MULLER Daniel and MUNROE Francis - Changing Rural Landscapes in Albania - 2008

Changing Rural Landscapes in Albania: Cropland Abandonment and Forest Clearing in the
Postsocialist Transition
Author(s): Daniel Müller and Darla K. Munroe
Source: Annals of the Association of American Geographers, Vol. 98, No. 4 (Dec., 2008), pp. 855-
876
Published by: Taylor & Francis, Ltd. on behalf of the Association of American Geographers
Stable URL: http://www.jstor.org/stable/25515164
Accessed: 07-03-2015 08:46 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.
Taylor & Francis, Ltd. and Association of American Geographers are collaborating with JSTOR to digitize, preserve and
extend access to Annals of the Association of American Geographers.
http://www.jstor.org
This content downloaded from 128.235.251.160 on Sat, 07 Mar 2015 08:46:29 UTC
All use subject to JSTOR Terms and Conditions
Changing Rural Landscapes in Albania: Cropland
Abandonment and Forest Clearing in the
Postsocialist Transition
Daniel Muller* and Darla K. Munroe^
"Leibniz InstituteofAgriculturalDevelopment inCentral and Eastern Europe (IAMO), German}', and Geomatics Lab,
Geography Department, Humboldt-Universitdt zu Berlin
^DepartmentofGeography, The Ohio State University
Postsocialist transitions in eastern Europe have focused on the establishment of private property rights,often to
the exclusion of other aspects of rural land-use systems.After the demise of socialism in 1991,Albania privatized
virtually all agricultural land by redistributing formerlycollective land on an equal per capita basis. This article
examines the dramatic effectsof the transition in general and the land reform in particular on Albania's rural
landscapes. A village-level survey was conducted to analyze household resources and constraints. The survey was
integratedwith data derived from satellite image interpretation and geographic information systems to develop
statistical models of two key land-cover changes of interest: the abandonment of cropland and forest-cover
loss. Statistical corrections were to control for cross-scale interactions and was
implemented geocomputation
employed to assess the goodness of fitof the models and the robustness of the results. Findings indicate that
most cropland abandonment at the onset of the transition period was concentrated inmarginal, less densely
areas. More was returns
populated recently, abandonment increasingly shaped by economic from cultivation
and the growing competition with nonfarm livelihood strategies. Land fragmentation, an outcome ofAlbania's
land reform, was associated with greater abandonment in the later stages of transition. Patterns of forest clearing
were subsistence driven and found around populated areas in the period immediately following the collapse of
socialism, whereas more commercial patterns over time. Because of the range of social and
clearing emerged
environmental impacts of large-scale shifts, studies of postsocialist necessitate a multiscale
policy landscapes
spatial framework.Key Words: Albania, integratedanalysis, land-use change, postsocialist, spatialmodel.
*mjsft&3iXtommn:?%GttJb:utom3:, mmmi?mmwumm^mo #tt??
MMftTo &wiXmmm%mx%Mi?M&%MTnfc^&mmmmm.XBm? mm
M.%mffi&%^&m^?mmm?g3Likto%LvrmMo mmmît^mmm^îm^Rmnm.
La transici6n al postsocialismo en el este de Europa se ha concentrado en el establecimiento del derecho a la

propiedad privada, y con frecuencia excluye otros aspectos de los sistemas de uso de tierras rurales.Despu^s de la
desaparicidn del socialismo en 1991, Albania privatizo virtualmente todas las tierras agricolas redistribuyendo
equitativamente per capita las anteriormente tierras colectivas. En este articulo se examinan los efectos
dramaticos de la transicidn en general, y la reformade la tierraen particular, en las tierras rurales de Albania. Se
realiz6 una encuesta a nivel de aldea con el objeto de analizar los recursos y las restricciones de las familias. La
encuesta se integro con datos derivados de sistemas de informaci6n e de
geografica interpretaci6n imagenes por
sat^lite con el finde desarrollar modelos estadisticos de dos cambios claves de internsen la cobertura del suelo: el
abandono de las tierras agricolas y la p^rdida de la cubierta forestal. Se implementaron correcciones estadisticas
Annals of theAssociation ofAmericanGeographers,98(4) 2008, pp. 855-876 ? 2008 byAssociation ofAmerican Geographers
Initial submission,August 2006; revised submissions,May andNovember 2007; final acceptance, December 2007
Published byTaylor ca Francis,LLC.
856 Miiller andMunroe
para controlar las interacciones interescalares y se us6 geocomputaci6n para evaluar la calidad del ajuste de los
modelos y la solidez de los resultados. Los resultados indican que la mayor parte del abandono de las tierras
agricolas al inicio del periodo de transicidn se concentr6 en areas marginales con poblaciones de menor densidad.
M&s recientemente, el abandono fue conformado cada vez m&s por el rendimiento economico del cultivo y la
creciente competencia con estrategias de sustento no La de la tierra, un resultado de la
agricolas. fragmentaci6n
reforma de tierrasde Albania, se asoci6 con un mayor abandono en las etapas posteriores de la transicidn. Los
patrones de desmonte forestal fueron impulsados por la subsistencia y se encontraron alrededor de areas pobladas
en el periodo inmediatamente posterior al colapso del socialismo, mientras que los patrones de desmonte mis
comerciales emergieron al transcurso del tiempo. Debido a la variedad de los impactos ambientales y sociales
por los cambios a gran escala de las polfticas, los estudios de las tierras necesitan un
provocados postsocialistas
marco espacial multiescalar. Palabras clave: Albania, analisis integrado,cambio del uso de la tierra,postsocialismo,
modelo espacial.
It has been more than fifteen years since the fall of persistent secondary consequences of the reform (e.g.,
the Berlin Wall marked the end of Soviet
hege land fragmentation).
mony and eastern Europe
in central (CEE) and Thepostsocialist transition and the accompany
of At the time land reform are inherently complex, multiscale
processes postsocialist transition1 began. ing
there was pervasive euphoria that the West had tri processes. Because of the heterogeneous social and envi
umphed over socialism.2 A second assertion accom ronmental impacts of large-scale policy shifts, studies of
panying this geopolitical shift was thatWestern-style postsocialist landscapes necessitate a spatially explicit,
institutions were (and had proven to be) the most ef multiscale framework. The land reform inAlbania was
fective. The countries ofCEE largely implemented sim constructed at the national
level. The decision of the
ilarmacrolevel policy changes following the collapse of Albanian government to redistribute the land on a per
socialism. Most countries discontinued the "command capita basis was framed by political and economic con
and control" role of a centralized government in favor siderations (Cungu and Swinnen 1999). The reformwas
of privatization, price and trade liberalization, and cur in turn implemented at lower levels of administration,
rency devaluation (Hamilton 1999). The establishment mediated by district-level investment and policy de
of private property rights in industry and agriculture was cisions, and finally implemented at the village level by
a key part of the privatization of economic assets, given village commissions (Stahl 2007). These policies touch
the belief in the central role of private property in the down in a spatially heterogeneous landscape, shaped
fabric of society (Blacksell and Born 2002). Large-scale by variations in agroclimatic suitability, biogeophysical
policy shifts led to fundamental reorganizations in eco conditions, market accessibility, and rural development
nomic practice, with myriad implications for land use levels, where land-use adjustments unfold over time.
(Lambin et al. 2001). InAlbania, the political and eco In this article, we present a multiscale, spatial
nomic changes of the postsocialist transition resulted in temporal framework to examine the local outcomes of
significant land-use and land-cover changes with asso the postsocialist transition on land cover in southeast
ciated environmental consequences. ern Albania between 1988 and 2003. Our empirical
Establishing private property rights for former agri focus is the abandonment of cropland and forest clear
cultural workers was themain objective of theAlbanian ing.Within these fifteen years, our study area lost a
land reform, although the policies were also intended to total of 258 km2 of cropland, equivalent to a 28 percent
reduction of the cropland from 1988. Cropland aban
realign land-use incentives to new market conditions.
its donment is arguably an important indicator of changing
With significant external support, Albania mapped
agricultural and residential land and set up
a unified, rural livelihood strategies in Albania. New practices
for the of real es effected by the transition can range from family farm
parcel-based system legal registration
tate. However, the registration system did not stimulate enterprises (Swinnen, Vranken, and Stanley 2006),
land transfers as expected, as the economic and social often supported by foreign remittances (Nicholson
incentives to sell or buy were lacking (Bloch 1998). 2004; Miluka et al. 2007); the abandonment of remote
areas concomitant with massive internal and external
Rural landscapes exhibit the effects of the macroeco
nomic shock of the transition, followed by short-term migration; and, finally, entrenchment and reversion to
reactions as well as more a gradual adaptation of local subsistence in the most remote areas (Miiller and Sikor
land-use practices to new market conditions, and the 2006). Total forest cover remained stable between
Changing Rural Landscapes inAlbania 857
1988 and 2003. However, 5 percent of the natural, periods: the initial period following the transition until
old-growth forest of 1988 was cleared by 2003, and both an economic shock in 1996 and a second period later
old-growth and secondary forest regenerated on former in the transition after 1996 until 2003. Data sources in
bush and grassland, as well as on 94 km2 of abandoned clude time series of satellite images, a large-scale survey
cropland. Thirty-nine percent of the landscape was implemented at the village level in 2004, and spatially
forestedin 1988; 202 km2 or roughly 13 percent of referenced biogeophysical data. The analysis in this ar
the land that was forest in 1988 was cleared by 2003. ticle follows in the tradition of statistical LUCC models
Substantial changes in forest integrity stem from the (Veldkamp and Lambin2001), drawing from prior work
continuing reliance on forest biomass for heating and to connect land-cover changes to hypothe
observed
cooking. In addition, forests and other communal sized land-use drivers via a statistical framework based
resources can suffergreatly when collective systems are on the notion of spatially varying land rents (Chomitz
replaced with market-based ones due to new collective and Gray 1996; Nelson and Hellerstein 1997).We
action problems that arise in transitional periods present several methodological improvements on prior
(Sikor 2004). work. We account for temporal dependence by consid
Examining a landscape before and after a major tran ering the state of land cover at the beginning of each
sition is useful in situations where an abrupt change period, excluding all locations that have no chance to
leads tomultiple adaptations, resulting in an evolution change (e.g., nonforested pixels cannot be deforested).
of rural land-use practices. The loss of cropland and for Village-level socioeconomic data were linked to pixel
est cover inAlbania indicates that the transition and level environmental data based on village boundaries.
the associated macroeconomic recession led to dramatic Contrary tomost other approaches, we use a robust esti
changes in the landscape. Declining revenues from agri mation technique that controls for intracluster correla
culture were accompanied by new livelihood strategies tions induced by integrating social and environmental
in a rapidly globalizing world. This sparked demographic data across scales, following Miiller and Munroe (2005)
changes and pushed people into other sectors of the and Gellrich et al. (2007). In that way, we account for
economy, leaving large tracts of cropland idle. The re the fact that multiple remote-sensing pixels fall within
sulting land-use changes are associated with interesting a common socioeconomic sphere of influence (in this
patterns such as the persistence of land fragmentation case, the village). In addition, unobserved effects at the
and the realignment of agricultural production as, over district level, varying policy implementations, or dif
time, predominantly more remote and less productive fering investment decisions are captured using dummy
areas fall out of cultivation. Forests have been threat variables. Spatial autocorrelation is controlled forwith
ened by the transition because a new delineation of a spatially stratified sampling approach based on dis
rights to forest uses remains to be implemented. In ad tance between non-neighboring locations. The sensi
dition, many households still rely on forest resources. tivity and robustness of the model fitwas assessed across
One major challenge of land-use and land-cover varying sampling distances.
change (LUCC) analysis relates to the necessary inte
gration of disciplinary perspectives, data, and methods
Changes in the Albanian Landscape
(Rindfuss et al. 2004). Land use occurs in a local bio
geophysical context and is shaped by social, political,
A Brief History of the Postsocialist
economic, and institutional processes at levels rang Albanian Economy
ing from the household to the globe (Moran, Ostrom,
and Randolph 2002); local land-use changes reflect the Before the transition, Albania was primarily an agri
influence of these processes, whether they are directly cultural country; more than one-third of gross domestic
observable in the region or not. Assessment of the rel product (GDP) came from agricultural production dur
ative effect of each of these processes demonstrates the ingthe 1990s (WorldBank 2006b), andmore thanhalf
complexity of LUCC (Leemans and Serneels 2004). of the population was employed in agriculture (Food
Mertens and Lambin (2000) underscore that land-cover andAgricultureOrganization [FAO] 2006). The share
are dynamic, often multidirectional; a spatially of agriculture inGDP
changes initially increased to 56 percent
and temporally explicit approach is required to link in 1995 and then gradually dropped to 25 percent in
land-cover change to land-use influences. 2004. The industrial sector contracted from 48 percent
We examine the changes that happened due to the in 1991 to 20 percent in 2005, as the share of the ser
transition and the reform by focusing on two key time vice sector gradually increased to 55 percent, most of
it being concentrated inTirana and the coastal region by variations in distance to the farmstead, soil fertility,
(WorldBank 2006b). and irrigation capacity. Village-level land distribution
The immediate consequence of the transition was a councils were formed to allocate plots, often in distant
sharp economic recession in 1991, followed by a slow locations within the village territory, to each farm fam
recovery. In 1996 there was a proliferation of bogus ily proportional to their household size, including the
investment opportunities with alleged high rates of re elderlyand smallchildren (StanfieldandKukeli 1995).
turn offered by several investment firms.These pyramid This egalitarian land reform resulted in the extreme
investment schemes attracted deposits up to roughly fragmentation of agricultural holdings and represented
half the size of Albania's GDP, finally collapsing in the starkest break with the collectivist past within the
early 1997, leading to a second sharp decline in GDP CEE region (Mathijs and Swinnen 1998).3
followed by a second recovery (Korovilas 2005). The The land reformestablished a legal basis for land
collapse of the pyramid schemes triggered a period of transfers and the sale of land was formally allowed in
anarchy from January to June 1997 inwhich the police 1995. A digital real estate registration system was de
and army were dissolved and 2,000 people were killed veloped with considerable international support. How
(Jarvis 2000). The anarchy led to the resignation of ever, land market activity has been limited and land
the government, a collapse of state revenues, and the owners were reluctant to exchange (Bloch 1998; Swin
depreciation of the national currency. nen, Vranken, and Stanley 2006; World Bank 2006a).
Migration outflows were sizable during the entire pe Swinnen, Vranken, and Stanley (2006) found that only
riod of transition, reaching peaks in the early period of 5 percent of households were reported to have sold land
transition at the beginning of the 1990s and after the fall and the share of rented land is around 10 percent in
of the pyramid investment schemes and the subsequent all of Albania, a
but only minority of the transactions
insecure period (Carletto et al. 2006). The Ministry of was registered in the land administration system. The
Labor and Social Affairs puts the number ofAlbanians aversion to register may be partially explained by the
living abroad at more than 700,000 out of a total of 3.1 corrupt practices of the land administration agency (Im
million in 1999 (Barjaba 2000). In 2004, thenumber movable PropertyRegistrationOffices [IPRO]; Stahl,
rose to 1 million, according to the Ministry of Labor Sikor, and Dorondel forthcoming). Precommunist own
and Social Affairs (cited inKing 2005). At present, in ers have continued unsuccessfully to demand state com
ternational migration rates inAlbania are among the pensation for the loss of their historical land claims,
highest in Europe, as are poverty rates (Zezza, Carletto, which has led to a lack of title security, impeding the
andDavis 2005). development of amore active land market (World Bank
Land Reform inAlbania 2006a).
Prior toWorld War II, the distribution of land in The of Land Use and Land
Fragmentation
Albania was unequal. The most productive agricultural
Ownership
areas were owned by a few families. Under socialism,
Albania was the only country in CEE that effectively The complete breakup of the agricultural collec
nationalized all land, based on its 1976 Constitution tives in Albania led to fragmentation of land owner
(Lerman 2001). After the demise of socialism, Alba ship (Bloch 1998;Cungu and Swinnen 1999). In 2000,
nia implemented a comprehensive land redistribution 440,000 farm families operated on approximately 1.8
program that established private property on virtually million parcels. An average farm household possessed
all cropland under use during the socialist period. Land 1.5 ha, spread over three to five parcels (Ministry of
redistribution was a politically feasible strategy; restitu Agriculture and Food 2002). The inactive land mar
tion to former owners would have resulted in less than ket with few land sales and rentals hinders land con
5 percent of the population owning the most produc solidation. Therefore, the fragmentation of both land
tive land (Swinnen 1999). The Albanian land reform ownership and land use resulting from the chosen land
of 1991, the Land Law 7501, was intended to redis reform strategy, the Land Law of 1991, largely persists
tribute all collectivized land to formermembers of the today.4
cooperatives on an equal per capita basis. Other rural Figure 1 depicts a "before and after" image of the land
residents who were not members of the cooperatives reform in a rural area in Albania. It is clear that the
were also awarded land but in smaller quantities (Law redistribution of collective agricultural land (Figure 1,
7501, Art. 6). Land to be redistributed was stratified left, shows the collective fields in 1988) greatly
From
to
Hj^^M^^^^^^^B^^^H^^^^^^Hl I^^^^^B^^^^^^^^^^^^^^^^^^B
Theimage
redistribution.^B|M^^^^^|^|^^^^^^^^^HHI ^^^^^^^^^^^^^^^^^^^^^^^^^H
a Thematic
Mapper!^^Kp9^^^^^^^V^^^^^^H^^^| ^^^^^^^^^^^^^^^^^^^^^^H^^H
with
(TM) !jP^*
a|'^^^^^^^^^^^| ^^^^^^^^^^^^^^^^^^^^^^^^^H
an
the Space-
iKA^^^H^HiJ^^^^^^^^^^^^H
^^^^^^^^^^^^^^^^^^^^^^^^H
borne
Thermal ^^Ê?^^^^^^^^^^^^^P^^^^H^1
^^^^^^^^^^^^^^^^^^^^^^^ÎB
Radiometer with
(ASTER) 15-m
ground j^^V^ F^^^^^^^^^^Hfl^^RI^H ^^^^K^^^^^^^^^^^^^^^^B^BM
w ^ ?'?'?|?
vê ^^B ^f% i^^H HBBhI^^^^^^^HI'? ?'?'?'
I MÊIm^Hlf^ #* ,j?
I %^ ilMBi EHHHHbHHHHHo_Kilometers_
increased the number and reduced the size of agricul household's residence (the total one-way distance to all
tural plots (Figure 1, right, illustrates the redistributed plots is almost 6 km).
land as of 2003). Figure 2 further demonstrates the spa
tial distribution of plots for four households within one
Reform of Forest Ownership
village (Figure 2, left). The actual Euclidean distances
between these plots are depicted on the right side of Fig Land privatization in Albania initially focused on
ure 2 for one household that possesses seven dispersed
agricultural land and only recently extended into other
that from 150 to 3,000 m2 in size (a total
plots range parts of the landscape, such as forest. At the time of
of 5,650 m2). These plots, which vary in steepness and collectivization in 1944, only 6.4 percent of forested
soil quality, are located at variable distances from the land inAlbania was private property; 92 percent and
^^^^^ h^^HII^HbI^^^^^^^^^H 2
Household \
H^HKlifl^HBilHi^l -^
J^B^^^H^^^^^^|^^^^^^^B^^H^^^@^1^^^M|9h 2.000""Yji"hi iV??.,

1,25? --?^".::;';^ \^
area^^^^^^^W^^^^mH
l^PBuilt-up
9?1
p/ofs Bj^^H^^^^^^^?y^^^^B . "T;""'*"-----^^^^:
II 1^^^^BÊhH|^^8H^h^h
Household ^^^^^H^^^Hj0Hj|^H^^HH^^^^Î
2 0
Household 0.25 0.5 Jj Plots(square .Distance
meters) (meters)# Residence
di {1H|^H
I-1-1-1-1
"~<>c^MH||^^^H
_ Household3
Kilometers_L^SEKpHIH
Figure 2. Farmland fragmentation in an Albanian village. Source: Authors, derived from participatory mapping facilitated by Johannes Stahl
and the authors using a panchromatic IKONOS image from 21 September 2003, with a ground resolution of 1m. The is located next
village
to a lake (dark area on the the plot locations on an overlaid
right). The households delineated transparency and reported the plot areas. Three
households are shown on the left. Straight-line distances from the residence to all plots of household 2 (right part of figure) are calculated in
a information system. Both maps have an identical scale.
geographic
860 Miiller and Munroe
1.6 percent were under state and communal ownership, tion for the years 1988 (Landsat Thematic Mapper
respectively(De Waal 2004). After the collapse of [TM]), 1996 (TM), and 2003 (TM and theTerra Ad
socialism, Albania's private forests were officially resti vanced Spaceborne Thermal Emission and Reflection
tuted to precollectivization owners and their heirs fol Radiometer [ASTER]).
lowingthe 1993LawNo. 7699 (Ministry
ofAgriculture On-screen interpretation was conducted in con
and Food 2002). In practice, forests largely remained junction with an Albanian forestry expert on a scale
in state ownership, and as of 2001 only 10,000 ha, or not larger than 1:40,000. Six spectral bands (at 30-m
1 percent, were returned to presocialist owners out of a resolution), excluding the thermal band, of the Landsat
total 1million ha of forestland (Ministry ofAgriculture TM images, and all three 15-m bands in the visible and
and Food 2002). Meanwhile, use rights to 40 percent near-infrared spectral range from the ASTER images
of the state-owned forests are in the process of being were used. Interpretation resulted in one land-cover
transferred to commune administrations (World Bank map for each of the respective years. Land-cover classes
2006a).5 Countrywide, forest thinning and loss of were defined following the Land Cover Classifica
forestland are among the most profound changes in tion System (LCCS) used in theAlbania National
land cover (Jansen et al. 2006). Forest health is a con Forestry Inventory (ANFI; Jansen et al. 2006). Before
cern for local livelihoods that strongly depend on forest and during interpretation in 2004 more than 300
resources in the absence of other energy sources. From a reference points on the ground were compared to
global perspective, the effects of forest degradation and the estimated land-cover classes and cross-checked
forest-cover loss on biodiversity may be significant, as with results from the ANFI. Using the year 2003 as a
Albania is located within theMediterranean Basin that base, interpretation of the historical images was only
is recognized as a global biodiversity hotspot in terms conducted for areas that had visibly changed to avoid
of endemic flora and fauna species (Myers et al. 2000). coregistration errors. Overlays of two land-cover maps
produced land-cover change maps that indicate the
extent and location of changes between 1988 and
Study Area
1996 (periodone) as well as between 1996 and 2003
Thisresearch focuses on an area of considerable eco abandonment was defined as
(period two). Cropland
nomic and agroecological diversity. The study area com those pixels with land cover that was categorized as
prises the four districts Elbasan, Gramsh, Librazhd, and cropland in the first year but not in the subsequent one.
Pogradec in the southeastern portion ofAlbania (Figure Cropland is distinct from pasture or grazing land6 and
3), covering a total land area of 3,800 km2. In the west consists of all land under annual and perennial crop
and east, flat, drier plains used intensively for crop pro orchards of fruits and olives.
production, including
duction are to be found, whereas the north, south, and Given that the primary agricultural land use in the
central parts of the region are humid and hilly uplands. socialist period was large collective farms producing
The fast-growing city of Elbasan, the capital of the grain crops, this category is a salient element of
district of Elbasan, is the third largest city in Alba the postsocialist transition. Second, forest-cover loss
nia with a population of 100,000. The district capital of was measured those that were forested in
by pixels
Pogradec in the east is another major market center that the first but not the subsequent year. The employed
attracts migrants from proximate areas. Lake Ohrid, definition for forest includes all forest and woodland
one of Europe's oldest lakes, is also found in Pogradec,
categories and does not distinguish between closed and
and is a major tourist destination. In contrast, the
open forest or between broadleaved, coniferous, and
less accessible mountains in Librazhd district contain mixed forest.All other land-cover categories including
poorer areas. Poverty estimates for 2001 place 37 per shrubland, coppice, or grassland that may be used
cent in Librazhd, 35 percent inGramsh, 31 percent in as pasture were excluded from the analysis. Changes
Pogradec, and 29 percent in Elbasan below the national within any of these categories as well as an expansion
poverty line ofUS$411 per capita per year (World Bank of cropland or the regeneration of forestland were not
2003). considered.
The spatial distribution of land-cover change in
Land-Cover Changes: 1988-2003
the study area reflects how the period of transition has
Changes in land cover were measured using time se reshaped the landscape (Figure 4). Net abandonment
ries of remotely sensed satellite data. Visual on-screen amounted to 339 km2 between 1988 and 1996 and
interpretation was used to derive land-cover informa 155 km2 between 1996 and 2003, which is a 36 percent
* / ? jj -^#^t( ^Jy S^^J^\ Kosovo

^M< *\ 1* ^VJ &J&/ ff \ (UNadministered)
/ / \ \C^ î^^y^4 f <^S ^ afea

StUdy
.^^/X^^ District
border
7 / j?"^ fl \ \ ^^s?*y\ r < w ^TnT^^x /^^ \"*" c-?
\ '^lJIH^^^^^^^^^^^^^^^^B National
yM'X....-road J YP^n^T^^^ f\
i i - '-'-1* \ ? ' I w E
i??-:?'?'?i ^^^B|^^^^^^^B {^1 P* y
Figure 3. Study area.
decrease of cropland from 1988 and a 21 percent forest in 1996, 13 percent (284 km2) were nonforest
decrease from 1996, respectively. Significant cropland by 2003 (Figure4). As much of forestbiodiversityis
abandonment was visible in Librazhd and Pogradec in likely associated with long-standing forest, we only
the firstperiod (Figure 4). Total forest cover (i.e., stable consider pixels that were cleared and ignore areas that
forest less forest clearing plus regrowth) remained regenerated during the fifteen years as these mainly
stable. However, there were significant spatial changes consist of young, secondary forest. Forest-cover loss was
within the forest category. Between 1988 and 1996, concentrated in Elbasan in the first and inGramsh and
forest cover was cleared on 16 percent of the forested Pogradec district in the second period, predominantly
pixels in 1988 (238 km2). Of the pixels covered by close to the district capitals (Figure 4).
862 Muller and Munroe
profound, and there are a variety of social and environ

mental consequences of such changes.
aor**
1988 - 1996 y^-^tW
; - Cropland abandonment:
>.
f' 339 sqkm We base our methodology in this article on a frame
^
work that has been used in other studies of LUCC (first
/<^-vv^^S^ ^ "Tli Forest
clearing:
/^%-i 1 \ 238 sqkm developed by Chomitz and Gray 1996; see also Nelson
^Vi?
and Hellerstein 1997; Mertens and Lambin 2000;
Munroe, Southworth, and Tucker 2002). Assuming
that the unit of analysis is the decision-maker (in
this case, an average household within a village), the
choice of land use to the potential utility or
relates
profit (land rent) that the decision-maker can expect to
receive from a particular land use compared to a set of
alternatives. Thegeographic, economic, demographic,
and social
contexts?spatial and aspatial?influence
the returns that a farm household can expect to
receive from this land use. Spatial factors include
the biogeophysical and
agroclimatic suitability of a
/^-/^ Cropland abandonment: particular parcel of land and cost of access to local,
1996-2003 *"
V 155sqkm
^d?*/ regional, and national markets. Road existence, road
^s/3?^!* ^ * 1 Forestclearing: quality, and market infrastructure are important factors
JLy^^T ^^ \ 284sqkm affecting land rent because transport costs to input

and output markets are a substantial portion of total
production costs. Other important influences include
farmer characteristics (age, education, household size,
income) and technical inputs (machinery, irrigation).
The Connection Between LUCC and Postsocialist

Transitions
Much of the interest in LUCC has focused on the

0 5 10 jp ?-<'V? and destruction of tropical forests and the
?^<jjptj* degradation
corresponding impacts on the carbon cycle and biodi
Legend Land-cover changes Inmodel
versity levels. A substantial literature exists discussing
Districtborder Croplandabandonment
dj 9& institutional arrangements and natural resource gover
Districtcapital Forestclearing
^ National road
f?
nance, challenging conventional notions that commu
Q p^_ nal forms of resource management (such as common
4. abandonment and forest Source:

property systems) lead to resource degradation (Os
Figure Cropland clearing.
trom 1990; Agrawal 2001; Lambin, Geist, and Lepers
Adapted from
Muller and Sikor (2006).
2003). Despite this large body of empirical, theoretical,
and analytical work addressing property rights, institu
Land-Cover and Land-Use Change tional reform, and land-cover change, there have been
in Transition surprisingly few integrated analyses of postsocialist cir

cumstances. Land reform is a major intervention that
Over the past couple of decades there has been sig redirects land-use incentives. Kaimowitz and Angelsen
nificant study and synthesis of the socioeconomic and (1998) and Lambin et al. (2001) point out thatperiods
political influences underlying land-use change that in of transition in such institutional contexts can exhibit
form studies of land-cover change (Lambin et al 2001). rapid, even drastic change of land-use strategies, par
Such influences are particularly relevant in the tran ticularly in times of risk and uncertainty. Analyses of
sition from command to market-oriented economies postsocialist landscapes can also provide insights re
(Global Land Project 2005). The scope and magni garding the multiscale effects of transition on land use
tude of rural changes in CEE since the transition are (Stahl 2007).
There is an established tradition within the LUCC Conceptual Framework

literature to test statistically for the influence of land
tenure on land-use incentives. Theoretically, land re Figure 5 summarizes our multiscale framework for sta
lations influence land-use change in a variety of ways. tistical modeling of cropland abandonment and forest
a
clearing following major economic and political trans
First, particular sets of land-use policies may allow cer
tain land uses and forbid others, as in the case of pro formation like the one
in Albania. Local land-cover
change happens in a spatially heterogeneous environ
tected areas (Cropper, Puri, and Griffiths 2001; Nelson,
ment and is embedded within broader social, political,
Harris, and Stone 2001; Miiller and Munroe 2005).
of tenure and economic contexts that range from the household
Second, insecurity land may decrease the in
to the globe. As one increases in scale (y-axis), the num
vestments that a farm household
iswilling tomake via
the discount rate or the length of time the user iswill ber of observations (N) on the x-axis decreases. At the
national and global level, such factors as markets, pre
ing to wait for returns to investment in land uses. Other
work has examined how land reform has shaped land vailing technology and infrastructure, and other poli
use change by identifying communal-level cies provide the external context that shapes land users'
influences
opportunities and constraints. These factors are framed
(Bray et al. 2004), variations in land-cover trajectories
across differing land reform programs (McCusker 2004), by the specific circumstances of the transition, which
or the influences of conservation programs in a context are associated with market liberalization, asset priva
of rapid infrastructural development et al. tization, and land reforms. It is important to illustrate
(Soares-Filho
that land reform is also a multiscale process. The legal
2004).
Scale f /-"" ^X
/ Postsocialist\ Global
transition\
.^^^^^^^msmmmmm^^^^^^
^HHHI^^Ki?8Hlii?iial^^^^HHl
^?-?^ / I developments
-liberalization
\ -
. x^\^^^^
^\ V privatization
J
.^^^
"Policies
.^-.-/-.y.T^W^^L^ ?.^^ V^^^^01^^^ J
/'''Land-use -^ '
decision-makingN ///f///M///M^^ \ ^X
V^Jland-use
chojce)^^ Ww^^^M^^^^ Vl,,age
"x\ \
'
\mmmm.H 1 Accessible
l Vv\ /\ **/ ^
^^^
||tM "i househo,ds
$v \i \
-n^^^Sm-^
~^mtm&*m^
lliil^Miiill " "" ?w^^*?fWIÂ^'T^/)/ v> \
w
^^^^^^^^^^ J- -^Xlli V<\ \
_^_ i j-1 I _^_"HdH?I _JB1_%^//WA' ^^mx\ \\
I- j-1_ i \
Forest
Cropland uiner "
abandonmentQh clearing _M__M_ W/MWk\r^ i
| | | ~jm^r~ WS0/sk "''~ ?T\
^
III? N
*
_ZIZZZZZ^^ Outcomes reaiized at various scales
"
variable I
I I Dependent ,
1 ^ , , 1 1 , . ' , . ' .
H^H /D^.,nd"ded'nmodje,x I LivestockI I Crop I Cultural I I Bio IC B,omass

(shading denotes scale) production
r-, '-' '-' '-' l-' '-' production landscape diversity
I_I Implicit(unobserved) process
Figure 5. Framework for a multiscale, spatially explicit analysis of cropland abandonment and forest clearing inAlbania. Source: Authors;
inspiredinpartbyKaimowitz andAngelsen (1998).
864 Muller and Munroe
framework is supported, designed, operationalized, and as a proxy for land managers1 decision units in statis
mediated at different levels from the international to tical LUCC models. In pixel-level statistical analyses,
the village, with outcomes also occurring at the plot the spatial resolution of the satellite data governs the
and pixel level. choice of the highest possible spatial resolution of the
District effects represent particular influences on land-cover data. Chomitz and Gray (1996) motivate
land use that originate from the district level, such observed pixel-level land cover as a function of unob
as unmeasured
biogeophysical variations (district served land-use decision-making occurring at the plot
boundaries may correspond to water catchments or soil level. To the degree that each pixel ismanaged accord
variations), differing governance strategies (district ing to its optimal use, this assumption is appropriate,
governments have their own budgets and can decide on but unobserved heterogeneity occurs across households
local taxes and fees; Hoxha 2001), and other proxies, and within plots.
such as the regional economy or social context. At the pixel level, factors describing both the bio
Below the district level are nucleated villages in our geophysical environment and local spatial accessibility
framework that are surrounded by cropland and pastures (to road infrastructure, markets, and dwellings [popu
(Muller and Sikor 2006). The village level is an impor lated places]) can be successfully integrated. Geophys
tant functional unit inAlbania and village land councils ical and agroclimatic suitability determine the natu
implemented the land reform. Socioeconomic data for ral potential inherent in a specific plot of land and
demography, education, agricultural technology, live include rainfall, soil suitability, elevation, and slope.
stock, forest utilization, and the impact of the land re Rainfall was expected to be negatively associated with
formwere measured at this level. Higher household den cropland abandonment water availability
because is a
sitywithin a village increases the demand for cropland constraining factor for crop production in the study
and for forest products. More reliance on the transfer of area. Rainfall is likely to be positively associated with
remittances, as an important contributor to rural liveli forest-cover loss because there ismore rainfall at higher
hoods, reduces the propensity of maintaining cropland altitudes where more forest tends to be located. More
under production and of the use of forest products. The abandonment and higher forest loss were expected at
influence of educational attainment may be unclear a steeper slopes and higher elevations. We further hy
priori, but one could expect that families with higher pothesized less abandonment on better soils. With re
education may be better equipped to adapt to market gard to accessibility, cropland abandonment is more
conditions and expand as well as intensify agricultural likely in remote places at the onset of the transition,
and more abandonment was expected closer to pop
production. On the other hand, higher education may
increase the chance to access employment opportuni ulated places in the second period when physically
ties elsewhere. Higher levels of capital-intensive inputs more remote parcels were already abandoned. Forest
(e.g., irrigation and tractors) may keep agricultural land clearing possibly expanded in the opposite direction,
under production. Other village-level factors influence from locations closer to dwellings tomore remote areas
forest density. Higher numbers of goats per village often over time. In this study, the two dependent variables
render afforestation projects and natural forest regener (cropland abandonment and forest clearing) result in
ation more expensive and less likely to thrive as goats a range of outcomes that can be potentially realized
forage on tree seedlings. Village-level firewood usage is at any level. The relevant but unmeasured outcomes
also important. Although the effects of firewood usage in the Albanian case may include livestock and crop
may have been disguised by the spatial resolution of the production, the cultural landscape, biodiversity, and
satellite data employed, we expect a measurable forest biomass.7
reducing effect because 95 percent of the households in

the sample rely on the use of firewood for cooking and
heating and use an average of 10 m3 of firewood per Data Sources and Methods
year.
The household is the fundamental unit of land-use To study the salient land-use changes since the land
reform of 1991, we incorporated remotely sensed land
decision-making, which is in turn implemented at the
the costs of obtaining plot-level cover information from 1988, 1996, and 2003; biogeo
plot level. However,
information can be prohibitive. Instead, the remotely physical and accessibility characteristics; and village
sensed data used to measure land cover are often used level survey data via a series of statistical models.
Biogeophysical Data political upheavals of 1997marked thebeginningof a

second phase of postsocialist transformation inAlbania
Rainfall data stem from twenty-three rainfall stations
(Mailer and Sikor 2006).
in and another twenty around the research area. Con
To avoid large biases arising from the long recall
tinuous surfaces were created from the station points via
periods, recall techniques connecting the past event to
a regularized spline interpolator (Mitas and Mitasova
major landmark events (Groves 1989) were used to cap
1999). Contour lines at 10-m intervals from 1:50,000 ture major retrospective changes since the collapse of
topographic maps were used to derive a digital elevation socialism. The questions were related to the collapse of
model fromwhich elevation inmeters and slope in de
socialismin 1991 and the collapse of pyramid invest
grees were calculated. A soil map was obtained from the ment schemes at the end of 1996. Villagers were highly
JointResearch Council and was reclassified into ordered motivated to talk about issues surrounding livelihood
suitability classes for agricultural production based on
changes, agricultural production, and land use. Discus
theWorld ReferenceBase forSoil Resources (WRB) of sions among the respondents, typically between five and
theFAO of theUnited Nations.8
ten, were facilitated by the enumerators to reach a broad
consensus.
Different land-use dynamics were hypothesized for

100-Village Survey
each period of postsocialist transformation. Additional
To study land-use change, one would ideally like to information to triangulate and cross-check village-level
have household data for a large region, but such data data was compiled from local officials and secondary so
are not publicly available and may not be financially cioeconomic and land-use statistics at the village, com
or logistically feasible to collect (Rindfuss et al. 2004). mune, and district level.
The strategy used in this analysis was to conduct key
informant interviews in a large number of villages to ob
Independent Variables
tain a statistically representative sample for the entire
study area to link a broad cross section of socioeconomic We included the distance to the national road net
data to a spatial framework. Because land redistribution work as a covariate. This road network was established
was implemented at the village level, studies at this long before the period under consideration and is, there
level capture the influence of such processes. Sixteen fore, assumed to be exogenous to the land-cover changes
Albanian enumerators were trained in survey and in observed during the postsocialist period. Further, the
terviewing techniques. Two enumerators administered distance from every pixel to the closest dwelling, that
structured questionnaires to a group of
in each village is, to all populated places in the study area, was in
respondents, including the village mayor or vice mayor cluded, as well as the transportation costs from each
and elderly villagers. pixel to the four district capitals. Village-level variables
The survey was carried out in the fall of 2004 in a include population density, measured as the number
stratified sample of 100 rural villages, drawn from a total of households per square kilometer. Migration returns,
of 425 rural villages in the four districts. Using propor proxied by the importance of remittances as an income
tional random sampling, fiftyvillages were selected from source for the village economy, were measured as the
two groups stratified by cost distance. Cost distance was percentage of villagers who cited remittances as their
measured as the distance from each village centroid to main source of cash income.
the closest market center, adjusting for present land Land fragmentation as an outcome of the village
cover, the density and quality of the road network, and level land redistribution was measured as the average
slope. number of redistributed plots per households within a
Data were collected for land-use
influences hypothe village. Due to the inactive land markets in Albania
sized to result in the two land-cover changes of interest. with very few rentals or sales, the number of plots
The survey focused on the years 1991, 1996, and 2004; owned by a household remained virtually constant in
1991 was used as a starting point of interest, as the year the study area since the land reform in 1991. More plots
when socialist distributional networks ceased to func per household correspond to a higher fragmentation of
tion and the land reform was implemented; 1996 was cropland because each household received a set of non
chosen as an intermediate point in time corresponding contiguous plots stratified by market distance and soil
to the peak of the pyramid investment schemes. Both attainment was measured as the
fertility. Educational
years were followed by a period of anarchy, and the percentage of villagers with a high school certificate.
Two indicators measuring input intensity, the irrigated Data Integration

area per village in hectares and the density of tractors
used for agricultural production, were included in the Analysis was conducted at the pixel level with a spa
differences across districts were tial resolution of 30 m. All pixel-based variables were
models. Unobserved
referenced to the Pulkovo 1942 Transverse Mercator
captured by the inclusion of three binary variables as
fixed effects forGramsh, Librazhd, and Pogradec, with projection that is common inAlbania. The integration
of the survey data at the village level with the spa
Elbasan as the reference district.
tially continuous land-cover change categories and the
Two additionalindependent variables were included
in the regressions for forest loss: the number of goats per biogeophysical and distance measures required informa
tion regarding the spatial extent of village influences on
village and firewood usage.
the landscape. However, digital village boundaries are
not available inAlbania, but cadastral boundaries for
for Statistical Analysis agricultural plots distributed in the land reform were.
Sampling Strategy extent of peripheralareas effectively managed at
The
the village covered and
Several sampling strategies were applied to reduce level?mainly by forest, shrub,
the size of the spatial data set, limit the estimations to grassland, but also including remote and unproductive
the pixels of interest, and reduce spatial autocorrelation agricultural land that was refused by villagers in the re
across observations. We accounted for inherent depen distribution process?was approximated by local tech
dencies in land-cover change on the previous state of nicians using expert knowledge, topographic base maps,
land cover by excluding all pixels that were not part of and remote-sensing information. These approximated
the respective class at the beginning of the period. That village boundaries were used to represent each village's
is, cropland abandonment between 1988 and 1996 or sphere of influence.11
1996 and 2003 could only take place on pixels that
were used to grow crops in 1988 or 1996, respectively.
StatisticalModel
Equally, forest-cover loss was confined to those pixels
that were forested at the start of the respective period. A useful land-use model is able to explain land
Thus, we included 64 percent of the pixels in the study cover changes in multidimensional space, accounting
areas in 1988 and 62 percent in 1996 in the sample for local variations in socioeconomic, biogeophysical,
(all land covered by either cropland or forest in the and accessibility characteristics (Veldkamp and Lambin
respective year). The inclusion of all pixels would un 2001). In this article, the hypothesized drivers of land
derestimate the coefficients as, for example, nonforested use change are linked to observed land-cover change
pixels cannot be deforested.9 via a binary logitmodel (following
Chomitz andGray
Spatial autocorrelation in the dependent variables or
1996; Mertens and Lambin 2000; Munroe, Southworth,
in the residuals may cause biased coefficients (Anselin and Tucker 2004). The logistic models regressed the
1988). The land-cover data certainly contain spatial au probability of observing a particular land-cover change
tocorrelation because the smallest patch of contiguous on the underlying land-use processes in a spatially ex
land cover in the study area is 19,000 m2 (roughly the to understand and explain
plicit environment land
area of two Landsat TM pixels). We controlled for such cover
change.
autocorrelation the spatial averages for
by calculating Binary logit models were estimated for the two peri
the geophysical variables elevation and slope in a five ods and both land-cover changes of interest. To assess
by-five window to reduce spatial autocorrelation and the robustness of the results we estimated the fourmod
noise (Nelson, Harris, and Stone 2001). We
in the data els fornine different sampling distances. Consequently,
further generated a series of spatially stratified samples the number of observations ranged from 426 selected
(following Besag 1974), as done inNelson and Heller observations of cropland in 1988 at a sampling distance
stein (1997), Miiller and Zeller (2002), andMunroe, of 600 m to 19,379 observations for forested pixels in
Southworth, and Tucker (2002), by selecting a subset 1988 at a distance of 120 m between selected neighbors.
of pixels separated by a specified orthogonal distance. Because socioeconomic data were collected at the
We expect that observations are less dependent the far level, pixels within a village are not necessarily
village
ther they are from the next selected pixel. To assess of each other but are more likely to
independent
the robustness of the results, the sampling distance was be independent across villages (Miiller and Munroe
varied from 120 m to 600 m in steps of 60 m.10 200c, Gellrich et al. 2007). Therefore, the assumption
of independence of observations within villages is and the two input measures) at a prior state. In addi
violated (Overmars and Verburg 2006). To relax the tion, all variables were tested formulticollinearity in
assumption of independence we used the Huber and each spatial sample. The descriptive statistics of all the
White sandwich estimator to adjust for within-village covariates and of the two dependent variables used for
correlations. This technique yields robust standard estimating the models are listed in Table 1 for a sam
errors without affecting the estimated coefficients pling distance of 480 m.
(Williams 2000;Gutierrez and Drukker 2005). As this
approach further allows for any arbitrary correlation
of observations within clusters, it also accounts for
Lessons from Postsocialist Albanian
potential spatial autocorrelation in the model residuals
(Gellrichetal.2007). Landscapes
Becauseland-cover changes are a response to the so What Influences the Abandonment
Cropland:
cioeconomic conditions at the beginning of the respec ofCropland?
tive period (Perz and Skole 2003), temporally lagged
independent variables are employed. In other words, In both periods, agricultural land was more likely
land-cover change over one period was regressed on the to be abandoned at lower elevations and farther from
value of the time-variant variables (household density, the national roads, as expected (Table 2 for a sampling
the importance of remittances, the educational proxy, distance of 480 m). Areas that are closer to the four
Table 1. Descriptive statistics; spatial sample drawn with spacing of 480 m between observations
1988-1996 1996-2003
Label M
Obs SD Min Max Obs M SD Min Max
Cropland abandonment 852 0.39 0.49 0 1 652 0.22 1 0.42 0

Spatial lag slope (degrees) 852 10.25 7.97 0 77 652 10.10 8.06
47 0
Spatial lag elevation (100 m) 852 538.37 376.22 24 2,195 652 522.85 413.55 40 2,195
Rainfall (100 mm) 852 12.16 2.83 7.63 23.29 652 12.44 2.85 7.63 22.98
Soil suitability 852 3.65 1.22 1 5 650 3.78 1.02
5 1
Distance tonational road (km) 852 4.51 3.40 0 15.74 652 4.70 3.61 0 21.23
Distance to dwellings (km) 852 0.70 0.42 0 2.14 652 0.70 0.44 0.03 2.91
Cost distance todistrictcapital 852 629.98 251.50 102 1,253 652 653.65 264.90 111 1,253
Households per km2, 1991 852 24.15 21.36 2.20 121.58 652 28.19 28.14 2.20 151.98
Most cash income fromremittances,1991 (% hh) 852 6.97 14.85 0 90 652 52.67 22.44 95 1
Number of redistributed plots/household 852 4.12 1.12 1 10 652 4.31 102 1.21
Villagers with high school certificate(%) 852 2.32 4.07 0 30 652 1.99 0
30 4.32
Irrigation(ha), 1991 852 101.23 141.66 0 657 652 82.93 146.54 637 0
Tractors per km2 cropland, 1991 852 1.57 4.25 0 40 652 0 1.35
1.14 6.25
Forest clearing 1,579 0.15 0.36 0 1 1,683 0.13 1

0.34 0
Spatial lag slope (degrees) 1,579 16.61 8.19 0 66 1,683 16.59 8.25
66 0
Spatial lag elevation (100 m) 1,579 963.74 420.57 24 1,960 1,683 960.38 416.17 24 1,960
Rainfall (100 mm) 1,579 14.80 3.35 7.52 23.5 1,683 14.67 3.470 7.52 23.5
Soil suitability 1,578 3.68 1.15 1 5 1,682 3.68 5
1.17 1
Distance tonational road (km) 1,579 8.11 4.56 0.03 21.65 1,683 7.76 4.648 0 21.65
Distance to dwellings (km) 1,579 1.44 1.00 0.03 6.32 1,683 1.43 1.013 0 6.32
Cost distance to districtcapital 1,579 700.71 259.13 155 1353 1,683 685.04 260.87 106 1353
Households per km2, 1991 1,579 12.25 11.43 2.20 121.58 1,683 11.55 11.41 2.20 151.98
Most cash income fromremittances,1991 (% hh) 1,579 8.02 15.43 0 90 1,683 48.73 27.07 90 1
Number of redistributed plots/household 1,579 4.11 1.18 2 10 1,683 2
4.09 1.16
10
Villagers with high school certificate(%) 1,579 3.45 4.36 0 30 1,683 3.33 0
304.06
Irrigation (ha), 1991 1,579 83.80 91.12 0 450 1,683 72.98 85.16 3150
Tractors per km2 cropland,
1991 1,579 4.10 10.12 0 40 1,683 0.55 1.06
6.25 0
Goats (heads), 1991 1,579 333.50 463.18 0 2,000 1,683 458.64 703.43 0 3,000
Firewood usage (%), 1991 1,579 99.85 3.43 5 100 1,683 97.80 13.65 1000
868 Muller andMunroe
Table 2. Estimated coefficients forbinary logitmodel of cropland abandonment and forestclearing, 1988-1996 and
1996-2003; spatial sample drawn with spacing of 480 m between observations
Cropland Cropland Forest

Forest
abandonment abandonment clearing clearing
1988-1996 1996-2003 1988-1996 1996-2003
Coefficient p value Coefficient p value Coefficient p value Coefficient p value
Spatial lag slope (degrees) -0.002 0.923 0.049* 0.012 -0.026* 0.033 -0.017 0.299
Spatial lag elevation (100m) -0.002*** 0.000 -0.002* 0.015 -0.002** 0.006 -0.001 0.142
Rainfall (100 mm) -0.001 0.994 0.282* 0.034 0.054 0.579 -0.190 0.174
Soil suitability -0.161 0.316 0.026 0.917 0.133 0.334 0.001 0.997
Distance tonational road (km) 0.221*** 0.001 0.251** 0.003 0.115** 0.004 0.323*** 0.000
Distance to dwellings (km) 1.119*** 0.001 0.490 0.095 -0.620** 0.001 -0.074 0.709
Cost distance to districtcapital -0.002** 0.006 -0.001 0.541 0.001 0.263 -0.003* 0.037
Households per km2, 1991/1996 -0.031** 0.007 -0.016 0.127 0.013 0.340 0.017 0.131
Households withmost cash -0.011 0.584 0.006 0.571 0.005 0.520 0.012 0.090
income fromremittances,1991/1996 (%)
Number of redistributed plots perhousehold -0.229 0.197 0.420* 0.012 0.246* 0.041 0.256 0.119
Villagers with high school certificate(%) -0.001 0.987 0.036 0.553 0.066 0.140 0.016 0.704
Irrigation(ha), 1991/1996 -0.002 0.394 0.001 0.819 -0.003 0.283 -0.009** 0.002
Tractors per km2 cropland, 1991/1996 -0.063 0.123 0.031 0.839 -0.016 0.331 0.176 0.522
Goats (heads), 1991/1996 0.000 0.431 0.000 0.619
Firewood usage (%), 1991/1996 -0.018 0.352 0.012 0.457
Gramsh district 1.240 0.074 0.421 0.607 1.135* 0.034 0.928 0.375
Librazhd district 3.154*** 0.000 -0.508 0.608 -0.028 0.945 -0.711 0.238
Pogradec district 2.299** 0.010 3.322* 0.020 2.147** 0.005 -1.205 0.411
Constant 1.348 0.465 -8.238** 0.005 -1.828 0.440 -1.440 0.627
PseudoR2 0.29
0.24 0.16
0.27
Bayesian informationcriterion 615.35
980.84 1133.99
1263.84
N 650
852 1,578
1,682
* ** ***
p < 0.05. p < 0.01. p < 0.001.
district capitals were more likely to be abandoned significant relationship with cropland abandonment.
in the first period, though the coefficient is small. Also, rainfall patterns and soil suitability were insignif
Abandonment increased with distance from populated icant determinants of cropland abandonment. The
places (dwellings) in the first period. Villages with documented results are robust to all sampling distances.
lower household density appeared to be reducing the
amount of cultivated cropland in the first period. A The Importance of Agricultural Land Rents In
positive and significant influence was exerted from the creased Over Time. Postsocialist transitions have
measure of cropland fragmentation on abandonment had strong effects on cropland use in the Albanian
in the second period. countryside. After the socialist system effectively col
Some of the district dummies were strongly signif lapsed, all the associated land institutions, agricultural
icant in both periods. The significance of the fixed processing, and distribution networks collapsed along
district effects points to substantial remaining differ with it.The land redistribution in 1991 triggered a first,
ences between the districts that were not captured by immediate wave of cropland abandonment, caused by
the other variables included in the models. Cropland the refusal of farmers to accept the most marginal plots
abandonment was more likely in Librazhd and Pogradec redistributed (Ministry ofAgriculture and Food 2002).
in the first, and in Pogradec in the second period com Over time, the importance of the physical suitability
pared to the omitted reference district Elbasan. In of a plot for production increased and abandonment
one of the two poor and remote districts (Gramsh), was initially higher in areas farther from populated
abandonment was not significantly different from places. Land-use patterns also provide an indication
Elbasan, whereas itwas higher in the other (Librazhd). of the changing relative importance of agriculture
The other village measures did not show a statistically within the national economy. At the onset of the
Table 3. Selected accuracy assessments across all spatial sampling schemes
Sampling distance,meters 120 180 240 300 360 420 480a 600
540 M
Cropland 0.433
Kappa 0.432 0.427 0.442 0.417 0.413 0.486 0.393 0.456 0.433
abandonment, Area underROC 0.795 0.792 0.793 0.802 0.791 0.794 0.815 0.800 0.794 0.797
1988-1996 Percent correct 74.21 74.37 73.94 74.65 73.17 73.19 75.12 73.72 75.51 74.21
N
9,681 4,990 3,012 1,996 1,449 1,089 852 666 542 2,697
Cropland 0.474
Kappa 0.463 0.498 0.464 0.529 0.466 0.466 0.455 0.471 0.476
abandonment, Area underROC 0.855 0.853 0.865 0.856 0.865 0.859 0.851 0.839 0.842 0.854
1996-2003 Percent correct 82.19 81.98 82.56 82.35 83.95 81.26 83.08 82.85 81.92 82.46
N
7,513 3,880 2,334 1,547 1,103 843 650 519 426 2,090
Forest clearing, Kappa0.107 0.086 0.086 0.081 0.105 0.097 0.106 0.107 0.201 0.108
1988-1996 Area underROC 0.784 0.780 0.782 0.776 0.783 0.775 0.783 0.806 0.831 0.789
Percent correct 86.25 86.11 86.21 86.04 86.16 85.62 85.81 86.16 87.61 86.22
N
18,296 9,328 5,679 3,781 2,680 2,010 1,579 1,264 1,065 5,075
Forest clearing, Kappa0.205 0.201 0.228 0.109 0.209 0.210 0.298 0.191 0.131 0.198
1996-2003 Area underROC 0.845 0.843 0.852 0.837 0.851 0.844 0.853 0.844 0.843 0.846
Percent correct 87.57 87.38 87.70 86.74 87.46 87.33 88.11 87.46 86.85 87.40
N
19,379 9,876 6,009 4,012 2,846 2,123 1,682 1,340 1,118 5,376
Note: ROC = receiver operating characteristic.

aResults presented inTable 1 and Table 2.
transition, economic returns from other sectors of the reduction than on output maximization. One way to
economy were significantly lower than at present; as reduce risks is to diversify production by maintaining
a result, cropland production was more likely to be plots of different conditions and in different locations.
an economically feasible livelihood option, even in This would imply that villages with more fragmented
relatively remote locations. These areas were largely fields tend to work smaller fields, as they achieve the
abandoned in the second period. same level of protection from risk as villages with less
Higher household density raised the demand to keep fragmented but larger fields.
land under production within a village. It seems ob On the other hand, fragmentation may have con
vious that more households cultivate more land, but tributed to lowering agricultural returns by increasing
careful interpretation is required. The young and eco labor demand and raising management requirements;
nomically active often embark on internal or interna as a result, a more fragmented distribution of cropland
tional migration and reduce the available on-farm labor would have caused cropland to fall out of production
pool (Miluka et al. 2007). The resulting remittances are in later stages of the transition, when market forces
rarely invested in intensifying or expanding farming, became more influential in shaping the landscape.
but are most likely used to purchase consumer goods Land consolidation via land market transactions could
and durables. Those who remain in the villages culti potentially ameliorate this abandonment, but in con
a
vate the land, but land-use intensity (the amount of text of uncertainty, poor access to credit, and collapses
labor or capital inputs per unit of land) may have stag in input and output markets, these transactions are not
nated or even decreased (cf. Stahl 2007). Our data do occurring. Sabates-Wheeler (2002) and Lerman, Csaki,
not directly capture this effect because we exclusively and Feder (2004) identify the high transaction costs
focus on land-cover
change (a change from cropland of formal arrangements, complex administrative pro
to noncropland) and have no spatially explicit data at cedures, and suboptimal land administration systems as
hand to analyze patterns of land-cover modifications major constraints of formal land consolidation inCEE
(e.g., intensification) at the pixel level. countries more generally. Rural households also view
Land fragmentation as a consequence of the land re land assets as a safety net in light of the past volatile
form could potentially have both positive and negative economic conditions and maintain strong social ties
effects on the profitability of crop production. In the to land. The low price of agricultural land in many
Albanian context, given the postsocialist experience of regions adds to these disincentives to become involved
several periods of turmoil and political volatility (Stahl in permanent land transactions, which may hinder the
2007), farm households may put greater weight on risk reduction in farm fragmentation (De Soto et al. 2002).
It is difficult to make conclusionsabout household farther from national roads, but closer to the dwellings,
level processes empirically with our village-level data. in villages where cropland was more fragmented, and
Nevertheless, the results indicate that in the firstpe in Pogradec in the first period (Table 2). Forest-cover
riod, there is no significant relationship between crop loss in the second period was concentrated far from the
land abandonment and land fragmentation, but in the national roads, in areas with better access to district
last decade, fragmentation was associated with greater capitals, and in villages with less irrigated area. The dif
abandonment. A more fragmented distribution of crop ferences between the two periods are interesting. In the
land involved the distribution of plots that are difficult firstperiod, forest clearing was more likely in areas that
to access and therefore were left uncultivated in later are located closer to the populated areas (dwellings) but
stages of the transition when economic rents became a not in the second period. That is in part due to themea
more decisive factor shaping land use. surement of the dependent variable; pixels deforested
There are significant spatial variations across dis during the firstperiod did not enter the sample used for
tricts,which are not otherwise captured in the statisti the estimations of the second period. Forest clearing was
cal models. Abandonment was more likely in Pogradec, more likely in Gramsh (a relatively poor district) and
which contains a major market center as well as lim Pogradec in the firstperiod but did not show statistically
ited tourist opportunities, than any other district in significant differences among districts in the second pe
the second period. In Pogradec, the growing market riod. Other variables did not exert an influence on forest
orientation led to little abandonment around the dis loss.
trict capital, whereas most of the remaining cropland

inmore remote areas of the district was abandoned in
the second (see also Figure 4). Pogradec also
period From Subsistence-Oriented Village Clearing to
experienced large emigration to Greece, which is less Commercial Timber Extraction* Early clearings con
than 50 km away. Elbasan contains a large and in centrated on locations that were easily accessible from
tensively used plain that has good road access to the the villages, whereas in later stages the retreating forests
district capital. In Gramsh, there is a greater reliability forced local people to travel farther to harvest forest
on subsistence production, preventing large-scale aban products. At the same time, rising incomes have en
donment. The district divergence was not attributable abled local people to purchase fuel wood on local mar
to the land reform, which was implemented identically kets, thereby creating an active local fuel wood trade.
in the four districts; variations in land reform outcomes As a result, a share of the forest extraction activities
should have only been realized locally where village shifted from predominantly subsistence-based extrac
councils decided on the land redistribution. tion driven by village demandto more commercially
oriented clearing, integrated into local and regional
Cropland Abandonment in a Nutshell: Retreat markets for fuel wood trade (Miiller and Sikor 2006;
from Remote Areas at the Onset of the Transi Stahl 2007).12
tion, Then Qradual Market Alignment. The sta The forested areas around the dwellings were largely
tistical results indicated an
increasing importance of cut in the period immediately following the collapse of
market influences over time that manifest in a signifi socialism. In contrast, the association of forest clear
cant reshaping of the landscape. In particular, peren ing with the distance to the nationalroad network was
nial crop cultivation fell behind compared to more considerably stronger in the second period with a co
profitable strategies within the agricultural sector like efficient three times as high as in the first period. The
livestock production, and, more important, to broader influence of the state of anarchy after the collapse of
changes in livelihood strategies away from agriculture the pyramid schemes may have led to such a strong re
toward other economic sectors. By the late 1990s, lationship by creating an open-access state that caused
land fragmentation induced by the village-level reform sharp increases in forest loss and much forest extraction
choice significantly accelerated the abandonment of close to roads, indicating that illegal sales and exports
cropland. of timber may have been an important strategy for rural
households in that period. Forest control and the en
forcement of forest regulations virtually ceased in this
Forests: Why Does Forest Cover Decrease?
period and forest resources were depleted rapidly (cf.
A decrease in forest cover (more forest clearing) was Stahl 2007). At the same time, higher clearing rates
likely to be observed at lower slopes, lower elevations, closer to the major market centers (district capitals)
indicate the emergence of a second, commercially ori Goodness of Fit and Validation
ented extraction strategy.
of forest loss are visible in both periods Despite the large variation in the number of selected
Hotspots
with distinct spatial clusters across the study area. In observations, the signs and the significance levels of
the second period, forest clearing has occurred in the all biogeophysical, distance, and access variables were
more remote and the poorer regions of the study area largely stable across the various sampling schemes.13 To
assess and validate further the robustness of the models,
(Figure 4), when more easily accessible forested areas
were already exploited. Much of the forest clearing in we calculated the area under the receiver operating
the second period was possibly caused by the period characteristic (ROC) curve, kappa statistic, and the
of anarchy that created an open-access state. Never percent correct metric (PCM) for all sampling dis
tances. The three metrics quantify the agreement be
theless, large old-growth forested areas seem to be well
tween observed and estimated
land-cover changes. All
protected by their inaccessible setting. In other parts,
three were derived from comparing two-by-two con
young secondary forest is regenerating on abandoned
and unused pastures. tingency tables of observed to fitted values (Pontius
cropland
2002). PCM is the percentage of correctly predicted
pixels divided by the total number of pixels. The stan

Forest Clearing in Brief: Firewood Trade Spurs dard kappa statistic quantifies the level of agreement
Commercial Clearing and Partly Replaces Subsis between the estimated and actual land-cover changes
tence Extraction. Patterns of forest clearing followed compared against a random map (Pontius 2002). The
different trajectories from the changes seen in cropland ROC curve plots the rate of true positives on the vertical
area. In the absence of other energy sources, rural Al against the rate of false positives on the horizontal axis
banians are still highly reliant on firewood extraction. (Metz 1978). The area under the curve was estimated
With almost constant wood demand since the start of using the nonparametric version of ROC and indicates
the transition, clearing patterns bifurcated into mainly how much the estimated probability deviates from ran
subsistence-oriented extraction at more remote loca dom. The greater the area under the ROC curve, the
tions and more commercially oriented clearing closer better is the model goodness of fit. Both kappa and the
to the major market centers. area under the ROC are excellent metrics for study ar
eas with large changes and clumped patterns of changes
(Pijanowski, Alexandridis, and Miiller 2006).

All of the metrics listed in Table 3 were consistent
Key Lessons
over the various
spatial sampling distances, although
Over time, market forces permeate farther into the the number of observations (N) decreases by a factor of
landscape, which is visible in the spatial patterns of eighteen. Across the two land-cover change processes,
both cropland abandonment and forest clearing. Initial the kappa measures indicated moderate fit for cropland
land-use strategies tended to be based on semisubsis abandonment with kappas above 0.4 but very low fit
tence food production and fuel wood extraction, forced for the forest loss equations, particularly in the first
on the majority of the rural population due to a lack period with an average kappa of 0.11. Conversely,
of nonfarm income opportunities and few connections the model fitmeasured with the area under the ROC
to migration networks. Access to a much more diverse curve and with the PCM ishigh for all fourmodels, in
set of livelihood opportunities shifted incentives away both periods and across all samples (Table 3). The area
from crop cultivation in many villages, partly toward under the ROC ranges from 0.77 to 0.87 and the PCM
higher valued production strategies within agriculture, from 73 percent to 88 percent, taking the highest prob
more
but important toward income-earning strategies ability as the predicted change. Overall, the statistical
in other sectors (cf. Stahl 2007). The temporal changes model showed relativelygood fitwith high PCMs but
we observe for forestry activities point in a similar direc poor kappa statistics for the forest models in the first
tion. Higher cash incomes have facilitated purchases of period. This result, although disappointing, is not that
firewood, thus commercializing the forestry sector and surprising because of the difficulty in predicting where
a share of
spurring higher commercial clearing activi forest would be cleared at the pixel level. The smallest
ties. In sum, both in agriculture and in forestry, growing patch of contiguous land cover in the study area is
market integration is increasingly shaping theAlbanian 19,000 m2. Therefore, there is likely to be some inde
landscape. terminacy in predicting the exact area of small-scale
clearing for firewood extraction by households because livestock production indexes almost doubled between
we do not directly observe the trees that are cleared. 1991 and 2005 (FAO 2006).
In a broader context, widespread abandonment of
areas has occurred across western
Conclusion agricultural Europe
(MacDonald et al. 2000) and acrossCEE after 1990
In this analysis, we controlled for the impacts ofmul (Brouwer, Baldock, and la Chapelle 2001; Kuemmerle
tiscale, spatially heterogeneous land-use incentives on et al. 2006) and will possibly continue on a larger scale
land cover to allow us to examine evolving land-cover (Verburg et al. 2006). InAlbania, further abandonment
transitionsfollowing the large-scale policy shifts im of cropland may continue as returns from internal and
mediately following the transition and the subsequent international migration will become the most impor
realignment of land-use incentives due to land reform. tant livelihood strategy for the younger generation and
We found that agricultural abandonment in Albania as many of the remaining elderly farmers decease. Fu
is strongly mediated by both the biogeophysical en ture abandonment in Albania may be aggravated by
vironment and transportation infrastructure. District the projected reductions in crop productivity caused by
level effects provide some evidence that abandonment high temperatures and drought in a region already vul
is more likely in some regions than others, but was nerable to climate variability (Intergovernmental Panel
most likely in relatively remote areas, or in the pres on Climate Change 2007).
ence of other economic opportunities, such as tourism. Bush encroachment and natural forest regeneration
Interestingly, the importance of remittance income was will result as secondary vegetation on abandoned crop
not a significant correlate of cropland abandonment, land with significant implications for the environment.
perhaps because a low share of remittance income is The impact of the successional vegetation on biodi
channeled into agricultural investments (Miluka et al. versity, soil conditions, or the carbon sequestration
2007). Forest-cover loss was highly sensitive to the potential depends on the prevailing natural conditions
time period. Forest clearing tended to shift from sub and will therefore vary across regions (Verburg et al.
sistence orientation in the firstyears after the collapse 2006). For example, to preserve biodiversity values
of socialism to more commercial extraction in later in areas where endemic biodiversity is connected to
stages. specific farming techniques, significant investments
Regarding the long-term impacts of cropland aban from the international community and the Albanian
donment inAlbania, the situation isnot clear-cut. On government are necessary tomaintain traditional rural
the one hand, virtual autarchy, and the socialist distri landscapes and their associated agricultural production
butional system more generally, kept a greater number systems. Furthermore, investments in rural infrastruc
of people working the land than can be sustained un ture, particularly in the dilapidated rural road network,
der a market economy, so the shedding of agricultural may offer competitive income opportunities in more
labor is inevitable. The abandonment of large areas of remote areas and may help reducing rural emigration,
cropland partly reflects the adjustment of the rural sec at the same time exerting positive influences on the
tor to the evolving market conditions and leads to a local, regional, and global environmental services
concentration of cultivation on more productive areas provided by rural landscapes in postsocialist Albania.
(Ioffe, Nefedova, and Zaslavsky 2004). However, the Rural landscapes will continue to evolve and change.
removal of land from production can in the short term Land reforms, particularly the establishment of pri
increase soil erosion and have other such undesirable vate property rights, are based on the logic that effi
effects, although in the longer term, the regeneration of ciency gains in agricultural production will occur as
vegetation may lead to increases in biomass (Houghton, a result. The reality is, over time and space, that
Hackler, and Lawrence 1999). Our data indicate that there is much variation in this outcome (Hamilton
abandonment ishighest in those areas that are neither 1999). Lerman (2000) provides an interesting sum
the most remote nor the most accessible. It may be mary of empirical evidence on
the success of land
the case that the poorest families in remote locations reform in terms of increasing farm output in CEE.
have moved toward subsistence production, which can He points out that starkly varying degrees of suc
mean greater rural poverty and higher inequality. Con cess in implementing such policies must relate to lo
versely, those farmers who can invest inmore market cal social and political contingencies. Moreover, he
oriented agriculture have shifted from cereal production notes that private property rights do not always trans
to higher valued products (cf. Stahl 2007). For instance, late into individual farming, and individual farming
can happen in the absence of strictly private property to facilitate comparative studies, within and outside of
eastern but not to encompass all socialist move
Europe,
rights.
ments, institutions, or practices.
In summary, the integrated, multiscale data set em
3. Despite this legislation a substantial amount of land has
ployed indicated that a large part of the spatial and tem been restituted by ad hoc village commissions. Kodder
poral variation in land-cover change was a function of itzsch (1999) estimates that in total about 15 to 20 per
and policy-related vari cent ofAlbania's agricultural lands were restituted.
biogeophysical, socioeconomic,
4. We concentrate on physical land fragmentation that is
ables across a diverse region in southeastern Albania.
defined as the spatial dispersion of fields into separate
Statistical analyses are important tools to combine mul and distinct parcels over a large area (Binns 1951).
tidimensional data relating to land use and land cover, 5. Communes are the first administrative unit with a le
in a spatially and temporally relevant context. Never gal jurisdiction inAlbania. Each commune consists of
several villages where the elected elders (kryeplaku) im
theless, considerable unexplained variance remains that
plement decisions of the commune. Districts represent
points to various factors influencing postsocialist land the second administrative level and are responsible for
scapes that were not a priori hypothesized. Other im coordination of activities, development, and the imple
portant factors cannot be captured in our probabilistic mentation of state policies (Hoxha 2001).
6. Land-use such as are also a relevant
framework. For example, the effect ofmigration may not practices grazing
be directly observable in a spatial framework. Social net change, but we could not distinguish pasture from
shrub or grassland with sufficient accuracy from the
works and connections that are difficult to quantify at
imagery and thus we limit our discussion to cropland
the village level may play a crucial role. In such circum abandonment.
stances, qualitative data collection and household-level 7. Feedback effects from outcomes to various levels exist
but are unobserved, omitted from the statistical analysis,
information can strengthen insights on such processes
and not included in Figure 5 for the sake of simplicity.
that are otherwise difficult to observe and to explain. 8. See theWRB home page at http://www.fao.org/ag/agl/
agll/wrb/default.stm(last accessed 5 June 2006).
9. decision processes can lead to sample selec
Sequential
tion bias that can
be corrected estimation
using other
Acknowledgment for example, a two-stage How
procedures; approach.
ever, such a selection bias is not present in this case,
Thisresearch was carried out with funding from the
because the empirical model is specified fora specific sub
German Research Foundation (Deutsche Forschungs set of the entire population; that is,all pixels. Therefore,
gemeinschaft) under the Emmy Noether-Programm. we could proceed using standard econometric methods
Darla K. Munroe received support from the Office of (Wooldridge 2002).
10. One reviewer cross-validation to as
International Affairs at The Ohio State University. suggested using
sess the goodness of fitwith observations (pixels) that
The authors are indebted to Thomas Sikor, Johannes not enter we
did the analyses. However, believe that
Stahl, and Kendra McSweeney for constructive com our
approach does not only address model goodness
ments that improved this article significantly. We also of fit but has the additional advantage that it deals
with autocorrelation across the various
thank Mario Gellrich for insightful statistical discus spatial sampling
distances.
sions, Ylli Hoxha for his work in the satellite image
11. The accuracy of these village boundaries was not verified
interpretation, and our many collaborators inAlbania. but to yield better estimates than
statistically appears
Finally, we are grateful for the helpful comments and other approaches to derive village boundaries employed
suggestions of three anonymous reviewers and the previously by, for example, Mertens et al. (2000) and
Miiller and Munroe (2005).
editor of the Nature and Society section.
12.We are grateful toThomas Sikor forpointing us in this
direction.
13. Tables containing the logit coefficients and p values at
Notes different sampling distances are omitted due to space
limitations but can be obtained from the authors upon
1. We use the term transition to refer to the transformation request.
of a centrally planned economy to a market-oriented
one. Transition specifically emphasizes the period of re
form,or the time horizon of the policies implemented,
and the resulting socioeconomic changes,
as is our focus. References
2. We use the term socialistcountryaccording to the defini
tion of JanosKornai (1992), which generally refers to a Agrawal, A. 2001. Common property institutions and sus
state under control of a Communist party. Synonyms in tainable governance of resources.World Development 29
clude Soviet-type system, centrally administered econ (10): 1649-72.
omy, centrally planned economy, command economy, Anselin, L. 1988. Spatial econometrics:Methods and models.
or state socialism (Kornai 1992). This term ismeant Dordrecht, The Netherlands: Kluwer Academic.
Bar jaba, K. 2000. Contemporary patterns inAlbanian emi Houghton, R. A., J.L. Hackler, and K. T. Lawrence. 1999.
gration. South-East Europe Review for Labour and Social The U.S. carbon budget: Contributions from land-use
Affairs 2:57-64. change. Science 285 (5427): 574-78.
Besag, J.E. 1974. Spatial interaction and the statistical analy Hoxha, A. 2001. Local government inAlbania. In Stabili
sis of lattice systems.Journal of theRoyal StatisticalSociety sation of heal governments, ed. E. Kandeva, 41-88. Bu
Series B 36:192-236. dapest, Hungary: Local Government and Public Service
Binns, B. O. 1951. The consolidation of fragmentedagricultural Reform Initiative/Open Society Institute.
holdings. FAO Agricultural Studies 11. Rome: FAO. Intergovernmental Panel on Climate Change (IPCC). 2007.
Blacksell, M., and K. M. Born. 2002. Private property resti Climate change 2007: Impacts, adaptation and vulner
tution: The geographical consequences of official gov ability.Working Group II Contribution to the Inter
ernment policies in Central and Eastern Europe. The governmental Panel on Climate Change Fourth Assess
Geographical Journal 168 (2): 178-90. ment Report, http://www.ipcc-wg2.org/index.html (last
Bloch, P. 1998. Picking up the pieces: Consolidation ofAl accessed 30 April 2007).
bania^ radical land reform. In Land reformin theformer Ioffe,G., T. Nefedova, and I. Zaslavsky. 2004. From spatial
Soviet Union and Eastern Europe, ed. S. K. Wegren, 189? continuity to fragmentation: The case of Russian farm
207. London: Routledge. ing.Annals of theAssociation ofAmerican Geographers 94
Bray, D. B., E. A. Ellis, N. Armijo-Canto, and C. T. Beck. (4): 913-43.
2004. The institutional drivers of sustainable landscapes: Jansen, L., G. Carrai, L. Morandini, P. Cerutti, and A.
A case study of the "Mayan Zone" inQuintana Roo, Spisni. 2006. Analysis of the spatio-temporal and
Mexico. Land Use Policy 21 (4): 333-46. semantic aspects of land-cover/use change dynamics
Brouwer, F., D. Baldock, and C. laChapelle. 2001. Agriculture 1991-2001 inAlbania at national and district levels.
and nature conservation in the candidate countries: Perspec EnvironmentalMonitoring andAssessment 119 (1-3): 107
tives in interaction.Amsterdam: Ministry ofAgriculture, 36.
Nature Management and Fisheries, The Netherlands. Jarvis,C. 2000. The riseand fall of thepyramid schemes inAlba
Carletto, C, B. Davis, M. Stampini, and A. Zezza. 2006. A nia. IMF Staff Papers Vol. 47, No. 1. http://www.
on the move: International in post
country migration imf.org/extemal/Pubs/FT/staffp/2000/00
communist Albania. InternationalMigration Review 40 01/pdf/jarvis.pdf(last accessed 22 November 2006).
(4): 767-85. Kaimowitz, D., and A. Angelsen. 1998. Economic models of
Chomitz, K. M., and D. Gray. 1996. Roads, lands use, and tropicaldeforestation:A review. Bogor, Indonesia: Center
deforestation: A spatial model applied to Belize. World for International ForestryResearch (CIFOR).
Bank Economic Review 10 (3): 487-512. King, R. L. 2005. Albania as a laboratory for the study of
Cropper, M., J. Puri, and C. Griffiths. 2001. Predicting the migration and development. Journal of Southern Europe
location of deforestation: The role of roads and protected and theBalkans 7(2): 133-55.
areas inNorth Thailand. Land Economics 77 (2): 172-86. Kodderitzsch, S. 1999. Reforms inAlbanian agriculture:Assess
Cungu, A., and J. F. M. Swinnen. 1999. Albania's radi inga sector in transition.
Washington, DC: The World
cal agrarian reform.Economic Development and Cultural Bank.
Change 47 (3): 605-20. Kornai, J. 1992. The socialist system:The political economy of
De Soto, H., P. Gordon, I.Gedeshi, and Z. Sinoimeri. 2002. socialism. Princeton, NJ: Princeton University Press.
Poverty inAlbania?A qualitative assessment. World Korovilas, J.P. 2005. Remittances and pyramid investment
Bank Technical Paper No. 520. Washington, DC: The schemes inAlbania. In Remittances: Development impact
World Bank. and future prospects, ed. S. M. Maimbo and D. Ratha,
De Waal, C. 2004. Post-socialist property rights and wrongs 175-92. Washington, DC: The World Bank.
inAlbania: An ethnography of agrarian change. Conser Kuemmerle, T., V. C. Radeloff, K. Perzanowski, and P.
vation& Society 2(1): 19-50. Hostert. 2006. Cross-border comparison of land cover
Food and Agriculture Organization (FAO). 2006. FAO and landscape pattern in Eastern Europe using a hybrid
STAT data. Food and Agriculture Organization (FAO). classification technique. Remote Sensing of Environment
http://faostat.fao.org/(last accessed 16May 2006). 103 (4): 449-64.
Gellrich, M., P. Baur, B. Koch, and N. E. Zimmermann. Lambin, E. F., H. J. Geist, and E. Lepers. 2003. Dynamics
2007. Agricultural land abandonment and natural forest of land-use and land-cover change in tropical regions.
re-growth in the Swiss mountains: A spatially explicit Annual Review of Environment and Resources 28:205
economic analysis. Agriculture, Ecosystems & Environ 41.
ment 118 (1-4): 93-108. Lambin, E. F., B. L. Turner II, H. J.Geist, S. B. Agbola,
Global Land Project. 2005. Science plan and implementation A. Angelsen, J.W. Bruce, O. T. Coomes, et al. 2001.
strategy.Stockholm, Sweden: IGBP Secretariat. The causes of land-use and land-cover change: Moving
Groves, R. M. 1989. Survey errorsand surveycosts.New York: beyond themyths. Global EnvironmentalChange 11 (4):
Wiley. 261-69.
Gutierrez, R., and D. M. Drukker. 2005. Reference for Leemans, R., and S. Serneels. 2004. Understanding land-use
cluster-correlated robust variance calculation. Stata to reconstruct, describe or in
Corp. change predict changes
http: //www.stata.com /support/faqs/stat/robust_ref.html land cover. Geojournal 61 (4): 305-07.
(last accessed 12 June 2006). Lerman, Z. 2000. From common heritage to divergence: Why
Hamilton, F. E. I. 1999. Transformation and space inCentral the transition countries are
drifting apart by
measures of
and Eastern Europe. The Geographical Journal 165 (2): agricultural performance? American Journal of Agricul
135-44. turalEconomics 82 (5): 1140-48.
-. 2001. in transition economies: From -. 2004. Modeling spatially and temporally complex
Agriculture
common heritage to divergence. Agricultural Economics land-cover change: The case of western Honduras. The
26 (2): 95-114. ProfessionalGeographer 56 (4): 544-59.
Lerman, Z., C. Csaki, and G. Feder. 2004. Evolving farm Myers, N., R. A. Mittermeier, C. G. Mittermeier, G. A. B.
structures and land use patterns in former socialist coun da Fonseca, and J.Kent. 2000. Biodiversity hotspots for
tries.Quarterly Journal of InternationalAgriculture 43 (4): conservation priorities.Nature 403 (6772): 853-58.
309-35. Nelson, G. C, V. Harris, and S. W. Stone. 2001. Deforesta
MacDonald, D., J.R. Crabtree, G. Wiesinger, T. Dax, N. tion, land use, and property rights: Empirical evidence
Stamou, P. Fleury, J. Gutierrez Lazpita, and A. Gi fromDarie'n, Panama. Land Economics 77 (2): 187-205.
bon. 2000. Agricultural abandonment inmountain areas Nelson, G. C, and D. Hellerstein. 1997. Do roads cause defor
of Europe: Environmental consequences and re estation? satellite images in econometric analysis
policy Using
sponse. JournalofEnvironmentalManagement 59 (1): 47 of land use. American JournalofAgricultural Economics 79
69. (1): 80-88.
Mathijs, E., and J. F. M. Swinnen. 1998. The economics Nicholson, B. 2004. Migrants as agents of development: Al
of agricultural decollectivization in East Central Europe banian return migrants and micro-enterprise. In New
and the formerSoviet Union. Economic Development and patterns of labourmigration inCentral and Eastern Eu
Cultural Change 47 (1): 1-26. rope, ed. D. Pop, 94-110. Cluj Napoca, Romania: AMM
McCusker, B. 2004. Land use change on recently redis Editura.
tributed farms in theNorthern Province, South Africa. Ostrom, E. 1990. Governing the commons: The evolution of
Human Ecology 32 (1): 49-75. institutionsfor collective action. Cambridge, U.K.: Cam
Mertens, B., and E. F. Lambin. 2000. Land-cover change tra bridgeUniversity Press.
jectories in southernCameroon. Annals of the
Association Overmars, K. P., and P. H. Verburg. 2006. Multilevel mod
of American Geographers 90 (3): 467-94. elling of land use fromfield to village level in the Philip
Mertens, B.,W. D. Sunderlin, O. Undoye, and E. F. Lambin. pines. Agricultural Systems89 (2-3): 435-56.
2000. Impact of macroeconomic change
on deforestation Perz, S. G., and D. L. Skole. 2003. Social determinants of
in South Cameroon: Integration of household survey secondary forests in the Brazilian Amazon. Social Science
and remotely-sensed data. World Development 28 (6): Research 32 (1): 25-60.
983-99. Pijanowski, B. C, K. T. Alexandridis, and D. Miiller. 2006.
Metz, C. E. 1978. Basic principles ofROC analysis. Seminars Modelling urbanization patterns in two diverse regions
inNuclear Medcine 8 (4): 283-98. of the world. Journal of Land Use Science 1 (2): 83
Miluka, J.,G. Carletto, B. Davis, and A. Zezza. 2007. The 108.
vanishingfarms?The impact of internationalmigration on Pontius, R. G. J.2002. Statistical methods to partition effects
Albanian family farming.Washington, DC: The World of quantity and location during comparison of categorical
Bank. maps atmultiple resolutions. PhotogrammetricEngineering
Ministry ofAgriculture and Food. 2002. Annual report2002. & Remote Sensing 68 (10): 1041-49.
Tirana, Albania: Ministry ofAgriculture and Food. Rindfuss, R. R., S. J.Walsh, I. B. L. Turner, J. Fox, and
Mitas, L., and H. Mitasova. 1999. Spatial interpolation. V. Mishra. 2004. Developing a science of land change:
In Geographical informationsystems, ed. P. A. Longley, Challenges and methodological issues. Proceedings of the
M. F. Goodchild, D. J.Maguire, and D. W. Rhind, 481 National Academy of Sciences 101 (39): 13976-81.
92. New York:Wiley. Sabates-Wheeler, R. 2002. Consolidation initiatives after
Moran, E. F., E. Ostrom, and J.C. Randolph. 2002. Ecological land reform:Responses tomultiple dimensions of land
systems and multi-tier human organization. In Knowl fragmentation in Eastern European agriculture. Journal
edgemanagement, organizational intelligenceand learning, of InternationalDevelopment 14 (7): 1005-18.
and complexity,ed. L. Douglas Kiel, in Encyclopedia of Sikor, T. 2004. The commons in transition: Agrarian and
Life Support Systems (EOLSS), developed under the aus environmental change inCentral and Eastern Europe.
pices of theUNESCO. Oxford, U.K.: EOLSS Publishers. EnvironmentalManagement 34 (2): 270-80.
http://www.eolss.net (last accessed 23 July2008). Soares-Filho, B., A. Alencar, D. Nepstad, G. Cerqueira, M. d.
Miiller, D., and D. K. Munroe. 2005. Tradeoffs between rural C. Vera Diaz, S. Rivero, L. Solorzano, and E. Voll. 2004.
development policies and forest protection: Spatially Simulating the response of land-cover changes to road
explicit modeling in theCentral Highlands ofVietnam. paving and governance along

a major Amazon highway:
Land Economics 81 (3): 412-25. The Santarem-Cuiab& corridor. Global Change Biology
Miiller, D., and T. Sikor. 2006. Effectsof postsocialist reforms 10 (5): 745-64.
on land cover and land use in south-eastern Albania. Stahl, J. 2007. The political ecology of postsocialist land use
Applied Geography 26 (3-4): 175-91. change: Case studies from three villages inSoutheastern
Muller, D., and M. Zeller. 2002. Land use dynamics in the Albania. Doctoral dissertation, Humboldt-Universitat
central highlands ofVietnam: A spatial model combin zu Berlin, Berlin.
ing village survey data and satellite imagery interpreta Stahl, J.,T. Sikor, and S. Dorondel. Forthcoming. Trans
tion.Agricultural Economics 27 (3): 333-54. parence de Vadministration fonciere en Albanie et en
Munroe, D., J. Southworth, and C. M. Tucker. 2002. The Roumanie (Transparency inAlbanian and Romanian land
dynamics of land-cover change in western Honduras: administration). Paris: CIHEAM.
Exploring spatial and temporal complexity. Agricultural Stanfield, D., and A. Kukeli. 1995. Consolidation of the
Economics 27 (3): 355-69. Albanian agricultural land reform through a program
for creating an immovable property registration system. based studies of future land use in Europe. Agriculture,
Computers, Environment and Urban Systems 19 (2): 131 Ecosystems& Environment 114 (1): 39-56.
40. Williams, R. L. 2000. A note on robust variance estimation
Swinnen, J. F. M. 1999. The political economy of land re for cluster-correlated data. Biometrics 56 (2): 645-46.
form choices inCentral and Eastern Europe. Economics Wooldridge, J.M. 2002. Econometric analysis of cross section
ofTransition 7 (3): 637-64. and panel data. Cambridge, MA: The MIT Press.
Swinnen, J.F. M., L. Vranken, and V. Stanley. 2006. Emerg World Bank. 2003. Albania povertyassessment.Washington,
ingchallenges of land rentalmarkets: A review of theavail DC: The World Bank.
able evidence forCentral and Eastern Europe and former -. 2006a. Status of the land reformand real propertymar
Soviet Union countries.Washington, DC: The World kets inAlbania. Tirana, Albania: The World Bank Office,
Bank. Tirana.
Veldkamp, A., and E. F. Lambin. 2001. Predicting land-use -. 2006b. World development indicators2006. Washing
change. Agriculture, Ecosystems& Environment 85 (1-3): ton,DC: The World Bank.
1-6. Zezza, A., G. Carletto, and B. Davis. 2005. Moving away from
Verburg, P. H., C. J. E. Schulp, N. Witte, and A. Veld poverty: A spatial analysis of poverty and migration in
kamp. 2006. Downscaling of land use change scenarios Albania. Journal of SouthernEurope and theBalkans 7 (2):
to assess the dynamics of European landscapes: Scenario 175-93.
Correspondence:Leibniz InstituteofAgriculturalDevelopment inCentral and Eastern Europe (IAMO), Theodor-Lieser-Str. 2, 06120 Halle
(Saale), Germany; or Geomatics Lab, Geography Humboldt-Universitat zu Berlin, Unter den Linden 6,10099
Department, Berlin, Germany,
e-mail: mueller@iamo.de (Miiller); Department of Geography, The Ohio State University, 1036 Derby Hall, 154 N. Oval Mall, Columbus,
OH 43210, e-mail: munroe.9@osu.edu (Munroe).

MULLER Daniel and MUNROE Francis - Changing Rural Landscapes in Albania - 2008

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

MULLER Daniel and MUNROE Francis - Changing Rural Landscapes in Albania - 2008

Uploaded by

Copyright:

Available Formats

Changing Rural Landscapes in Albania: Cropland Abandonment and Forest Clearing in the

*mjsft&3iXtommn:?%GttJb:utom3:, mmmi?mmwumm^mo #tt??

La transici6n al postsocialismo en el este de Europa se ha concentrado en el establecimiento del derecho a la

J^B^^^H^^^^^^|^^^^^^^B^^H^^^@^1^^^M|9h 2.000""Yji"hi iV??.,

* / ? jj -^#^t( ^Jy S^^J^\ Kosovo

/ / \ \C^ ^i^^y^4 f <^S ^ afea

Figure 3. Study area.

profound, and there are a variety of social and environ

JLy^^T ^^ \ 284sqkm affecting land rent because transport costs to input

The Connection Between LUCC and Postsocialist

Much of the interest in LUCC has focused on the

4. abandonment and forest Source:

in Transition surprisingly few integrated analyses of postsocialist cir

There is an established tradition within the LUCC Conceptual Framework

H^H /D^.,nd"ded'nmodje,x I LivestockI I Crop I Cultural I I Bio IC B,omass

reducing effect because 95 percent of the households in

Biogeophysical Data political upheavals of 1997marked thebeginningof a

Different land-use dynamics were hypothesized for

Two indicators measuring input intensity, the irrigated Data Integration

Cropland abandonment 852 0.39 0.49 0 1 652 0.22 1 0.42 0

Forest clearing 1,579 0.15 0.36 0 1 1,683 0.13 1

Cropland Cropland Forest

Coefficient p value Coefficient p value Coefficient p value Coefficient p value

Table 3. Selected accuracy assessments across all spatial sampling schemes

Note: ROC = receiver operating characteristic.

trict capital, whereas most of the remaining cropland

pixels divided by the total number of pixels. The stan

(Pijanowski, Alexandridis, and Miiller 2006).

explicit modeling in theCentral Highlands ofVietnam. paving and governance along

You might also like