Contents lists available at ScienceDirect

Geoforum

journal homepage: www.elsevier.com/locate/geoforum

The politics of uneven smallholder cacao expansion: A critical physicalgeography of agricultural transformation in Southeast Sulawesi, IndonesiaLisa C. KelleyGeography & Environment Department, University of Hawai’i, Mãnoa, 445 Saunders Hall, Honolulu, HI 96822, United States

A R T I C L E I N F O

Keywords:Commodity crop expansionLUCCIndonesiaCacaoPolitical ecologyCritical physical geographyRemote sensing

A B S T R A C T

This article reconstructs and explains variability in pathways of smallholder cacao expansion and land use andcover change (LUCC) over the past four decades in Southeast Sulawesi, Indonesia. I first draw on approachesfrom remote sensing (RS) to reconstruct cacao expansion in two top-producing districts (1972–2014), high-lighting the variegated environment changes shaped by smallholder cacao plantings, both inside and outsideforested lands. I then integrate these data with theories and methods from political ecology (PE) and criticalphysical geography (CPG) to document the uneven politics of land and capital access mediating this variability.Using these data, I argue that variability in crop expansion and LUCC should not be considered as an exceptionbut as the norm; integral to any generalizable explanation of LUCC. I also show how a focus on variability cannot only supplement but also shift dominant accounts of LUCC.

1. Introduction

Land acquisitions and conversions for export-oriented agriculturalcommodities have rapidly increased in recent decades (Phalan et al.,2013). This is particularly true in Indonesia, where the land held incacao, oil palm and rubber increased from an estimated 1,835,548 ha to14,673,925 ha over the period 1980–2016 (FAOSTAT, 2018). The es-tablishment of export-oriented agricultural economies has reshapedagrarian societies and livelihoods, more fully incorporating people intoboth nation states and global markets (Hall, 2011; Sikor, 2012; Fox andCastella, 2013). It has also driven profound and conflictual changes inthe biophysical landscape, including the conversion of many forestedand agro-forested lands into monotypic tree crop commodities (Gibbset al., 2010; Carlson et al., 2012; Davis et al., 2015).

In this context, a growing body of work has sought to understandpathways of commodity crop expansion and associated land use andcover change (LUCC). In Indonesia, recent analyses have drawn uponsatellite imagery to document the spatio-temporal linkages betweenindustrial agricultural concessions, smallholder crop adoption andforest cover loss (Miettinen and Liew, 2010; Lee et al., 2014; Gaveauet al., 2014; Abood et al., 2015). Related work has quantified the causalrelevance of “proximate” and “underlying” drivers of LUCC: identifyingthe linkages between LUCC and road construction (Miyamoto, 2006),transportation costs and agricultural rents (Brun et al., 2015), and priorspatial patterns of clearance (Cushman et al., 2017). Spatially-explicitmodels have also been used to predict the impact of policies on LUCC in

key sites of commodity crop production (Koh and Ghazoul, 2010; Lawet al., 2015, Van der Laan et al., 2017).

Strikingly little work, however, has grappled with the high varia-bility of commodity crop expansion and LUCC at more local scales ofanalysis. As one recent analysis in Indonesia notes, “most studies ofdeforestation drivers have been based on macro-level regional or na-tional data, thus presenting difficulties in addressing the complexity oflocal situations” (Nugroho et al. 2017: 2). Meta-analyses suggest thatwhile the macro-economic, demographic and agronomic correlates ofcrop expansion are often well understood, this is far less true of thesocio-political and historical dynamics mediating how these factorsshape LUCC sub-regionally (Lambin and Geist, 2008; Turner andRobbins, 2008; Hettig et al., 2016). This gap constrains potential policyresponses, limiting an understanding of associated variability in liveli-hoods, biodiversity conservation and land-based carbon emissions(Hall, 2011; Chaplin-Kramer et al., 2015).

This paper engages with these debates in Sulawesi, Indonesia, anapparent exemplar of the tensions between agricultural expansion andforest conservation. Sulawesi’s cacao (Theobroma cacao L.) boom hasbeen celebrated as a remarkable development success: from almost noproduction in the 1980s, more than 600,000 ha of land were convertedinto cacao by the early 2000s, 92% by smallholders operating on anaverage 1.75 ha (Direktorat Jenderal Perkebunan, 2014). Existingwork, however, suggests that smallholder cacao plantings have oftenbeen closely linked to deforestation (Ruf and Siswoputranto, 1995;Erasmi et al., 2004; Steffan-Dewenter et al., 2007). In this regard, cacao

https://doi.org/10.1016/j.geoforum.2018.10.006Received 31 October 2017; Received in revised form 3 October 2018; Accepted 4 October 2018

E-mail address: lckelley@hawaii.edu.

Geoforum 97 (2018) 22–34

0016-7185/ © 2018 Elsevier Ltd. All rights reserved.

T

expansion in Sulawesi has been seen to uphold a consistent patternthroughout history (Ruf and Siswoputranto, 1995; Clough et al., 2009).Drawing on secondary data, for example, Neiston et al. (2004) havesuggested that most of the seven million hectares of land in cacaoglobally was previously held in forest cover.

In explaining cacao-deforestation linkages, scholarly work typicallyemphasizes the yield advantages growers can obtain by planting cacaoin ‘virgin,’ ‘primary,’ or ‘undisturbed’ forested lands. Francois Ruf’sforest rent theory of cacao expansion argues that the Ricardian rentsafforded by freshly cleared forested land—high soil fertility, low diseaselevels and protective shade for newly planted seedlings—have com-bined with rapid in-migration and moments of opportunity in worldmarkets to propel cacao expansion into forests, in Sulawesi and else-where. Forest rents also beget what Clarence-Smith and Ruf (1996: 1)suggest is the basic problem of cacao economies: “their need for freshsupplies of primary forest to maintain themselves.” Because forest rentseventually expire under intensive production, growing material andlabor inputs will be required to maintain the same yields. Assuming nopremium to offset these costs, forest rent theory posits that forestedland will out-compete existing sites of production.

Beginning in the late 1990s, smallholder cacao producers inSulawesi began to experience widespread pest and pathogen losses.Informed by ideas of forest rent, economic and conservation analystsargued that further forest clearance in the sector was likely. In 2000, forexample, Ruf and Yoddang speculated that despite yield losses, pro-duction had nonetheless continued to climb between 1998 and 1999“owing to the massive scale on which new planting continues whileforests last” (108). In recent years, sustained declines in yields1 havealso led analysts to argue that yield intensification is necessary to“spare” further forest from clearance (Ruf and Yoddang, 2001; Neilson,2007; Clough et al., 2009). Clough et al. (2009), for example, has ar-gued that:

“a failure to sustain production in current cacao-growing areas inSulawesi likely means a shift to the forest margins” while Neilson(2007: 2) has noted that “Left to market forces alone, the ‘mining’ ofcocoa regions will in all likelihood continue unabated across tropicalfrontiers until all potential cocoa lands have been physically ex-hausted.”

Anchored in an analysis of variability, the objective of this study isto re-read recent processes of cacao expansion and LUCC in SoutheastSulawesi province, the site of 16% of all production in Indonesia and afocal site of debates surrounding cacao intensification. I begin bydrawing on approaches from remote sensing (RS) to reconstruct cacaoexpansion and LUCC in two top-producing districts over the past fourdecades (1972–2015), highlighting the variegated environmentalchanges shaped by smallholder cacao plantings, both inside and outsideforested lands. I then integrate LUCC mapping with theories andmethods from political ecology (PE) and critical physical geography(CPG) to document the uneven politics of access to land and capitalmediating this variability. Below, I show how these approaches can beused in tandem to better capture and explain variability than has beentrue in past work. I also show how an analytical emphasis on variabilitycan not only supplement but also shift dominant accounts of LUCC;shifting persistent assumptions guiding policy.

2. Known and unknown dimensions of smallholder cacaoexpansion in Sulawesi, Indonesia

Consistent with the forest rent theory of expansion, most work toexplain smallholder cacao boom in Sulawesi has emphasized three keyfactors. First, Sulawesi’s ample forest reserves, particularly in the fertile

alluvial lowlands and in sites of abundant rainfall (Durand, 1995; Rufand Siswoputranto, 1995; Ruf and Yoddang, 1996). Second, the in-migration of people of Bugis ethnicity into forested regions from the1970s onward, many of whom had experience with cacao from anearlier regional boom in Malaysia (Jamal and Pomp, 1993; Pomp andBerger, 1995; Steffan-Dewenter et al., 2007). And third, the lack oftrade or taxation policies in the sector, which allowed smallholders tobenefit from some of the highest free-on-board prices globally(Akiyama and Nishio, 1996; Ruf and Yoddang, 1996; Neilson, 2007).

Past analysts have emphasized that forest access, in-migration andmarket opportunity intersected with diverse socio-spatial and culturalrelations to shape the course of cacao expansion (Durand, 1995;Neilson, 2007). The experience of many Buginese individuals as mi-grants to agricultural frontiers elsewhere, for example, has commonlybeen noted, as has the relative savvy and experience of Buginese peoplewith trading, commodity production and Green Revolution technolo-gies (Pomp and Berger, 1995; Ruf and Yoddang, 2004). These analyses,however, have often treated such relations as add-ons to the primaryanalytical frame of forest rent. Much like more positivist or abstractingmodels of LUCC, local historical, social and spatial relations generallyappear as contingencies rather than as central components of the story(Leiter and Harding, 2004).

Tania Li’s work complements these analyses by foregrounding theneed to understand market and governance conditions as co-con-stituted. In the Central Highlands, for example, and mirroring thefindings of Ruf et al. (2001) elsewhere, she shows that “rather thanhave the headman take their land from them and sell it off or allocate itto someone else, villagers have been tempted to sell land to migrantswhile they can” (Li, 2002: 428). Li’s analysis also powerfully capturesthe “everyday” processes of accumulation and dispossession (Hall et al.,2011: 145) that squeezed many people off of ancestral land holdings,shaping uneven pathways of expansion and market incorporation (Li,2014). Successful crop adoption for some people has hinged on theexclusion of others, often in ways that have shifted land ownershipaway from indigenous land claimants.

Li’s work on its own, however, insufficiently explains variability inexpansion and LUCC for two reasons. First, Li’s work focuses on thedifferent conceptualizations of land, territory and modernity whichhave catalyzed cacao adoptions; associated LUCC pathways, or themanifestation of these dynamics across broader spatial scales, are not afocus of her analysis. Second, Li’s analysis has largely focused on anindigenous highlander frontier where the autonomy of resource users“was only slightly modified by state rule and taxation” (2014: 7). Suchregions are distinct from lowland regions where most cacao productionexpanded and which are generally characterized by much longer andmore direct histories of state engagement (Li, 2005; Scott, 2009). This isparticularly true in Indonesia where the decades immediately precedingcacao expansion saw tremendous upheaval associated with the estab-lishment of Suharto’s authoritarian New Order regime (1967–1998).

Throughout the 1960s, the New Order regime built army basesacross lowland regions in an effort to control various insurgencies (deKoninck, 1996; Peluso and Vandergeest, 2011). By means of the 1967Basic Forestry Law, the state also claimed as property thousands ofhectares of lowland forests, croplands and grasslands. This law, in-formed by European ideas of eminent domain introduced under Dutchcolonialism, granted the newly independent Indonesian Government dejure rights over all lands perceived to be ‘un-used’ or ‘under-utilized’(Vandergeest and Peluso, 1995; Peluso and Vandergeest, 2001). Thedesignation of state forests provided the juridical basis for colonizationschemes, agricultural development initiatives and various forest in-dustries (Booth, 1992; McCarthy et al., 2012). Land use prohibitionswithin state forests also recast customary practices as “criminal prac-tices in huge chunks of the rural landscape” (Peluso and Vandergeest,2001: 766).

In reconstructing and explaining cacao expansion and LUCC below,I am guided by two specific objectives. The first is to reconsider the

1 From a high of 1132 metric tons/hectare in 2002, average yields in thesector had dropped to 386 metric tons/hectare by 2016 (FAOSTAT, 2018).

L.C. Kelley Geoforum 97 (2018) 22–34

23

linkages between cacao expansion and LUCC with greater attention tovariability than has been possible in studies reliant on secondary dataor localized observations (e.g., Ruf and Yoddang, 1999; Erasmi et al.,2004; Steffan-Dewenter et al., 2007). These methods impose variousselection biases, particularly as past work has focused on known sites ormoments of rapid expansion. A longer term and more spatially robustanalysis of cacao expansion and LUCC linkages can nuance an under-standing of how cacao expansion has influenced landscape change,particularly in the context of recent production crisis. Such analysis isespecially important in light of evidence that smallholder cacao plant-ings also facilitated revegetation in Sulawesi (Ruf and Zadi, 1998;Kelley, 2017).

My second objective is to document the social, spatial and historicalrelations that have mediated variability in cacao expansion and LUCCin Southeast Sulawesi. Guided by existing work and initial researchfindings, my analysis focuses on three policies under-explored in prioraccounts of expansion:

(i) The Ministry of Forestry’s declaration of state claims over600,000 ha of land for production forestry, limited productionforestry and conservation from 1967 onwards (BPS, 2017);

(ii) The sponsored resettlement of over 260,000 people from elsewherein Indonesia through the Transmigration Program from the 1960sonward (Departemen Tenaga Kerja dan Transmigrasi, 2015); and

(iii) Programs to promote tree crop development from the 1980s on-wards, including the Rejuvenation and Rehabilitation of ExportCrops Program (late 1980s) and the Sulawesi Rain-Fed AgriculturalDevelopment Program (1998–2004) (Asian Development Bank,2004).

3. Bridging remote sensing and political ecology to produce acritical physical geography of landscape change

To address the above research objectives, this paper integrates ap-proaches to mapping LUCC from RS with theories and methods from PEand CPG. This section briefly introduces how these approaches facil-itate an analysis of variability in commodity crop expansion and LUCC,particularly when used in conjunction.

3.1. LUCC mapping in smallholder commodity frontiers

Smallholder tree crop frontiers are challenging to map for a numberof reasons. Smallholder tree crop plantings are often produced in plotsof land smaller than the resolution of openly accessible satellite data(e.g., MODIS), with many tree crop plantings closely resembling ad-jacent patches of forested or agro-forested land in visible/near-infraredwavelengths (Kelley et al., 2018). Many parts of the tropics where cacaoproduction is concentrated are also characterized by rugged topo-graphy, chronic cloud cover and high overall rates of inter-annual LUCC(Broich et al., 2011; Hansen et al., 2013). These dynamics limit theaccuracy and reliability of LUCC maps at the scales necessary to capturevariability in crop expansion, particularly given the cost and difficultyof acquiring and processing satellite imagery historically.

Recent developments within RS address many of these challenges,however, enabling increasingly accurate analyses of commodity cropexpansion and LUCC across broad temporal and spatial scales.Developments important to this study include open access to Landsatimagery (Wulder et al., 2012); the emergence of cloud-based computingplatforms (Gorelick et al., 2017); and the application of “ensemble”learning algorithms to difficult classification challenges (Belgiu andDrăguţ, 2016; Hurni and Fox, 2018). The release of the Landsat satelliteimagery archive is particularly critical with respect to an analysis ofvariability because it has enabled no-cost access to repeat satellite ob-servations at most locations on Earth’s surface at 30–60m resolutionfrom 1972-onwards. This, coupled to improved algorithmic capabilitiesand cloud-based techniques for processing and manipulating RS

imagery, enables regional, national and even global analyses of LUCC(Hansen et al., 2013; Xiong et al., 2016).

While these advancements provide a stronger basis for re-constructing variability in cacao-LUCC linkages, RS data are of limitedvalue in isolation. Despite the growing accessibility of historical sa-tellite imagery, for example, much available imagery is coarse in spatialresolution, impeding nuanced analysis without supplementary data tocontextualize, validate or interpret observed changes. RS data are alsoof minimal use in policy and management decisions where abstractedfrom the complex on-the-ground relations informing LUCC (Denniset al., 2005; Goldstein, 2014). Some critics suggest that a reliance onmodeling correlates of LUCC with available secondary datasets has attimes led analysts of RS data to reproduce simplistic tropes surroundingenvironmental change and/or misinterpret underlying causal dynamics(Robbins, 2001; Turner, 2003).

3.2. Political ecologies and critical physical geographies of landscapechange

The above perspectives highlight the importance of integratingtechniques from RS with reflexive and embedded analytical ap-proaches. Work in the field of PE and CPG is useful in this regard be-cause it begins from the assumption that commodity markets are deeplyembedded in local social and environmental relations (Polanyi, 1957);as such, there is no ex ante reason to expect that different locations willexperience similar change trajectories (Bebbington, 2001). This shapesan analytical emphasis on how particular causal dynamics are con-stituted and come to function rather than on how much particular dri-vers of change matter, per se, supplementing more quantitative orcorrelative assessments of change (Blakie and Brookfield, 1987;Robbins, 2011). It also informs an emphasis on historically-oriented,field intensive research methods capable of grounding and con-textualizing RS data (Fairhead and Leach, 1996; Arce-Nazario, 2007;Lukas, 2014).

Relevant to the case of cacao expansion, work in PE and CPG hascommonly focused on how the politics of access to and control overland influence agrarian land use and livelihood practices (Ribot andPeluso, 2003; Peluso and Lund, 2011). While rarely coupled to ananalysis of LUCC, this work has shown how state development regimesconstructed some regions’ “comparative advantage” for crop expansionin the post-WWII period, introducing colonization schemes, land useconcessions and targeted crop development initiatives (Hecht, 1985;Hirsch, 1989; de Koninck, 1996; Hall et al., 2011). Important to ananalysis of variability, this work also consistently theorizes states andstate land claims as uneven, contested, and produced through on-the-ground relations (Vandergeest and Peluso, 1995; Corson, 2011).

Through attention to a diversity of economy-society-state relations,past work in PE has also commonly explained the spatially and tem-porally uneven dimensions of commodity expansion (Vandergeest et al.,1999; Hall, 2004; Sikor et al., 2005; McCarthy, 2010). Drawing on acomparative case study, for example, Vandergeest et al. (1999) showhow spatially uneven processes of shrimp aquaculture expansion inSouthern Thailand relate to the timing of expansion and uneven stateregulations and territorialisation processes. McCarthy (2010) explainsvariegation in oil palm expansion in Kalimantan, Indonesia as a resultof roads, topography, and the idiosyncratic contingencies of earlierstate-agribusiness interventions. These analyses reiterate the wayscommodity booms also shape social difference, driving favorable aswell as “adverse” forms of market incorporation (du Toit and Hickey,2007).

Finally, this paper is closely inspired by recent efforts to integratePE and RS to produce a CPG of landscape change (Lukas, 2014; Galvan-Miyoshi et al., 2015; Arce-Nazario, 2016). Work in CPG responds to thecritique that PE has left behind its ecological basis too fully (Walker,2005), losing ground in “the real states of nature and the scientificdebates which surround them” (Blaikie, 1999: 134). Work in CPG

L.C. Kelley Geoforum 97 (2018) 22–34

24

maintains an emphasis on the multi-stranded power relations struc-turing human-environmental change while drawing on techniques fromthe environmental sciences, including RS, that enable a more explicitassessment of environmental changes and conditions (Lave et al., 2014;Lave et al., 2018). Such analyses have enriched understandings basedsolely on RS or PE, unsettling, for example, the apparent stability ofcartographic categories (e.g., Robbins, 2001) and assumed but un-examined linkages between state engagement and landscape change(e.g., Lukas, 2014).

4. Methods

This study adopts an embedded and comparative case study ap-proach. I begin by reconstructing trajectories of LUCC in the districts ofKolaka and Konawe. I then focus on those dynamics influencing cacaoexpansion and LUCC in four lowland villages selected for comparativeanalysis. The following section introduces the study area and describesall data, most of which was collected over a fourteen-month periodspent living in the area (05–07/2012, 08/2014–2015).

4.1. Study area and site selection

Kolaka and Konawe districts of Southeast Sulawesi province (Fig. 1)are characterized by a wet–dry climate and encompass a variety ofhabitats, including peat swamp, lowland, montane and karst forest,mangroves, grasslands and diverse agricultural lands (Whitten et al.,1987). This landscape diversity has long fostered a wide variety of landuse practices. Early accounts suggest that Tolaki people, indigenous tothe region, historically engaged in swidden or shifting hillside riceproduction, rain-fed lowland rice production, swamp fishing, buffaloraising and hunting, sago palm cultivation and forest product collection(Tarimana, 1989; de Jong, 2011). From the mid-1960s onward, manyagrarian and forest land uses were re-oriented around the production ofirrigated wet rice production and export-oriented tree crops such ascacao, cashew and coconut (Direktorat Jenderal Perkebunan, 2014).

Following initial RS analyses, I selected four lowland villages forfurther analysis by triangulating across multiple lines of evidence (lit-erature review, initial site visits and interviews, secondary data andinitial remote sensing analyses) (Table 1). These villages share im-portant similarities: all have similar lowland forest ecologies, longhistories of Tolaki settlement and land use, and numerous householdswhich have re-organized their livelihood strategies around cacao pro-duction at some point in the past three decades. They also providegenerative contrasts, as each village is characterized by a differenthistory of cacao expansion and state engagement. This allows me, forexample, to analyze the early demarcation of state forests in one villagevs. the removal of use restrictions in another while noting similarities inthe orientation of forest policy enforcement in both.

As the brief settlement histories in Table 1 capture, all four villageshave also been shaped by various political reconfigurations and asso-ciated forms of political violence over the past century. These includethe onset and elaboration of Dutch colonial rule (1907–1942); a briefbut violent occupation by the Japanese army during the Second WorldWar (1942–1945); political violence and warfare associated with theDarul Islam Indonesia/Tentara Indonesia (DI/TII) insurgency(1950s–1960s); and the formal establishment of Southeast Sulawesiprovince in 1964 under the newly established Indonesian government,then led by Sukarno (1945–1967). Suharto's New Order regime(1967–1998) assumed control of the government shortly thereafter,remaining in power until overthrown amidst economic and politicalcrisis in the late 1990s (Potter and Lee, 1998; de Jong, 2011). After thefall of the Suharto dictatorship in 1998, President Habibie initiated thedecentralization of many decisions surrounding natural resource man-agement and governance from central to provincial and district levels.

4.2. Reconstructing pathways of cacao expansion and LUCC

This study approximates past cacao-LUCC linkages by analyzing theextent to which current production overlaps with prior forest cover loss,informed by recent analyses by Abood et al. (2015) and Gaveau et al.(2014) in the oil palm, mining and timber economies. I began bycompositing roughly 200 individual Landsat 8 scenes (30m) into asingle gap-free image for the year 2014, following compositing tech-niques elaborated in Kelley (2017). I then trained this image withinGoogle Earth Engine (Gorelick et al., 2017), a cloud-based computingplatform that facilitates automated cloud-masking and filtering. Totrain land cover classes, I used over 500 field-collected data points anda Random Forest “ensemble” classifier (1000 iterations), a learningalgorithm which averages across multiple decision trees to ensure re-latively unbiased results (Belgiu and Drăguţ, 2016).

The final analysis maps current cacao production against existingdatasets on forest cover loss over five time periods: 1972–1995,1995–2000, 2000–2005, 2005–2010, and 2010–2014 (Hansen et al.,2013; Kelley, 2017). These analyses, importantly, do not capturewhether or not all land that was deforested was immediately broughtinto production, do not indicate those locations where cacao has beenbrought into production only to be abandoned, and do not reveal thestatus of converted lands. To assess the validity of overlap analyses, andto gain insight on how best to interpret LUCC shifts, I integrated LUCCmapping with (i) geo-located oral historical and survey data on land useand conversions from village case studies, detailed below and (ii) offi-cial government data on cacao production at the sub-district level(Direktorat Jenderal Perkebunan, 2014). Throughout, I also draw onsupplemental datasets on elevation, soil distribution, and in-migrationto help explain observed patterns (Table 2).

Fig. 1. Study Area. Map of Indonesia with Southeast Sulawesi denoted in red and the administrative districts of Kolaka and Konawe (the site of the four village casestudies and the focus of subsequent regional analyses) denoted with a black box.

L.C. Kelley Geoforum 97 (2018) 22–34

25

4.3. Explaining observed cacao expansion and LUCC

To reconstruct processes of cacao expansion at both village anddistrict scales, I relied on oral histories, in-depth interviews andhousehold surveys. I began by conducting oral histories with approxi-mately 25 individuals in each of the four case study villages, roughlysplit between older Tolaki inhabitants and earlier in-migrants.Respondents were asked a set of questions about their personal historiesfollowed by open-ended discussion about their lives and key events.Various open-ended questions were also asked to understand villageand settlement history, including questions surrounding key events (e.g.moments of significant in- or out-migration). The goal of these historieswas to better understand the social, environmental and political con-texts in which cacao expansion had occurred, as well as to identifyconstitutive differences in crop expansion across the four villages.

To better understand the design and implementation of policies andprograms captured by oral histories, I conducted in-depth interviewswith government officials at the district and provincial levels. Key

informant interviews were designed to understand the motivations forspecific programs as well as their extent and importance at the regionalscale of analysis. In total, I conducted 36 in-depth key informant in-terviews, mostly with district- and provincial-level officials from theDirectorate of Estate Crops, the Forestry Department, and the PlanningDepartment. Whenever possible, both oral histories and in-depth in-terviews were conducted during field visits, allowing for qualitativefindings to be cross-referenced with LUCC maps. Respondents for oralhistories and in-depth interviews were selected using targeted subjectsand referral sampling, and most oral histories and interviews weretaped and transcribed.

Finally, I used household surveys to reconstruct trajectories of cacaoexpansion and LUCC in each village. Surveys were randomized with50–60 households surveyed in each village (N=207 in total).Household surveys detailed all land holdings, including (i) year andmode of acquisition; (ii) prior land uses and covers; (iii) the tenurialclaims associated with land; (iv) its location within the village andgeneral quality; (v) price of purchase, if applicable and (vi) prior

Table 1Overview of the four villages selected for comparative case study analysis.

1a. Agro-Ecological and Demographic Overview

Polia Taowengga Unahoa Besulutu

District Kolaka Kolaka Konawe KonawePopulation (2012) 4202 948 769 731Area (km2) 72.32 28.73 13.87 9.91Pop. Density (ppl/km2) 58 33 55 74% Bugis In-Migrants 96% 34% 26% 55%First Bugis In-Migration 1970s 1970s 1980s 1980s% In-Migrants 100% 66% 43% 57%Min-Max (Avg) Elevation (m) 56–725 (305) 55–693 (216) 31–421 (170) 25–67 (50)% Alluvial Soils 0% 0% 36% 40%

1b. Brief Settlement Histories

Polia is a Tolaki settlement predating 1900.During the DI/TII insurgency in the late1950s, settlers were relocated to a nearbyrural town. Tolaki land claimants beganto return by the late 1970s but wereforcibly resettled into what is nowTaowengga village. Though some Tolakipeople still claim land in the area, Poliawas predominantly re-occupied by Bugisin-migrants from the 1970s onward, whowere granted land by a Buginese ex-armygeneral in the region

Taowengga is a Tolaki settlement alsopredating 1900 and adjoining Polia.People from Taowengga were alsorelocated to a rural town during the late1970s, and were resettled alongsideTolaki land claimants from Polia whenreturning to the area. Through the mid-1970s, Bugis in-migrants as well asJavanese transmigrants were also settledin the area through resettlement schemessponsored by village and districtofficials.

Unahoa village was originally the site oftwo Tolaki settlements. Unlike the othervillages, Unahoa was a key site ofadministrative activity during both Dutchand Japanese occupations and was a keyarmy base for the Indonesian Governmentduring the DI/TII insurgency. Followingthe insurgency, households from bothTolaki settlements were permanentlyresettled in lowland regions, withtransmigration into the area beginning inthe mid-1970s.

Besulutu has intermittently been usedfor swidden agriculture throughout thepast century. Several familiespermanently re-settled Besulutu in 1958while fleeing violence associated withthe DI/TII insurgency to the west.Through the 1960 and 1970s, the firstsettlers were joined by families fromnearby areas also fleeing violence. Aresettlement program in 1978 alsosettled 200 Javanese households in thearea. Lacking irrigation, these familiesleft by the late 1980s.

An overview of basic demographic and agro-ecological characteristics of the four village case studies, and a brief overview of each area's settlement history pre-cacaoexpansion. Pseudonyms are used for all villages. Sources: household surveys, oral histories and supplemental datasets from the Shuttle Radar Topography Mission andCannon et al., 2007.

Table 2Supplemental Data Sources.

Variable Spatial Res. Yrs. Available Source

Forest Cover 60m 1972, 1995, 2015 Kelley (2017)Forest Cover Loss 60m 1972–1995, 1995–2000 Kelley (2017)Forest Cover Loss 15m 2000–2014 Hansen et al. (2013)Area Cacao District, Sub-District 1979–2013 Direktorat Jenderal PerkebunanSoils 15m 2014 Cannon et al. (2007)Elevation 15m 2014 Shuttle Radar Topography MissionOverall In-Migration District 1980–2010 Ruggles et al. (2017)Transmigration District 1968–2010 Dept. Tenaga Kerja dan Transmigrasi

This table describes assessed variables and data sources used in the following discussion of cacao expansion and forest cover loss. Data on in-migration were drawn from two sources, enabling in-migration from South Sulawesi, the key site of Bugis in-migrants, to be isolated fromformal transmigration initiatives.

L.C. Kelley Geoforum 97 (2018) 22–34

26

support for production on the lands in question. Household survey dataare the basis for the reconstructed patterns of cacao expansion pre-sented in Fig. 4 and provide supplemental information on land accessand control throughout. A proportionally smaller number of householdswere surveyed in Polia; I account for this by weighting observed landconversions accordingly.

5. Results

5.1. Reconstructing trajectories of cacao expansion and LUCC

Previous investigations show that cacao production in SoutheastSulawesi began in 1979 and accelerated through the early 2000s. Mostcacao plantings have followed existing roads, concentrated in pocketsof rich alluvial soil, and many have been managed in either sparseshade or monotype on holdings under two hectares in size. Since theoutbreak of pest and pathogens, most cacao has also been managedwith heavy reliance on fertilizers, pesticides and herbicides. Despiteregional consistencies in production, however, the timing, extent andeffects of expansion have been highly variable, even within areas ofsimilar road proximity, in-migration and agro-ecological advantage.

Table 3 contrasts key facets of cacao expansion in the four studyvillages. Despite enabling conditions in each location, cacao expansionwas most rapid in Polia, beginning in the 1980s and acceleratingthrough the early 1990s. In contrast, cacao expansion did not accelerateuntil the mid-2000s in Taowengga and Unahoa or until the late-1990sin Besulutu. Fewer people also adopted the crop in these latter threevillages and more people have exited the sector since initial adoptions.Today, while 94% of households Polia continue to rely on cacao forincome, only 58%, 43% and 61% do in Taowengga, Unahoa and Be-sulutu, respectively. Regionally, 48.6% of cacao production is located injust ten of 54 villages within the top four producing sub-districts, apattern that bears little relationship to associated soils or topography(Supplemental Materials 1 and 2).

Fig. 2a and b connect these observations to broader LUCC dynamics.Fig. 2a presents the results of overlap analyses for two areas locatedroughly 25 miles from each other, both spanning the primary roadnetwork constructed in the early 1900s and both focal sites of statecolonization initiatives in the 1970s. The top image, capturing severalsub-districts surrounding Polia and Taowengga, documents the closeconnections between cacao and forest cover loss between 1972 and2000 before a general decoupling of these linkages between 2000 and2015. The lower image, covering several sub-districts surroundingUnahoa, captures only muted levels of cacao expansion into forestedlands either before or after 2000. It also reveals the extent of cacaoplantings in areas of< 25% tree cover. Fig. 2b highlights how thesetrends manifested in temporally uneven patterns of expansion in ag-gregate, particularly over the period 1995–2000.

Land use genealogies and oral histories build on RS data by cap-turing changes difficult to detect through mapping alone. Oral histories,for example, suggest that many lands converted into cacao from<25%tree cover were previously managed as grasslands, settlements andswidden rice fields. They also suggest that the extraction of timber andrattan was common in forested lands in the decades preceding cacao

expansion. Land use genealogies highlight how the fallowing of formerswidden settlements and fields contributed to revegetation in manylandscapes, in some cases, producing the secondary forest later con-verted to cacao (Fig. 3). Genealogies also reveal the diverse processesdriving forest cover loss outside the cacao sector, including the estab-lishment of forest industries, irrigated wet rice fields and, in recentyears, a growing number of corporate nickel mining and oil palm es-tates.

These findings suggest that rather than being an exception, varia-tion in cacao expansion and associated LUCC has been constitutive ofregional trends over the past four decades. This is also true with respectto smallholder livelihoods. Those individuals that had preferential ac-cess to land or capital in early decades of expansion have generally

Table 3Cacao adoption and management in the four village case studies.

Polia Taowengga Unahoa Besulutu

First Adoption of Cacao 1985 1986 1989 1985Peak Cacao Adoptions 1997–1998 2003–2004 2003–2004 1999–2000% of HHs Adopting Cacao 94% 64% 54% 73%% of HHs Still Producing Cacao (2015) 94% 58% 43% 61%Avg. Cacao Holding (2015) (ha) 2.02 0.92 0.91 1.35Avg. Land Managed (2015) (ha) 3.00 1.55 1.76 2.05

Fig. 2. Cacao and LUCC, 1972–2014. (a) (Above) captures two areas of roughly625 square kilometers, highlighting uneven pathways of cacao expansion andforest cover loss in two areas located only 25 km apart. The top image, spanningseveral sub-districts in Kolaka, documents close connections between cacao anddeforestation, particularly pre-2000. The lower image, spanning severalneighboring sub-districts in Konawe, captures the extent of deforestation out-side the smallholder cacao economy. Both images also highlight the spatially-disparate but meaningful extent of cacao plantings inside areas of< 25% treecover. Overlap analyses are overlaid on recent satellite imagery with dark greendenoting forest. (b) (Below) presents data on overlap in aggregate, illustratingthe extent of cacao expansion into the forest from 1995 to 2000, as well as therelative concentration of production in Kolaka.

L.C. Kelley Geoforum 97 (2018) 22–34

27

accumulated wealth and land, particularly if they were already tradingcacao against the dollar when the Rupiah collapsed during the1997–1998 Asian Financial Crisis. Many other individuals, however,have experienced the formation of cacao markets adversely, including,for example, people who adopted the crop shortly after productioncrisis began, eventually abandoning their fields amid rising debt.Attrition in the sector and land consolidation continues to inform dis-parities in land access across producing vs. non-producing households(Polia: 3.04 ha vs. 1.20 ha; Taowengga: 1.96 ha vs. 1.10 ha; Unahoa:3.40 ha vs. 1.72 ha; Besulutu: 3.63 ha vs. 1.27 ha).

As the following two sections detail, uneven processes of cacaoexpansion and landscape transformation have been closely intertwinedwith and constitutive of uneven processes of smallholder market in-corporation. They have also been shaped throughout by the influence ofland use policies and politics. The next section elaborates this argumentby providing a general overview of the several decades preceding cacaoexpansion, highlighting how state land claims shaped the availability ofthose factors centered in the forest rent model: forests, in-migrants and

markets. It also highlights the uneven ways state land claims and as-sociated development initiatives reshaped the politics of access to landand capital, a dynamic central to later variability in cacao expansionand LUCC.

5.2. Priming the region for cacao expansion

In the decades preceding cacao expansion, Southeast Sulawesi’sforests were commonly inhabited by Tolaki people in settlements or-ganized around kin groups practicing rotational swidden rice produc-tion. Within this system, land was typically cultivated for a period oftwo to three years before being fallowed again. Diverse “inscriptions”(Bailey and Bryant, 1997) marked Tolaki claims in the landscape:groves of sago palms planted near settlements; managed fruit, teak andbamboo species used to delineate swidden boundaries; producedgrasslands used for hunting and buffalo grazing; and secondary re-growth in recently fallowed rice fields. Beginning in the late 1950s,however, Tolaki settlement and land use practices were displaced in asequence of events that would ultimately “free” many inscribed land-scapes for later conversion to cacao.

The initial displacement of Tolaki people was shaped by conflictfollowing the arrival of DI/TII fighters aiming to consolidate territory fora planned Indonesian Islamic State (Negara Islam Indonesia) in the late1950s and early 1960s. Interviews suggest that Tolaki settlements wereoften spontaneously abandoned as people fled violence, property de-struction and risk of theft or rape. Forced relocations were also common.Army divisions sent to control the insurgency often resettled Tolakipeople in nearby towns to cut DI/TII fighters off from food and suppliesin hillside settlements (Lukas, 2014). Shortly after the re-establishment ofpeace in the mid-1960s, these relocations provided the basis for deemingprior settlements and land holdings ‘under-utilized’ or ‘unused,’ and assuch, property of the state (Peluso and Vandergeest, 2011).

State land claims encompassed much of the landscape. However, thecapacity of state actors and institutions to realize and enforce theseclaims varied considerably, shaped by idiosyncracies of prior settle-ment, the specificity of local resistances, and the exact configuration ofinstitutions, actors and interests present. Besulutu, for example, was farremoved from main roads, re-opened in the early 1960s by Tolakipeople fleeing DI/TII violence that arrived in the area by raft. No of-ficials from the Forestry Department would visit Besulutu until 1981,leaving the governance of the forest in the hands of local leaders. As thevillage head remembers telling inhabitants, “If you don’t inhabit theland [i.e. if you continue to farm the forest], go ahead and depart, otherpeople will be found to inhabit it again.” These threats, however, weregenerally only modestly effective.

State capacity to enforce land claims tended to be stronger in siteslocated along the main road, including Polia, Taowengga and Unahoa.Even within such sites, however, forest governance was highly variable.Tolaki swiddeners trying to re-enter Polia and Taowengga following tenyears of forced resettlement were at first turned away by district andsub-district officials. By marching in a group of over 100 through thenight, however, these individuals successfully forced the hand of thesub-district head, re-asserting control over much grassland and someforested land in the area. Many of these individuals also successfullyestablished swidden rice fields despite prohibitions, locating their plotsfar from other swiddeners to diminish the threat of arrest. By 1975 inUnahoa, in contrast, district police had burned the swidden fields andtemporary dwellings (rumah kebun) of Tolaki swiddeners to force theirreturn to lowland settlements.

Despite distinctions in their realization, state claims over land werefoundational to the course of cacao expansion for several reasons. First,they eroded prior claims over land by asserting new property regimesand by facilitating regrowth and clearance in former settlements andfields. Second, they made state actors and institutions the key arbiter ofland transactions, shifting the politics of access to land in ways thatboth impeded and enabled cacao expansion in later years. Third, state

Fig. 3. Former Swidden Landscapes from Above. This image overlays GPS pointdata collected during roughly a half-day’s walk from the edge of Unahoa’scurrent settlement area (denoted by the points to the far right) into hillsideforests formerly used for swidden rice under a rotational production system.Many households planted fruit trees such as durian or rambutan to mark theirland claims, and to supplement household consumption. The presence of manyplanted fruit trees between 50 and 100 years old throughout the forest, in-dicated by black dots, provides evidence of these claims, and of former swiddenagricultural production in what are now densely vegetated secondary forestlands.

Fig. 2. (continued)

L.C. Kelley Geoforum 97 (2018) 22–34

28

land claims served as the basis for the establishment and elaboration offorest industries and racialized settlement programs designed to po-pulate “backwards” areas with commodity-oriented lowland peoples.These developments worked alongside direct prohibitions on swiddenagriculture to reinforce ideological distinctions between agriculture andthe forest, thwarting other visions of land use and society.

Table 4 reconstructs in-migration relative to total population in theregion from 1968 to 2010, capturing the extent to which in-migrationreconfigured rural demographics. Oral histories suggest colonizationschemes often enabled further in-migration, in part by motivating theconstruction of roads and irrigation networks. Colonization schemesalso drove significant landscape change. In Besulutu, over 100 familiesfrom flooded areas and failed transmigration schemes were resettled inforested lands between 1978 and 1981 by the Social Welfare Division.In and around Polia, Unahoa and Taowengga, hundreds of hectares offormer Tolaki sago palm groves were converted into irrigated wet riceproduction over the period 1975–1984 through repeated transmigra-tion initiatives. The infusion of labor into rural areas simultaneouslyaccelerated timber and rattan harvest in nearby areas.

Land use policies and politics pre-dating initial cacao expansion in1979 thus not only displaced prior claimants from many forestedlandscapes. They also brokered and extended social and market rela-tions integral to later cacao expansion, a point I revisit in greater spe-cificity below. These dynamics reiterate the impossibility of fully un-derstanding cacao expansion in the reductionist terms of profitmaximization or forest availability. Much as cacao expansion was notreducible to agronomic conditions, however, neither was it reducible toland use policies. The next section draws on oral histories, in-depthinterviews and household surveys to connect the more general historytraced here with a more detailed reconstruction of cacao expansion ineach of the four villages.

Drawing on these data, below, I argue that while many factorsplayed a role in shaping variability in cacao expansion pathways, twodynamics were most central: (i) the uneven politics of smallholder ac-cess to land and capital and (ii) the labor and capital intensity of cacaoproduction at the time growers gained access to land.

5.3. Variable pathways of expansion into forested lands

By the late 1980s, the Plantation Department was pursuing specificefforts to develop the region’s cacao industry, consistent with a broadernational thrust to diversify government export revenues away from oilproduction (McCarthy et al., 2012). Durand (1995) has shown thatthese programs had the effect of catalyzing adoptions through a com-bination of low-interest loans, hybrid “F1” seedlings, and advice onproduction in early stages of crop establishment. I found that the Re-juvenation and Rehabilitation of Export Crops Program (Peremajaan

dan Rehabilitasi Tanaman Ekspor, PRTE), which distributed cacaoseedlings in three of the four case study villages, had little effect oncacao expansion. As Ruf and Yoddang (2004) and Neilson (2007) havehighlighted, this was partly due to a lack of state capacity. More fun-damentally, however, it reflected the persistence of state claims overforested land and the growing strength of the Forestry Department bythe late 1980s and 1990s.

Fig. 4 situates the following analysis with a detailed reconstructionof cacao adoption and LUCC in each of the four villages. As the diagramreflects, cacao expansion typically depended on a shift in the politics ofaccess to state-claimed forests. This occurred earliest in Polia, withgrowers gaining access to forested lands as early as the mid-1980s. Incontrast, people in Taowengga, Unahoa and Besulutu remained unableto access forested lands for cacao until the upheaval in forest controlthat accompanied the fall of the Suharto regime in the late 1990s.Drawing on data from each of the four villages, this section highlightshow these differences in timing worked alongside earlier legacies ofsmallholder capital accumulation, sustained efforts to promote cropexpansion, and the shifting ecology of production to shape uneven ex-pansion pathways.

5.3.1. PoliaPolia and Taowengga, originally part of the same settlement, were

both primed for cacao expansion by the early to mid-1980s. Over thepreceding decade, the ex-army general that governed the area had re-cruited and settled dozens of Bugis migrants, many of whom had familyrelations linking the area to established sites of cacao production inNorth Kolaka. Irrigation and road networks established during the1970s had also elaborated trade connections, while adjoining timberconcessions diminished the economic value of adjacent forests in waysthat would justify the allocation of these lands to smallholder produ-cers. By the mid-1980s, a 3000 ha cacao plantation had also been es-tablished in a neighboring sub-district, a development that worked inconjunction with the later PRTE initiative in Taowengga to increasepeoples’ access to knowledge and inputs for the crop.

In this context, the core factor that enabled rapid cacao expansion inPolia but not Taowengga was a letter of land clarification (SuratKeterangan Tanah, SKT) encompassing a broad tract of forested land inPolia. While the SKT was not commensurate with a land title and didnot fully extricate land from Ministry of Forestry claims, it conferredrelatively high tenurial security and provided the means for the villagehead to allocate lands to in-migrants and local elites who could affordassociated “measurement fees” (pembayaran ukuran) of 10,000 Rupiah(∼$0.75 in 2018 prices). As one Forestry Department official workingduring these years reflected, “Land claimants would say: why would theBupati [district head] have issued this if it wasn’t legal?” A first wave ofcacao adoptions quickly began in Polia.

Table 4In-migration to Southeast Sulawesi as proportion of total population, 1968–2010.

Kolaka Konawe

S. Sulawesi in-migrants Trans- migrants Total population S. Sulawesi in-migrants Trans- migrants Total population

1968–1975 7682 13,4641975–1980 16,611 0 97,284 6615 23,974 226,7451980–1985 13,670 13,208 139,892 7115 48,566 267,4041985–1990 44,954 0 176,534 10,348 2916 338,5361990–1995 1619 38211995–2000 18,150 1054 259,700 14,360 13,064 536,4202000–2005 7364 391 311,965 7839 2363 713,2352005–2010 13,090 707 367,390 12,180 3619 800,180

Total 113,839 24,661 367,390 58,457 111,787 800,180

This table presents all available data on in-migration from South Sulawesi, the key site from which Bugis in-migrants would have traveled, and formal governmenttransmigration programs from the late 1960s onward in both focal districts. The higher concentration of transmigrants into Konawe reflects the greater presence ofthe state in that region given its proximity to the provincial capital. Blank cells indicate those years for which migration data from either source is missing.

L.C. Kelley Geoforum 97 (2018) 22–34

29

One official from the district Forestry Department helped to situatethe SKT regionally:

Even though this was before Reformasi [the fall of Suharto and po-litical decentralization], it in fact happened that the district leaderwas brave enough to issue SKTs, even within the kawasan [stateforest]. Even today, if we talk with people from the PlantationDepartment, they say that the SKT has legal strength only one stepbelow a land certificate and that it’s reliable for however many years

moving forward, even though that land remains in the state forest.At that time, SKTs weren’t only issued in this area but in many areas.This only happened in Kolaka though, other district leaders wouldn’tissue these.

This history refutes an interpretation of the regional disparity incacao production captured in Fig. 2 as solely reflecting higher rates ofBugis in-migration or Kolaka’s relatively more favorable coastal cli-mate.

Fig. 4. Reconstructed Sequences of Cacao Expansion in the Four Case Study Villages. Reconstructs smallholder clearances for cacao from the early 1980s through 2014 ascaptured through a randomized household survey. These findings reiterate district-level LUCC mapping, highlighting strong variability in expansion pathways. Theyalso highlight the pivotal role shifting access to state forests played in enabling cacao expansion in each location.

L.C. Kelley Geoforum 97 (2018) 22–34

30

As the case of Polia suggests, the accumulation of land and capitalamong growers with early access to forested land often provided itsown engine of expansion, facilitating further in-migration and landclearance for cacao. In Polia, growers used the initial profits fromproduction to accumulate hectares of fallowed forest for later use, oftenpurchasing two hectare blocks (kaplings). Land was also transferred tofamily members using land sharing or bagi tanah2 practices or by sellingthe land for “family prices” (harga keluarga). In all of these ways, boomconditions tended to extend boom conditions. Accelerated land con-versions through the mid-1990s were also enabled by the pest andpathogen-free context in which growers planted the crop; growersplanting forested land in cacao joke that the only production input theyused during this time was “5-5,” their two hands.

Although the SKT enabled rapid cacao expansion, most individualsable to afford land through this scheme were Bugis migrants that hadarrived in the region with capital from elsewhere, typically from landsales in South Sulawesi or from work in Malaysia or elsewhere inIndonesia. Capital from outside the area was almost always necessarygiven the measurement fees charged; laborers logging state forestsduring these years, for example, made only 40 Rupiah/day. The spreadof cacao thus solidified the shift in land ownership away from Tolakipeople, cementing earlier processes of displacement. By the 1990s,Polia had split from Taowengga to become its own settlement, nearlyfully comprised of Bugis in-migrants with some connection to the cacaoeconomy.

5.3.2. Taowengga and UnahoaLike Polia and Taowengga, Unahoa was located along the primary

road network through the province, which by the mid-1980s was wellconnected to nearby towns and to a growing network of cacao traders.The areas surrounding Unahoa had also been the site of repeat colo-nization initiatives beginning in roughly 1978, with growers in Unahoagaining access to seedlings for cacao by the late 1980s as part of thesame smallholder tree crop development initiative that had operated inTaowengga. In both Taowengga and Unahoa, however, the persistenceof state claims over forested lands would largely suppress cacao adop-tions until the early 2000s. This, taken in conjunction with the shiftingecology of cacao production and the general economic hardship thatcharacterized this time period, depressed the magnitude of expansion inboth locations.

The strength of state land claims in both locations was long-standing, linked to earlier DI/TII occupations and villigization efforts aswell as the subsequent establishment of forest industries. Following amajor forest fire in the mid-1970s, most forests in Taowengga wereincluded within two successive pine reforestation schemes. The limitedcapacity of the Forest Department in Polia and Taowengga thus focusedon these lands, leading to the exclusion of Taowengga forests from theSKT issued for Polia. In Unahoa, state claims over land were initiallyenforced by the Police Department as part of an effort to ensure people’ssettlement in the lowlands. By the late 1980s, the Forestry Departmenthad also established a state water company in the foothills immediatelysurrounding Unahoa. Following the establishment of the water com-pany, the Forestry Department regularly policed state-claimed lands forsmallholder agriculture.

In this context, only modest cacao plantings took place in eitherlocation by the late 1990s despite the Plantation Department’s efforts topromote cacao (see Fig. 4). Mirroring what Padwe (2011) has observedof cashew tree plantings in Cambodia’s northeast highlands, most cacaoplantings before the 2000s were undertaken by Tolaki individualsplanting cacao seedlings in swidden rice fields and grasslands becausethey perceived tree crops to be less vulnerable to expropriation. These

initial plantings could nonetheless be destroyed, with planters subjectto arrest. Three people planting the crop in Taowengga in the 1990s hadtheir fields destroyed in a day that also resulted in their arrest. As oneofficial recounted: “I gave them an example that would be easy to un-derstand. I told them, the government has a farm (kebun); the moneyenters the government’s reserves the same way the money enters yourpocket when you plant cacao.”

Suharto's fall in 1998 was critical in enabling cacao expansion be-cause it inspired the assertion of decades-long grievances, shaping an“ambiguity of authority” over forested resources that persisted throughthe mid-2000s (Barr et al., 2002; McCarthy, 2004). In Taowengga,prolonged protests resulted in the arrest of the village head that hadsecured the SKT in Polia and led to the reoccupation and clearing offorested lands, including most lands previously targeted for reforesta-tion. A sub-district official from the Forestry Department subsequentlytraveled to the provincial capital to enroll Taowengga in a communityforestry initiative that secured 35-year lease rights for Tolaki claimantsin two-hectare plots. In Unahoa, village elites capitalized on Suharto’sfall in 1998 by selling 20 ha of forested land to a group of Bugis in-migrants with family connections in the area. District police confiscatedmigrants’ chainsaws within the first several days of land clearance.However, these transactions provided later basis for provisional landuse rights following a shift in the district leadership in 2004.

While these developments enabled access to the forested land oftenpreferred for cacao, the ecology and economics of cacao had eroded bythe late 1990s. While the Asian Financial Crisis produced windfallprofits for some, most individuals not yet engaged in export-orientedproduction experienced economic hardship. The latter resulted fromdiminished off-farm work opportunities, higher expenses for rice andother foods (Sunderlin, 1999; Breman and Wiradi, 2002), and in 1997,a severe drought followed by heavy La Niña rains (Ruf and Yoddang,2004). Cacao production was a difficult proposition for many in-dividuals: cacao trees do not yield for three to four years, and by thelate 1990s, pests and pathogens (the Cacao Pod Borer in particular)emerged in many areas. Many growers perceived that expensive syn-thetic inputs were necessary to ensure the success of their crop. Si-multaneously, the price of fertilizers rose by 100–200% between 1997and 1998 with the removal of Green Revolution-era subsidies (Ruf andYoddang, 1999).

Two dynamics informed the modest cacao expansion that occurredin both locations. First, cacao was planted by Tolaki individuals pri-marily as a means of solidifying prior claims over the land. In Unahoa,for example, perennial crops such as cacao were considered more ef-fective for securing use rights than were seasonal vegetable plantings.This perception was linked to Forestry Department concerns sur-rounding the effects of tree clearance on hillside erosion and watercompany operations. Second, cacao was planted by Bugis in-migrantswho used the sale of established cacao fields to finance expansion intoless established (and expensive) land markets. One family, for example,used profits from the sale of 2 ha of productive cacao in Polia to financethe acquisition of 12 ha of unplanted land in Unahoa. Not everyone waswilling to do this, however, given sustained conflict and confusion overthe tenurial status of forested lands and fields.

5.3.3. BesulutuBesulutu is interesting in relation to the other three villages because

it provides a context in which the enforcement of state claims over landwas relatively weak. Because earlier colonization schemes and forestindustries were not as successfully established in Besulutu as in theother villages, access to irrigation, roads and markets also remainedlimited. These conditions deterred cacao expansion, even after a locallandlord with connections to the provincial government began to pro-mote cacao expansion through a land sharing scheme in 1997, makingpermanent access to land more secure. As one of the first migrants thatplanted cacao through this scheme remembered: “The people hereweren’t interested in tree crop production when we arrived, they were

2 Through this practice, if four hectares of land are opened, two of theseremain with the claimant after production is established and two with theperson or family that opened the land.

L.C. Kelley Geoforum 97 (2018) 22–34

31

still looking for what could be eaten right away (i.e. short term treecrops or money from off-farm work). This was the time of krismon [theAsian Financial Crisis].”

A key facilitator of more rapid cacao expansion emerged two yearslater with the Sulawesi Rain-Fed Agricultural Development Program(SRADP). Between 1998 and 2004, SRADP operated on some 52,000 hathroughout the province (Asian Development Bank, 2004). The pro-gram represented a post-decentralization shift in development politicsaround non-state actors and was implemented through a partnershipbetween the Asian Development Bank and the provincial PlantationDepartment. As such, it provided a staggering $43.8 million dollarbudget to support the expansion of cacao, peppercorn and coconut,then the most important estate crops in the region. While this programgreatly exceeded the forms of support for agricultural developmentmade available during Suharto years, it shared a similar goal. This wasto further boost commodity crop production in lowland areas, sup-planting purportedly destructive swidden farming systems while“freeing” upland forests for conservation (Asian Development Bank,2004).

Besulutu was selected for participation in SRADP because com-peting Forestry Department claims were weak in the area and becausean employee of the Plantation Department had married in Besulutu,subsequently enlisting the village in SRADP. SRADP provided an initial250,000 Rupiah to compensate growers for opening forested lands forproduction. It also provided title to the land, salary for planting the landin cacao, and seedlings, fertilizers and pesticides. The availability ofSRADP support encouraged further Buginese in-migration, includingamong people learning of the program from their kin already estab-lished in the area. Program supports also triggered the return of Tolakiinhabitants who had previously abandoned the area; many returneesdid so to gain title to previously managed lands. As one woman re-membered, “People were happy again to manage the lands themselves.It was guaranteed support. People were paid by the government to doit.”

6. Discussion

Cacao and deforestation have often been seen as inextricably linkedthroughout history. As Ruf and Clough (2004: 108) write, for example:

Where [cacao boom and bust cycles] started, they led to the openingup of new forests, sometimes at a tremendous speed. Where theyended, they left behind, in the best cases, disease-infested groves oflow productivity in a secondary forest environment but often onlypoor fallows and pastures.

Findings from this study push back against such a declensionistinterpretation by highlighting the diverse environmental changes as-sociated with cacao. They also recast the implicit politics of such areading, showing that much of what has looked like deforestation forcacao has had as much or more to do with the politics of access to landand capital.

Cacao produced in Southeast Sulawesi from the 1970s onward wasnot a crop which required forested land; it was an opportunity thatemerged in the wake of forced displacements and uneven patterns ofsmallholder access to land and capital. From roughly the 1950s throughthe 1980s, state land claims and forced evacuations of indigenousTolaki swiddeners helped to free forested land for conversion to cacaoproduction. Simultaneously, transmigration programs and forest in-dustries populated frontier areas with in-migrants and drove initialencroachments into the forest, deepening market linkages betweenrural areas and nearby towns. In some cases, support for production andagricultural development initiatives then provided growers with inputs,information and tenurial security as they planted forested lands incacao.

These dynamics, however, manifested unevenly across space, timeand agrarian society, shaping observed variability in cacao expansion

pathways. Oral histories and in-depth interviews suggest that the pol-itics of access to land within state-claimed forests were particularlycritical in determining when and where cacao expansion occurred.They also re-iterate the role of Suharto’s fall and subsequent decen-tralization debates, policies and practices in shifting the politics of ac-cess to and control over forested land. While these dynamics shapedincreasing access to forested lands by the late 1990s, they coincidedwith generalized economic depression and rising production costs.These factors converged to shape a strong early adopters’ effect in thesector, creating opportunities for well-capitalized in-migrants to expandland holdings while reinforcing limited market access for most otherindividuals.

One result of these processes has been highly heterogeneous path-ways of environmental change. Findings corroborate existing researchin revealing a significant connection between cacao and forest coverloss, particularly over the period 1995–2000. However, findings alsoreveal the extent of forest cover loss outside the cacao economy and thesimultaneity of tree cover gain and forest regrowth, much shaped bydisplacements of Tolaki forest settlements and smallholder tree cropplantings in grasslands and swidden fallows. These findings nuance aninterpretation of the environmental legacy of cacao expansion, parti-cularly given the forest industries that often degraded forested lands inthe years preceding cacao adoption.

Taken collectively, these dynamics also reiterate the path-depen-dent ways earlier histories of land use and control influence long-termtrajectories of LUCC. As findings from all four villages show, cacao notonly depended upon a reconfiguration of forest property relations, itmaterially and jurisdictionally inscribed these shifts into the landscape.These property inscriptions remain integral to LUCC. In Polia, recentforest cover loss is predominantly fueled by smallholder expansion intopreviously accumulated blocks of forested land. In Besulutu, it is fueledby the sale of remaining forest and fallowed cacao fields to oil palminvestors, a process facilitated by the removal of lands from state claimsat the time of SRADP. Some covert smallholder plantings continue instate-claimed forests surrounding Taowengga and Unahoa, often withthe goal of solidifying claims over land before final policies are set.Most recent clearance, however, reflects sustained logging pressure.

From a land use policy perspective, these dynamics caution againstthe notion that yield increases will be effective in addressing forestcover loss. While cacao and deforestation have largely been decoupledsince the mid-2000s, new commodity frontiers are rapidly forming,particularly where cacao expansion was most significant historically.Agricultural subsidies to land-wealthy growers could further fuel forestcover loss while exacerbating existing inequities and conflicts overaccess to land and markets.

7. Conclusion

The lack of work exploring heterogeneity in commodity crop ex-pansion and LUCC is an artefact of two tendencies in the broader lit-eratures on these themes. The first is a slippage between scales ofanalysis and analytical generalizability. In this slippage, so-called re-gional, national or global processes or factors (e.g., macro-economicincentives, topography, land availability) are taken to be synonymouswith greater generalizability whereas local processes are considered tobe irreducibly complex and/or idiosyncratic (Turner, 1999; Munroeet al., 2014). The second is the assumption that markets and policies areexogenous/external to local social relations; and as such, possible tomodel and predict even when abstracted from these relations (Turnerand Robbins, 2008; Munroe et al., 2014).

Contributing to a growing body of work, this paper illustrates howintegrating methods from RS with approaches from PE and CPG canmove beyond these tendencies. In this paper, RS data provided the basisfor moving beyond existing reliance on localized observations of LUCCto scale and contextualize variability in processes of cacao expansion.RS data also helped to situate case studies regionally. Methods and

L.C. Kelley Geoforum 97 (2018) 22–34

32

theories from PE and CPG, in turn, uncovered the political, historicaland ecological dynamics shaping variability in crop expansion. Theyalso helped to capture environmental changes invisible to the RS re-cord; for example, the variegated agrarian ecologies both pre-datingand produced by cacao expansion.

Taken together, these approaches suggest two central insights forthe literature on commodity crop expansion and LUCC. The first is thatvariation in processes of crop expansion and LUCC should not be con-sidered as an exception but as the norm. This implies that historical,social, and spatial relations are not just an add-on to the analyticalframe, but a central piece of the analytical puzzle. It also implies that ananalysis of variability is integral to any generalizable explanation ofLUCC. The second is the way in which an analysis of variation can helpto re-read more reductionist accounts of crop expansion and LUCC.While mono-causal explanations of LUCC have largely been debunked,they live on in the often invisible assumptions guiding land use policies.Foregrounding and reconstructing variability in commodity crop ex-pansion can shift the terms of these debates, uncovering more relevantsites of policy attention.

Acknowledgments

I thank the many people in Indonesia who shared their time andexperiences during field research, enabling these analyses. I am alsograteful to Nancy Peluso for her mentorship and scholarship, both ofwhich were central to the approach adopted here. Matthew Potts andMatthew Luskin provided substantive feedback on early drafts of thismanuscript, as did Isha Ray, Hekia Bodwich, Katy Seto, Annie Shattuck,Emily Reisman and members of UC Berkeley’s Land Lab and Potts Labs.I thank the Fulbright US Student Program, the National ScienceFoundation Graduate Research Fellowship Program, and researchgrants from the World Resources Institute, the American Institute forIndonesian Studies, and the World Agroforestry Centre’s “LinkingKnowledge to Action” project funded by the Department of ForeignAffairs, Trade and Development in Canada. Support from these in-stitutions made this research possible.

Appendix A. Supplementary material

Supplementary data to this article can be found online at https://doi.org/10.1016/j.geoforum.2018.10.006.

References

Abood, S.A., Lee, J.S.H., Burivalova, Z., Garcia-Ulloa, J., Koh, L.P., 2015. Relative con-tributions of the logging, fiber, oil palm, and mining industries to forest loss inIndonesia. Conserv. Lett. 8 (1), 58–67.

Akiyama, T., Nishio, A. 1996. Indonesia’s Cocoa Boom: Hands-Off Policy EncouragesSmallholder Dynamism. SSRN Scholarly Paper No. ID 620579. Social ScienceResearch Network, Rochester, NY.

Arce-Nazario, J.A., 2007. Human landscapes have complex trajectories: reconstructingPeruvian Amazon landscape history from 1948 to 2005. Landscape Ecol. 22 (1),89–101.

Arce-Nazario, J.A., 2016. Translating land-use science to a museum exhibit. J. Land UseSci. 11 (4), 417–428.

Asian Development Bank, 2004. Project Completion Report on the Sulawesi Rain-FedAgricultural Development Project in Indonesia. Asian Development Bank, Manila,Philippines.

Barr, C., Resosudarmo, I.A.P., Dermawan, A., McCarthy, J., 2002. Decentralisation ofForest Administration in Indonesia: Implications for Forest Sustainability,Community Livelihoods, and Economic Development. Center for InternationalForestry Research (CIFOR), Bogor, Indonesia.

Bebbington, A., 2001. Globalized Andes? Livelihoods, landscapes and development.Ecumene 8 (4), 414–436.

Belgiu, M., Drăguţ, L., 2016. Random forest in remote sensing: a review of applicationsand future directions. ISPRS J. Photogramm. Remote Sens. 114, 24–31.

Blaikie, P., Brookfield, H., 1987. Land Degradation and Society. Routledge.Booth, A. (Ed.), 1992. The Oil Boom and After: Indonesian Economic Policy and

Performance in the Soeharto Era. Oxford University Press, USA.Blaikie, P., 1999. A review of political ecology. Zietschrift fur Wirtschaftsgeographie 43,

131–147.BPS. 2017. Badan Pusat Statistik. Available at: https://www.bps.go.id/ Accessed 15

August 2017.Breman, J., Wiradi, G., 2002. Good Times and Bad Times in Rural Java. KITLV Press,

Leiden.Broich, M., Hansen, M.C., Potapov, P., Adusei, B., Lindquist, E., Stehman, S.V., 2011.

Time-series analysis of multi-resolution optical imagery for quantifying forest coverloss in Sumatra and Kalimantan, Indonesia. Int. J. Appl. Earth Observ. Geoinf. 13,277–291.

Brun, C., Cook, A.R., Lee, J.S.H., Wich, S.A., Koh, L.P., Carrasco, L.R., 2015. Analysis ofdeforestation and protected area effectiveness in Indonesia: a comparison of Bayesianspatial models. Global Environ. Change 31, 285–295.

Bryant, R.L., Bailey, S., 1997. Third World Political Ecology. Routledge, New York, USA.Cannon, C.H., Summers, M., Harting, J.R., Kessler, P.J., 2007. Developing conservation

priorities based on forest type, condition, and threats in a poorly known ecoregion:Sulawesi, Indonesia. Biotropica 39 (6), 747–759.

Carlson, K.M., Curran, L.M., Ratnasari, D., Pittman, A.M., Soares-Filho, B.S., Asner, G.P.,Trigg, S.N., Gaveau, D.A., Lawrence, D.L., Rodrigues, H.O., 2012. Committed carbonemissions, deforestation, and community land conversion from oil palm plantationexpansion in West Kalimantan, Indonesia. Proc. Natl. Acad. Sci. 109 (19),7559–7564.

Chaplin-Kramer, R., Sharp, R.P., Mandle, L., Sim, S., Johnson, J., Butnar, I., i Canals, L.M.,Eichelberger, B.A., Ramler, I., Mueller, I.C., McLachlan, I.N., 2015. Spatial patterns ofagricultural expansion determine impacts on biodiversity and carbon storage. Proc.Natl. Acad. Sci. 112 (24), 7402–7407.

Clarence-Smith, W.G., Ruf, F., 1996. Cocoa pioneer fronts: the historical determinants. In:Cocoa Pioneer Fronts Since 1800. Palgrave Macmillan, London, pp. 1–22.

Clough, Y., Faust, H., Tscharntke, T., 2009. Cacao boom and bust: sustainability ofagroforests and opportunities for biodiversity conservation. Conserv. Lett. 2 (5),197–205.

Corson, Catherine, 2011. Territorialization, enclosure and neoliberalism: non-state in-fluence in struggles over Madagascar's forests. J. Peasant Stud. 38 (4), 703–726.

Cushman, S.A., Macdonald, E.A., Landguth, E.L., Malhi, Y., Macdonald, D.W., 2017.Multiple-scale prediction of forest loss risk across Borneo. Landscape Ecol. 1–18.

Davis, K.F., Yu, K., Rulli, M.C., Pichdara, L., D’Odorico, P., 2015. Accelerated defor-estation driven by large-scale land acquisitions in Cambodia. Nat. Geosci. 8, 772–775.

Direktorat Jenderal Perkebunan, 2014. Unpublished data obtained in Kendari, SulawesiTenggara, Indonesia, 15 August 2014.

Departemen Tenaga Kerja dan Transmigrasi, 2015. Unpublished data obtained inKendari, Sulawesi Tenggara, Indonesia, 30 March 2015.

Durand, F., 1995. Farmer strategies and agricultural development: the choice of cocoa inEastern Indonesia. In: Ruf, F., Siswoputranto, P.S. (Eds.), Cocoa Cycles: TheEconomics of Cocoa Supply. Elsevier, Cambridge, UK.

Erasmi, S., Twele, A., Ardiansyah, M., Malik, A., Kappas, M., 2004. Mapping deforestationand land cover conversion at the rainforest margin in Central Sulawesi, Indonesia.EARSeL eProceedings 3 (3), 388–397.

Fairhead, J., Leach, M., 1996. Misreading the African Landscape: Society and Ecology in aForest-Savanna Mosaic. Cambridge University Press.

FAOSTAT, 2018. Production: Crops. Food and Agricultural Organization of the UnitedNations. Available at: http://faostat.fao.org/ Accessed 11 August 2018.

Fox, J., Castella, J.C., 2013. Expansion of rubber (Hevea brasiliensis) in MainlandSoutheast Asia: what are the prospects for smallholders? J. Peasant Stud. 40 (1),155–170.

Galvan-Miyoshi, Y., Walker, R., Warf, B., 2015. Land change regimes and the evolution ofthe maize-cattle complex in neoliberal Mexico. Land 4 (3), 754–777.

Gaveau, D.L.A., Sloan, S., Molidena, E., Yaen, H., Sheil, D., Abram, N.K., Ancrenaz, M.,Nasi, R., Quinones, M., Wielaard, N., Meijaard, E., 2014. Four decades of forestpersistence, clearance and logging on Borneo. PLoS One 9, e101654. https://doi.org/10.1371/journal.pone.0101654.

Gibbs, H.K., Ruesch, A.S., Achard, F., Clayton, M.K., Holmgren, P., Ramankutty, N., Foley,J.A., 2010. Tropical forests were the primary sources of new agricultural land in the1980s and 1990s. Proc. Natl. Acad. Sci. 107 (38), 16732–16737.

Gorelick, N., Hancher, M., Dixon, M., Ilyushchenko, S., Thau, D., Moore, R., 2017. Googleearth engine: planetary-scale geospatial analysis for everyone. Remote Sens. Environ.

Goldstein, J.E., 2014. The afterlives of degraded tropical forests. Environ. Soc.: Adv. Res.5, 124–140.

Hall, D., 2004. Explaining the diversity of Southeast Asian shrimp aquaculture. J.Agrarian Change 4 (3), 315–335.

Hall, D., 2011. Land grabs, land control, and Southeast Asian crop booms. J. Peasant Stud.38 (4), 837–857.

Hall, D., Hirsch, P., Li, T., 2011. Powers of Exclusion: Land Dilemmas in Southeast Asia.NUS Press.

Hansen, M.C., Potapov, P.V., Moore, R., Hancher, M., Turubanova, S.A., Tyukavina, A.,Thau, D., Stehman, S.V., Goetz, S.J., Loveland, T.R., Kommareddy, A., Egorov, A.,Chini, L., Justice, C.O., Townshend, J.R.G., 2013. High-resolution global maps of21st-century forest cover change. Science 342, 850–853. Data available on-line at:http://earthenginepartners.appspot.com/science-2013-global-forest.

Hecht, S.B., 1985. Environment, development and politics: capital accumulation and thelivestock sector in eastern Amazonia. World Dev. 13 (6), 663–684.

Hettig, E., Lay, J., Sipangule, K., 2016. Drivers of households’ land-use decisions: a criticalreview of micro-level studies in tropical regions. Land 5 (4), 32.

Hickey, S., Du Toit, A., 2007. Adverse Incorporation. Social Exclusion and ChronicPoverty. In: Chronic Poverty Research Centre Working Papers series, vol. 81.

Hirsch, P., 1989. The state in the village: interpreting rural development in Thailand. Dev.Change 20 (1), 35–56.

Hurni, K., Fox, J., 2018. The expansion of tree-based boom crops in mainland SoutheastAsia: 2001 to 2014. J. Land Use Sci. 1–22.

Jamal, S., Pomp, M., 1993. Smallholder adoption of tree crops: a case study of cocoa in

L.C. Kelley Geoforum 97 (2018) 22–34

33

Sulawesi. Bull. Indonesian Econ. Stud. 29 (3), 69–94.De Jong, C.G.F., 2011. Nieuwe hoofden Nieuw goden: Geschiedenis van de Tolaki en

Tomoronene, twee volkeren in Zuidoost-Celebes (Indonesie), tot ca. 1950 (Dutchlanguage). LAP Lambert Academic Publishing.

Dennis, R.A., Mayer, J., Applegate, G., Chokkalingam, U., Colfer, C.J.P., Kurniawan, I.,Lachowski, H., Maus, P., Permana, R.P., Ruchiat, Y., Stolle, F., 2005. Fire, people andpixels: linking social science and remote sensing to understand underlying causes andimpacts of fires in Indonesia. Human Ecol. 33 (4), 465–504.

Kelley, L.C., 2017. Richer histories for more relevant policies: 42 years of tree cover lossand gain in Southeast Sulawesi, Indonesia. Glob. Change Biol. 23, 830–839.

Kelley, L.C., Pitcher, L., Bacon, C., 2018. Using Google earth engine to map complexshade-grown coffee landscapes in Northern Nicaragua. Remote Sens. 10 (6).

Koh, L.P., Ghazoul, J., 2010. Spatially explicit scenario analysis for reconciling agri-cultural expansion, forest protection, and carbon conservation in Indonesia. Proc.Natl. Acad. Sci. 107 (24), 11140–11144.

de Koninck, R., 1996. The peasantry as the territorial spearhead of the state in SoutheastAsia: the case of Vietnam. Sojourn: J. Soc. Issues SoutheastAsia 231–258.

Lambin, E.F., Geist, H.J., 2008. Land-Use and Land-Cover Change: Local Processes andGlobal Impacts. Springer Science and Business Media.

Law, E.A., Meijaard, E., Bryan, B.A., Mallawaarachchi, T., Koh, L.P., Wilson, K.A., 2015.Better land-use allocation outperforms land sparing and land sharing approaches toconservation in Central Kalimantan, Indonesia. Biol. Conserv. 186, 276–286.

Lave, R., Wilson, M.W., Barron, E.S., Biermann, C., Carey, M.A., Duvall, C.S., Johnson, L.,Lane, K.M., McClintock, N., Munroe, D., Pain, R., 2014. Intervention: critical physicalgeography. Can. Geogr./Le Géographe canadien 58 (1), 1–10.

Lave, R., Biermann, C., Lane, S. (Eds.), 2018. The Palgrave Handbook of Critical PhysicalGeography. Palgrave Macmillan.

Lee, J.S.H., Abood, S., Ghazoul, J., Barus, B., Obidzinski, K., Koh, L.P., 2014.Environmental impacts of large-scale oil palm enterprises exceed that of smallhold-ings in Indonesia. Conserv. Lett. 7 (1), 25–33.

Leiter, J., Harding, S., 2004. Trinidad, Brazil, and Ghana: three melting moments in thehistory of cocoa. J. Rural Stud. 20 (1), 113–130.

Li, T.M., 2002. Local histories, global markets: Cocoa and class in upland Sulawesi. Dev.Change 33, 415–437.

Li, T., 2005. Transforming the Indonesian Uplands. Routledge.Li, T.M., 2014. Land's End: Capitalist Relations on an Indigenous Frontier. Duke

University Press, Durham, NC, USA.Lukas, M.C., 2014. Eroding battlefields: land degradation in Java reconsidered. Geoforum

56, 87–100.McCarthy, J.F., 2004. Changing to gray: decentralization and the emergence of volatile

socio-legal configurations in Central Kalimantan, Indonesia. World Dev. 32,1199–1223.

McCarthy, J.F., 2010. Processes of inclusion and adverse incorporation: oil palm andagrarian change in Sumatra, Indonesia. J. Peasant Stud. 37 (4), 821–850.

McCarthy, J.F., Vel, A.C., Afiff, S., 2012. Trajectories of land acquisition and enclosure:development schemes, virtual land grabs, and green acquisitions in Indonesia’s OuterIslands. J. Peasant Stud. 29 (2), 521–549.

Miettinen, J., Liew, S.C., 2010. Degradation and development of peatlands in PeninsularMalaysia and in the islands of Sumatra and Borneo since 1990. Land Degrad. Dev. 21(3), 285–296.

Miyamoto, M., 2006. Forest conversion to rubber around Sumatran villages in Indonesia:comparing the impacts of road construction, transmigration projects and population.Forest Policy Econ. 9 (1), 1–12.

Munroe, D.K., McSweeney, K., Olson, J.L., Mansfield, B., 2014. Using economic geo-graphy to reinvigorate land-change science. Geoforum 52, 12–21.

Neilson, J., 2007. Global markets, farmers and the state: sustaining profits in theIndonesian cocoa sector. Bull. Indonesian Econ. Stud. 43 (2), 227–250.

Nieston, E.T., Rice, R.E., Ratay, S.M., Paratore, K., Hardner, J.J., Fearnside, P.M., 2004.Commodities and Conservation: The Need for Greater Habitat Protection in theTropics. Conservation International, Washington, DC, USA.

Nugroho, H.Y.S.H., van der Veen, A., Skidmore, A.K., Hussin, Y.A., 2017. Expansion oftraditional land-use and deforestation: a case study of an adat forest in the KandiloSubwatershed, East Kalimantan, Indonesia. J. For. Res. https://doi.org/10.1007/s11676-017-0449-9.

Padwe, J., 2011. Cashews, cash and capitalism in Northeast Cambodia. Cambodia’s Econ.Trans. 123–153.

Peluso, N.L., Vandergeest, P., 2001. Genealogies of the political forest and customaryrights in Indonesia, Malaysia, and Thailand. J. Asian Stud. 60 (03), 761–812.

Peluso, N.L., Vandergeest, P., 2011. Political ecologies of war and forests: counter-insurgencies and the making of national natures. Ann. Assoc. Am. Geogr. 101 (3),587–608.

Peluso, N.L., Lund, C., 2011. New frontiers of land control: introduction. J. Peasant Stud.38 (4), 667–681.

Phalan, B., Bertzky, M., Butchart, S.H., Donald, P.F., Scharlemann, J.P., Stattersfield, A.J.,Balmford, A., 2013. Crop expansion and conservation priorities in tropical countries.PloS One 8 (1), 51759.

Polanyi, K., 1957. The Great Transformation: The Political and Economic Origin of OurTime. Beacon Press.

Pomp, M., Burger, K., 1995. Innovation and imitation: adoption of cocoa by Indonesian

smallholders. World Dev. 23 (3), 423–431.Potter, L., Lee, J., 1998. Tree Planting Trends in Indonesia: Trends, Impacts and

Directions. Occasional Paper No. 18. The Center for International Forestry Research,Bogor, Indonesia.

Ribot, J.C., Peluso, N.L., 2003. A theory of access. Rural Sociol. 68 (2), 153–181.Robbins, P., 2001. Fixed categories in a portable landscape: the causes and consequences

of land-cover categorization. Environ. Plann. A 33 (1), 161–179.Robbins, P., 2011. Political Ecology: A Critical Introduction Vol. 16 John Wiley & Sons.Ruf, F., Schroth, G., 2004. Chocolate forests and monocultures: a historical review of

cocoa growing and its conflicting role in tropical deforestation and forest conserva-tion. In: Schroth, G., da Fonseca, G.A.B., Harvey, C.A., Gascon, C., Vasconcelos, H.L.,Izac, A.N. (Eds.), Agroforestry and Biodiversity Conservation in Tropical Landscapes.Island Press, Washington, USA, pp. 107–134.

Ruf, F., Siswoputranto, P.S., 1995. Cocoa Cycles: The Economics of Cocoa Supply.Elsevier, Cambridge, UK.

Ruf, F., Yoddang, 1996. Smallholder Cocoa in Indonesia: why a Cocoa Boom in Sulawesi?In: Clarence-Smith, W.G. (Ed.), Cocoa Pioneer Fronts Since 1800. PalgraveMacmillan, UK, pp. 212–231.

Ruf, F., Yoddang, C.T., 1999. The Impact of the Economic Crisis on Indonesia’s CocoaSector. ACIAR Indonesia Research Project. Working Paper 99.20.

Ruf, F., Yoddang, Ardhy, W., 2001. Transmigrants and the Cocoa Windfall: 'Paradise isHere, not in Bali’. In: Gérard, F., Ruf, F. (Eds.), Agriculture in Crisis: People,Commodities and Natural Resources in Indonesia, 1996-2000. Cirad, Montpellier,France.

Ruf, F., Yoddang, 2001. Cocoa migrants from boom to bust. In: Gérard, F., Ruf, F. (Eds.),Agriculture in Crisis: People, Commodities and Natural Resources in Indonesia 1996-2001. Routledge, New York, NY, USA.

Ruf, F., Yoddang, 2004. Pod Borer and Pesticides. From Slash-and-Burn to Replanting:Green Revolutions in the Indonesian Uplands, pp. 147–172.

Ruf, F., Zadi, H., 1998. Cocoa: From Deforestation to Reforestation. Accessible at: http://agritrop.cirad.fr/390123/1/document_390123.pdf.

Ruggles, S., Genadek, K., Goeken, R., Grover, J., Sobek, M., 2017. Integrated Public UseMicrodata Series: Version 7.0. University of Minnesota, Minneapolis. https://doi.org/10.18128/D010.V7.0.

Scott, J.C., 2009. The Art of Not Being Governed. An Anarchist History of Upland. YaleUniversity Press, New Haven.

Sikor, T., Vi, T., Thi, P., 2005. The dynamics of commoditization in a Vietnamese uplandsvillage, 1980–2000. J. Agrarian Change 5 (3), 405–428.

Sikor, T., 2012. Tree plantations, politics of possession and the absence of land grabs inVietnam. J. Peasant Stud. 39 (3–4), 1077–1101.

Steffan-Dewenter, I., Kessler, M., Barkmann, J., Bos, M.M., Buchori, D., Erasmi, S., Faust,H., Gerold, G., Glenk, K., Gradstein, S.R., Guhardja, E., Harteveld, M., Hertel, D.,Höhn, P., Kappas, M., Köhler, S., Leuschner, C., Maertens, M., Marggraf, R., Migge-Kleian, S., Mogea, J., Pitopang, R., Schaefer, M., Schwarze, S., Sporn, S.G.,Steingrebe, A., Tjitrosoedirdjo, S.S., Tjitroseomito, S., Twele, A., Weber, R.,Woltmann, L., Zeller, M., Tscharntke, T., 2007. Tradeoffs between income, biodi-versity, and ecosystem functioning during tropical rainforest conversion and agro-forestry intensification. Proc. Natl. Acad. Sci. 104 (12), 4973–4978.

Sunderlin, W.D., 1999. Between danger and opportunity: Indonesia and forests in an eraof economic crisis and political change. Soc. Nat. Resour. 12 (6), 559–570.

Tarimana, A., 1989. Kebudayaan Tolaki (Indonesian Language). Balai Pustaka, Indonesia.Turner, Matthew D., 1999. Merging local and regional analyses of land-use change: The

case of livestock in the Sahel. Ann. Assoc. Am. Geogr. 82 (2), 191–219.Turner, M.D., 2003. Methodological reflections on the use of remote sensing and geo-

graphic information science in human ecological research. Human Ecol. 31 (2),255–279.

Turner, B.L., Robbins, P., 2008. Land-change science and political ecology: similarities,differences, and implications for sustainability science. Annu. Rev. Environ. Resour.33, 295–316.

Van der Laan, C., Wicke, B., Verweij, P.A., Faaij, A.P., 2017. Mitigation of unwanteddirect and indirect land-use change–an integrated approach illustrated for palm oil,pulpwood, rubber and rice production in North and East Kalimantan, Indonesia.Global Change Bioenergy 9 (2), 429–444.

Vandergeest, P., Peluso, N.L., 1995. Territorialization and state power in Thailand.Theory Soc. 24 (3), 385–426.

Vandergeest, P., Flaherty, M., Miller, P., 1999. A political ecology of shrimp aquaculturein Thailand. Rural Sociol. 64 (4), 573–596.

Walker, Peter, 2005. Political ecology: where is the ecology? Prog. Hum. Geogr. 29 (1),73–82.

Whitten, T., Mustafa, M., Henderson, G.S., 1987. The Ecology of Sulawesi. Gadjah MadaUniversity Press, Yogyakarta, Indonesia.

Wulder, M.A., Masek, J.G., Cohen, W.B., Loveland, T.R., Woodcock, C.E., 2012. Openingthe archive: how free data has enabled the science and monitoring promise ofLandsat. Remote Sens. Environ. 122, 2–10.

Xiong, J., Thenkabail, P.S., Tilton, J.C., Gumma, M.K., Teluguntla, P., Oliphant, A.,Congalton, R.G., Yadav, K., Gorelick, N., 2016. Nominal 30-m cropland extent map ofcontinental Africa by integrating pixel-based and object-based algorithms usingSentinel-2 and Landsat-8 data on Google Earth Engine. Remote Sens. 9 (1065).https://doi.org/10.3390/rs9101065.

L.C. Kelley Geoforum 97 (2018) 22–34

34