Concepts & Trends

A Revised Conservation Assessment of Dipterocarps in Sabah

A Revised Conservation Assessment of Dipterocarps in Sabah
of 9
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  A Revised Conservation Assessment of Dipterocarps in Sabah Colin R. Maycock  1,3,5 , Chris J. Kettle 2 , Eyen Khoo 1 , Joan T. Pereira 1 , John B. Sugau 1 , Reuben Nilus 1 , Robert C. Ong 1 ,Nazahatul A. Amaludin 3 , Mark F. Newman 4 ,  and  David F.R.P. Burslem 31 Forest Research Centre, Sabah Forest Department, Sabah 90714, Malaysia 2 Institute of Terrestrial Ecosystems, ETH Zu¨rich, CHN G 73.1, Universita¨tstrasse 16, Zu¨rich 8092, Switzerland 3 Institute of Biological and Environmental Sciences, University of Aberdeen, Cruickshank Building, St Machar Drive, Aberdeen,AB24 3UU, U.K. 4 Royal Botanic Garden Edinburgh, 20A Inverleith Row, Edinburgh, EH3 5LR, U.K. ABSTRACT Borneo has experienced a rapid decline in the extent of forest cover, which has reduced the amount of habitat available for many plant and animal species. The precise impact of habitat loss on the conservation status of dipterocarp trees is uncertain. We use three con-trasting techniques, the extent of occurrence, area of occupancy and ecological niche models derived using   MAXENT , in conjunction witha current land-use map of Sabah, to derive estimates of habitat loss and infer a regional IUCN Red List conservation status for 33Sabah dipterocarp species. Estimates of habitat loss differed signi fi cantly according to the methods employed and between species ondifferent habitat types. Proportion of habitat loss determined from the ecological niche models varied from 21 percent for  Shorea micans  to 99.5 percent for  Dipterocarpus lamellatus  . Thirty-two of the 33 dipterocarp species analyzed in this study would have their Sabah popu-lations classi fi ed as Threatened (equal to a habitat loss of   >  30%) under the A2 IUCN Red List criterion. Dipterocarps that occur inlowland forests have experienced greater habitat loss than upland/lower montane or ultrama fi c species. In addition, species with thelowest predicted area within their historic distributions had the highest proportion of habitat lost, which provides a rationale for target-ing conservation effort on the species with narrow distributions. We recommend the ecological niche modeling approach as a rapidassessment tool for reconstructing species ’  historic distributions during conservation assessments of tropical trees. Abstract in Malay is available in the online version of this article. Key words  : Borneo; conservation status; Dipterocarpaceae; ecological niche modeling; forest restoration; IUCN Red List;  MAXENT ; Sabah. T HE  G LOBAL   S TRATEGY FOR   P LANT  C ONSERVATION MAKES CLEAR THAT HALTING THE DECLINE  of plant diversity is one of thegreatest challenges facing the global community (Secretariat of the Convention on Biological Diversity 2002). Nowhere is thismore urgent than in the forests of Southeast Asia, which havethe greatest proportion of threatened vascular plant species inthe tropics and the highest annual rates of deforestation (Sodhi et al.  2010). In the Malaysian state of Sabah on the island of Borneo, there has been a rapid decline in the extent of the nat-ural forests, from about 86 percent in 1953 to  <  50 percent currently in recent estimates (McMorrow & Talip 2001,Reynolds  et al.  2011). This decline is disproportionately high inlowland coastal regions, particularly around the west and northcoasts of Sabah, and has varied among the different forest typesrepresented in the state (Table S1). The reduction in forest cover is leading directly to the endangerment, and local extinc-tion, of many plant and animal species, including many ecologi-cally and economically important timber trees within the family Dipterocarpaceae (Sodhi  et al.  2010). Previously, it was not pos-sible to determine the impact of habitat loss on the conserva-tion status of most species in Sabah because there has been noattempt to reconstruct historic species distributions prior to therecent loss of forest cover. In this article, we use ecological niche models (ENMs) to reconstruct these historic distributionsfor 33 dipterocarp species and compare these maps to the cur-rent distribution of forest cover of Sabah. We use these com-parisons to develop new conservation assessments of the Sabahpopulations of the 33 dipterocarp species and contrast ourassessments to the current IUCN conservation assessmentsderived by other authors using different methods.In Sabah, there is currently little emphasis on the conserva-tion of individual plant species, which re fl ects a lack of informa-tion about their distribution and conservation status (Ministry of Science, Technology & the Environment 1998). This emphasismay change following the creation of the Aichi targets during therecent COP 10 meeting of the Convention on Biological Diver-sity. Target 12 calls for the development of national Red Listsand conservation strategies for threatened species by 2016, aspart of the Aichi target  ’ s wider goal of preventing the extinction,and improving the conservation status, of known threatened spe-cies (Secretariat of the Convention on Biological Diversity 2010). A change in the IUCN Red List status is the proposed indicatorfor assessing progress toward meeting this target. Received 23 March 2011; revision accepted 10 October 2011. 5 Corresponding author; e-mail: ª  2012 The Author(s) 649Journal compilation  ª  2012 by The Association for Tropical Biology and ConservationBIOTROPICA 44(5): 649–657 2012 10.1111/j.1744-7429.2011.00852.x S PECIAL  S ECTION  Estimating a species ’  extent of occurrence (EOO) or area of occupancy (AOO) is the currently accepted approach for con-ducting IUCN Red List assessments (IUCN Standards & Peti-tions Subcommittee 2010). The EOO is de fi ned as the areacontained within a polygon drawn to create the shortest bound-ary solution encompassing all known sites of occurrence of aspecies. The AOO measures the area within the EOO that isactually occupied by a given species (IUCN Standards & PetitionsSubcommittee 2010). The AOO is determined by counting thenumber of occupied cells in a uniform grid covering the speciesrange and then multiplying the number of occupied cells by thesize of the grid cell adopted (IUCN Standards & Petitions Sub-committee 2010). While these methods alone may provide a suit-able means of estimating extinction risk for range-restrictedspecies (Brummitt   et al.  2008), for the majority of plants they need to be coupled with estimates of habitat loss (McIntyre 1992,Nic Lughadha  et al.  2005) and population size (Brummitt   et al. 2008) to provide meaningful conservation assessments.Estimating the extent of habitat loss for plants is not straight-forward. Most tropical plants are known from only a few collec-tions and are often highly habitat-speci fi c (Ashton 2004, Cannon& Leighton 2004, Davis  et al.  2005). In Borneo, habitat associa-tions are particularly common among dipterocarps, as many spe-cies display a strong degree of speci fi city to soil type (Paoli  et al. 2006, Sukri  et al.  2011) and they are often naturally fragmentedinto an archipelago of habitat islands (Ashton 2004, 2010). Ashton(2004, 2008) considers that the IUCN assessment criteria may not be suitable for assessing the conservation status of these long-lived, highly habitat-speci fi c trees, a view which is supported by previous work showing that neither the EOO nor the AOO iseffective for estimating distributions of species with disjunct distri-butions (Solano & Feria 2007). The EOO overestimates the distri-bution of species with disjunct distributions because it fails toaccount for the patchy distribution of habitat (Solano & Feria2007), whereas the AOO can underestimate species distributionsbecause of sampling constraints (Solano & Feria 2007). A new and alternative solution is to use ENMs as a meansof estimating the potential distribution of threatened species(Solano & Feria 2007, López-Toledo  et al.  2011). This approach usesalgorithms to generate maps of a species ’  potential distribution onthe basis of a mathematical representation of its distributionin environmental space (Phillips  et al.  2006). Numerous tech-niques for generating ENMs have been proposed (Elith  et al. 2006), and the relative merits of the different techniques aredebated (Elith  et al.  2006, Graham  et al.  2008). We use the pro-gram  MAXENT  to model the distributions of 33 dipterocarp spe-cies in Sabah using high-resolution soil data. Because this methodexplicitly incorporates the strong habitat associations of diptero-carps, we predicted that it would result in a more realistic esti-mate of habitat loss than the IUCN ’ s methods based on EOOand AOO, which fail to capture this important element of dip-terocarp ecology. This point is illustrated by mapping historic dis-tributions against EOO and AOO (Fig. S1), which suggest that the IUCN ’ s methods may underestimate habitat loss and ulti-mately lead to inappropriate conservation assessments in somecases. The current IUCN Red List guidelines do not allow forENMs to be used to estimate a species ’  AOO and subsequently its conservation status (IUCN Standards & Petitions Subcommit-tee 2010).The ecological dominance, and commercial importance, of the Dipterocarpaceae in the forests of Southeast Asia have stimu-lated a signi fi cant quantity of ecological and taxonomic researchon the family (Ashton 2004, 2008, 2010). As a result, diptero-carps are the best studied component of the  fl ora of Borneo, andserve as an appropriate model for developing and testing rapidassessment strategies. Approximately 58 percent of all diptero-carps have been evaluated for the IUCN Red List, of which 94percent are listed as endangered (Nic Lughadha  et al.  2005),although there is uncertainty over the accuracy of these assess-ments (Chen 2004). In this study, we compared estimates of habi-tat loss based on reductions in the EOO, AOO and predicteddistributions determined by ecological niche modeling for 33 dip-terocarp species in Sabah. Based on these estimates of habitat loss, we re-classi fi ed the threat status of the Sabah populations of each species using the A2 criterion (IUCN Standards & PetitionsSubcommittee 2010). These new threat assessments were com-pared with previous assessments that were implemented using the extent of each species ’  geographical range (Criterion B; Chua et al.  2010) and non-quantitative expert assessments (Ashton2004, IUCN 2010). We tested the following speci fi c hypotheses:(i) the extent of estimated habitat loss predicted using ENMsexceeds that predicted by the EOO or area of occurrence meth-ods, which generates a potential contrast in species conservationassessments; (ii) habitat loss has been greater for species of low-land than upland/montane or ultrama fi c environments; and(iii) habitat loss is more signi fi cant, as a proportion of total habi-tat area, for range-restricted and/or habitat-specialist species. METHODS Ecological niche models were generated using   MAXENT  v. 3.3.1(; Phillips  et al. 2006) based on locality data obtained from herbarium specimens,research plots and  fi eld surveys (Table S2).  MAXENT  has beenshown to outperform other modeling techniques (Elith  et al. 2006) in the sense that it is least affected by errors in locality dataand gives more robust outcomes when there are few collectionlocalities (Graham  et al.  2008, Wisz  et al.  2008). This method usesmaximum entropy density estimations to represent the distribu-tion of a species as a probability distribution over the study area(Phillips & Dudík 2008). The models are derived from locality data (presence-only data) and a set of environmental variablesthat are considered the most important for de fi ning the suitability of the environment for the species. These data are used todevelop a model that predicts environmental suitability as a func-tion of the environmental variables. The model is then projectedinto geographic space to predict the distribution for the species(Phillips  et al.  2006). Model validation is conducted by testing the prediction success of the model on an independent dataset (Cayuela  et al.  2009). 650 Maycock  et al. S PECIAL  S ECTION  Collection locality data were sourced from herbarium speci-men labels held in the Sandakan herbarium. These collections were obtained over a 73-yr period (1938  –  2011), with the majority (75%) having been collected during the 1950s and 1960s, whichpredates the major loss of forest cover in Sabah. All herbariumspecimen locality records without geographical coordinates weregeo-referenced by consulting 1:250,000 soil maps and 1:50,000forest stratum maps. Samples that could not be placed con fi -dently were excluded from the study to avoid the use of impre-cise distributional information. Supplementary locality data werecollected from research plots established predominantly during the1950s and 1960s and during more recent   fi eld surveys (2009  –  2010). All dipterocarps encountered during these surveys wereidenti fi ed and their position recorded with a GPS (Garmin GPS-map CSX60, Olathe, Kansas, USA). Where multiple specimens/records had been collected from the same soil association in onelocality (corresponding to a 2  9  2 km grid cell on a map of Sabah), only one record was included in the analysis to reducesampling bias (Hijmans  et al.  2000). Conversely, collections frommore than one soil association within a 2  9  2 km grid cell wereincluded as independent items of data for constructing speciesdistributions.The ENMs were developed using bioclimatic, altitudinal andedaphic variables. Bioclimatic data for Sabah at 30 arc-second(  ~  1 km 2  ) resolution were obtained from WorldClim (http://www. Shuttle Mission Radio Telemetry altitudinal dataat 3 arc-second resolution (  ~  90 by 90 m) were obtained from theCGIAR-CSI GeoPortal ( The slopeof the terrain was calculated from the altitudinal data using Arc- View 3.3 (ESRI, Redlands, California, USA). GIS shape  fi les of soil association, landform, soil suitability and soil parent material (digitised from the Soils of Sabah maps; Acers  et al.  1975) wereobtained from the Sabah Forestry Department. Electronic ver-sions of these maps are available from Selvaradjou  et al.  (2005).The shape  fi les were converted to ESRI ASCII grid format using  ArcView Spatial Analyst. All spatial data were on the WGS84projection. As the edaphic variables are categorical and cannot beaggregated into larger mapping units, it was necessary to set thegrid size to the smallest common size of 6 arc seconds(  ~  0.04 km 2  ).For the 20 species with more than 12 collecting localities, weran two separate models. The  fi rst model was developed using the full set of locality data to provide best estimates of the spe-cies ’  potential distributions, and the second model was developedusing a random selection of 75 percent of the locality data at each iteration as a training dataset, with the remaining 25 percent reserved for testing the resulting model. For species with fewerthan 12 collection localities, no model testing was conducted. Thearea under the curve of the receiver-operating characteristic foreach model testing run is given in Table S2. In all cases, we ran100 replicated runs of   MAXENT  on the locality data with 10,000background points randomly selected per run (Phillips  et al. 2006). We used  DIVA - GIS to estimate the percentage of habi-tat loss for each species (Hijmans  et al.  2005). The average pre-dicted distribution from 100 replicated runs of   MAXENT  was usedas an estimate of the total historic distribution of each species.This distribution map was overlaid with a grid  fi le provided by the Sabah Forestry Department showing the land-use in Sabah in2008, which enabled us to calculate the percent of the historicdistribution that falls into protected areas, production forests,under natural forest management or on land that has been, or isunder conversion to alternative land-uses (Fig. S2). Habitat loss was calculated as the historic distribution minus the area retainedin protected areas or production forests. This approach assumesthat no extant populations of the species exist outside the perma-nent forest estate, and that all protected areas and productionforests under natural forest management still maintain popula-tions of the study species. The validity of these assumptions isdiscussed below. Species were classi fi ed into IUCN threat catego-ries based on the percent of habitat loss using the A2 criterion(Table S3; IUCN Standards & Petitions Subcommittee 2010).Criterion A is designed to highlight species that have undergonea signi fi cant decline, measured either as a decline in populationsize or habitat area, in the recent past or species that are pro-jected to experience a signi fi cant decline in the near future. Thiscriterion is more relevant for a regional assessment than CriterionB, which is the size of the geographic range of a species (Abeli et al.  2009). Our approach also assumes that all loss of forest cover has occurred within three generations of the study species. We selected the A2 criterion as the habitat/population reductionhas neither ceased nor is reversible in the majority of cases. All locality data were used to determine EOO, which is ameasure of the spatial spread of the area occupied by the speciesand is not intended to be an estimate of the amount of potential habitat or range size (IUCN Standards & Petitions Subcommittee2010). EOO was calculated by the minimum convex polygonmethod using the Conservation Assessment Tools (CATS) pro-duced by the Royal Botanic Gardens Kew (Moat 2007), andincludes discontinuities in habitat (Gaston & Fuller 2009). Habitat loss as a percentage of EOO was calculated using   DIVA - GIS by converting the EOO polygon to a grid  fi le, overlaying the current land-use grid  fi le, and calculating the percentage of EOO outside protected areas and the permanent forest estate(Hijmans  et al.  2005). While this does not conform to the IUCNStandards and Petitions Subcommittee (2010) recommendedpractice for determining a reduction in EOO, it was the only approach that was possible with the available data. The AOO was calculated using CATS based on a 2  9  2 km grid cell as rec-ommended by the IUCN Standards and Petitions Subcommittee(2010). Habitat loss based on a reduction in AOO was calculatedusing the herbarium collections and research plot data only fol-lowing Callmander  et al.  (2007).The species were classi fi ed as lowland, largely ultrama fi c orupland/montane based on Ashton (2004), to allow comparisonbetween species occupying different habitats. ANOVA was usedto test the signi fi cance of differences in percent habitat lossamong species occupying these three habitat types and among the different assessment methods using transformed percent hab-itat loss data. Linear regression models were used to investigate Conservation Assessment of Dipterocarp 651 S PECIAL  S ECTION  the relationship between habitat loss and distribution areapredicted from the AOO, EOO and ENMs. Finally, to allow comparison between this work and previous assessments (Ashton2004, Chua  et al.  2010, IUCN 2010), we assigned a numeric valuefrom 0 to 5 to each threat category, from 0 corresponding to anassessment of   ‘ least concern ’ , to 5 representing   ‘ extinct  ’ . Paired Wilcoxon tests were used to test for a difference in medianscores between the assessments. All analyses were conductedusing   R   2.12.0 (R Development Core Team 2010). RESULTS C OMPARISON OF ASSESSMENT METHODS AND HABITAT TYPES .  —  Habi-tat loss determined from the ENMs varied from 21 percent for Shorea micans   to 99.5 percent for  Dipterocarpus lamellatus  , with 32of the 33 dipterocarp species analyzed in this study (97%) classi- fi ed as threatened (equal to a habitat loss of   >  30%) under the A2 criteria (Fig. 1A and Fig. S2A-i). Mean (  ±  SEM) habitat lossfor the six upland/montane and ultrama fi c species (37.6  ±  5.7%) was signi fi cantly lower than for the 27 lowland (68.6  ±  2.3%)species (  F  2,30  =  13.3,  P   <  0.0001). Among the lowland species,the proportion of the habitat lost was greatest in species with thelowest predicted area within the historic distribution (Fig. 1A, R  2adj.  =  0.57,  P   <  0.0001).Estimated habitat loss calculated from reduction in EOOranged from 39 to 74 percent, and all 33 dipterocarp species were classi fi ed as threatened (VU or EN: Fig. 1B). We foundno signi fi cant difference in percentage habitat loss among species of lowland, ultrama fi c or upland/montane environments(  F  2,29  =  1.74,  P   =  0.192; Fig. 1B), and no signi fi cant relationshipbetween the EOO of the lowland species and the proportion of habitat lost (  R  2adj.  =  0.016,  P   =  0.534). By contrast, only 63 per-cent of the species examined were classi fi ed as threatened basedon reduction in AOO (Fig. 1C), and there were no signi fi cant differences in estimated habitat loss between species of lowland, upland/montane and ultrama fi c environments(  F  2,30  =  2.164,  P   =  0.132). For the lowland species, there was asigni fi cant negative relationship between AOO and the propor-tion of habitat lost based on reduction in AOO (Fig. 1C,  R  2adj.  = 0.36,  P   =  0.0006).Estimates of habitat loss differed signi fi cantly according to whether historic distributions were estimated using EOO, AOOor ENMs (  F  2,89  =  18.41,  P   <  0.0001) and between species of dif-ferent habitat types (  F  2,89  =  9.92,  P   =  0.0001). The interactionbetween computational method and habitat types was also signi fi -cant (  F  4,89  =  3.71,  P   =  0.0076; Fig. 2). For lowland forest species,estimated habitat loss was higher when based on ENMs(mean  ±  SEM: 68.6  ±  2.3%) than either reduction in AOO(mean  ±  SEM: 44.8  ±  4.2%) or EOO (mean  ±  SEM: 52.1  ± 1.4%). For ultrama fi c and upland/montane species, the methodbased on reduction in EOO estimated highest habitat loss(Figs. 2A and C). There was closer agreement in the threat catego-ries obtained when comparing between the methods based onENMs and reduction in either the EOO or AOO (Figs. 2A and FIGURE 1. Relationships between habitat loss and (A) predicted area within the historic distribution, (B) extent of occurrence and (C) area of occupancy for 33dipterocarp species in Sabah. The IUCN threat categories under criteria A2 are listed at the side of the  fi gure. These categories correspond to the percentagedecline in habitat with vulnerable (VU), endangered (EN), critically endangered (CR) and extinct (EX) corresponding to a 30, 50, 80 and 100 percent decline inhabitat respectively. The solid lines indicate the signi fi cant relationship between habitat loss and predicted area of distribution or area of occupancy for the 27 low-land species. 652 Maycock  et al. S PECIAL  S ECTION  B) than when comparing between the methods based on reduc-tion in EOO and AOO (Fig. 2C).Habitat loss estimated from reduction in the AOO was sig-ni fi cantly lower than the estimated forest cover loss (  ~  50%) forSabah as a whole (mean  =  41%, single sample  t  -test:  t   =   2.36,df   =  32,  P   =  0.025). Mean habitat loss estimates based on reduc-tion in the EOO and ENMs were signi fi cantly higher than theestimated forest cover loss for Sabah as a whole (means of 58%and 63%, respectively; single sample  t  -test:  t   =  4.66, df   =  32, P   =  0.00006 and  t   =  4.36, df   =  32,  P   =  0.0001, respectively).The estimates of habitat loss determined as a function of reduc-tion in EOO varied less among species (Fig. 1B) than estimatesobtained from either ecological niche modeling (Fig. 1A) or fromreduction in AOO (Fig. 1C).C OMPARISONS WITH PREVIOUS ASSESSMENTS .  —  There were signi fi cant differences in the distribution of species among categories of threat based on the six assessments available for comparison:those based on the three methods adopted in our study, the cur-rent global assessments (IUCN 2010), the Malaysia Plant Red List project assessments (Chua  et al.  2010) and Ashton ’ s (2004) assess-ments (Table 1). Of the 23 species examined in this study that  were also assessed for the IUCN Red List, 20 species had ahigher global threat category on the IUCN Red List (Table 1). By contrast, all 11 species examined by Chua  et al.  (2010) for theMalaysia Plant Red List had lower threat categories on theMalaysia Plant Red List. We found no signi fi cant differencesbetween our assessments based on EOO and ENMs and Ashton ’ s (2004) assessments based on expert opinion (Table 1),but Ashton ’ s (2004) assessments were signi fi cantly lower than theglobal assessments (IUCN 2010) and signi fi cantly higher than theMalaysia Plant Red List project assessments (Table 1). DISCUSSION C OMPARISON OF METHODS FOR COMPUTING HABITAT LOSS .  —  Ourestimates of percent loss of habitat within species ’  historic distri-butions were generally greater when calculated using ENMs thaneither of the methods recommended by the IUCN. This isimportant because the Aichi 2020 targets call for the preventionof extinction and the improvement in conservation status of threatened species based on the IUCN ’ s categories of threat (Sec-retariat of the Convention on Biological Diversity 2010). Achiev-ing these targets will require a better understanding of speciesthreat status at both regional and global scales. However, ourresults suggest that current methods for assigning categoriesof threat may be inadequate or inappropriate because they areinsensitive to aspects of species ’  ecology. These methods may be FIGURE 2. Comparison of estimates of percentage habitat loss determined by the ecological niche modeling technique compared to (A) reduction in the extent of occurrence (EOO) and (B) area of occupancy (AOO), and (C) a comparison of the two IUCN methods. The boxes indicate the thresholds for IUCN threat categories under criteria A2; points outside these boxes indicate where the two assessment methods estimate different categories of threat. The dotted line(slope  =  1) is used to indicate differences in the estimates of percent habitat loss and codes correspond to IUCN threat categories (LC, least concern; VU, vulner-able; EN, endangered; CR, critically endangered). Conservation Assessment of Dipterocarp 653 S PECIAL  S ECTION
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks