are relatively
small,compared to other geographical areas.However,this
result might be an artifact,due to the spatial representation
of water bodies as line objects,and riparian buffers
immediately adjacent to water bodies may include a sig-
nificant portion of the waterways.Overall,the results do not
support the hypothesis that forest disturbances and changes
vary with distance from protected areas and water features.
Plot/resource conditions (PR)
Slope The study area is characterized by very low relief.
Most slopes are less than 7 %.Approximately 68 % of
forestlands are located in the lowest relief areas,with slope
less than 2 %,and 30 % in areas with slopes of 2–5 %.
About 40 % of the forests in the low (0–2 %) slope areas
have been disturbed,in contrast to 26 % in the middle
(2–5 %) slope areas.Overall,lower slope areas have
experienced more cumulative disturbances.
Forest species and other vegetation Forest species
information is derived from FIA and MS-GAP data,which
show that pine plantations have increased since 1987
(Table 4).The correlation coefficients (r) between pine
plantation and forest area at the county level,using 1977,
1987,1994,2006 FIA data,are in the range of 0.86–0.89
(p\0.0001).This positive relationship indicates that pine
plantations increase in counties with more abundant forest
resources.Unlike artificially regenerated pines,natural
pine and pine-oaks display a decreasing trend since 1987.
The cross-correlation between artificial pines and natural
pines indicates a strong positive relationship,suggesting
that more plantations occur in areas with natural pines.
This relationship is stronger if the pine plantation is cor-
related with natural pine data in the previous period
Fig.6 Net change in forest land cover by management status.
Management status is based on the four categories of the US national
GAP standards (MS-GAP 2012).Lands with GAP status codes 1 and
2 have the highest degree of management for conservation.They are
assigned to land tracts that have permanent protection and require a
mandated management plan to maintain their natural state.Status 2
allows for less than 5 % of anthropogenic land use.Status 3 lands
support multiple uses (including forestry,mining) and status 4 lands
are either unprotected or of unknown management intent.Land tracts
in status 1–3 belong to the protected land category,and status 4 lands
are unprotected private lands
Forest dynamics in Mississippi
123
(r = 0.60–0.62,p\0.0001),suggesting replacement of
natural stands of pines with plantations.
Spatial location of pine sites provides the best expla-
nation for forest disturbance pattern among all the social
and physical factors.The cross-sectional correlation anal-
ysis of 1987,1994,and 2006 FIA data with forest distur-
bance data over 1987–1989,1999–1995,and 2005–2007
displays a strong positive relationship between pine sites
and forest disturbance (r = 0.61–0.66;p\0.0001).More
frequent forest disturbances (i.e.,‘‘multiple’’ disturbances)
are also related to pine trees,which may indicate active
silvicultural management or multiple harvesting (C2) from
pine sites.On the other hand,hardwood sites are negatively
correlated with forest disturbances,indicating fewer dis-
turbances in hardwood sites.
Locations of pine trees were identified from MS-GAP.
Forest mapping products were analyzed to estimate ages of
trees.This analysis shows that pine tree sites identified
from MS-GAP are generally associated with a high (stands
of 5–12 years in age) or a medium (stands of 12–20 years
in age) pine density.Map comparisons of MS-GAP data,
NLCD 2001,and forest mapping products indicate that the
shrub/scrub land-use category has similar spectral charac-
teristics as a high-density pine site.Therefore,we analyzed
forest disturbance and forest regain from the land class of
pine/shrub and hardwood.The results show the most dra-
matic changes (including gain/loss) from pine/shrub lands,
with the most disturbances.Forest disturbance and gain/
loss from hardwood sites remain relatively constant.
Results are summarized in Tables 5 and 6.
Regional land settlement patterns in MS clearly reflect
physical production conditions and soil conditions,as
shown on geologic soil maps.Agricultural lands are con-
centrated in the most productive soils,and the remaining
areas are largely dedicated to forests.Because we previ-
ously provided a detailed analysis of regional land,a dis-
cussion on the physical production and soil condition is
omitted.
Discussion
This study has uncovered important changes and dynamics in
secondary forests.The historical record of FIApoints out that
MS forestland areas have continued to expand over the last
two decades (Table 4).This increase occurred in the later
stages of forest transition,even as MS population increased.
This expansion has occurred throughout the state,reverting
agricultural and barren lands to forest,except in the most
productive agricultural zones.MS forests have experienced
very frequent and extensive changes,and as a result,there has
been little appreciable change in the total forest cover
(Table 2).These changes are spatially and temporally corre-
lated and systematic,even though MS forests are highly
fragmented.These short-term fluctuations have led to large
cumulative disturbances,and,as a result,40 % of the forest
cover was disturbed during the last two decades.Recent
observations indicate a long-term rotation or stabilization of
secondary forests.Although MS forests were reestablished
mainly to provide timber resources,they rarely experienced a
large-scale cuttingdue torotational timber harvesting.Rather,
frequent and coordinated harvesting has occurred on a short-
term basis and regeneration has followed within 10 years.
This has prevented large-scale disturbances and improved the
sustainability of the environment.More research based on a
Table 4 Pine plantation area by land ownership (in 1,000 km
2
) and
total forest land area (%) derived from data provided by Southern
Research Stations (1977,1987,1994) and 2006 FIA data
Years Government Industry NIPF
owners
Total
plantation
area
Total forest
area (%)
from FIA
1977 113.7 488.9 541.3 1,143.8 55.2
1987 112.7 845.5 585.9 1,544.2 56.2
1994 194.3 1,384.6 1,384.5 2,963.5 61.9
2006 – – – 4,608.5 65.3
Total forest land area (%) estimated from FIA is different from total
forest cover from satellite data (Table 2).The majority of the varia-
tion was likely due to differences in the definition of forested land,as
well as measurement scales and methodology.The total forest area is
relative to total state area.See Yeo and Huang (2012) for further
discussion
Table 5 Forest disturbance (%) by broad categories of forest species
Forest land cover 1987–1989 1990–1992 1993–1995 1996–1998 1999–2001 2002–2004 2005–2007
Pine/shrub 66.29 63.6 63.95 60.63 52.02 62.63 63.98
Pine 55.82 52.2 44.72 29.69 21 53.67 56.51
Shrub 10.47 11.4 19.23 30.94 31.02 8.96 7.47
Hardwood 15.48 17.18 16.12 15.99 17.46 19.59 23.23
Relative disturbance (%) 10 10 16 12 23 12 16
Forest disturbances (%) are calculated by land-use category,normalized by the total disturbance observed during each monitoring period
Relative disturbance (%) is the amount of biennial disturbance,normalized by the total cumulative disturbance over 1987–2007
I.-Y.Yeo,C.Huang
123
longer time span is needed to confirmthis finding.This study
is based on a mid-termtime scale (*25 years),which is too
short to capture the entire timber harvesting cycle.
A follow-on analysis was conducted to understand the
drivers for forest changes,including overall trajectory,short-
termdynamics,and spatial change pattern.The results show
that the overall trend was mainly affected by two factors:the
market value of timber products and the ecological charac-
teristics of forests.These two factors are closely related to
each other,as the ecological characteristics of secondary
forests clearly reflect the economic motivation behind forest
regeneration.Despite population growth,MS remains largely
rural,and its rural economy depends primarily onforestry and
agriculture (Fox et al.2007;Henderson et al.2008;MFC
2010).An increase in timber price and government assistance
programs to support forestry has increased plantation on pri-
vate lands (Zhang and Polyakov 2010).Pine timber products
have the highest market value and their prices have increased
at the highest rate (Fig.3).MS biophysical characteristics
(e.g.,subtropical climate,soil,and slope) allow pine trees
(such as loblolly pines) to reach maturity and become eco-
nomically marketable in 10–15 years (MFC 2010).These
factors are very favorable for pine plantation,allowing for a
rapid spread and turnover.The results show pine sites expe-
rienced the largest changes,in terms of forest gain and loss.A
time series analysis indicates a 10–15 year of disturbance,
which coincides with a rotation of loblolly pines,the most
frequently planted pine in the region.
MS forests are characterized by frequent fluctuation and
extensive changes.The change patterns were examined using
key social and biophysical factors at various spatial and
temporal scales.The selected variables reflect market,policy,
characteristics of local community,and physical condition of
plots.The results showthat most of the social and biophysical
variables remain stable,reflecting unique characteristics of
local communities and their physical conditions.However,
the time series of forest disturbances and cumulative distur-
bances do not display much geographical variations across
local communities.The rate of forest change (in terms of gain
and loss) and its spatial variation are highly variable,and its
cumulative impacts are spatially extensive.Its widespread
and rapid change patterns make an interpolation or spatial
aggregation meaningless to characterize forest changes.As a
result,most of the selected social and physical variables fail
to explain forest changes.The spatial patterns of forest
change were affected mostly by the spatial distribution of
pine plantations and age,highlighting the importance of
forestry to the local economy.In order to improve produc-
tivity and profitability,pine plantation requires site-specific,
integrated,and intensive management that optimizes
resources availability through the rotation periods (Fox et al.
2007).Our study shows that most communities are actively
engaged in intensive forest management,regardless of their
socioeconomic status,cultural characteristics,or urbanity/
rurality.Systematic and coordinated forest change maps
indicate successful silvicultural practices to sustain timber
resources,which would lead to high economic returns.
Conclusion and areas for future research
This case study provides empirical evidence supporting
‘‘forest scarcity pathway’’ of FTT (Rudel et al.2005) to
explain forest cover change,even in the later stages of forest
transition.It also emphasizes the importance of biophysical
factors (Perz 2007) and their linkage to the local economy.In
addition,this study highlights the importance of forest policy,
whichplays a critical role inmaintainingthe recoveryof forest
cover.As a result,there was systematic and coordinated
management among individual land owners,producing
unique spatial and temporal patterns of forest cover change.
As this study was designed to investigate multiple factors,it
can be expanded to undertake multivariate analyses with
selected variables.This kind of analyses may provide a
quantitative measure of the relative strength of each variable
affecting forest cover change,and can clarify any spurious
relationship.However,the complexity of the observed forest
change could not be explained by commonly used social and
physical variables.The key question is how to select the
appropriate scale of analysis,at which both forest dynamics
Table 6 Forest gain (%) by broad categories of forest species
Land use 1987–1989 1990–1992 1993–1995 1996–1998 1999–2001 2002–2004 2005–2007
Pine/shrub 34.42 71.46 105.19 -45.39 -20.78 60.5 -65.37
Pine 14.84 62.02 160.07 16.62 52.47 8.44 -60.38
Shrub 19.58 9.44 -54.88 -62.01 -73.25 52.06 -4.99
Hardwood 65.16 13.61 19.49 -15.69 -12.84 11.88 -24.01
Relative gain/loss (%) 1.1 17.6 6.2 -11.1 -14.8 19.1 1.1
The share of forest disturbance is calculated as forest disturbance by land-use category,normalized by the total forest net gain/loss during each
monitoring period
Relative gain and loss (%) is calculated by normalizing biennial forest gain/loss by the total forest gains/loss over 1987–2007
Forest dynamics in Mississippi
123
and other physical and social variables become meaningful.
For example,some geographical variables (e.g.,distance to
road/river,management status) are spatially explicit and
available at the 30-m cell level.The results showed the
importance of distance factors in assessing forest disturbance.
However,theycannot besimplyaggregatedtomatchthe scale
of other social or physical variables tobeusedinamultivariate
analysis,as this geographical information would be lost.The
social and physical variables cannot be disaggregated at the
30-m level to match the scale of other geographical factor.
Rather,this study has demonstrated the hierarchical nature of
landchanges,influencedbythe importance of forest resources
in the local economy (as assumed by the forest scarcity
pathway),the linkage to biogeophysical characteristics,and
forest policy and management in the local community.A
multivariate analysis should reflect this hierarchical structure
with better quality data that can show the characteristics of
landowners,and time series information on the spatial distri-
bution of forest species and age.These data need to be spa-
tially explicit.Detailed ground surveys and landowner
interviews would be necessary to better understand the
importance of the characteristics of land owners and to refine
the structure of the quantitative model.Such a model would
integrate three researchareas (i.e.,econometric analysis,land-
use sciences,and community planning),and quantify the
importance of the four factors and related variables over the
mid-termscale.This would improve our current understand-
ing of forest cover change,linking the short-term,cross-sec-
tional analysis to long-termtrajectory studies.
This study provides valuable lessons on management of
secondary forests and points to opportunities for enhancing
ecosystem services and ecological functioning.It shows that
MS forests have fully recovered with intensive planting of
native species (Tables 4,5,6).Such planting may eliminate
the potential risks posed by introducing fast-growing exotic
species,and reduce the uncertainty in predicting future eco-
logical impacts (Chornesky et al.2005).The productivity of
pine trees has increased dramatically to provide wood pro-
duction on a sustainable basis.This success is largely due to
cooperativeresearchandtechnologytransfer efforts across the
public and private sectors.Industry and government have
implemented a number of research projects to develop pine
plantation silviculture over the last 60 years (Fox et al.2007).
The scientific research results have been made quickly
available to the public,through incentives and assistance by
government,university,and industry (Fox et al.2007).Forest
management has been successfully coordinated across local
communities to maximize productivity and financial returns,
without large-scale deforestation.
Despite their successful restoration,secondary MS forests
offer reduced ecological benefits.Because the reforestation
effort largely stems from protecting the economic value of
forest resources,it has not incorporated key principles of
ecosystem management to improve ecosystem services and
functions.The results showclear differences in forest change
by ownership type.Protected areas with high conservation
status have experienced the least disturbances and high net
gains.These areas may provide valuable ecosystem services
over the long term,but their overall benefits are likely to be
insignificant,given their small size.In contrast,the prevalent
private forestlands hold a greater potential to improve eco-
logical functions,but theseforests havedisturbancethat donot
vary with distances to critical ecological features,such as
water bodies or protected areas.Accordingly,the US EPA
reports that a large portion of MS rivers fails to meet water
quality standards (Wear and Greis 2002).As forest resources
shouldprovide the greatest protectionfor water resources,this
finding is very surprising.This finding highlights the need to
integrate land and water conservation practices into forest
management,rather than treating them as separate issues.In
addition,this study shows that secondary forests undergo
highly dynamic short-term changes,which result in homo-
geneous forest systems with dominant species and even age
distribution.As rotation lengths have been considerably
reduced (Fox et al.2007),pine sites are more frequently
harvested and undergo intensive site preparation treatments.
They are generally associated with a high- to medium-density
pine (stands of 12–20 years in age).This management prac-
tice has important effects on above-ground carbon storage.
Careful assessment is needed to evaluate the impact of
intensive plantationat various growthstages andmanagement
scheduling.This assessment needs to consider the long- as
well as the short-termecological benefits of secondaryforests,
to better determine and coordinate the timing,intensity,and
spatial extent of forest management.As forest dynamics vary
by forest species,policies and programs should provide dif-
ferent levels of incentives and supports specific to forest
species and their main uses (e.g.,timber production,habitat),
and set management thresholds for harvesting to enhance
ecosystemservices and functioning.
Acknowledgments Thanks are due to Mr.A.Islam (Geography,
University of Maryland) for technical assistance on data processing,
and to Drs.Kasischke and Riter (Geographical Sciences,University
of Maryland),and Guldmann (City and Regional Planning,The Ohio
State University) for valuable suggestions.Comments made by three
anonymous reviewers are greatly appreciated.
References
Arano KG,Munn IA (2006) Evaluating forest management intensity:
a comparison among major forest landowner types.For Policy
Econ 9:237–248
Beach RH,Pattanayak SK,Yang J,Murray B,Abt R (2005)
Econometric studies of non-industrial private forest manage-
ment:a review and synthesis.For Policy Econ 7(3):261–281
Best C,Wayburn LA (2001) America’s private forests:status and
stewardship.Island Press,Covelo
I.-Y.Yeo,C.Huang
123
Birch TW (1996) Private forestland owners of the United States,
1994.USDA Forest Service Resource Bulletin NE-134
Chazdon RL (2008) Beyond deforestation:restoring forests and
ecosystem services on degraded lands.Science 320:1458.doi:
10.1126/science.1155365
Chornesky EA,Bartuska AM,Gregory GH,Britton KO,Cummings-
Carlson J,Davis FW,Eskow J,Gordon DR,Gottaschalk KW,
Haack RA,Hansen AJ,Mack RN,Rachel FJ,Shannon MA,
Waigner LA,Wigley TB (2005) Science priorities for reducing
the threat of invasive species to sustainable forestry.Bioscience
55(4):335–348
Defries S,Foley J,Asner G (2004) Land-use choices:balancing
human needs and ecosystem function.Front Ecol Environ
2(5):249–257
Duffy PD,Ursic SJ (1991) Land rehabilitation success in the Yazoo
basin,USA.Land Use Policy 8(3):196–205
Fickle JE (2001) Mississippi forests and forestry.University Press of
Mississippi,Jackson
Fox TR,Jokela E,Allen HE (2007) The development of pine
plantation silviculture in the southern United States.J For
105(7):337–347
Grainger A (1995) The forest transition:an alternative approach.Area
27(3):242–251
Henderson JE,Munn IA,Perez-Verdin G,Grebner DL (2008)
Forestry in Mississippi:the impact of the forest products industry
on the post-Katrina Mississippi economy—an input–output
analysis.Forest and Wildlife Research Center,Research Bulletin
FO374,Mississippi State University.31 pp.Available:www.
fwrc.msstate.edu/pubs/forestryinmississippi06.pdf
Huang C,Goward SN,Masek JG,Gao F,Vermote EF,Thomas N,
Schleewis K,Kennedy RE,Zhu Z,Eidenshink JC,Townshend
JRG (2009) Development of time series stacks of Landsat
images for reconstructing forest disturbance history.Int J Digit
Earth 2(3):195–218
Huang C,Goward SN,Masek JG,Nancy T,Zhu Z,Vogelmann JE
(2010) An automated approach for reconstructing recent forest
disturbance history using dense Landsat time series stacks.
Remote Sens Environ 114(1):183–198
Hyberg B,Holthanusen D (1989) The behavior of non-industrial
private forest landowners.Can J For Res 19:1014–1023
Joshi S,Arano KG (2009) Determinants of private forest management
decisions:a study on West Virginia NIPF landowners.For Policy
Econ 11(2):118–125
Karppinen H (1998) Values and objectives of non-industrial private
forest owners in Finland.Silva Fennica 32(1):43–59
Lambin EF (1997) Modelling and monitoring land-cover change
process in tropical regions.Prog Phys Geogr 21(3):375–393
Lambin EF,Turner BL et al (2001) The causes of land-use and land-
cover change:moving beyond the myths.Glob Environ Change
11(4):261–269
Le HD,Smith C,Herbohn J,Harrison S (2012) More than just trees:
assessing reforestation success in tropical developing countries.
J Rural Sci 28:5–19
Liverman D,Moran EF,Rindfuss RR,Stern PC (eds) (1998) People
and pixels:linking remote sensing and social science.Natl.
Acad,Washington
Mather AS (1992) The forest transition.Area 24(4):367–379
Mather AS (2007) Recent Asian forest transitions in relation to forest-
transition theory.Int For Rev 9(1):491–502
Mississippi Forestry Commission (MFC) (2010) Mississippi’s assess-
ment of forest resources and forest resource strategy.Mississippi
Forestry Commission,Jackson
Mississippi GAP Analysis Project (MS-GAP) (2012) http://www.gap.
uidaho.edu/bulletins/13/Mississippi.htm.Accessed 2012
Munn IA,Barlow SA et al (2002) Urbanization’s impact on timber
harvesting in the south central United States.J Environ Manag
64(1):65–76
Nagendra H,Southworth J (2010) Reforesting landscapes:linking
pattern and process.Springer’s Landscape Series,p 396
Perz SG (2007) Grand theory and context-specificity in the study of
forest dynamics:forest transition theory and other directions.
Prof Geogr 59(1):105–114
Perz SG,Skole DL (2003) Secondary forest expansion in the
Brazilian Amazon and the refinement of forest transition theory.
Soc Nat Resour 16:277–294
Rudel TK (2009) Tree farms:driving forces and regional patterns in
the global expansion of forest plantations.Land Use Policy
26(3):545–550
Rudel TK,Bates D et al (2002) A tropical forest transition?
Agricultural change,out-migration,and secondary forests in
the Ecuadorian Amazon.Ann Assoc Am Geogr 92(1):87–102
Rudel TK,Coomes OT,Moran E,Achard F,Angelsen A,Xu J,Lambin
E (2005) Forest transitions:towards a global understanding of land
use change.Glob Environ Change Part A 15(1):23–31
Schaaf KA,Broussard SR (2006) Private forest policy tools:a
national survey exploring the American public’s perceptions and
support.For Policy Econ 9(4):316–334
Schaaf KA,Ross-Davis AL,Broussard S (2006) Exploring the
dimensionality and social bases of the public’s timber harvesting
attitudes.Landsc Urban Plan 78(1–2):135–146
Smith WB (2002) Forest inventory and analysis:a national inventory
and monitoring program.Environ Pollut 116(Supplement 1):
S233–S242
Sun X,Sun C,Munn IA,Hussain A (2009) Knowledge of three
regeneration programs and application behavior among Missis-
sippi nonindustrial private forest landowners:a two-step sample
selection approach.J For Econ 15(3):187–204
USDA Census of Agriculture-National Agricultural Statistic Service
(USDA-NASS) (2010) http://www.agcensus.usda.gov.Accessed
2010
USDA Economic research Service (2010) http://www.ers.usda.gov/.
Accessed 2010
USDA Forest Service (USFS) (2010) National Forests in Mississippi,
http://www.fs.fed.us/r8/mississippi/homochitto/.Accessed 2010
Wear DN,Greis JG (2002) Southern forest resource assessment—
summary report.Gen.Tech.Rep.Asheville,NC,US Department
of Agriculture,Forest Service,Southern Research Station,
103 pp
Yeo I,Huang C (2012) Revisiting the forest transition theory with
historical records and geospatial data:a case study from
Mississippi (USA).Land Use Policy (under review)
Zhang D,Mehmood SR (2001) Predicting nonindustrial private forest
landowner’s choices of a forester for harvesting and tree planting
assistance in Alabama.South J Appl For 25(3):101–107
Zhang D,Polyakov M (2010) The geographical distribution of
plantation forests and land resources potentially available for
pine plantations in the US South.Biomass Energy 34(12):1643–
1654.doi:10.1016/j.biombioe.2010.05.006
Forest dynamics in Mississippi
123