Informing Decisions on an Extremely Data Poor Species Facing Imminent Extinction

Informing decisions on an extremely data poor species facing imminent extinction

M ATTHEW J. GRAINGER,DUSIT N GOPRASERT P HILIP J.K. MC G OWAN and T OMMASO S AVINI

Abstract Some of the species that are believed to have the Convention on Biological Diversity’s Strategic Plan for highest probability of extinction are also amongst the Biodiversity – (Secretariat of the Convention on most poorly known, and this makes it extremely difficult Biological Diversity, ) and target . of the to decide how to spend scarce resources. Assessments of UN’s Sustainable Development Goals (UN Sustainable conservation status made on the basis of loss or degradation Development Knowledge Platform, ). Our understand- of habitat and lack of records may provide compelling indi- ing of how high the probability of extinction is for individual cations of a decline in geographical range and population species is variable, as is our ability to identify places that size, but they do not help identify where conservation action should be priorities for action. In some instances, species might be best targeted. Methods for assessing the probabil- are well known and easily detectable, meaning that there ity of extinction and for modelling species’ distributions is a sound basis for identifying where and how to act. For exist, but their data requirements often exceed the informa- other species, however, it is extremely difficult to be confi- tion that is available for some of the most urgent conserva- dent about their proximity to extinction, let alone decide tion cases. Here we use all available information (localities, where searches should be focused or where conservation in- expert information, climate and landcover) about a high- terventions should be implemented. This variation in our priority Vietnamese bird species (Edwards’s pheasant understanding is typically a result of variable information Lophura edwardsi) to assess objectively the probability of about species, across both space and time, which, in turn, its persistence, and where surveys or other conservation ac- is attributable to factors such as detectability (Bibby et al., tion should be targeted. It is clear that the species is on the ), search effort (Boakes et al., ) and how well infor- threshold of extinction and there is an urgent need to survey mation is documented and made accessible (Boakes et al., Bach Ma National Park (including the extension) and to b). consider surveying Ke Go Nature Reserve. This approach South-east Asia has been highlighted as a region where has potential to help identify where conservation action there is both a high risk of extinction of many vertebrate should be targeted for other Critically Endangered species species (e.g. Hoffmann et al., ) and a severe lack of infor which there is an extreme scarcity of information. formation on where and how to act to prevent their extinction (Duckworth et al., ). These concerns led to a call for Keywords Data, Edwards’s pheasant, extinction, Lophura urgent action to address the threats facing tropical Asia’s edwardsi, optimal linear estimator, Siamese fireback, silver species, at the  World Conservation Congress (IUCN, pheasant ), and the establishment of the Asian Species Action Partnership (ASAP, ). For many of these species the available data on location or ecology are few, and they are often considered to be of Introduction poor quality, which typically refers to old records for pecies that are considered to be close to extinction which the date and location are uncertain. Using these re- Sare often a target for conservation action. This may in- cords without critical appraisal of the nature of this uncer- volve dedicated action by conservation organizations, tainty could result in subjective assessments of where a working nationally or internationally (e.g. the Alliance for species may still occur, what its habitat is and where searches Zero Extinction, and BirdLife International through its should be focused. ’ Preventing Extinctions Programme), or the establishment Since Edwards s pheasant Lophura edwardsi was recate- of global policy targets, such as Aichi Target  of the gorized as Critically Endangered on the IUCN Red List in  there has been increasing attention to its conservation. It is vital, therefore, that as much information as possible, MATTHEW J. GRAINGER and PHILIP J.K. MCGOWAN (Corresponding author) School even if of unknown quality, is used as the basis for defining of Biology, Newcastle University, Newcastle upon Tyne NE1 7RU, UK ’ E-mail [email protected] the species status and deciding what conservation action should be undertaken. Here we make use of all available in- DUSIT NGOPRASERT and TOMMASO SAVINI Conservation Ecology Programme, King ’ Mongkut’s University of Technology Thonburi, Bangkok, Thailand formation to model the potential for extinction of Edwards s Received  January . Revision requested  February . pheasant and compare this with the two other Lophura spe- Accepted  May . First published online  August . cies inhabiting the same area. We then examine the spatial

Oryx, 2019, 53(3), 484–490 © 2017 Fauna & Flora International doi:10.1017/S0030605317000813 Downloaded from https://www.cambridge.org/core. IP address: 170.106.40.139, on 28 Sep 2021 at 22:46:57, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0030605317000813 Data poor species facing extinction 485

uncertainty associated with the location data to determine if Modelling time to extinction a species distribution model could be produced to guide sur-  vey effort. Finally, we produce a Bayesian model to predict The optimal linear estimator (Cooke, ; Roberts & Solow,   ’ current habitat availability and identify sites where the spe- ; Solow, ), or Cooke s estimator (Collen et al.,  cies may still occur. ), is a non-parametric extinction date estimator. The approach is based on the Weibull distribution, a two- parameter model that has its origin in engineering risk analysis (Solow, ; Collen et al., ). The technique The study species: Edwards’s pheasant is considered robust where the probability of observing a In  Edwards’s pheasant was uplisted to Critically species is low, and it does not assume that sighting effort  Endangered (BirdLife International, ) because of the has been equal over time (Rivadeneira et al., ). Even lack of recent records (the last being a poached individual when the assumptions are not met fully because of the real- in ; one record in  is unconfirmed and another ities of search effort and data availability, the optimal linear  is of a captive individual with an unknown history; S.P. estimator is broadly accurate (Collen et al., ; Clements  Mahood & J.C. Eames, pers. comm.), extremely high hunt- et al., ). Its prediction of time to extinction (TE) based  ing pressure, and habitat fragmentation and degradation on the k most recent sightings is described by Solow ( ). throughout its known range (BirdLife International, ). There is uncertainty regarding how best to determine k. All of this led to increasing concern for the survival of the In theory it should be only the most recent sightings (Solow,   species. Since , searches have been conducted at some ), but Collen et al. ( ) showed that increasing the  potential sites but no evidence of the species’ existence has number of sightings used (tested to a maximum of sight- been found, although other Galliformes species have been ings) increases the accuracy of prediction. However, the ’ recorded (Pham & Le, ). First described in , records large gap in the sighting record of Edwards s pheasant dur- of the species are restricted to central Vietnam (Ha Tinh, ing the First Indochina War and the subsequent Vietnam – Quang Binh, Quang Tri, Thua Thein Hue Provinces), an War ( ) invalidates the assumptions of the optimal area long considered to be of conservation concern because linear estimator if applied to a series of records that spans   of high endemism and the high level of threat to which spe- this gap (C. Clements, pers. comm., December ) and  cies are subjected (Eames et al., ). Two other forms of so we used only the most recent records ( onwards). The Lophura were thought to be closely related species until re- data for this analysis consisted of the year of each confirmed  cently, but are now considered to be conspecific: the imper- observation and a test year ( ). We used the package     ial pheasant L. imperialis has been shown to be a naturally sExtinct (Clements, )inRv. . . (R Development  occurring hybrid of Edwards’s pheasant and the silver Core Team, ) to calculate the optimal linear estimator ’ pheasant L. nycthemera (Hennache et al., ), and the for Edwards s pheasant and two congeneric species, the sil- Vietnamese pheasant L. hatinhensis is now considered to ver pheasant and the Siamese fireback L. diardi, which are be an inbred form of Edwards’s pheasant (Hennache extant in the region. Data for these other Lophura pheasants  et al., ). Henceforth we refer to all forms of the species were also extracted from Boakes et al. ( a), with more re- as Edwards’s pheasant. The species went unrecorded be- cent records extracted from the Global Biodiversity  tween the early s and late s, during which time Information Facility (GBIF, ). much of its suspected habitat was further defoliated and de-  graded (BirdLife International, ). Spatial uncertainty

We suspect that there is positional uncertainty associated Methods with some, if not all, of the location points and that for some records this is up to  km (some locations were re- Location data ported as the nearest commune, village or district centre). Positional uncertainty in species distribution models has Geo-referenced location data for Edwards’s pheasant were been evaluated for cases in which errors were known and extracted from the Galliformes database of Boakes et al. relatively small (,  km), and found to have little effect (a). As noted above, records previously ascribed to L. (Graham et al., ; Johnson & Gillingham, ). hatinhensis and L. imperialis were extracted for inclusion Naimi et al. (, ) showed that high levels of spatial in the study, in addition to those of L. edwardsi. The records heterogeneity in environmental predicator variables leads to consist of reported locations from historical notebooks, reduced model performance. We used a distance of  km peer-reviewed publications, books and specimen records (the maximum suspected error in point locations) to deter- (Mahood & Eames (in press) provide a detailed assessment mine the reference values (using the usdm package (Naimi, of the records, including those without spatial locations). )inR) and compared these to each of the  location

points. K values .  imply that spatial similarity is lower TABLE 1 Conditional probabilities of habitat suitability for than expected (high spatial heterogeneity) and values ,  Edwards’s pheasant Lophura edwardsi, based on the IUCN Red  imply that spatial similarity is higher than expected (low List account of the species (BirdLife International, ). spatial heterogeneity). Probability of habitat The ability of the model to discriminate between occu- suitability (%) pied and unoccupied areas was estimated from the area under the curve (AUC) of the receiver operating character- Elevation Low Medium High istics (Phillips et al., ). We used , random points Forest cover; monthly rainfall never , 30 mm within the BirdLife–NatureServe shapefile for Edwards’s , 100 m 20 40 40 – pheasant and the Vietnamese pheasant (a single shapefile 100 300 m 50 20 30 – for the species has not yet been produced) as background 300 500 m 70 20 10 – points. We executed the MaxEnt procedure in the dismo 500 700 m 80 10 10 . package (Hijmans et al., )inR. 700 m 90 5 5 Forest cover; monthly rainfall never , 40 mm , 100 m 10 40 50 Belief network 100–300 m 20 40 40 300–500 m 50 20 30 We developed a Bayesian belief network to account for 500–700 m 70 20 10 the suspected uncertainty in the spatial locations. The re- . 700 m 80 10 10 sulting Bayesian model provides a logical expert (IUCN Forest cover; monthly rainfall never , 50 mm ’ Red List) derived map of potential Edwards s pheasant , 100 m 10 30 60 habitat, albeit one that cannot be evaluated empirically, 100–300 m 20 30 50 based on the habitat description in the account of the spe- 300–500 m 20 40 40  cies on the IUCN Red List (BirdLife International, ), 500–700 m 50 20 30 namely: . 700 m 70 20 10 ‘It was said to inhabit exceedingly damp mountain forests up to an es- Forest cover; monthly rainfall never , 60 mm timated 600 m, favouring thick underbrush and lianas. However, all , 100 m 10 20 70 early collecting localities were in the forested level lowlands, and – there is no evidence that it can live above 300 m. It is most abundant 100 300 m 10 20 70 in areas with thick undergrowth and liana covered hillsides (N. Brickle, 300–500 m 10 30 60 in litt., 2004). Records in the 1990s came from lowland areas which 500–700 m 20 30 50 have been selectively logged (N. Brickle, in litt., 2004).’ . 700 m 50 30 20 We interpreted this as increased probability of habitat suit- No forest; monthly rainfall never , 30 mm ability for Edwards’s pheasant in areas that were forest, in , 100 m 70 20 10 areas that were at low elevation and had high monthly rain- 100–300 m 70 20 10 fall (Table ). A review of published literature and assess- 300–500 m 70 20 10 ment of the substantial body of grey literature generated 500–700 m 70 20 10 since ecological fieldwork restarted in Indochina in the . 700 m 70 20 10 late s (Brickle et al., ) has provided no information No forest; monthly rainfall never , 40 mm on habitat suitability that altered this understanding. , 100 m 70 20 10 To parameterize the model we extracted monthly rainfall 100–300 m 70 20 10 values from the WorldClim climatic dataset (WorldClim, 300–500 m 70 20 10  – ) version ., which has a spatial resolution of  km 500 700 m 70 20 10 . (for more details see Hijmans et al., ) and to this 700 m 70 20 10 No forest; monthly rainfall never , 50 mm added an elevation and forest coverage layer using the , Raster package in R (Hijmans, ). 100 m 70 20 10 – A raster dataset at  km resolution combining data on the 100 300 m 70 20 10 – precipitation of the driest month (WorldClim, ), eleva- 300 500 m 70 20 10 500–700 m 70 20 10 tion and forest cover was developed in R using the Raster . 700 m 70 20 10 package. Values for each layer at each raster pixel in the re- No forest; monthly rainfall never , 60 mm gion were then exported to be used as a case-file in Netica . , 100 m 70 20 10 (Norsys Corp, ). The case-file was then run through the 100–300 m 70 20 10 belief network and the probability of high habitat suitability 300–500 m 70 20 10 calculated. This was then converted back into a raster in 500–700 m 70 20 10    R and displayed graphically in ArcGIS . . (ESRI, . 700 m 70 20 10 Redlands, USA).

Results

Modelling time to extinction Using the optimal linear estimator, Edwards’spheasantisestimatedtohavegone extinct in , with a lower confidence interval (CI; i.e. earliest estimated date of extinction) of  and upper CI (latest estimated date of extinction if no further sightings are made) of . With the upper interval falling post  (the test year), we can interpret this result as showing that it is probable, given the nature of the historical records, that Edwards’s pheasant is still extant in the wild (Rivadeneira et al., ). Our low sample size may, however, have inflated the value of the upper CI (Strauss & Sadler, ) and thus our estimated date of extinction. The two congeners known to be extant in the region had estimated extinction dates post , with an upper CI of  and  for the Siamese pheasant and silver pheasant, respectively.

Species distribution model Within the area bounded by the BirdLife–NatureServe () extent of occurrence for Edwards’s pheasant the AUC was ., meaning that the model was worse than random in predicting the presence of Edwards’s pheasant. Therefore, no further analyses were conducted that sought to link locations of Edwards’s pheasant to environmental variables.

FIG. 1 Spatial representation of the results of our belief network Belief network Maximum habitat suitability values in ’   based on the IUCN Red List habitat description for Edwards s the belief network did not exceed . (because of the pheasant Lophura edwardsi (see Methods). High probability of uncertainty expressed in the conditional probabilities). occupancy based on altitude, climatic conditions (monthly Areas with probability of habitat suitability of . . were rainfall) and presence of forest is indicated in warmer colours. found in Khe Net and Ke Go Nature Reserves in the The species’ range according to the BirdLife–NatureServe north, and there were few grid squares of high probability shapefile () is delineated in black. of suitable habitat located in the south (Fig. ).

Discussion There is very little suitable habitat remaining for the species. Exceedingly damp forest falls into two blocks, which Edwards’s pheasant may still survive in the wild, but the largely coincide with the distribution of locality records of small number of records that exist for this species may Edwards’s pheasant. The form previously known as the mean that our assessment is optimistic. Whether the latter Vietnamese pheasant was reported mostly from the nor- date for re-sighting (the upper CI, in this case ) is opti- thern block and only during – (Hennache et al., mistic or not, what is clear is that the existing records indi- ), the only exception being a single record south of cate that Edwards’s pheasant is on the threshold of Hue (Mahood & Eames, in press). As the Vietnamese pheas- extinction. The low detectability of the species offers hope ant is now considered to be an inbred form of Edwards’s that it may exist but be recorded rarely, but our most pes- pheasant, these few observations suggest that the population simistic prediction is that the species went extinct in . has been declining for some time and is likely to have suf- Given the uncertainty in the spatial locations for the spe- fered considerably from heavy deforestation in Central cies we could not be confident that we would produce a Vietnam from the early s (Müller & Zeller, ). meaningful species distribution model. The belief model, Other ecological knowledge suggests that the survival which is based on the qualitative description given in the prospects of Edwards’s pheasant are very poor. Species IUCN Red List account, clearly suggests that the most suit- with small ranges tend to be scarce within those ranges able habitat will be in the northern part of the species’ range. (Brown, ), making them more susceptible to hunting

compared with sympatric, widely distributed species. The time a decision is to be made must be gathered, assessed congeneric silver pheasant and Siamese fireback both have and, where possible, used to make inference. Bayesian net- larger geographical ranges and would be expected to suffer works have been shown to be effective in determining the similar levels of hunting given that none are particular tar- distribution of species when there is little available ecological gets for poachers. Furthermore, narrow endemism in verte- information (e.g. Smith et al., ) and few resources for brate species is considered to be an indicator of limited conservation planning (Tantipisanuh et al., ). The opti- flexibility to habitat disturbance (Wijesinghe & Brooke, mal linear estimator has also been shown to be broadly ac- ), mainly as a result of edge effects of fragmented habi- curate in the face of data realities (poor search effort and data tat (Williams & Pearson, ; Brooks et al., ). The claim availability; Collen et al., ; Clements et al., ). that Edwards’s pheasant inhabits degraded habitat has not Species distribution models based on MaxEnt have been been confirmed (Eames, ) and could be a misinterpret- shown to be effective at determining distribution accurately ation of its presence in bamboo patches in pristine habitat even when using few location records (– records; van with abundant palms and rattan understorey (Robson Proosdij et al., ); however, there is some evidence that et al., ). The Siamese fireback is commonly found models using fewer than  locations are less accurate, in heavily degraded acacia and eucalyptus plantation and caution must be taken in using these (Wisz et al., (Suwanrat et al., ) and was found during the  survey ). MaxEnt is probably better suited to species for in Central Vietnam when Edwards’s pheasant was not, sug- which there are more ecological data available and a greater gesting that the habitat may have been too degraded for understanding of the most appropriate environmental vari- Edwards’s pheasants but not for its more tolerant congener. ables than we have for Edwards’s pheasant at present. Following our analysis, the fragments that should be Researchers and managers who are faced with making searched as a matter of urgency are Ke Go Nature Reserve, decisions about what actions may be appropriate for a spe- Bach Ma National Park (site of the most recent confirmed cies for which there is only low-quality and uncertain data and unconfirmed records) and the extension to Bach Ma should consider taking the following approach to inform National Park. Although camera-trap surveys for other spe- their decision. Firstly, gather all available data on locations cies have yielded little evidence of Galliformes at these sites from historical records, scientific sources, local communi- (Willcox, ), and at Ke Go the most recent records were ties, and any other available sources (i.e. from all stake- of the inbred Vietnamese form, these sites offer the best pro- holders). Secondly, assess critically this information to spects for conservation action because of the most recent re- identify potential biases and uncertainties, bearing in cords and the relative suitability of remaining habitat. mind that it may not be possible to address these through The approach that we have used to critically and object- modelling but they need to be highlighted. Thirdly, build ively assess the data that exist on Edwards’s pheasant, a a Bayesian belief network (or networks) based on the avail- poorly known and highly threatened species, has brought able data and assess these critically, ideally involving all sta- temporal and spatial focus to the need for action. The spe- keholders in this step whenever possible. At the same time, cies may already be extinct, and if not it is surely close. use an optimal linear estimator to assess the likelihood that Historical records and remaining habitat that is thought to the species still survives. Fourthly, use the model and the op- be suitable make it clear where effort should be targeted. timal linear estimator prediction to determine whether a Combining all available evidence within these temporal survey should be conducted and, if so, at which sites. and spatial frameworks provides direction for where Finally, either carry out the survey or propose another searches should be conducted and conservation action con- course of action, such as categorization as Extinct (in the sidered for this species. Given the crisis facing many similar- Wild), reintroduction or other, as appropriate. ly poorly known species that are believed to be on the verge of extinction, in South-east Asia and elsewhere, we believe that this approach may prove useful in distilling conserva- Acknowledgements tion direction from limited data. The modelling approaches we have used here have poten- We thank King Mongkut’s University of Technology Thonburi, tial to be useful for other species in the region and across the Royal Golden Jubilee Fund (Thailand) and Newcastle other regions where data are scarce. All models are only as University Research Committee for funding the opportunities good as the data on which they are based, and it is important to collaborate on this work. to recognize that the data available in this case and many others (e.g. the saola Pseudoryx nghetinhensis) may not fit well with the particular assumptions of any model. Despite Author contributions this limitation, conservation managers cannot afford to wait until all of the desired data become available, particularly All authors conceived the study, discussed and developed when funds are in short supply. All data available at the the analyses and wrote the article.

References EAMES, J.C., EVE, R., TORDOFF, A.W. () The importance of Vu Quang Nature Reserve, Vietnam, for bird conservation, in the ASAP () Asian Species Action Partnership (ASAP). Https://www. context of the Annamese Lowlands Endemic Bird Area. Bird iucn.org/about/work/programmes/species/our_work/asian Conservation International, , –. speciesactionpartnership/ [accessed  November ]. GBIF () Global Biodiversity Information Facility: Free and Open  BIBBY, C.J., BURGESS, N.D., HILL, D.A. & MUSTOE, S.H. Access to Biodiversity Data. Http://www.gbif.org/ [accessed June () Bird Census Techniques. nd edition. Academic Press, ]. London, UK. GRAHAM, C.H., ELITH, J., HIJMANS, R.J., GUISAN, A., PETERSON, A.T.  BIRDLIFE INTERNATIONAL () Archived  discussion: &LOISELLE, B.A. ( ) The influence of spatial errors in species Edwards’s pheasant (Lophura edwardsi) and Vietnamese pheasant occurrence data used in distribution models. Journal of Applied (L. hatinhensis) are considered to be the same species and are treated Ecology, , –. as L. edwardsi: list as Critically Endangered? BirdLife’s Globally HENNACHE, A., MAHOOD, S.P., EAMES, J.C. & RANDI,E.() Threatened Bird Forums. Http://www.birdlife.org/globally- Lophura hatinhensis is an invalid taxon. Forktail, , –. threatened-bird-forums///edwards%E%%s-pheasant- HENNACHE, A., RASMUSSEN, P., LUCCHINI, V., RIMONDI,S.&RANDI, lophura-edwardsi-and-vietnamese-pheasant-l-hatinhensis-are- E. () Hybrid origin of the imperial pheasant Lophura imperialis considered-to-be-the-same-species-and-are-treated-as-l-edwardsi- (Delacour & Jabouille, ) demonstrated by morphology, hybrid list-as-critically-endangered/ [accessed  June ]. experiments, and DNA analyses. Biological Journal of the Linnean  – BIRDLIFE INTERNATIONAL () Lophura edwardsi.InThe IUCN Society, , . Red List of Threatened Species : e.TA. Http:// HIJMANS, R.J. () Introduction to the ‘raster’ package (version .–). dx.doi.org/./IUCN.UK..RLTS.TA.en Http://cran.r-project.org/web/packages/raster/vignettes/Raster.pdf. [accessed  May ]. [accessed  May ]. BOAKES, E.H., FULLER, R.A., MCGOWAN, P.J.K. & MACE, G.M. () HIJMANS, R.J., CAMERON, S.E., PARRA, J.L., JONES, P.G. & JARVIS,A. Uncertainty in identifying local extinctions: the distribution of () Very high resolution interpolated climate surfaces for missing data and its effects on biodiversity measures. Biology Letters, global land areas. International Journal of Climatology, , , . Http://dx.doi.org/./rsbl... –.  BOAKES, E.H., MCGOWAN, P.J.K., FULLER, R.A., CHANG-QING, D., HIJMANS, R.J., PHILLIPS, S., LEATHWICK,J.&ELITH,J.( ) dismo: CLARK, N.E., O’CONNOR,K.&MACE, G.M. (a) Data from: Species Distribution Modeling. Https://cran.r-project.org/web/ Distorted views of biodiversity: spatial and temporal bias in species packages/dismo/index.html [accessed  May ]. occurrence data. Dryad Digital Repository, http://dx.doi.org/./ HOFFMANN, M., HILTON-TAYLOR, C., ANGULO, A., BÖHM, M., dryad.. BROOKS, T.M., BUTCHART, S.H.M. et al. () The impact of ’  BOAKES, E.H., MCGOWAN, P.J.K., FULLER, R.A., CHANG-QING,D., conservation on the status of the world s vertebrates. Science, , – CLARK, N.E., O’CONNER,K.&MACE, G.M. (b) Distorted . views of biodiversity: spatial and temporal bias in species occurrence IUCN () WCC--Res- Conservation of tropical Asia’s data. PLoS Biology, (), e. threatened species. Http://portals.iucn.org/library/sites/library/files/     BRICKLE, N.W., DUCKWORTH, J.W., TORDOFF, A.W., POOLE, C.M., resrecfiles/WCC_ _RES_ _EN.pdf. [accessed May ].  TIMMINS,R.&MCGOWAN, P.J.K. () The status and JOHNSON, C.J. & GILLINGHAM, M.P. ( ) Sensitivity of conservation of Galliformes in Cambodia, Laos and Vietnam. species-distribution models to error, bias, and model design: an Biodiversity and Conservation, , –. application to resource selection functions for woodland caribou.  – BROOKS, T.M., PIMM, S.L. & OYUGI, J.O. () Time lag between Ecological Modelling, , . deforestation and bird extinction in tropical forest fragments. MAHOOD, S.P. & EAMES, J.C. (in press) Is Edwards’s pheasant Lophura Conservation Biology, , –. edwardsi extinct in the wild? Forktail.  BROWN, J.H. () On the relationship between abundance and MÜLLER,D.&ZELLER,M.( ) Land-use dynamics in the Central distribution of species. The American Naturalist, , –. Highlands of Vietnam: a spatial model combining village survey CLEMENTS, C.F. () sExtinct: Calculates the historic date of data with satellite imagery interpretation. Agricultural Economics, extinction given a series of sighting events. R package version .. , –. Https://CRAN.R-project.org/package=sExtinct. NAIMI,B.() usdm: Uncertainty analysis for species distribution    CLEMENTS, C.F., WORSFOLD, N.T., WARREN, P.H., COLLEN,B.,CLARK, models. R package version . - . Https://CRAN.R-project.org/ N., BLACKBURN,T.M.&PETCHEY,O.L.() Experimentally testing package=usdm. the accuracy of an extinction estimator: Solow’s optimal linear NAIMI,B.,HAMM, N.A.S., GROEN, T.A., SKIDMORE, A.K. & estimation model. Journal of Animal Ecology, , –. TOXOPEUS, A.G. () Where is positional uncertainty  COLLEN, B., PURVIS,A.&MACE, G.M. () When is a species really a problem for species distribution modelling? Ecography, , extinct? Testing extinction inference from a sighting record to –. inform conservation assessment. Diversity & Distributions, , NAIMI,B.,SKIDMORE, A.K., GROEN, T.A. & HAMM, N.A.S. () –. Spatial autocorrelation in predictors reduces the impact of COOKE,P.() Optimal linear estimation of bounds of random positional uncertainty in occurrence data on species distribution variables. Biometrika, , –. modelling. Journal of Biogeography, , –.      DUCKWORTH, J.W., BATTERS, G., BELANT, J.L., BENNETT, E.L., NORSYS CORP ( ) Netica . for MS Windows ( to ). BRUNNER, J., BURTON, J. et al. () Why South-east Asia should Https://www.norsys.com/netica.html. be the world’s priority for averting imminent species extinctions, PHAM, T.A. & LE, T.T. (compilers) () Action Plan for the and a call to join a developing cross-institutional programme to Conservation of the Edwards’s Pheasant Lophura edwardsi – tackle this urgent issue. Sapiens, , –.  with vision to . Viet Nature Conservation Centre, Hanoi, EAMES, J.C. () Ke Go Nature Reserve, the place of wood. World Vietnam. Http://thiennhienviet.org.vn/ep/wp-content/uploads/ Birdwatching, , –. //EP-Action-Plan-VN-EPWG.pdf [accessed  May ].

PHILLIPS, S.J., ANDERSON, R.P. & SCHAPIRE, R.E. () Maximum conservation planning with scarce resources. Ecological entropy modelling of species geographic distributions. Ecological Applications, , –. Modelling, , –. UN SUSTAINABLE DEVELOPMENT KNOWLEDGE PLATFORM () RDEVELOPMENT CORE TEAM () R: A Language and Environment Sustainable Development Goals. Https://sustainabledevelopment. for Statistical Computing. R Foundation for Statistical Computing, un.org/?menu= [accessed  November ]. Vienna, Austria. VAN PROOSDIJ, A.S.J., SOSEF, M.S.M., WIERINGA, J.J. & RAES,N. RIVADENEIRA, M.M., HUNT,G.&ROY,K.() The use of sighting () Minimum required number of specimen records to develop records to infer species extinctions: an evaluation of different accurate species distribution models. Ecography, , –. methods. Ecology, , –. WIJESINGHE, M.R. & BROOKE,M.DE L. () Impact of habitat ROBERTS, D.L. & SOLOW, A.R. () Flightless birds: when did the disturbance on the distribution of endemic species of small dodo become extinct? Nature, , . mammals and birds in a tropical rain forest in Sri Lanka. Journal of ROBSON, C., EAMES, J.C., WOLSTENCROFT, J.A., NGUYEN,Cu& Tropical Ecology, , –. TRUONG, V.L. () Recent records of birds from Vietnam. WILLCOX,D.() The conservation status of small carnivores in the Forktail, , –. Ke-Go-Khe Net Lowlands, Central Vietnam. Small Carnivore SECRETARIAT OF THE CONVENTION ON BIOLOGICAL DIVERSITY Conservation, , –. () Decision Adopted by the Conference of the Parties to the WILLIAMS, S.E. & PEARSON, R.G. () Historical rainforest Convention on Biological Diversity at its Tenth Meeting: X/. The contractions, localized extinctions and patterns of vertebrate Strategic Plan For Biodiversity – and the Aichi Biodiversity endemism in the rainforests of Australia’s wet tropics. Proceedings of Targets. Https://www.cbd.int/doc/decisions/cop-/ the Royal Society B, , –. cop--dec--en.pdf [accessed  March ]. WISZ, M.S., HIJMANS, R.J., LI, J., PETERSON, A.T., GRAHAM, C.H., SMITH, C.S., HOWES, A.L., PRICE,B.&MCALPINE, C.A. () Using GUISAN,A.&NCEASPREDICTING SPECIES DISTRIBUTIONS a Bayesian belief network to predict suitable habitat of an WORKING GROUP () Effects of sample size on the performance endangered mammal—the Julia Creek dunnart (Sminthopsis of species distribution models. Diversity & Distributions, , douglasi). Biological Conservation, , –. –. SOLOW, A.R. () Inferring extinction from a sighting record. WORLDCLIM () WorldClim climatic dataset Version .. Http:// Mathematical Biosciences, , –. www.worldclim.org/. STRAUSS,D.&SADLER, P.M. () Classical confidence intervals and Bayesian probability estimates for ends of local taxon ranges. Biographical sketches Mathematical Geology, , –. SUWANRAT, S., NGOPRASERT, D., SUTHERLAND, C., SUWANWAREE,P. MATT GRAINGER is interested in developing pragmatic solutions to &SAVINI,T.() Estimating density of secretive terrestrial birds complex conservation and sustainability problems. DUSIT (Siamese fireback) in pristine and degraded forest using camera NGOPRASERT is focused on the conservation and population ecology traps and distance sampling. Global Ecology and Conservation, , of mammalian carnivores and birds. PHILIP MC G OWAN seeks to –. understand how species conservation can be achieved efficiently and TANTIPISANUH, N., GALE, G.A. & POLLINO,C.() Bayesian effectively and TOMMASO SAVINI is interested in behavioural ecology networks for habitat suitability modeling: a potential tool for and landscape use of mammals and birds.