Contents lists available at ScienceDirect

Land Use Policy

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

Spatial and temporal patterns of land clearing during policy change

B. Alexander Simmonsa,⁎, Elizabeth A. Lawa, Raymundo Marcos-Martinezb,c, Brett A. Bryand,Clive McAlpinee, Kerrie A. Wilsona

a Australian Research Council Centre of Excellence for Environmental Decisions, School of Biological Sciences, The University of Queensland, St. Lucia, QLD, 4072,Australiab CSIRO Land and Water, Black Mountain, Canberra, ACT, 2601, Australiac School of Commerce, University of South Australia, Adelaide, SA, 5002, Australiad Centre for Integrative Ecology, Deakin University, Burwood, VIC, 3125, Australiae School of Earth and Environmental Sciences, The University of Queensland, St. Lucia, QLD, 4072, Australia

A R T I C L E I N F O

Keywords:Environmental policyDeforestationPanic clearingPattern analysisPolicy responseVegetation Management Act

A B S T R A C T

Environmental policies and regulations have been instrumental in influencing deforestation rates around theworld. Understanding how these policies change stakeholder behaviours is critical for determining policy im-pact. In Queensland, Australia, changes in native vegetation management policy seem to have influenced landclearing behaviour of landholders. Periods of peak clearing rates have been associated with periods precedingthe introduction of stricter legislation. However, the characteristics of clearing patterns during the last twodecades are poorly understood. This study investigates the underlying spatiotemporal patterns in land clearingusing a range of biophysical, climatic, and property characteristics of clearing events. Principal component andhierarchical cluster analyses were applied to identify dissimilarities between years along the political timeline.Overall, aggregate landholders’ clearing characteristics remain generally consistent over time, though noticeabledeviations are observed at smaller regional and temporal scales. While clearing patterns in some regions haveshifted to reflect the policy’s goals, others have experienced minimal or contradictory changes following reg-ulation. Potential ‘panic’ or ‘pre-emptive’ effects are evident in the analysis, such as spikes in clearing for pastureexpansions, but differ across regions. Because different regions are driven by different pressures, such as landavailability and regulatory opportunity, it is imperative that the varying spatial and temporal behavioural re-sponses of landholders are monitored to understand the influence of policy and its evolution. Future policyamendments would benefit from monitoring these regional responses from landholders to better assess the ef-fectiveness of policy and the potential perversities of policy uncertainty.

1. Introduction

Deforestation, with its consequential effects of habitat loss, frag-mentation, and degradation, is a well-recognized threat to biodiversityand ecosystem function (McAlpine et al., 2002; Lindenmayer et al.,2005; Bradshaw, 2012). Environmental regulations and other policyinstruments greatly influence deforestation rates around the world,whether directly or indirectly (Meyfroidt and Lambin, 2011). A numberof countries have directly reduced deforestation rates using conserva-tion policies incorporating logging bans (Southworth and Tucker, 2001;Mather, 2007), mandated reforestation or afforestation (Klock, 1995;Wang et al., 2007), and land use restrictions (Fox et al., 2009; Assunçãoet al., 2012). Other countries such as Costa Rica and India have ex-perienced a decline in deforestation indirectly, due to economic andideological pressures (Kull et al., 2007; Daniels, 2009) and forest

management decentralisation (Agrawal, 2007; DeFries and Pandey,2010), respectively. Angelsen and Kaimowitz (1999) argue that policyinstruments and macroeconomic variables represent the underlyingcauses of deforestation, and these factors will directly influence moreimmediate causes of deforestation, such as institutions, infrastructure,markets, and technology.

Policy instruments can thus modify the dynamics of the human-environment system, but by doing so, may not always work as intended.Policies can even lead to the potential for perverse, or unintentional,outcomes to emerge (Miteva et al., 2012). Some conservation policyinstruments have resulted in leakage effects, whereby deforestation isdisplaced from the regulated region into unregulated areas (Wear andMurray, 2004; Oliveira et al., 2007; Gaveau et al., 2009). Other policiesmeant to indirectly curb deforestation, like those reducing agriculturalrents or removing clear-to-own property laws, may also result in

https://doi.org/10.1016/j.landusepol.2018.03.049Received 6 July 2017; Received in revised form 26 March 2018; Accepted 26 March 2018

⁎ Corresponding author.E-mail address: b.simmons@uq.edu.au (B.A. Simmons).

Land Use Policy 75 (2018) 399–410

Available online 11 April 20180264-8377/ © 2018 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).

T

http://www.sciencedirect.com/science/journal/02648377https://www.elsevier.com/locate/landusepolhttps://doi.org/10.1016/j.landusepol.2018.03.049https://doi.org/10.1016/j.landusepol.2018.03.049mailto:b.simmons@uq.edu.auhttps://doi.org/10.1016/j.landusepol.2018.03.049http://crossmark.crossref.org/dialog/?doi=10.1016/j.landusepol.2018.03.049&domain=pdf

accelerated clearing rates when they are poorly implemented, lackpublic support, or introduce high levels of legislative uncertainty(Kaimowitz et al., 1998; Angelsen, 2009).

Vegetation management policies directly affect the livelihoods ofagricultural landholders, placing constraints on economic growth,property rights, and potentially tenure security (Alston et al., 2000;Sant’Anna and Young, 2010; Aldrich, 2012). The link between defor-estation and issues of property rights and tenure security has been mostobvious in developing nations, where landholders clear forest to laytheir claim on the land, prevent squatters from infiltrating, and receivefinancial incentives (as well as property legitimacy) from the govern-ment (Alston et al., 2000; Aldrich et al., 2012; Brown et al., 2016).When this sense of security and autonomy is threatened by incomingpolicies dictating how landholders are permitted to manage their land,some may react by pre-emptively clearing vegetation. If controversialpolicies are complemented by high political instability, regime changes,or legislative ambiguity, the reactions from landholders to this un-certainty can significantly increase deforestation rates over time(Deacon, 1994; Barbier and Burgess, 2001).

Deforestation patterns in Australia are broadly reflective of therapid rate of modern deforestation globally (Lindenmayer et al., 2005).In its relatively brief colonial history, Australia has seen agriculturalexpansion reduce forest cover by nearly 15%, with 7.2 million ha (7%)of primary forests cleared in the last 40 years alone (Bradshaw, 2012;Evans, 2016). Since the 1970s, the State of Queensland has lost 9.7million ha of total forest from land clearing, accounting for more than60% of clearing in the entire country over this period (Evans, 2016),with native vegetation cover reduced by at least 50% over the last 200years (Australian Bureau of Rural Sciences, 2010). Like many devel-oping countries, the first century of development in Queensland wasmarked by heavy governmental encouragement for landholders to clearas much vegetation as possible in an effort to raise economic prosperity(Braithwaite, 1996; Bradshaw, 2012). It was not until the end of the20th century that public opinion began to change regarding the value ofnative vegetation, and the Queensland Government entered into aperiod of land clearing policy reform, which brought about the firststrict regulations on vegetation management practices with the Vege-tation Management Act (VMA) 1999. Landholders in Queensland havesince experienced considerable evolutions in state vegetation manage-ment policy.

The most infamous period of land clearing in Queensland in recenthistory involved the rapid increase in clearing rates during 1999–2000and 2002–2003, likely in response to the initial enactment of the VMA1999 and subsequent stricter implementation in 2003—periods com-monly referred to as panic clearing (Productivity Commission, 2004;Lindenmayer et al., 2005; Taylor, 2015). The definition of panicclearing, however, is unclear; it has been used to describe rushedclearing activities (i.e. future plans that were expedited) or unplannedclearing activities (i.e. activities that the landholder had no future in-tentions of executing), though evidence of both types of panic clearinghave been reported anecdotally in this case (Senate Inquiry, 2010).Further, it is unclear whether panic clearing constitutes increasedbusiness-as-usual clearing (i.e. clearing locations similar to locationscleared in the past), increased atypical clearing (i.e. clearing locationsdissimilar to those cleared in the past), or a combination of both.Identifying how these different characterisations of panic clearingcontributed to the increased volume of clearing across regions is im-perative to our understanding of how landholders make reactive, short-term land clearing decisions. One attribute of panic clearing remainsconsistent, however, which is that panic clearing is pre-emptive, due toexpected clearing limitations imposed by future regulations (McIntyreet al., 2002; Productivity Commission, 2004; McGrath, 2007). Suchperverse pre-emptive responses from landholders and stakeholders canalso be found elsewhere in the conservation realm, following listings onthe U.S. Endangered Species Act (Lueck and Michael, 2003) and tradebans under the Convention on International Trade of Endangered

Species (CITES; Rivalan et al., 2007).The convoluted introduction of strict vegetation management reg-

ulations in Queensland led to landholder uncertainty regarding futureproperty rights and tenure security (Productivity Commission, 2004;Senate Inquiry, 2010), which has also been observed in developingcountries undergoing substantial policy evolution (Alston et al., 2000;Aldrich et al., 2012). Queensland thus serves as an important and re-levant global case study to highlight how these clearing behaviours maychange over time amidst continual (and sometimes contradictory)changes to a single vegetation management policy. Further, the avail-ability of quality data on the characteristics of clearing in Queenslandallows for more thorough investigations that may not be present inother cases.

To date, the extent of state-wide vegetation clearing in Queenslandhas been widely publicised in the literature (e.g. Evans, 2016;Queensland Department of Science, Information Technology andInnovation, 2016), yet minimal attention has been placed on thecharacteristics of clearing over time in this case. This provides associa-tive evidence of how vegetation management policy has affected ag-gregate landholder actions (i.e. the ‘what’), rather than using thecharacteristics of clearing to investigate the dynamic and differentialbehaviours of landholders (i.e. the ‘how’). Such temporal analyses haverecently been used to assess global patterns of deforestation to identifydrivers of clearing behaviour and forest transition (Hosonuma et al.,2012; Sandker et al., 2017), but these same concepts can be applied atfiner scales. Previous investigations into the trends of land clearing inQueensland have also relied upon state- or national-level drivers ofdeforestation (e.g. Evans, 2016; Marcos-Martinez et al., 2018), despitethe global recognition of regionally dependent deforestation driversand landholder responses (Geist and Lambin, 2002). Thus these studiesmay produce generalised patterns and policy recommendations thatmay not adequately capture or identify regional landholders’ beha-viours and potential motivations.

This study investigates the underlying spatial and temporal char-acteristics and patterns of land clearing within the context of evolvingvegetation management policies, using Queensland as a case study.Using a range of biophysical, climatic, and property characteristics toidentify underlying patterns in clearing events across the politicaltimeline, we analyse (1) how the observable biophysical, socio-economic, and property characteristics of clearing events change overtime, (2) what principle components can be derived from the spatialcharacteristics of clearing events, and (3) how these components differbetween key policy periods. Further, we focus on periods described aspanic clearing and assess how their clearing characteristics differ fromprevious years. To compare the potentially different spatial patterns,our analysis is undertaken at multiple scales: (1) an aggregate state-level analysis, (2) contrasting bioregion analyses, within a historicalclearing hotspot (Brigalow Belt South), a relatively intact frontier forclearing (Cape York Peninsula), and an area of dense urban sprawl(South Eastern Queensland), and (3) a composite region of particularcurrent environmental concern (Great Barrier Reef catchment).

2. Methods

2.1. Study areas

Queensland (QLD, 2.04M km2) is the most climatically diverse statein Australia, with 50–3000mm mean annual precipitation dependingon the region, including equatorial, tropical, subtropical, temperate,grassland, and desert bioregions. We examined clearing patterns acrossthe entire State of Queensland, Australia, with a focus on four sub-regions of interest (Fig. 1): the Brigalow Belt South bioregion (BBS),Cape York Peninsula bioregion (CYP), South East Queensland bioregion(SEQ), and the Great Barrier Reef catchment (GBRC) as defined by theformer Department of Environment and Resource Management(Rollason and Howell, 2012). The BBS bioregion (267,000 km2), in

B.A. Simmons et al. Land Use Policy 75 (2018) 399–410

400

south-central QLD, is a subtropical (500–750mm) biodiversity hotspotthat has been extensively historically cleared for agricultural develop-ment, especially pasture expansion (Cogger et al., 2003; Fensham andFairfax, 2003). The CYP bioregion (131,000 km2), in contrast, is one ofthe most intact of the QLD bioregions, though interest in grazing isincreasing in this tropical savannah and rainforest (1000–2000mm)region (Crowley and Garnett, 1998). SEQ bioregion (77,500 km2) is themost densely populated bioregion in QLD. This subtropical region ishighly diverse in climate and production, and, despite being alreadyone of the most extensively cleared in QLD, is currently experiencingsubstantial expansion of urban areas (Wilson et al., 2002; Petersonet al., 2007). The GBRC (388,000 km2) spans multiple tropical andsubtropical bioregions (Rollason and Howell, 2012), where the expan-sion and intensification of agricultural industries continues to causeconcern for the health of the Great Barrier Reef (Great Barrier ReefMarine Park Authority, 2014).

2.1.1. Timeline of vegetation management policy in QueenslandQueensland’s entrance into strict command-and-control land

clearing regulation began following the enactment of the IntegratedPlanning Act (IPA) 1997 and the Vegetation Management Act (VMA) 1999(Fig. 2). The VMA 1999 serves to (1) identify and define the differenttypes of vegetation and their conservation value/protection status, and(2) outline the policy framework underlying clearing permits, whichthen guides the IPA 1997’s requirements for assessing and enforcingthese permits (Productivity Commission, 2004; McGrath, 2010). Thiscontroversial legislation was hastily developed, according to membersof Parliament during debate (Kehoe, 2009), and approved by the end of1999, but the official proclamation of the Act was delayed by the pre-mier until financial assistance could be provided from the Common-wealth. By September 2000, the VMA was proclaimed and the bulk ofstatutory provisions commenced (Kehoe, 2009). This period betweenassent and proclamation provided a window of uncertainty amongstlandholders as to what the future held for their ability to clear vege-tation on their property, and is associated with peak clearing ratesacross the state. In an attempt to halt the rate of pre-emptive clearings,the centrist government passed an unexpected moratorium on clearingapplications in 2003, followed by a pivotal amendment to the VMA in2004, which declared that broad-scale clearing would be banned by thebeginning of 2007. The following interim period (2004–2007) aimed toease the transition for landholders prior to the broad-scale clearing ban,as the government offered some monetary compensation packages andallowed a transitional cap of 5000 km2 of broad-scale clearings to becarried out across the state before 2007 (McGrath, 2007; Kehoe, 2009).The period following the ban saw additional amendments to the VMA,

further restricting the clearing activities permissible on freehold andleasehold lands. In 2012, political power switched to the more con-servative party, beginning a period of clearing policy relaxations in anattempt to reduce the burdens felt by landholders. Despite the return ofa centrist premier in 2015, efforts to reinstate previous clearing re-strictions have been unsuccessful due to the party’s lack of majorityrepresentation in Parliament.

2.2. Clearing data and characterisation variables

Land clearing data was obtained from the annual StatewideLandcover and Trees Study (SLATS) dataset (Queensland SpatialCatalogue, 2016g), for fiscal-year periods 1988–1991, 1991–1995,1995–1997, 1997–1999, and annually from 1999–2000 to 2014–2015.We denote the fiscal year periods by their end year, and use a per yearaverage for those spanning multiple years. The SLATS data are based onthe supervised classification of multiple Landsat satellite images anddigital terrain models at a resolution of approx. 30m (Macintosh,2007). We defined policy periods as (Fig. 2): unregulated (1988–1995),reform (1995–1999), regulation (1999–2004), interim (2004–2007), re-striction (2007–2012), and relaxation (2012–2015). Each incident ofclearing is coded according to the clearing type (see Scarth et al., 2008).To focus on private landholder-driven clearing, which represents 97%of total deforestation in the state during this time period, we excludednatural tree loss, as well as tree loss within natural resource manage-ment areas (national parks, reserves, state forests, forest reserves) andon aquatic tenures (ports, harbours, water resources). To estimate an-nual remnant vegetation clearing, the earliest and most accurate map ofremnant vegetation was obtained from the Queensland Herbarium for1997 (QSC, 2016f). Annual remnant forest maps were developed byfirst excluding the grassland ecosystems from the 1997 remnant extent,and then removing cells cleared in each period. High-value regrowthwas not distinguished in this study. Because maps of remnant vegeta-tion do not exist for the unregulated and early reform periods, we do notdistinguish vegetation according to remnant status for the subsequentpattern analysis in order to facilitate comparability between all policyperiods. When discussing the extent of clearing (Section 3.1), however,we identify remnant vegetation clearing events for the period in whichdata is available (1997–2015).

To characterise clearing events, nine temporally constant variableswere selected: clearing type, slope, elevation, remoteness, rainfallvariability, historical drought declarations, parcel size, tenure, and re-gional ecosystem threat status (Table 1). Henceforth, these variables arecollectively referred to as the characteristics of land clearing. Thesevariables represent biophysical, climatic, property, and legal

Fig. 1. Current land use and remnant vegetation extent within the four study regions of Queensland. Excluded areas include national and regional parks, state forests,reserves, forest reserves, timber reserves, and water resources. Land uses are derived from the Australian Land Use and Management (ALUM) Classification system.

B.A. Simmons et al. Land Use Policy 75 (2018) 399–410

401

characteristics that, together, describe suitability for agricultural con-version, the most common driver of deforestation across the globe(Joppa and Pfaff, 2011; Busch and Ferretti-Gallon, 2017; Marcos-Martinez et al., 2018), including Queensland (Evans, 2016). Parcel sizewas used as a proxy for property size, often identified as a correlate ofdeforestation (e.g. Pichón, 1997; Geoghegan et al., 2001; Seabrooket al., 2007, 2008). Tenure was included as a property and legal char-acteristic, which designates the legislation and benefits of clearing andother property management practices (Turner et al., 1996; Fenshamet al., 1998; Lindenmayer et al., 2005). Regional ecosystem threat statusprovides a measure of the percent of clearing within each cell that fallson the pre-clearing extent of ‘least concern’ regional ecosystems. Re-gional ecosystems (REs) are vegetation communities characterized byunique geology, soil, and landform combinations (Sattler and Williams,1999), and are designated a threat status of either ‘endangered,’ ‘ofconcern,’ or ‘least concern.’ This measure is thus a representation ofboth ecosystem rarity and vulnerability status under the VMA, as ‘leastconcern’ REs are most prevalent throughout the State and are under lesspressure from traditional clearing trends.

2.3. Data analyses

2.3.1. Summarising clearing events over timeThe historical extent of land clearing was calculated in ArcGIS using

SLATS clearing polygons for total clearing (1989–1997) and for rem-nant and non-remnant clearings (1998–2015) after removing naturaltree loss (i.e. ‘natural tree death’ and ‘natural disaster damage’) andclearings within protected areas and other resource management areas(see Fig. 1). Annual clearing summaries were generated to report theproportion of clearing area consisting of the following clearing purposesidentified by SLATS: pasture, crop, settlement, mine, infrastructure,timber plantation, and thinning. The frequency of clearing events perparcel of land was calculated by adding the number of SLATS reportingperiods in which a given parcel experienced at least one clearing event,for a maximum of 20 time periods.

2.3.2. Comparing components of clearing spatial characteristics acrosspolicy periods

The SLATS clearing polygons were converted to raster pixels with a100m resolution and attributed with the overlapping clearing

Fig. 2. The evolution of vegetation management policy in Queensland. Political timeline of pivotal legislation and regime shifts, categorised according to six policyperiods between 1990 and 2016: unregulated, reform, regulation, interim, restriction, and relaxation. State elections of 2001, 2004, 2006, and 2009 where the centristgovernment maintained power are not shown.

Table 1Variables used to characterise clearing events in the principal component analysis (PCA).

Variable Description Source

Clearing type Percent of clearing attributed to ‘pasture’ SLATS clearing descriptions (QSC, 2016g)Elevation Elevation (m.a.s.l.) QSC (2016d)Historical drought declarations Percentage of time between 1963 and 2011 in which the relevant Local Government

Area was declared in drought by the State of QueenslandQSC (2016a)

Parcel size Area (km2) of individual parcels of land, used as a proxy for property size QSC (2016c)Rainfall variability Standard deviation of average rainfall from 1890 to 2013 QSC (2016b)Regional ecosystem (RE) threat

statusPercent of clearing on ‘least concern’ regional ecosystems QSC (2016e)

Remoteness Measure of a location’s proximity to the nearest urban centre (0= highlyaccessible, 15=very remote)

Accessibility/Remoteness Index of Australia (ARIA),(Atlas of Living Australia, 2016)

Slope Slope (deg.) QSC (2016d)Tenure Percent of clearing on ‘freehold’ land QSC (2016c)

B.A. Simmons et al. Land Use Policy 75 (2018) 399–410

402

characteristics for Queensland, resulting in over 9 million clearing ob-servations (pixels) for the entire timeframe. The dataset was sampledwithout replacement in R version 2.15.3 (R Core Team, 2016) using thepackage ‘kimisc’ (Müller, 2014) giving 100,000 observations for each ofthe QLD, BBS, SEQ, and GBRC case studies. The entire CYP regionconsisted of only 28,263 observations, so no sampling was performed.The validity and representativeness of the QLD sample was confirmedby comparing the variable distributions of the samples and populations,and testing that two independent samples produced equivalent results.

2.3.2.1. Principal component analysis. Principal component analyses(PCAs) were performed for the entire State of Queensland, as well asfor each case study region, in order to identify principle components ofthe nine spatial characteristics of clearing events. These principlecomponents represent key dimensions of clearing patterns. We applythis PCA only to the attribute space of the spatial data and do not adaptthe analysis for effects of heterogeneity or autocorrelation, as we arenot producing spatial predictions (Demšar et al., 2013). Prior toanalysis, Bartlett’s test of sphericity (Bartlett, 1950) and the Kaiser-Meyer-Olkin (KMO) test of sampling adequacy (Kaiser, 1970) wereapplied to each case study to assess the usefulness of PCA for thedatasets. Keeping with standard practice, we determined a PCA wasacceptable for the dataset when KMO > 0.5 (Tabachnick and Fidell,2001). Initial PCAs were performed using the R package ‘FactoMineR’(Husson et al., 2017). The number of principal components (PCs)selected was based upon the commonly used Kaiser Criterion (Kaiser,1958) and verified by a scree test (Cattell, 1966). In this analysis, wefound that the scree test generally verified use of the Kaiser Criterion(Fig. A.1), though we modestly extended the cut-off point (E > 0.98,see Section 3.2).

When variables produced significant contributions to more than onePC, we applied varimax rotation (Kaiser, 1958) to the selected PCsusing the R package ‘psych’ (Revelle, 2017) to produce rotated com-ponents (RCs) with simple structure. Here, we use the term ‘rotated

components’ in lieu of the more traditional term ‘factors’ in order toeliminate confusion between principal components and factors. In caseswhen variables exhibited cross-loading on two or more RCs, the vari-ables were removed from the dataset, and the analysis was performedagain until rotation produced the desired simple structure. The finalselection of RCs was made based upon interpretability and variablerepresentation. While an RC with at least three high-loading variables isgenerally satisfactory (Costello and Osborne, 2005), we selected RCswith two high-loading variables when the variables represented a si-milar conceptual construct, thus enhancing the interpretability of thecomponents.

The variables chosen for interpretation of each selected RC werebased upon the strength of their loading according to the scale providedby Liu et al. (2003): strong (> 0.75), moderate (0.50–0.75), weak(0.30–0.49). We thus interpreted each RC according to all variableswith moderate or strong loadings. Each RC was interpreted based uponthe direction and strength of the characteristic variables’ loadings anddefined according to the relationship of the variables within the com-ponent. We classify the final RCs as the components driving landclearing characteristics within the respective region, and compare an-nual scores across the RCs to identify temporal changes and patterns inthe strength of each component on land clearing. For more detail on thePCA methodology, see Appendix B.

2.3.2.2. Component score clustering. To visually estimate how theinfluence of these components differ between key policy periods, weproduced a hierarchical cluster scheme based on the values of eachobservation from their respective RC for each study region. Ward’sclustering method was used to maximise intra-group similarity andanalyse observations according to variance rather than distance (Ward,1963). Clustering was performed for all study areas by meancomponent score per policy period. To identify the final number ofdistinct clusters for each study area, we defined a dendrogram cut offheight of 0.40, which was determined post hoc in an effort to select a

Fig. 3. Annual clearing extent across policy periods, broken down by remnant and non-remnant vegetation clearing (where data is available) for (a) the State ofQueensland (QLD), (b) Brigalow Belt South (BBS) bioregion, (c) Great Barrier Reef catchment (GBRC), (d) South Eastern Queensland (SEQ) bioregion, and (e) CapeYork Peninsula (CYP) bioregion. Data excludes natural tree loss and clearing within protected areas and other resource management areas. (*) Clearing extentrepresents an average annual estimate.

B.A. Simmons et al. Land Use Policy 75 (2018) 399–410

403

moderate yet conservative height that was relevant to all study regions.

3. Results

3.1. Comparative clearing characteristics across Queensland

Land clearing has declined across the study period by approximately61% throughout Queensland (QLD), from an average of 7425 km2 peryear during 1988–1991 to 2922 km2 in 2015, though with considerablefluctuation (Fig. 3a). Similarly, the proportion of land clearing con-stituting remnant vegetation has decreased substantially since 1997throughout the state, from 69% to 40%. Clearing trends at the statelevel are most similar to those of the Brigalow Belt South (BBS), whichhas also seen a 62% reduction in clearing and a larger reduction inremnant clearing (Fig. 3b). Trends in the Great Barrier Reef catchment(GBRC) are also similar to QLD and BBS, though this region has ex-perienced a greater decline in clearing (76%) (Fig. 3c). South EasternQueensland (SEQ) and Cape York Peninsula (CYP) are most dissimilarfrom other regions, with SEQ maintaining similar cyclical clearing ratesover time and the CYP experiencing a 304% increase in remnant ve-getation clearings (Fig. 3d,e). The uncharacteristic peak of clearingobserved at the state level in 2000 is only reflected in the GBRC. For

most regions, the interim and restriction periods saw a gradual decreasein clearing extent, though all regions experienced increasing clearingrates during the subsequent relaxation period.

Clearing purposes within the BBS, GBRC, and SEQ have primarilybeen for pastures (Fig. A.2). For all regions, the proportions of clearingfor cropping was substantial during the unregulated and reform periods,but has since become marginal. Trends in the BBS have remained re-latively consistent, though infrastructure and thinning purposes haveincreased slightly since restriction. The GBRC has also shown con-sistency over time, but clearing purposes are more diverse than the BBS,including a noticeable increase in mining clearance during 2005–2013.Given the population density of SEQ, this region has the highestclearance rates for settlements and consistent clearing rates for infra-structure. Initially moderate, pasture clearings in SEQ increased indominance to peak rates during the interim period, beyond which theproportion of clearing for settlements and timber plantations began toincrease, and the frequency of pasture clearing diminished. Clearingpurposes in CYP have seen mining replace infrastructure developmentover time, particularly following the interim period. Like SEQ, pastureclearings gained an atypical dominance during regulation.

Historical clearing in the BBS consists of patchy hot- and cold-spotsin the landscape, where some properties are responsible for a largelydisproportionate share of the region’s clearing extent during this time(Fig. A.3). This is in contrast to the GBRC, which exhibits a morehomogeneous extent of moderate- to high-frequency clearings acrossthe landscape, where more landholders have contributed more equallyto the overall clearing extent. Few properties in SEQ have frequentlycleared in the last 26 years, suggesting that many clearings are one-offevents, as would be expected given the frequency of clearing for set-tlements and infrastructure in this region. Despite the small amount ofclearing occurring in the CYP, a number of properties are clearingfrequently. Though the greatest frequency of clearing is found onmining sites, most properties undergoing continual small-scale clearingregimes are in relatively natural landscapes, indicative of new pastureclearings or necessary infrastructure. It should be noted that land-holders will often need to re-clear previous areas due to the resilience ofmany tree species. While some regrowth is old enough to produce aheight and biomass signature detectable by satellite imagery and clas-sifiable by SLATS, we suspect the majority of regrowth is cleared beforereaching such a state. Thus, it is expected that re-clearing of previousareas is continual, and largely unidentifiable, but some landholdersclear new areas on their property more frequently than others.

3.2. Principal component analyses and clustering

3.2.1. State of QueenslandThe first rotated component (RC) produced a strong loading for

remoteness (0.78) and moderately strong loadings for parcel size (0.73)and tenure (−0.71), indicating that high values on this RC reflectclearing events on more remote areas, larger parcels of land, andgreater frequency on leasehold properties (Tables C.1, C.2). Thus weinterpreted this first RC, which captures 24% of the variance, as ameasure of property characteristics, deemed the Property Component.The second RC captures 20% of the variance and produced strongloadings for rainfall variability (−0.89) and drought declarations(0.76), reflecting clearing events in regions with more consistent rain-fall patterns that are historically more prone to droughts, which weinterpreted further as a Climate Component. The third RC (16% of var-iance) exhibited a strong loading for elevation (0.82) and a moderatelystrong loading for slope (0.73), reflecting geographic characteristics ofclearing events (higher elevations and steeper slopes), interpreted as aTerrain Component. The fourth RC only exhibited strong loadings on REthreat status (0.91) and was thus removed from further analysis(Table 2). Cumulatively, the Property, Climate, and Terrain Compo-nents captured 60% of the variance in the data.

When plotting the component scores by policy period, large

Table 2Composition of each rotated component, ordered according to the variancecaptured (v), and subsequent interpretation of the components’ measurement.Composition is represented as variable (correlation with component, r). Variablesare included when r > |0.34|.

Region Component v Composition Interpretation

QLD 1 24% Remoteness (0.78) Property ComponentParcel size (0.73)Tenure (−0.71)

2 20% Rainfall variability(−0.89)

Climate Component

Drought declarations(0.76)

3 16% Elevation (0.82) Terrain ComponentSlope (0.73)

BBS 1 37% Remoteness (0.88) Property ConditionsComponentElevation (0.79)

Drought declarations(−0.71)Parcel size (0.71)Tenure (−0.69)

2 17% Slope (0.71) Pasture ExpansionComponentRE threat status (0.70)

Clearing type (0.51)

GBRC 1 29% Remoteness (0.85) Property ComponentTenure (−0.78)Parcel size (0.78)

2 18% Rainfall variability(0.81)

Climate GuidanceComponent

Clearing type (−0.77)

SEQ 1 34% Drought declarations(0.90)

Ecosystem Component

Rainfall variability(−0.83)Elevation (0.79)RE threat status(−0.40)

2 16% Tenure (0.83) Pasture GrowthComponentClearing type (0.60)

CYP 1 28% Remoteness (0.78) Pastoral SuitabilityComponentElevation (0.68)

Parcel size (0.68)Clearing type (0.61)

2 26% RE threat status(−0.74)

Minority FeaturesComponent

Slope (0.67)Tenure (0.67)

B.A. Simmons et al. Land Use Policy 75 (2018) 399–410

404

overlaps between policy periods occurred across all RCs, with the mostdistinguishable deviations occurring during regulation/interim and re-striction periods (Fig. 4). The restriction period was most distinguishable,scoring lowest on the Property and Climate components and highest onthe Terrain Component, indicating a greater proclivity toward clearingon atypical properties (those less frequently cleared in the past) that areunder biophysical conditions more classically suitable for agriculture.This is most closely related to patterns during reform, but opposes thetrends during regulation/interim periods, which saw a return to clearingson previously common properties but in slightly less optimal locationsfor agriculture. Unregulated and relaxation periods remained similar andscored more modestly along all RCs. At the state level, clearing patternswere thus largely consistent over time, though some dissimilarities canbe identified. The cluster analysis confirmed this pattern, distinguishingonly two clusters based on the mean component scores of the policyperiods: (1) reform, restriction, and relaxation, (2) unregulated, reg-ulation, and interim (Fig. A.4). Periods composing the second clusterrepresent some of the highest clearing rates in the state, indicating thatlandholders have responded similarly according to how much theyclear and where they clear.

3.2.2. Brigalow Belt SouthThe first RC in the BBS represented the same relationships as the

Property Component for QLD, with the additional influence of drought-prone locations and elevation. Thus we interpreted this component as aProperty Conditions Component. High values along the second RC re-presented clearing events on steeper slopes, targeting of ‘least concern’(unthreatened) vegetation, and conducted primarily for pastures. Weinterpreted this component as a measurement of landholders’ motiva-tion and ability to clear, whereby landholders would be expanding theirpastures to less-attractive locations (steeper slopes) without removingvegetation under VMA protection. We defined this RC as a PastureExpansion Component (Table 2).

The unregulated period was the most distinct in the BBS across thetwo RCs, scoring lowest on Pasture Expansion and highest on PropertyConditions (Fig. 5a). As expected, clearing during this period wasmarked by unrestrained preferences, likely guided by choosing the mostsuitable locations regardless of the type of vegetation present. Theperiod of reform signalled a noticeable decline along the PropertyConditions Component, indicating an increase in the proportion ofclearing in historically infrequent locations. Although regulation saw thefirst increase in Pasture Expansion, it continued to share commoncharacteristics with the reform period. The remaining policy periodsgenerally consisted of negative Property Conditions scores and highPasture Expansion scores, and exhibited considerable year-to-yearoverlap (Fig. A.5). The cluster analysis confirmed this relative ambi-guity between years, resulting in two clusters where only the initialunregulated period was distinguished: (1) unregulated, (2) interim, re-form, regulation, restriction, and relaxation (Fig. A.6). Despite the highvariation in clearing extent in the BBS (Fig. 3b), it appears that clearingcharacteristics since policy reform have remained relatively similar.

3.2.3. Great Barrier Reef catchmentLike QLD, the primary RC identified in the GBRC was the Property

Component. The second RC was associated with clearing events forpastures in areas with more consistent rainfall patterns—a possible risk-management strategy; we interpreted this as a Climate GuidanceComponent (Table 2). The unregulated, regulation, and restriction periodsare the most distinct across these two components (Fig. 5b), re-presenting a gradual shift from pasture clearings on typical ruralproperties in more climatically stable locations, toward more non-pas-toral clearings (such as mining and settlements) in areas with greaterrainfall variability. While the majority of the regulation years scoredhighest along the Property Component (and neutral along the ClimateGuidance Component), the years immediately following reform andpreceding interim were less distinct (Fig. A.5). These distinctions re-sulted in three clusters: (1) unregulated and regulation, (2) restriction,(3) reform, interim, and relaxation (Fig. A.6).

3.2.4. South Eastern QueenslandThe dominant RC for SEQ was interpreted as an Ecosystem

Component, a representation of both climatic and topographic qualitiesinfluencing ecosystem characteristics (Table 2). Less influential was aPasture Growth Component, a measurement of new pasture clearings onfreehold properties, related to (though distinct from) the Pasture Ex-pansion Component in the BBS. Prior to the broad-scale clearing ban in2007, clearing patterns gradually increased along the two RCs, signal-ling an atypical increase in the proportion of pasture developments onfreehold tenures in higher and drier environments (Fig. 5c). Restrictionbrought about a return to declining Pasture Growth, as settlements andinfrastructure development began to replace the previous pastoralclearings. This eventually led to the most distinct period, relaxation,which saw a stark decline in the frequency of pasture clearings onfreehold lands in historically common ecosystems in exchange for in-creased timber plantations and thinning (Fig. A.5). These patternsproduced three clusters: (1) relaxation, (2) unregulated and reform, (3)regulation, interim, and restriction (Fig. A.6).

3.2.5. Cape York PeninsulaClearing patterns within CYP were distinguished by two final RCs

(Table 2). The Pastoral Suitability Component included the propertiesmost beneficial to pasture conversion (high remoteness, larger parcelsof land, relatively higher elevations, and more pasture clearings). Thesecond RC, the Minority Features Component, represented more clearingof threatened vegetation on steeper slopes, within primarily freeholdproperties. Because threatened REs, steep slopes, and freehold proper-ties are less common in the CYP landscape, this RC captures land-holders’ proclivity for clearing locations within the minority of the re-gion’s spatial features. The reform, regulation, and interim periods aremost distinct. Minority Features peaked during reform, when clearingfor cropping was prevalent, and eventually declined to previous levelsfollowing regulation, when the dominant clearing purposes shifted frompastures to mining and infrastructure. Pastoral Suitability, however,

Fig. 4. Mean component scores and standard deviation ellipses by policy period across the final three rotated components for the State of Queensland.

B.A. Simmons et al. Land Use Policy 75 (2018) 399–410

405

continued to increase until restriction (Fig. 5d). Characteristics remainedsimilar during the unregulated period and the most recent periods, re-striction and relaxation, indicating a largely cyclical trend in clearingcharacteristics over time, despite shifting clearing purposes (Fig. A.5).Four distinct clusters were identified: (1) regulation, (2) reform, (3)interim, (4) unregulated, restriction, and relaxation (Fig. A.6).

3.3. Synthesis of regional findings

Clearing during the unregulated years was characterised by ‘frontier’clearing—carried out in large, remote areas with low drought risk,where rural landholders had the freedom to clear large areas of rela-tively abundant remnant vegetation. At most spatial scales, the reformperiod coincided with the beginning of a shift in the proportion ofclearing occurring in locations less suitable for agriculture as morelandholders cleared available vegetation on freehold properties nowthat leasehold lands were under new restrictions (Fig. 5). After assent ofthe VMA (regulation), all regions experienced atypical shifts in clearingcharacteristics, with most turning to larger, remote areas for theirpasture expansions, largely irrespective of regional climatic constraints(though at the state level these areas were often in drier regions). Thesetrends remained relatively similar in most regions during the interimperiod, though the allowance of some broad-scale clearing permits bythe government likely facilitated regions like the BBS and GBRC toexpand their pastures to property-level characteristics reflective ofthose during reform.

The restriction period exhibited some of the most distinct changes in

clearing patterns, with aggregate clearing events shifting towardsmaller, more accessible freehold properties where climatic and terrainconditions were less extreme (Fig. 4). This state-wide trend is mirroredin the GBRC, though clearing patterns in smaller regions like SEQ andCYP exhibited their own unique deviations from previous norms. Forexample, restriction was characterised by a decline in pasture clearingson freehold lands in SEQ, which was uncharacteristic of the previousdecade, and clearing regimes in CYP dramatically receded from mar-ginal lands, beyond what had characterised the region’s clearing historysince 1991. While the BBS continued to exhibit high inter-year varia-bility, clearing characteristics of landholders in this region have re-mained generally consistent since regulation. With the notable exceptionof CYP, the reduction of pasture developments during this period is alikely result of the increasing policy restrictions, and more intensiveland uses in private, semiurban areas began to rise. Aside from SEQ, thefinal relaxation period was characterised by a return to more moderatecharacteristics (near-zero component scores), reflective of previouspolicy periods, in all regions. The aforementioned trends are most ap-plicable to the large clearing hotspot regions of the BBS and GBRC,which are most representative of state level patterns.

4. Discussion

4.1. Highlighting the heterogeneity of clearing patterns

This study investigated the characteristics of spatial and temporalpatterns of land clearing at state and regional levels amidst

Fig. 5. Mean component scores and standard deviation ellipses by policy period across the final rotated components for (a) Brigalow Belt South (BBS), (b) GreatBarrier Reef catchment (GBRC), (c) South Eastern Queensland (SEQ), and (d) Cape York Peninsula (CYP).

B.A. Simmons et al. Land Use Policy 75 (2018) 399–410

406

considerable policy development and uncertainty. The results highlightthe importance of identifying regional-scale patterns across multiplefacets of land clearing, as we observed some similarities, but also manyregional differences in landholder clearing patterns relative to regula-tion, in terms of the extent, frequency, timing, and characteristics ofclearing events. Overall, aggregate landholders’ clearing characteristicsremain generally consistent over time—areas that are most suitable forpasture development and expansion are primarily cleared. Clearing ineach region, however, is likely driven by different factors based upontheir unique characteristics. The results of this analysis illustrate howlandholders in different regions were differentially altering theirclearing behaviours amidst policy change and what different factorscharacterised their clearing regimes. This improved understanding ofstate and regional level responses surrounding clearing regulation islikely to benefit future policy development, as goals and targets can beadapted to reflect the relevant clearing dynamics across the state.Conclusions drawn from these results thus present a cautionary tale forresearchers and policy-makers around the world: using coarse measuresto understand deforestation patterns and landholder responses to con-servation policy may not result in the intended outcomes if they over-look more locally-relevant drivers of clearing.

In the BBS and GBRC, clearing behaviours prior to regulation largelyrepresented ‘frontier’ clearing, reflecting historical broad-scale clearingpatterns driven primarily by targeting remote lands with greater agri-cultural suitability and potential profitability (Kirkpatrick, 1999;Lindenmayer et al., 2005). This apparent economic rationality dimin-ished as new regulations developed, affecting these two clearing hot-spots in different ways. The importance of potential profitability ap-pears strongest for BBS landholders, yet the region’s extensive clearinghistory has reduced the number of profitable clearing options available.BBS landholders likely showed minimal variation in their clearingspatial characteristics in the years following policy reform due to theircontinued reliance on selecting the most profitable, available areas tocompensate for growing policy restrictions and declining terms of trade(Seabrook et al., 2006). These landholders may thus be selectivelyclearing with a focus on the quality rather than the quantity of clearedland. Non-pastoral land uses in the GBRC, however, are less dispersedthroughout the region and more vegetation is available for clearing(Fig. 1), allowing for pasture clearings to be less concentrated intohotspots (Fig. A3). Landholders in the GBRC thus are limited more byregulatory opportunity than spatial constraints; this is reflected in theirtendency for frontier clearing prior to the broad-scale clearing ban andthe subsequent shift during policy restriction, where clearing for exemptor permissible purposes became more prominent, such as mining, set-tlements, and infrastructure. These motivations reflect documentedtestimonials in Productivity Commission (2004), which highlighted BBSfarmers’ proclivity for targeting as much productive vegetation aspossible, as well as GBRC sugarcane farmers’ unwillingness to conserveunproductive lands due to clearing bans on productive, vegetated lands.

Because the CYP and SEQ, in contrast, are less representative ofstate clearing patterns in terms of extent and characteristics, we do notdiscuss their patterns at length. SEQ represents a region of dense urbansprawl surrounded by pastoral land, and thus landholders have re-sponded by taking advantage of new pasture additions on their propertyuntil the broad-scale clearing ban, when pasture clearings droppedsubstantially and intensive uses like settlements and infrastructure be-came the more dominant purposes for clearing. Similarly, the CYP mayhave tried to capitalise on developing new pastures during the earlyyears of regulation, which meant identifying rarer locations in the re-gion that could prove suitable for these new developments. Followingthe clearing ban, clearing for more intensive purposes increased, suchas mining and infrastructure—purposes less constrained by the influ-ences of agricultural suitability. These regions’ tendencies to capitaliseon pasture developments during early regulation may represent a per-verse response to the VMA. Thus, even in regions where pastoralclearings are less common, the uncertainty of future property

management restrictions may have provoked pre-emptive and atypicalclearings typically seen in areas with greater reliance on pastoral andagricultural land uses, like the BBS and GBRC (Mendelsohn, 1994;Zhang, 2001).

4.2. Landholder responses to policy: effectiveness, panic, and uncertainty

The introduction and modification of vegetation managementpolicy has had a significant influence on landholders’ clearing beha-viours in the last 25 years, though the impact may vary substantiallybetween regions. This impact can be characterised in two ways: theextent of clearing and the spatial characteristics of clearing. Thestrengthening of the VMA and the introduction of the broad-scaleclearing ban coincided with a large reduction in remnant clearing ex-tent across most regions, a reduction in pasture developments, and lessfrontier clearing. When coupled with the subsequent rise in totalclearing extent following policy relaxation, some have argued that thispresents (associative) evidence for the effectiveness of the restrictionsset forth by the VMA (e.g. Maron et al., 2015; Evans, 2016). In relationto the clearing characteristics, however, this effectiveness may be lim-ited. Landholders in regions like the GBRC and SEQ have changed theirclearing behaviours in ways more reflective of the goals of the VMA (i.e.less pastoral clearing in historical locations), which is more reflective ofsuccessful forest conservation policies (Fox et al., 2009). Yet in the BBS,clearing patterns have generally been consistent following initial policyreform, and in the CYP, the immunity of mining activities from VMAregulations has likely minimised the effectiveness of the policy. TheVMA may thus have created general change in the extent of clearingevents, yet its ability to change what landholders are clearing variesthroughout Queensland. Indeed, this has recently been identified byRhodes et al. (2017), who found the VMA to be ineffective at reducingclearing rates of the most threatened types of vegetation relative to non-threatened vegetation, despite an overall reduction in clearing events.

Our results shed additional light on the perverse nature of panicclearing following the introduction of the VMA. While anecdotal evi-dence reveals unplanned clearing occurred (Productivity Commission,2004; Senate Inquiry, 2010), our analysis suggests this is more reflectedin the amount of clearing rather than the characteristics of clearing:clearing characteristics during the peak period of panic clearing(1999–2000) did not differ substantially from the previous years in allregions except SEQ (Fig. A.5). While these clearing events may not havebeen planned in the short-term (Productivity Commission, 2004), thelack of significant departures from previous clearing norms suggestslandholders may have reasonably anticipated clearing these areas fur-ther in the future. Panic clearing may be best represented as a four-yearperiod, in which landholders responded to anticipated regulations byimmediately increasing previous clearing practices on their land, fol-lowed by a progression of selective pasture expansion into moreavailable, potentially viable locations. Given the characteristics of panicclearing, we hypothesise that the initial panic clearing response mayrepresent expedited plans by landholders, while the progression mayrepresent unplanned clearings guided by landholders’ existing knowl-edge of potentially profitable locations for expansion. This behaviour isreflective of other pre-emptive behaviours immediately following theintroduction of policies restricting property management rights (e.g.Alston et al., 2000; Aldrich et al., 2012). Despite considerable policyuncertainty still present throughout the timeline, similar incidences ofpanic clearing in Queensland have likely been avoided due to the im-plementation of more retrospective policies and amendments, thoughsome amendments have continued to be rapidly ascended (Kehoe,2009).

While the ongoing changes to vegetation management policy mayhave effectively produced a continuous blanket of uncertainty forlandholders, we would expect peak uncertainty to occur immediatelyprior to (and potentially following) the introduction of new policies orpolitical shifts—or most notably at the transitions between policy

B.A. Simmons et al. Land Use Policy 75 (2018) 399–410

407

periods. Particularly considering the haste with which the VMA 1999and additional amendments were enacted (Kehoe, 2009), effects of thisuncertainty could result in dramatic short-term shifts in clearing char-acteristics during these two-year windows. The potential effects of peakpolicy uncertainty are most observable in the BBS, where (in addition topanic clearing) the frequency of clearing by leasehold landholders im-mediately dropped following the Land Act 1994, and clearings forpasture expansion dramatically increased in the lead-up to the newconservative government in 2012. This region is likely most sensitive touncertainty given the importance of potential profitability and spatiallimitations for expansion. Because policy uncertainty often results inincreased deforestation (Zhang, 2001), the effects of uncertainty in ahotspot such as the BBS have the potential to significantly diminish theeffects of the VMA. SEQ and CYP also have frequent short-term de-viations, but these may be due to the influence of outliers in these re-gions where significantly less clearing activity has occurred. The GBRC,in contrast, has experienced minimal changes surrounding peak periodsof policy uncertainty, potentially due to the region’s suspected relianceon regulatory opportunity rather than availability.

4.3. Implications for policy and future directions

During the recent push by the centrist political party to reinstateprevious land clearing restrictions, AgForce—Queensland’s primarylobby group representing beef, sheep, wool, and grain produ-cers—argued that the inconsistency of the VMA “severely impacts onthe ability of landholders to plan and implement effective long-termproperty and business management decisions” (AgForce, 2016, p. 2).When coupled with arguments of a lack of transparency in the VMA’sgoals, inflexibility of assessment codes, and inconsistencies in land-holders’ interpretations of the Acts (Productivity Commission, 2004;Senate Inquiry, 2010), it stands to reason that the degree of politicaland legislative uncertainty over the last 20 years may have dramaticallyaffected landholder’s clearing regimes. This is most suggestive in thehistorical clearing hotspot of the BBS, where pastures have continued toexpand despite suboptimal biophysical conditions—a trend thatemerged after the first clearing regulations were introduced.

To date, scientists and conservation groups have primarily reliedupon the mere extent of land clearing to quantify the impact of theVMA. This measure, however, does not account for the large degree ofcomplexity of ecological systems nor the behavioural responses oflandholders (Dovers, 2005; Knill et al., 2011). In some cases, the changein clearing extent coincides with expected changes in clearing char-acteristics, such as in the GBRC during the restriction period. Otherregions, such as the BBS, do not exhibit such similarities between extentand characteristics. Most notably, the large similarities in land clearingcharacteristics during regulation and pre-regulation are disconcerting,as landholders have continued to clear in areas where vegetation hasbeen extensively cleared already, thus jeopardising biodiversity andincreasing the potential for land degradation in these areas—two otherpurposes for which the Act was meant to address. In regions with suchconsistent clearing characteristics, initial attempts at regulatingclearing were likely negated by both the large volume of panic clearingand the characteristics of the areas targeted. Indeed, numerous ac-counts from landholders and industry representatives have argued thatthe early impacts of the VMA were negligible at the regional level dueto panic clearing and market or industry conditions, such as the lack ofexpansion of the sugar industry (Productivity Commission, 2004).Possibly more comforting is the degree of similarity between clearingduring restriction and relaxation periods. That is, despite concerns overincreasing rates of clearing since policy relaxation (Maron et al., 2015),landholders largely have not returned to targeting previous locationsobserved during pre-regulation, and the proportion of remnant clearingremains relatively small compared to previous years of similar clearingrates. Successful future policy amendments would thus benefit fromexplicit considerations and monitoring of these heterogeneous,

spatiotemporal variations in landholder clearing behaviours, allowingfor more accurate indicators of the successes and potential perversitiesof vegetation management policy.

This study provides the first step in illuminating the complexities ofland clearance, allowing us to identify regional patterns of clearing anddevelop hypotheses that can be tested at a finer scale and with morerobust causal inference methods. While we focus on Queensland, thefindings are relevant to other regions implementing legislation to reg-ulate deforestation. This study outlines the patterns in the extent of landclearing across Queensland (the ‘what’) and provides the most in-depthspatiotemporal analysis of the characteristics of land clearing patternsto date (the ‘how’). However, the reasoning behind land clearing be-havioural patterns (the ‘why’) is yet to be fully understood.Recognisably, a number of potential influential factors are not includedin this analysis that may explain why landholders are expanding pas-toral lands, why they may preferentially target marginal lands to clear,or why they continue to clear amidst unfavourable biophysical condi-tions. Using the spatially-constant biophysical and property variables asaggregate measures of agricultural suitability accounted for 47–60% ofthe variance in the data, depending upon the region. Additional factorsmust then contribute to a considerable amount of the variation ob-served in clearing locations, and these may include a number of de-mographic, economic, cultural, and personality characteristics thathave noticeably influenced landholders’ environmental behaviours inQueensland (e.g. Seabrook et al., 2006, 2008; Moon and Cocklin, 2011;Emtage and Herbohn, 2012).

Some studies have investigated the various spatiotemporal driversof deforestation in Queensland (e.g. Seabrook et al., 2007; Evans, 2016;Marcos-Martinez et al., 2018), yet as is evident from this study, gen-eralising drivers often misses important regional drivers of deforesta-tion. Using coarse, state-level resolutions for pattern analysis mayidentify overarching trends, but important localised effects may gounnoticed (Dong et al., 2015). Future studies looking at regional andeven sub-regional scale drivers of clearance would greatly benefit ve-getation management policy, with the potential to guide local in-itiatives, programs, or interventions that may be more successful andsustainable. These finer scales have been the backbone of natural re-source management programs in Australia (Hajkowicz, 2009), wherepolycentric or multi-level approaches encourage positive environ-mental outcomes at multiple scales (Ostrom, 2010). An important pieceto the land clearing puzzle also lies in the effectiveness of the VMA.Recent investigations have been made into the effectiveness of thispolicy by analysing the impact on threatened remnant vegetation (e.g.Rhodes et al., 2017), yet there is little account for the myriad of con-founding factors that may be masking the direct causal link betweenregulation and observed clearing. Identifying causal linkages, however,requires a thorough understanding of regional patterns, their char-acteristics, and how much variation is captured by relevant con-founding factors, and this study will provide the first step for suchcausal inference analyses. Future studies seeking more comprehensiveevaluations of the causal impacts of historical vegetation managementpolicy in Queensland will be crucial to developing new policies that canachieve greater long-term, bipartisan stability.

5. Conclusions

Analysis of deforestation patterns based simply upon the extent oftree loss ignores the characteristics of deforestation and the factors thatmay be driving landholders’ responses to deforestation policies. Inparticularly large regions or countries, aggregate trends may not cap-ture localised drivers of deforestation. When policies utilise these gen-eralisations to implement regulations and restrictions, this could elicituncertainties for landholders, who may not be able to reconcile thesepolicies with their own motivations and rationales for clearing. Morestudies investigating the role of conservation policy, and particularlythe consistency of such policies, are needed to enhance our

B.A. Simmons et al. Land Use Policy 75 (2018) 399–410

408

understanding of the effectiveness of conservation policy instruments.Further, the effectiveness of these policy instruments needs to extendbeyond the sheer numbers, where policy-makers can determine if therealised characteristics of clearing are aligning with the intended goalsand objectives of the policy. If regional patterns and drivers of defor-estation can be identified, more relevant and explicit interventions canbe developed that may ultimately prove most effective and sustainableover time.

Conflicts of interest

None.

Acknowledgements

This work was supported by the Discovery and Future Fellowshipprograms of the Australian Research Council, and the AustralianResearch Council Centre of Excellence for Environmental Decisions(CE11001000104), funded by the Australian Government.

Appendix A. Supplementary data

Supplementary material related to this article can be found, in theonline version, at doi:https://doi.org/10.1016/j.landusepol.2018.03.049.

References

AgForce, 2016. Submission to: Vegetation Management (Reinstatement) and OtherLegislation Amendment Bill 2016. Submission No. 598. Queensland Parliament,Brisbane.

Agrawal, A., 2007. Forests, governance, and sustainability: common property theory andits contributions. Int. J. Commons 1 (1), 111–136.

Aldrich, S.P., 2012. Contested groves: forest reserves and land conflict in the easternAmazon. J. Lat. Am. Geogr. 11 (2), 73–101. http://dx.doi.org/10.1353/lag.2012.0042.

Aldrich, S., Walker, R., Simmons, C., Caldas, M., Perz, S., 2012. Contentious and changein the Amazon’s arc of deforestation. Ann. Assoc. Am. Geogr. 102 (1), 103–128.http://dx.doi.org/10.1080/00045608.2011.620501.

Alston, L.J., Libecap, G.D., Mueller, B., 2000. Land reform policies, the sources of violentconflict, and implications for deforestation in the Brazilian Amazon. J. Environ. Econ.Manag. 39 (2), 162–188. http://dx.doi.org/10.1006/jeem.1999.1103.

Angelsen, A., 2009. Policy options to reduce deforestation. In: Angelsen, A. (Ed.),Realising REDD+: National Strategy and Policy Options. CIFOR, Bogor, Indonesia.

Angelsen, A., Kaimowitz, D., 1999. Rethinking the causes of deforestation: lessons fromeconomic models. World Bank Res. Obs. 14 (1), 73–98. http://dx.doi.org/10.1093/wbro/14.1.73.

Assunção, J., Gandour, C., Rocha, R., 2012. Deforestation Slowdown in the Legal Amazon:Prices or Policies? Climate Policy Initiative Working Paper. Pontifica UniversidadeCatólica, Rio de Janeiro.

[dataset] Atlas of Living Australia, 2016. Accessibility/Remoteness Index of Australia[Data set]. Retrieved from http://spatial.ala.org.au/layers/.

Australian Bureau of Rural Sciences, 2010. Australia’s Forests at a Glance 2010.Australian Government Department of Agriculture, Fisheries and Forestry, Canberra.

Barbier, E.B., Burgess, J.C., 2001. Tropical deforestation, tenure insecurity, and un-sustainability. For. Sci. 47 (4), 497–509.

Bartlett, M.S., 1950. Tests of significance in factor analysis. Br. J. Psychol. 3 (2), 77–85.http://dx.doi.org/10.1111/j.2044-8317.1950.tb00285.x.

Bradshaw, C.J.A., 2012. Little left to lose: deforestation and forest degradation inAustralia since European colonization. J. Plant Ecol. 5 (1), 109–120. http://dx.doi.org/10.1093/jpe/rtr038.

Braithwaite, L.W., 1996. Conservation of arboreal herbivores: the Australian scene. Aust.J. Ecol. 21 (1), 21–30. http://dx.doi.org/10.1111/j.1442-9993.1996.tb00582.x.

Brown, D.S., Brown, J.C., Brown, C., 2016. Land occupations and deforestation in theBrazilian Amazon. Land Use Policy 54, 331–338. http://dx.doi.org/10.1016/j.landusepol.2016.02.003.

Busch, J., Ferretti-Gallon, K., 2017. What drives deforestation and what stops it? A meta-analysis. Rev. Environ. Econ. Policy 11 (1), 3–23. http://dx.doi.org/10.1093/reep/rew013.

Cattell, R.B., 1966. The scree test for the number of factors. Multivar. Behav. Res. 1 (2),245–276. http://dx.doi.org/10.1207/s15327906mbr0102_10.

Cogger, H., Ford, H., Johnson, C., Holman, J., Butler, D., 2003. Impacts of land clearingon Australian wildlife in Queensland. WWF Aust. Rep. 48.

Costello, A.B., Osborne, J.W., 2005. Best practices in exploratory factor analysis: fourrecommendations for getting the most from your analysis. Pract. Assess. Res. Eval. 10(7), 1–9.

Crowley, G.M., Garnett, S.T., 1998. Vegetation change in the grasslands and grassy

woodlands of east-central Cape York Peninsula, Australia. Pac. Conserv. Biol. 4 (2),132–148. http://dx.doi.org/10.1071/PC980132.

Daniels, A.E., 2009. Forest expansion in Northwest Costa Rica: conjecture of the globalmarket, land-use intensification, and forest protection. In: Nagendra, H., Southworth,J. (Eds.), Reforesting Landscapes: Linking Pattern and Process. Springer, Dordrecht.

Deacon, R.T., 1994. Deforestation and the rule of law in a cross-section of countries. LandEcono. 70 (4), 414–430. http://dx.doi.org/10.2307/3146638.

DeFries, R., Pandey, D., 2010. Urbanization, the energy ladder and forest transitions inIndia’s emerging economy. Land Use Policy 27 (2), 130–138. http://dx.doi.org/10.1016/j.landusepol.2009.07.003.

Demšar, U., Harris, P., Brunsdon, C., Fotheringham, A.S., McLoone, S., 2013. Principalcomponent analysis on spatial data: an overview. Ann. Assoc. Am. Geogr. 103 (1),106–128. http://dx.doi.org/10.1080/00045608.2012.689236.

Dong, M., Bryan, B.A., Connor, J.D., Nolan, M., Gao, L., 2015. Land use mapping errorintroduces strongly-localised, scale-dependent uncertainty into land use and eco-system services modelling. Ecosyst. Serv. 15, 63–74. http://dx.doi.org/10.1016/j.ecoser.2015.07.006.

Dovers, S., 2005. Environment and Sustainability Policy: Creation, Implementation,Evaluation. Federation Press, Sydney.

Emtage, N., Herbohn, J., 2012. Assessing rural landholders diversity in the wet tropicsregion of Queensland, Australia in relation to natural resource management pro-grams: a market segmentation approach. Agric. Syst. 110, 107–118. http://dx.doi.org/10.1016/j.agsy.2012.03.013.

Evans, M.C., 2016. Deforestation in Australia: drivers, trends and policy responses. Pac.Conserv. Biol. 22 (2), 130–150. http://dx.doi.org/10.1071/PC15052.

Fensham, R.J., Fairfax, R.J., 2003. A land management history for central Queensland,Australia as determined from land-holder questionnaire and aerial photography. J.Environ. Manag. 68 (4), 409–420. http://dx.doi.org/10.1016/S0301-4797(03)00110-5.

Fensham, R.J., McCosker, J.C., Cox, M.J., 1998. Estimating clearance in Acacia-domi-nated ecosystems in central Queensland using land-system mapping data. Aust. J.Bot. 46 (2), 305–319. http://dx.doi.org/10.1071/BT96129.

Fox, J., Fujita, Y., Ngidang, D., Peluso, N., Potter, L., et al., 2009. Policies, political-economy, and swidden in Southeast Asia. Hum. Ecol. 37 (3), 305–322. http://dx.doi.org/10.1007/s10745-009-9240-7.

Gaveau, D.L.A., Epting, J., Lyne, O., Linkie, M., Kumara, I., et al., 2009. Evaluatingwhether protected areas reduce tropical deforestation in Sumatra. J. Biogeogr. 36(11), 2165–2175. http://dx.doi.org/10.1111/j.1365-2699.2009.02147.x.

Geist, H.J., Lambin, E.F., 2002. Proximate causes and underlying driving forces of tropicaldeforestation. BioScience 52 (2), 143–150. http://dx.doi.org/10.1641/0006-3568(2002)052[0143:PCAUDF]2.0.CO;2.

Geoghegan, J., Villar, S.C., Klepeis, P., Mendoza, M.P., Ogneva-Himmelberger, Y.,Chowdhury, R.R., et al., 2001. Modeling tropical deforestation in the southernYucatan peninsular region: comparing survey and satellite data. Agric. Ecosyst.Environ. 85 (1–3), 25–46. http://dx.doi.org/10.1016/S0167-8809(01)00201-8.

Great Barrier Reef Marine Park Authority, 2014. Great Barrier Reef Outlook Report 2014.GBRMPA, Townsville.

Hajkowicz, S., 2009. The evolution of Australia’s natural resource management programs:towards improved targeting and evaluation of investments. Land Use Policy 26 (2),471–478. http://dx.doi.org/10.1016/j.landusepol.2008.06.004.

Hosonuma, N., Herold, M., De Sy, V., De Fries, R.S., Brockhaus, M., Verchot, L., Angelsen,A., Romijn, E., 2012. An assessment of deforestation and forest degradation drivers indeveloping countries. Environ. Res. Lett. 7 (4). http://dx.doi.org/10.1088/1748-9326/7/4/044009. Article 044009.

Husson, F., Josse, J., Le, S., Mazet, J., 2017. FactoMineR: Multivariate Exploratory DataAnalysis and Data Mining With R. R Package Version 1.35. http://CRAN.R-project.org/package=FactoMineR.

Joppa, L.N., Pfaff, A., 2011. Global protected area impacts. Proc. R. Soc. B 278 (1712),1633–1638. http://dx.doi.org/10.1098/rspb.2010.1713.

Kaimowitz, D., Byron, N., Sunderlin, W., 1998. Public policies to reduce inappropriatetropical deforestation. In: Lutz, E. (Ed.), Agriculture and the Environment:Perspectives on Sustainable Rural Development. The World Bank, Washington, DC.

Kaiser, H.F., 1958. The varimax criterion for analytic rotation in factor analysis.Psychometrika 23 (3), 187–200. http://dx.doi.org/10.1007/BF02289233.

Kaiser, H.F., 1970. A second generation little jiffy. Psychometrika 35 (4), 401–415.http://dx.doi.org/10.1007/BF02291817.

Kehoe, J., 2009. Environmental law making in Queensland: the vegetation managementact 1999 (Qld). Environ. Plann. Law J. 26 (5), 392–410.

Kirkpatrick, J.B., 1999. A Continent Transformed: Human Impact on the NaturalVegetation of Australia. Oxford University Press, Melbourne.

Klock, J., 1995. The advancement of South Korean forestry: national development andforest policy from 1910 to present. Korea J. 35 (1), 119–132.

Knill, C., Schulze, K., Tosun, J., 2011. Measuring Environmental Policy Change:Conceptual Alternatives and Research Implications. Political Science Series, Institutefor Advanced Studies, Vienna.

Kull, C.A., Ibrahim, C.K., Meredith, T.C., 2007. Tropical forest transitions and globali-zation: neo-liberalism, migration, tourism, and international conservation agendas.Soc. Nat. Resour. 20 (8), 723–737. http://dx.doi.org/10.1080/08941920701329702.

Lindenmayer, D., 2005. Vegetation loss and degradation. In: Lindenmayer, D., Burgman,M. (Eds.), Practical Conservation Biology. CSIRO Publishing, Melbourne.

Liu, C.W., Lin, K.H., Kuo, Y.M., 2003. Application of factor analysis in the assessment ofgroundwater quality in a blackfoot disease area in Taiwan. Sci. Total Environ. 313(1–3), 77–89. http://dx.doi.org/10.1016/S0048-9697(02)00683-6.

Lueck, D., Michael, J.A., 2003. Preemptive habitat destruction under the endangeredspecies act. J. Law Econo. 46 (1), 27–60. http://dx.doi.org/10.1086/344670.

Macintosh, A., 2007. The National Greenhouse Accounts and Land Clearing: Do the

B.A. Simmons et al. Land Use Policy 75 (2018) 399–410

409

https://doi.org/10.1016/j.landusepol.2018.03.049https://doi.org/10.1016/j.landusepol.2018.03.049http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0005http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0005http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0005http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0010http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0010http://dx.doi.org/10.1353/lag.2012.0042http://dx.doi.org/10.1353/lag.2012.0042http://dx.doi.org/10.1080/00045608.2011.620501http://dx.doi.org/10.1006/jeem.1999.1103http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0030http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0030http://dx.doi.org/10.1093/wbro/14.1.73http://dx.doi.org/10.1093/wbro/14.1.73http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0040http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0040http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0040http://spatial.ala.org.au/layers/http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0050http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0050http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0055http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0055http://dx.doi.org/10.1111/j.2044-8317.1950.tb00285.xhttp://dx.doi.org/10.1093/jpe/rtr038http://dx.doi.org/10.1093/jpe/rtr038http://dx.doi.org/10.1111/j.1442-9993.1996.tb00582.xhttp://dx.doi.org/10.1016/j.landusepol.2016.02.003http://dx.doi.org/10.1016/j.landusepol.2016.02.003http://dx.doi.org/10.1093/reep/rew013http://dx.doi.org/10.1093/reep/rew013http://dx.doi.org/10.1207/s15327906mbr0102_10http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0090http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0090http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0095http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0095http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0095http://dx.doi.org/10.1071/PC980132http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0105http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0105http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0105http://dx.doi.org/10.2307/3146638http://dx.doi.org/10.1016/j.landusepol.2009.07.003http://dx.doi.org/10.1016/j.landusepol.2009.07.003http://dx.doi.org/10.1080/00045608.2012.689236http://dx.doi.org/10.1016/j.ecoser.2015.07.006http://dx.doi.org/10.1016/j.ecoser.2015.07.006http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0130http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0130http://dx.doi.org/10.1016/j.agsy.2012.03.013http://dx.doi.org/10.1016/j.agsy.2012.03.013http://dx.doi.org/10.1071/PC15052http://dx.doi.org/10.1016/S0301-4797(03)00110-5http://dx.doi.org/10.1016/S0301-4797(03)00110-5http://dx.doi.org/10.1071/BT96129http://dx.doi.org/10.1007/s10745-009-9240-7http://dx.doi.org/10.1007/s10745-009-9240-7http://dx.doi.org/10.1111/j.1365-2699.2009.02147.xhttp://dx.doi.org/10.1641/0006-3568(2002)052[0143:PCAUDF]2.0.CO;2http://dx.doi.org/10.1641/0006-3568(2002)052[0143:PCAUDF]2.0.CO;2http://dx.doi.org/10.1016/S0167-8809(01)00201-8http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0175http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0175http://dx.doi.org/10.1016/j.landusepol.2008.06.004http://dx.doi.org/10.1088/1748-9326/7/4/044009http://dx.doi.org/10.1088/1748-9326/7/4/044009http://CRAN.R-project.org/package=FactoMineRhttp://CRAN.R-project.org/package=FactoMineRhttp://dx.doi.org/10.1098/rspb.2010.1713http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0200http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0200http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0200http://dx.doi.org/10.1007/BF02289233http://dx.doi.org/10.1007/BF02291817http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0215http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0215http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0220http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0220http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0225http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0225http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0230http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0230http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0230http://dx.doi.org/10.1080/08941920701329702http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0240http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0240http://dx.doi.org/10.1016/S0048-9697(02)00683-6http://dx.doi.org/10.1086/344670http://refhub.elsevier.com/S0264-8377(17)30860-8/sbref0255

Numbers Stack Up? Research Paper No. 38. Australia Institute, pp. 1–22.Marcos-Martinez, R., Bryan, B.A., Schwabe, K.A., Connor, J.D., Law, E.A., 2018. Forest

transition in developed agricultural regions needs efficient regulatory policy. For.Policy Econ. 86, 65–75. http://dx.doi.org/10.1016/j.forpol.2017.10.021.

Maron, M., Laurance, B., Pressey, B., Catterall, C.P., Watson, J., Rhodes, J., 2015. Landclearing in Queensland triples after policy ping pong. Conversat. Aust (Accessed 15December 2016). Retrieved from. http://theconversation.com/land-clearing-in-queensland-triples-after-policy-ping-pong-38279.

Mather, A.S., 2007. Recent Asian forest transitions in relation to forest-transition theory.Int. For. Rev. 9 (1), 491–502. http://dx.doi.org/10.1505/ifor.9.1.491.

McAlpine, C.A., Fensham, R.J., Temple-Smith, D.E., 2002. Biodiversity conservation andvegetation clearing in Queensland: principles and thresholds. Rangel. J. 24 (1),36–55. http://dx.doi.org/10.1071/RJ02002.

McGrath, C., 2007. End of broadscale clearing in Queensland. Environ. Plann. Law J. 24(1), 5–13.

McGrath, C., 2010. Does Environmental Law Work? How to Evaluate the Effectiveness ofan Environmental Legal System. Lambert Academic Publishing, Saarbrücken,Germany.

McIntyre, S., McIvor, J.G., Heard, K.M. (Eds.), 2002. Managing and Conserving GrassyWoodlands. CSIRO, Melbourne.

Mendelsohn, R., 1994. Property rights and tropical deforestation. Oxford EconomicPapers 46. pp. 750–756.

Meyfroidt, P., Lambin, E.F., 2011. Global forest transition: prospects for an end to de-forestation. Annu. Rev. Environ. Resour. 36, 343–371. http://dx.doi.org/10.1146/annurev-environ-090710-143732.

Miteva, D.A., Pattanayak, S.K., Ferraro, P.J., 2012. Evaluation of biodiversity policy in-struments: what works and what doesn’t? Oxf. Rev. Econ. Policy 28 (1), 69–92.http://dx.doi.org/10.1093/oxrep/grs009.

Moon, K., Cocklin, C., 2011. Participation in biodiversity conservation: motivations andbarriers of Australian landholders. J. Rural Stud. 27 (3), 331–342. http://dx.doi.org/10.1016/j.jrurstud.2011.04.001.

Müller, K., 2014. Kimisc: Kirill’s Miscellaneous Functions. R Package Version 0.3. http://CRAN.R-project.org/package=kimisc.

Oliveira, P.J.C., Asner, G.P., Knapp, D.E., Almeyda, A., Galván-Gildemeister, R., et al.,2007. Land-use allocation protects the Peruvian Amazon. Science 317 (5842),1233–1236. http://dx.doi.org/10.1126/science.1146324.

Ostrom, E., 2010. Polycentric systems for coping with collective action and global en-vironmental change. Glob. Environ. Change 20 (4), 550–557. http://dx.doi.org/10.1016/j.gloenvcha.2010.07.004.

Peterson, A., McAlpine, C.A., Ward, D., Rayner, S., 2007. New regionalism and natureconservation: lessons from South East Queensland, Australia. Landsc. Urban Plan. 82(3), 132–144. http://dx.doi.org/10.1016/j.landurbplan.2007.02.003.

Pichón, F.J., 1997. Settler households and land-use patterns in the Amazon frontier: farm-level evidence from Ecuador. World Dev. 25 (1), 67–91. http://dx.doi.org/10.1016/S0305-750X(96)00091-5.

Productivity Commission, 2004. Impacts of Native Vegetation and BiodiversityRegulations (Report No. 29). Productivity Commission, Melbourne.

Queensland Department of Science, Information Technology and Innovation, 2016. LandCover Change in Queensland 2014-15: A Statewide Landcover and Trees Study(SLATS) Report. DSITI, Brisbane.

[dataset] Queensland Spatial Catalogue (QSC), 2016a. Agricultural land audit – historicdrought declarations [Data set]. Retrieved from http://qldspatial.information.qld.gov.au/catalogue/.

[dataset] Queensland Spatial Catalogue (QSC), 2016b. Agricultural land audit – standarddeviation of rainfall – all years – Queensland [Data set]. Retrieved from http://qldspatial.information.qld.gov.au/catalogue/.

[dataset] Queensland Spatial Catalogue (QSC), 2016c. Cadastral data – Queensland series[Data set]. Retrieved from http://qldspatial.information.qld.gov.au/catalogue/.

[dataset] Queensland Spatial Catalogue (QSC), 2016d. Digital elevation model – 3 second– Queensland [Data set]. Retrieved from http://qldspatial.information.qld.gov.au/catalogue/.

[dataset] Queensland Spatial Catalogue (QSC), 2016e. Pre-clearing Broad VegetationGroups of Queensland (Version 3) [Data set]. Retrieved from http://qldspatial.information.qld.gov.au/catalogue/.

[dataset] Queensland Spatial Catalogue (QSC), 2016f. Remnant 2013 Vegetation Cover ofQueensland [Data set]. Retrieved from http://qldspatial.information.qld.gov.au/catalogue/.

[dataset] Queensland Spatial Catalogue (QSC), 2016g. Statewide Landcover and TreesStudy Queensland series [Data set]. Retrieved from http://qldspatial.information.qld.gov.au/catalogue/.

R Core Team, 2016. R: a Language and Environment for Statistical Computing. RFoundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/.

Revelle, W., 2017. Psych: Procedures for Personality and Psychological Research. RPackage Version 1.6.12. http://CRAN.R-project.org/package=psych.

Rhodes, J.R., Cattarino, L., Seabrook, L., Maron, M., 2017. Assessing the effectiveness ofregulation to protect threatened forests. Biol. Conserv. 216, 33–42. http://dx.doi.org/10.1016/j.biocon.2017.09.020.

Rivalan, P., Delmas, V., Angulo, E., Bull, L.S., Hall, R.J., et al., 2007. Can bans stimulatewildlife trade? Nature 447, 529–530. http://dx.doi.org/10.1038/447529a.

Rollason, S.N., Howell, S., 2012. Aquatic Conservation Assessments (ACA), UsingAquaBAMM, For the Non-Riverine Wetlands of the Great Barrier Reef Catchment,Version 1.3. Department of Environment and Resource Management, Brisbane.

Sandker, M., Finegold, Y., D’Annunzio, R., Lindquist, E., 2017. Global deforestationpatterns: comparing recent and past forest loss processes through a spatially explicitanalysis. Int. For. Rev. 19 (3), 350–368. http://dx.doi.org/10.1505/146554817821865081.

Sant’Anna, A.A., Young, C.E.F., 2010. Direitos de propriedade, desmatamento e conflitosrurais na Amazônia [property rights, deforestation and rural conflicts in theAmazon]. Economia Aplicada 14 (3), 377–387. http://dx.doi.org/10.1590/S1413-80502010000300006.

Sattler, P.S., Williams, R.D., 1999. The Conservation Status of Queensland’s BioregionalEcosystems. Environmental Protection Agency and Queensland National ParksAssociation, Brisbane.

Scarth, P., Gillingham, S., Muir, J., 2008. Assimilation of spectral information and tem-poral history into a statewide woody cover change classification. In: Proceedings ofthe 14th Australasian Remote Sensing and Photogrammetry Conference. Darwin, NT,Australia. September 29-October 3.

Seabrook, L., McAlpine, C., Fensham, R., 2006. Cattle, crops and clearing: regional driversof landscape change in the Brigalow Belt, Queensland, Australia, 1840–2004. Landsc.Urban Plann. 78 (4), 373–385. http://dx.doi.org/10.1016/j.landurbplan.2005.11.007.

Seabrook, L., McAlpine, C., Fensham, R., 2007. Spatial and temporal analysis of vegeta-tion change in agricultural landscapes: a case study of two brigalow (Acacia harpo-phylla) landscapes in Queensland, Australia. Agric. Ecosyst. Environ. 120 (2–4),211–228. http://dx.doi.org/10.1016/j.agee.2006.09.005.

Seabrook, L., McAlpine, C., Fensham, R., 2008. What influences farmers to keep trees? Acase study from the Brigalow Belt, Queensland, Australia. Landsc. Urban Plann. 84(3–4), 266–281. http://dx.doi.org/10.1016/j.landurbplan.2007.08.006.

Senate Inquiry, 2010. Native Vegetation Laws, Greenhouse Gas Abatement and ClimateChange Measures. The Senate Finance and Public Administration ReferencesCommittee, Canberra.

Southworth, J., Tucker, C., 2001. The influence of accessibility, local institutions, andsocioeconomic factors on forest cover change in the mountains of western Honduras.Mt. Res. Dev. 21 (3), 276–283. http://dx.doi.org/10.1659/0276-4741(2001)021[0276:TIOALI]2.0.CO;2.

Tabachnick, B.G., Fidell, L.S., 2001. Using Multivariate Statistics, 4th ed. Allyn andBacon, Boston.

Taylor, M., 2015. Bushland Destruction Rapidly Increasing in Queensland. WWF Briefing,Sydney: WWF Australia.

Turner, M.D., Wear, D.W., Flamm, R.O., 1996. Land ownership and land-cover change inthe Southern Appalachian Highlands and the Olympic Peninsula. Ecol. Appl. 6 (4),1150–1172. http://dx.doi.org/10.2307/2269599.

Wang, G., Innes, J.L., Lei, J., Dai, S., Wu, S.W., 2007. China’s forestry reforms. Science318 (5856), 1556–1557. http://dx.doi.org/10.1126/science.1147247.

Ward, J.H., 1963. Hierarchical grouping to optimize an objective function. J. Am. Stat.Assoc. 58 (301), 236–244. http://dx.doi.org/10.2307/2282967.

Wear, D.N., Murray, B.C., 2004. Federal timber restrictions, interregional spillovers, andthe impact of U. S. softwood markets. J. Environ. Econ. Manag. 47 (2), 307–330.http://dx.doi.org/10.1016/S0095-0696(03)00081-0.

Wilson, B.A., Neldner, V.J., Accad, A., 2002. The extent and status of remnant vegetationin Queensland and its implications for statewide vegetation management and legis-lation. Rangel. J. 24 (1), 6–35. http://dx.doi.org/10.1071/RJ02001.

Zhang, D., 2001. Faustmann in an uncertain pol