Assessment of the conservation status of partridges and pheasants in South East Asia

Biodiversity and Conservation 6, 1321±1337 (1997) Assessment of the conservation status of partridges and pheasants in South East Asia PHILIP MCGOWAN* à and MICHAEL GILLMAN Ecology and Conservation Research Group, Biology Department, The Open University, Walton Hall, Milton Keynes MK7 6AA, UK Received 16 October 1996; accepted 20 January 1997 The recent compilation of IUCN Action Plans for partridges and pheasants has highlighted the need for locality data to be used in conservation evaluation. Here we used locality data to measure changes in the proportion of localities from which each of 25 South East Asian species has been recorded, changes in their geographic extent and the amount of habitat that is potentially available to them. These are measures that use the data which are most readily available. Single landmass endemics from both montane and lowland forests appear to have fared badly with peacock pheasants Polyplectron and hill-partridges Arborophila su ering particularly large declines, although for montane forest species, these apparent declines are probably the result of relatively less survey e ort in highland areas. Species previously widespread that give cause for concern include the green peafowl, Pavo muticus, and the blue-breasted quail, Coturnix chinensis. Conservation priorities include action on behalf of lowland forest specialists, some of which have su ered substantial reductions in the area of habitat that is potentially available. Surveys are required on the hills and mountains of Borneo, the Malay Peninsula and Sumatra, and in the Lesser Sundas. Increased resolution of vegetation data would permit more accurate prediction of species ranges. Keywords: partridges; pheasants; South East Asia; conservation status; tropical forest. Introduction The status of partridge and pheasant species has recently been assessed qualitatively through the compilation of Action Plans for partridges, quail and francolin (McGowan et al., 1995), and pheasants (McGowan and Garson, 1995), and the revision of the BirdLife International list of threatened bird species (Collar et al., 1994). Both assessments used draft versions of the recently revised IUCN Red List system to apply threat categories to individual taxa. The result has been a more systematic approach to considering extinction risk than has previously been the case. In addition it has led to a realization of the types of data that are most useful when discussing the conservation status of poorly known species. Knowledge of individual localities from which a species has been recorded through time is required for the objective determination of priorities centred on both species and areas. We address that need for partridges and pheasants of the Sunda subregion of South East Asia. Here we measure changes in the proportion of localities from which each of 25 species has been recorded, changes in their geographic extent and the amount of habitat that is potentially available to them. From this we identify habitats and landmasses most in need of attention, either in terms of conservation action or additional data gathering. * To whom correspondence should be addressed. à Present address: Ecoscope Applied Ecologists, 9 Bennell Court, Comberton, Cambridge, CB3 7DS, UK. 0960-3115 Ó 1997 Chapman & Hall

1322 McGowan and Gillman The Sunda subregion The Sunda subregion has its northern boundary in the Malay Peninsula at the Isthmus of Kra at about 10 N (e.g. Walker, 1982; Treherne and Cranbrook, 1988), although Whitmore (1984a) indicates that there is a oristic boundary near the Thai±Malaysian border. From this north-western limit, the subregion covers land that lies on the Sunda shelf: Malaysia, Brunei, Singapore and Indonesia east to Bali, and north to the Palawan group of islands in the Philippines. Vertebrate zoogeographers tend to consider that in the northwest the subregion also contains the peninsular parts of Thailand and Burma (e.g. birds: Medway and Wells, 1976; mammals: Cranbrook, 1988). Considering the natural distributions of South East Asian partridges and pheasant species, we include the Lesser Sunda islands and Sulawesi here. As the distribution data for the Palawan peacock pheasant lack much detail, we have omitted the Philippines from this study. Pheasants and partridges in the Sunda subregion An area of high pheasant species (Johnsgard, 1986) and partridge genus (Johnsgard, 1988) richness, the Sundas contain 13 species in six genera that occur nowhere else. An additional seven species, including those belonging to four monotypic genera, have isolated records as far north as 15 N in Peninsular Thailand and Burma, but the majority of their distributions lie within the Sunda subregion. These seven `near subregional endemics' appear as reliant on the forest formations of the Sunda subregion as do the 13 endemic species. Loss of forest cover in the Sunda subregion It is widely believed that rainforest was the natural climax vegetation across virtually all of Malesia (Richards, 1952; Whitmore, 1984b), a vegetation type that has been subjected to high levels of attrition this century. For example, Peninsular Malaysia covers 131 598 square kilometres of land (Collins et al., 1991), virtually all of which was originally covered with tropical forest (Myers, 1988). One hundred years ago 90% of the land, or around 120 000 square kilometres, was believed to be still under natural cover. At the start of this century the development of tobacco and rubber plantations on the west coast began in earnest, to be followed by the commencement of widespread logging extraction in the 1950s using the Malayan Uniform System and, most recently, the establishment of palm oil estates throughout the peninsula since the 1960s (Wells, 1985). Forest cover has reduced from 90% at the end of the last century, to 68% in 1966, 63% by 1970±2 and 47% by the end of 1985, extrapolated to 43% by 1990 (Collins et al., 1991). In 1985, 47 500 square kilometres (77%) of the land under forest was designated as Permanent Forest Estate, of which 19 000 square kilometres is set aside for protection and amenity. Hidden further within these gures is the scale of forest loss in the level lowlands. Much of the land in the peninsula lies over at or gently undulating terrain in the lowlands and which is therefore easily accessible to logging operations. Consequently, most of the forest that is still undisturbed is found only on the slopes and ridges of the Main Range, where forest clearance activities are economically less rewarding. Indeed, most of the designated protection forest is in this mountain range (Collins et al., 1991). Whilst the rate of forest loss has declined from the late 1970s when the annual rate was 2500 square kilometres, to just under 1000 square kilometres per year during 1985±90, this seems to be too late to conserve much rainforest in the level lowlands. At present around 5700 square kilometres are

Partridge and pheasant status in South East Asia 1323 contained within the protected area network (Collins et al., 1991), little of which is in large blocks in the lowlands. The loss of the forests of the level lowlands is so acute that its avifauna is considered to be the most seriously threatened bird community throughout the Sunda subregion (Wells, 1985). Methods Localities Data on localities were collected from a wide variety of journals, reports and eldworkers active in the regions. In some cases co-ordinates were presented in these source papers. In most cases, however, gazetteers and maps were required. Some information was not used either because places could not be traced or the citing of too large areas as the locality, such as the `Upper Mahakam', a large river in east Borneo. Places could not be traced for one of three reasons, namely, references to small areas that do not appear on maps or in gazetteers; changes in place names since collections were made; or inappropriate transliteration between languages (for example, Dutch spelling of Indonesian names, or English spelling of Malay names). The vast majority of co-ordinates is believed accurate to within 10 of the collection or observation site and are included here. All localities used are thought to be accurate to within one degree. Where sites could not be determined to within one degree they have not been used. The dataset was split into two time periods, pre-1970 and 1970 onwards to access changes in information available and the status of species. From this dataset, the number of localities for each species on each landmass can be determined. Furthermore, the number of localities as a proportion of the total number of localities on that landmass can be calculated for each species to allow for variation in search e ort between landmasses and time periods. The spatial distribution of the localities is accounted for by determining the extent of occurrence (Gaston, 1994; Species Survival Commission, 1994). This is measured by connecting the outermost points of the distribution such that no internal angle is greater than 180, except where this would result in areas of sea being included within the extent of occurrence. The Lesser Sundas were excluded from this analysis as the data obtained from these islands were inadequate. Habitat Habitat maps were supplied by the World Conservation Monitoring Centre from its Biodiversity Map Library, and derived from the sources listed in Collins et al. (1991) where they are presented. These maps were used to estimate the amount of habitat potentially available to each species on each landmass. This was done by measuring the area of each habitat polygon that the species occurred in. Therefore, if there was a species record in a lowland forest polygon, it was assumed that the whole of that block was potentially available to the species. If there was now a non-forest polygon that contained an old record for a species, the whole of that polygon was treated as potentially available to the species pre-1970. Recent records of species in southern Thailand from Holmes (1973) and Holmes and Wells (1975) were not included in this analysis as they simply listed occurrence in each province. Given the small size and wide dispersion of all habitat polygons comprising forest in southern Thailand, the locality data were not detailed enough to allow con dent assignation to any particular habitat polygon. The Lesser Sundas were omitted from this

1324 McGowan and Gillman analysis for the same reason: indication that the species simply occurred on the island was not su cient to place the record within a habitat polygon with con dence. In order to conduct comparisons between habitat types, all species have been assigned to one of three broad habitat categories. For most partridges and pheasants this was relatively straightforward as they are considered dependent upon intact forest. The birdhabitat categories used were lowland forest specialists, montane forest specialists and not forest specialists. Species were assigned to the rst two categories following Wells (1985), and the species that he does not mention as forest specialists are treated as `not forest specialists'. Caloperdix oculea has been omitted from some analyses as it is considered a lowland forest specialist on Borneo and Sumatra and a montane forest specialist on the Malay Peninsula. Mean changes were calculated for habitats on each landmass and also for each landmass for each of the three measures. These means were calculated from raw data, not as means of proportionate changes for each species. Results Changes in the proportion of localities Since 1970 the lack of recent information for each species is acute, especially when considered by landmass (see Appendix). For example, there are records from fewer than 10 localities for 13 of the 14 species that have been recorded on Sumatra. There are no recent records for four species on one landmass that they inhabited prior to 1970: Arborophila charltonii and Coturnix chinensis on Sumatra, Coturnix ypsilophorus in the Lesser Sundas, and Pavo muticus on the Malay Peninsula. The change in the proportion of localities from which each species was recorded (essentially its detection rate) is given in Table 1. When comparing the proportions between the two periods, the only species for which there was a signi cant di erence across the whole subregion, in fact an increase over time, was Argusianus argus (G ˆ 17:80, p < 0:0001). This overall agreement provides some support for the robustness of the sampling: if there were substantial di erences in sampling e ectiveness between the two periods, more variation between the two data sets would be expected. Both Arborophila charltonii (G ˆ 3:48) and Coturnix chinensis (G ˆ 3:05), however have high reductions which are approaching signi cance ( p values between 0.05 (G ˆ 3:84) and 0.1 (G ˆ 2:71)). Five of the seven species with the highest proportional decrease are restricted to one landmass and are subregional endemics or near-endemics. Considering all three habitats across the subregion as a whole, there is no change in the detection rate of species (ˆ 1:00; Table 1). This implies that each locality visited has yielded as many species records since 1970 as they did pre-1970. When considering each habitat separately, however, there are changes between the two periods. More lowland forest species have been detected at sites visited since 1970 than before 1970 (Table 1). For the subregion as a whole, this proportional increase is 1.22. The other two habitats show overall decreases in mean changes, being bigger in montane forest (0.78). Treating the landmasses separately, the proportion of sites from which species were recorded since 1970 was higher than before, although not signi cantly (G test), on both Borneo (1.03) and the Malay Peninsula (1.26) (Table 1). These general patterns, however, hide considerable variation in the change in detection rate within each habitat on each landmass. The only signi cant change is a 1.56 increase in

Partridge and pheasant status in South East Asia 1325 Table 1. Change over time in the proportion of localities from which each species was recorded on each landmass; numbers are the post 1970 proportion of localities on any given landmass that a species has been recorded from divided by the pre-1970 proportion Species Habitat Borneo Sumatra Java Lesser Sundas Malay Peninsula All Sunda subregion Arborophila charltonii L 0.41 0 0.52 0.44 Arborophila hyperythra M 1.29 1.29 Arborophila javanica M 1.77 1.77 Arborophila orientalis M 0.42 1.11 0.83 0.66 Arborophila rubrirostris M 0.58 0.58 Argusianus argus L 1.54 4.24 3.11 2.32 Caloperdix oculea L/M 0.56 0.64 1.04 0.88 Coturnix chinensis N 0.90 0 0.55 2.25 0.29 0.60 Coturnix ypsilophorus N 0 0 Gallus gallus N 1.82 0.87 1.5 1.68 1.38 Gallus varius N 0.72 0.9 0.73 Haematortyx sanguiniceps M 0.94 0.94 Lophura bulweri L 0.56 0.56 Lophura erythrophthalma L 1.43 0.53 2.54 1.68 Lophura ignita L 1.12 1.41 1.17 1.05 Lophura inornata M 0.40 0.40 Melanoperdix nigra L 1.12 1.06 0.50 0.79 Pavo muticus N 0.86 0.00 0.74 Polyplectron chalcurum M 0.72 0.74 Polyplectron inopinatum M 0.64 0.64 Polyplectron malacense L 0.54 0.54 Polyplectron schleiermacheri L 0.32 0.32 Rheinardia ocellata L 1.12 2.33 2.33 Rhizothera longirostris L 0.50 0.42 1.75 1.14 Rollulus rouloul L 1.33 0.53 1.68 1.31 Means by habitat L 1.04 1.19 1.56 1.22 a M 1.06 0.57 1.60 0.82 0.78 a N 0.90 0.91 0.77 0.94 0.86 0.86 Means by landmass 1.03 0.87 0.85 0.94 1.26 1.00 Habitat classes follow Wells (1985) and are denoted by L (lowland forest specialist), M (montane forest specialist) and N (not a forest specialist). Mean changes are also given for each habitat on each landmass and for each landmass. a Caloperdix oculea omitted from calculation for the whole subregion as it is considered a lowland forest specialist on Sumatra and Borneo and a montane forest specialist on the Malay Peninsula by Wells (1985). the proportion of lowland species recorded per site on the Malay Peninsula (G ˆ 4:25, p <0:05). Also of note, however, is the striking decrease in detection rate of Sumatran montane species of which 0.57 were recorded since 1970 for each species recorded before (G ˆ 2:70, p ˆ 0:1). Montane forest species also show a decrease in detection rate on the Malay Peninsula (0.82). Species that are not forest specialists show a reduction in detection on all landmasses, being most marked on Java (0.77). As noted above, detection rates

1326 McGowan and Gillman of lowland forest species have increased on all landmasses and there is a large increase in the detection rate of montane forest species on Java (1.60). Changes in the extent of occurrence The paucity of locality data for species on many landmasses means that changes in extents of occurrence could only be calculated across the subregion for 15 of the 25 species. There are three cases where a species has not been recorded recently on a landmass. These are Arborophila charltonii and Coturnix chinensis on Sumatra and Pavo muticus on the Malay Peninsula. In a further eight instances a species has recently been recorded from only one or two localities on a landmass on which it was more widespread prior to 1970 (see Appendix for number of localities) and extents of occurrence could not be constructed. Where extents of occurrence could be calculated for both periods, there are 13 out of 34 cases in which a species' extent of occurrence on one landmass is now less than half the area formerly occupied (excluding the apparent landmass extinctions noted above). Eight of these reductions are by more than three-quarters. Three of the biggest decreases occur on Sumatra, where the extents of occurrence of Arborophila rubrirostris, Lophura inornata and Rollulus rouloul are all less than 10% of those prior to 1970. The biggest decrease of all is Melanoperdix nigra on the Malay Peninsula which has seen a reduction of 98%. Across the subregion as a whole, it is single landmass endemics that are faring worst. The ve species that now have global extents of occurrence of less than a quarter of that prior to 1970 (Table 2) are all found on only one landmass: Arborophila rubrirostris, Haematortyx sanguiniceps, Lophura inornata, Polyplectron chalcurum, and Polyplectron malacense. Four species now have larger global extents of occurrence than before 1970, two of which are montane species restricted to one landmass (Arborophila hyperythra and Arborophila javanica), Argusianus argus which is a lowland forest specialist that occurs on three landmasses and Gallus gallus which is not a forest specialist and is fairly widespread throughout the subregion. The last species, however, has su ered a contraction on Java of two-thirds and a single record on Sumatra is responsible for the more than fourfold increase in the extent of occurrence there. Considering the habitats across the whole subregion, there is a massive reduction in the mean extents of occurrence for montane species, which were recently recorded from 27% of the area previously inhabited. Among the means for montane species on each landmass, the Javan result (1.60) stands out, not just as the only increase, but as such a large increase. In contrast, the extents of occurrence of Sumatran montane species have decreased by over 85% on average. The means for lowland forest species show a decrease only on Borneo. The species that are not forest specialists vary from a substantial increase in mean extent of occurrence on Sumatra (probably due to the Gallus gallus result noted above) to a two-thirds reduction on the Malay Peninsula. There has been a reduction in the mean extent of occurrence on all four landmasses. Across the subregion as a whole the mean reduction is 15% for the 15 species (Table 2). The biggest mean reduction per landmass is more than one-quarter for the 12 species on the Malay Peninsula. Species on Java have seen a reduction of one- fth on average, and those on Sumatra of nearly one- fth. The mean extent of occurrence on Borneo has decreased by one-tenth (Table 2).

Partridge and pheasant status in South East Asia 1327 Table 2. Change in the extent of occurrence of species in South East Asia; numbers are the post 1970 occurrence expressed as a proportion of the pre-1970 extent of occurrence Species Habitat Borneo Sumatra Java Malay Peninsula All Sunda subregion Arborophila charltonii L ± 0 0.57 ± Arborophila hyperythra M 1.14 1.14 Arborophila javanica M 1.60 1.60 Arborophila orientalis M ± ± 0.64 ± Arborophila rubrirostris M 0.07 0.07 Argusianus argus L 1.05 1.40 1.43 1.32 Caloperdix oculea L/M ± 0.85 0.59 ± Coturnix chinensis N 1.62 0 0.46 ± ± Gallus gallus N 4.42 0.33 1.14 1.58 Gallus varius N 0.85 0.85 Haematortyx sanguiniceps M 0.23 0.23 Lophura bulweri L 0.82 0.82 Lophura erythrophthalma L 0.50 ± 2.34 ± Lophura ignita L 0.95 1.06 0.31 0.90 Lophura inornata M 0.05 0.05 Melanoperdix nigra L 0.20 ± 0.02 ± Pavo muticus N 1.02 0 0.49 Polyplectron chalcurum M 0.22 0.22 Polyplectron inopinatum M 0.84 0.84 Polyplectron malacense L 0.10 0.10 Polyplectron schleiermacheri L ± ± Rheinardia ocellata L ± ± Rhizothera longirostris L ± ± 0.31 ± Rollulus rouloul L 0.94 0.07 0.85 0.78 Means by habitat L 0.80 1.00 1.00 0.94 M 0.37 0.14 1.60 0.62 0.27 N 1.62 2.84 0.73 0.59 1.05 Means by landmass 0.90 0.83 0.80 0.72 0.85 The Lesser Sundas are omitted here because of the poor resolution of the locality data with respect to island size. ± indicates that the species is present, but there are insu cient data to calculate percentage change (only one or two records for one or other of the periods). Habitat classes follow Wells (1985) and are denoted by L (lowland forest specialist), M (montane forest specialist) and N (not a forest specialist). Caloperdix oculea is considered a lowland forest specialist on Sumatra and Borneo and a montane forest specialist on the Malay Peninsula. Changes in the availability of potential habitat Across the four landmasses treated, eight species appear to have undergone reductions in potentially available habitat of three-quarters or more (Table 3). The possible restriction of the Bornean endemic Polyplectron schleiermacheri to 0.2% of the habitat which it may have had available to it pre-1970 is the biggest overall reduction recorded, followed by the

1328 McGowan and Gillman Table 3. Change in the amount of habitat potentially available to species in South East Asia; numbers are the extent of post 1970 habitat available as a proportion of that pre-1970 Species Habitat Borneo Sumatra Java Malay Peninsula All Sunda subregion Arborophila charltonii L 0.33 0 0.39 0.19 Arborophila hyperythra M 0.95 0.95 Arborophila javanica M 0.01 0.01 Arborophila orientalis M 0.24 243.32 0.98 1.35 Arborophila rubrirostris M 0.25 0.25 Argusianus argus L 0.92 1.27 0.38 0.84 Caloperdix oculea L/M 0.95 0.11 0.29 0.60 Coturnix chinensis N 0.88 0 1.00 0.90 0.76 Gallus gallus N 2.34 1.00 1.39 1.47 Gallus varius N 1.00 0.99 Haematortyx sanguiniceps M 0.10 0.10 Lophura bulweri L 0.89 0.79 Lophura erythrophthalma L 0.90 0.0004 0.26 0.51 Lophura ignita L 0.85 0.46 0.48 0.64 Lophura inornata M 0.94 0.94 Melanoperdix nigra L 0.93 0.08 0.74 0.63 Pavo muticus N 1.07 0.00 0.51 Polyplectron chalcurum M 0.11 0.11 Polyplectron inopinatum M 0.15 0.15 Polyplectron malacense L 0.25 0.25 Polyplectron schleiermacheri L 0.002 0.002 Rheinardia ocellata L 1.30 1.30 Rhizothera longirostris L 0.85 0.09 0.45 0.68 Rollulus rouloul L 0.88 0.15 0.58 0.59 Means by habitat L 0.72 0.30 ± 0.37 0.57 M 0.52 0.23 1.00 0.28 0.50 N 0.78 1.19 1.00 0.78 0.89 Means by landmass 0.70 0.38 1.00 0.44 0.61 The Lesser Sundas are omitted here because of the poor resolution of the locality data with respect to habitat fragment size. Habitat classes follow Wells (1985) and are denoted by L (lowland forest specialist), M (montane forest specialist) and N (not a forest specialist). Caloperdix oculea is considered a lowland forest specialist on Sumatra and Borneo and a montane forest specialist on the Malay Peninsula. Javan montane endemic Arborophila javanica (1.0%). Both of these species, along with Rheinardia ocellata now appear to have less than 1500 square kilometres of potential habitat available in the subregion. The presence of ve montane species among the eight species with the largest reductions in available habitat is of note. Considering that the widespread disruption of habitat in the Sunda subregion has, so far, mostly occurred in the lowlands this result seems counter-intuitive and is discussed below. Ten of the eleven species considered by Wells (1985) to be lowland forest specialists across the whole subregion (i.e. excluding Caloperdix oculea) now have less habitat

Partridge and pheasant status in South East Asia 1329 potentially available to them than before 1970. Eight have witnessed reductions greater than 30% (Table 3). The mean reduction in potential habitat available was greatest on Sumatra where more than 60% of the habitat available to these partridges and pheasants appears to have been lost (Table 3). The species on the Malay Peninsula have su ered a mean reduction of 56% which is only marginally less severe. On both landmasses, the means are raised by high values for species that are not forest specialists. Discussion Changes in status Single landmass endemics appear to have undergone the biggest reductions according to all three measures. The largest reductions are evident among both highland and lowland species and on all landmasses. A lack of recent records for Polyplectron schleiermacheri is clearly a major cause for concern. Other species that deserve attention according to all measures are Arborophila charltonii, Arborophila rubrirostris, Haematortyx sanguiniceps, Polyplectron inopinatum, Polyplectron chalcurum and Polyplectron malacense. There are several other noteworthy results. Coturnix chinensis has not been recorded recently on Sumatra and across the subregion it has recently been recorded at 40% fewer sites that have been searched compared with pre-1970. This quail has always been considered to be widely distributed throughout Asia, but the decrease in records across the subregion is striking for a species still perceived as common and widespread in standard texts (e.g. Johnsgard, 1988). There appears to have been relatively little recent ornithological activity in the Lesser Sundas, resulting in a lack of records for Coturnix ypsilophorus, a species which is considered widespread further east (Johnsgard, 1988), outside the subregion. In addition to Coturnix ypsilophorus and Coturnix chinensis, there are no recent records for two species from one landmass on which they were previously reported: Arborophila charltonii on Sumatra, and Pavo muticus on the Malay Peninsula. The Sumatran form of Arborophila charltonii has been assigned to the subspecies A. c. atjenensis, which is now considered to be of dubious validity (G. Mees in McGowan et al., 1995). Nonetheless, given uncertainty over its exact habitat and altitude distribution, the concentration of records in two widely separated localities and the widespread loss of forest, the taxon merits urgent attention. Pavo muticus is thought to have become extinct on the Malay Peninsula, with the last con rmed reports from Malaysia in the early 1960s. The species is thought to be extinct in Peninsular Thailand (McGowan and Garson, 1995). Although not a subregional endemic, and recently found at a number of sites on Java (van Balen et al., 1995), the species is a conservation priority. Formerly widespread throughout Asia, south and east from northeast India/south-west China, and readily seen throughout its range (Delacour, 1977), the species is now restricted to isolated localities in most countries (McGowan and Garson, 1995; Evans and Timmins, 1995). It is likely that few of the sites on Java hold large enough populations to be considered viable. Rheinardia ocellata is only known from two sites within the subregion. Con rmed from mid-altitude on two mountains, and almost certainly to be found on other hills within the same complex (McGowan and Garson, 1995), the Peninsular Malaysia population falls almost entirely within Taman Negara National Park. The species is also found in Vietnam

1330 McGowan and Gillman (see sites listed in Robson et al., 1989, 1993a, b). Consequently, the small number of sites is not cause for immediate concern, although some monitoring is advisable. Across the subregion as a whole, lowland forest and montane forest specialists both appear to have witnessed reductions in the amount of habitat available to them. Whilst the montane forest species appear to have fared badly according to all measures, for lowland forest specialists, this habitat result contradicts those measures based on localities only. The interaction between these measures is discussed below, but the decline in the amount of habitat available to all lowland forest species is clearly of concern. Methodological constraints on interpretation Interpreting these results as assessments of conservation status requires caution. This is because, as for most taxa in tropical countries, the assessments are bedevilled by a lack of information. For example, much of the forest loss on Sumatra is in the lowlands (Collins et al., 1991) where the tall dipterocarp forests lie over relatively level land, making extraction economically attractive. According to locality only measures, the montane species appear to be su ering large declines: that is there are relatively few recent records. For example, Lophura inornata emerges as having one of the largest declines according to extent of occurrence and proportion of localities, but the extent of its potentially available habitat seems to have decreased relatively little (94.4% of pre 1970 habitat was available post 1970). An interpretation of this result is that montane areas of Sumatra have been searched only on few occasions since 1970 and that there are thus few records of any montane species. During the same period, the montane forests in which L. inornata lives have been relatively little touched and, whilst the absence of records for this species is of concern, it is unlikely to have witnessed the scale of decline predicted by the locality data. Conversely, the montane species Arborophila javanica appears to be faring satisfactorily on Java according to the two locality based measures, but has seen the second biggest decline according to the habitat measure. This is the result of two factors: the intensive searching for this species recently (van Balen, 1992) and deforestation on Java being so extensive that montane areas are now a ected. The intensive searching produced a reasonable number of contemporary records and the deforestation has been so extensive that reasonable estimates of the former boundary between lowland and montane forest are not possible from this habitat dataset. Therefore, old sites that are now deforested lie in massive non-forest polygons giving the impression of having had unrealistically large areas of habitat potentially available historically that are now not available. We believe that this situation is currently limited to Java and that deforestation is not yet so far advanced on the other landmasses as to obscure the boundary between what was lowland forest and what was montane forest, and thus result in excessive predictions of the amount of habitat that was formerly potentially available. Once the probable recent under-surveying of montane areas is acknowledged, the interaction between the measures used can be very revealing about their utility. When the proportions of localities were compared for each habitat there was an increase in the detection of lowland forest species. the increase indicates that each site visited in the subregion contained 1.22 more lowland species recently than pre-1970 and could be due to one of two factors: one, a real increase in the number of species per site, and the other, a consequence of sampling. In the post 1970 period there were either more species records per locality in the lowlands, or a higher proportion of sites containing lowland species were searched than previously. Given our assertion that there has been little e ort in the

Partridge and pheasant status in South East Asia 1331 highlands, the increase in the detection of lowland species is probably simply a re ection of relatively less survey e ort in the more remote areas. The relatively higher survey e ort in the lowlands is probably due to ease of access, both as montane sites are normally in more interior locations and thus further from human habitation, roads etc., and also because the fragmentation of the lowland forest has opened up many more areas which are physically separate from each other. This may suggest that for lowland forest species consideration of the habitat is of key importance. The declines in the amount of lowland forest that is potentially available is likely to be a more indicative assessment of their status. Most of the historical records of species are based on collections made by museum workers and using leg snares. Snaring appears to be unbiased in the species that it records. Recent records, however, are based on both visual and vocal observations and the likelihood of detection varies from species to species. Argusianus argus has one of the most distinctive and far carrying calls in South East Asian rainforest and as such is far more easily encountered than all other species. As it appears that historical accounts usually did not include vocal registrations, the inclusion of vocal records, which are more readily obtained than snare records in contemporary lists, will a ect the interpretation of status based on locality information alone. Consequently, for this species, consideration of the status of its habitat is especially important in understanding the species' status. Conservation priorities The foregoing indicates that translating results from our measures of change into conservation recommendations for this group of species is not straightforward. Nonetheless, several priorities are clear in terms of both information needs and direct action. (i) The decline in the amount of lowland forest habitat available across the subregion is a major conservation concern. Lowland forest specialists on Sumatra and the Malay Peninsula appear to be faring extremely badly. Not only is this habitat being reduced in size, but the attendant fragmentation may result in blocks of habitat that are too small to support viable populations. Therefore, whilst there may be lowland forest present in these small blocks, it is e ectively useless for these species. The situation facing this habitat is exempli ed by three species which are among the highest priorities in the subregion. Polyplectron schleiermacheri is endemic to the lowland forest of Borneo and remains di cult to con rm at any site. Several unsubstantiated reports have been received (McGowan, unpublished data), but none of these have been con rmed. The species does appear to be di cult to record. Lowland forest has been disrupted along many major waterways on Borneo and may be causing acute problems for this species should it use similar forests to the closely related Polyplectron malacense of the Malay Peninsula. P. malacense itself is worthy of attention. A species that previously was found in forest from southern Peninsular Thailand and extreme southern Burma through most of Peninsular Malaysia, the species is now found in only a few forest blocks. Given that the species does not appear to be evenly distributed throughout all available lowland forest (McGowan, 1994), the area actually occupied by this species is likely to be smaller than the amount of lowland forest in the species' range. Furthermore, the species may well be extinct in Thailand (Round, 1988) and as lowland forest is still being logged in Peninsular Malaysia, the species may only survive in the few well protected areas within its range, such as Taman Negara and the Krau Wildlife Reserve. It is not known whether the populations are fragmented within these areas or if they are viable.

1332 McGowan and Gillman A nal lowland species which appears to be poorly known is Melanoperdix nigra. This is a species that is rarely seen and little has been added to our knowledge of this species since early collections reported by, amongst others, Blasius (1883), BuÈ ttikofer (1900), Chasen and Kloss (1932) and Sharpe (1877). Determining the distribution of this species on Sumatra, Borneo and the Malay Peninsula, and understanding its habitat use, must be a conservation priority in the subregion. The apparent plight of Melanoperdix nigra exempli es that of the suite of lowland forest species that are to be found on these three landmasses (including the very poorly known Arborophila charltonii). With increasing fragmentation of the once continuous forest blocks, the identi cation of fragments that are large enough to hold viable populations is needed. This situation faces many bird species in addition to the 13 endemic and seven near-endemic partridges and pheasants included here. (ii) There is a lack of recent information on species inhabiting hill and mountain areas, especially on Sumatra, Borneo and Peninsula Malaysia. This lack of information is presumably due to the remoteness of many of these areas, although the di culty of detection in the absence of trapping must also hinder data collection. Whilst these areas are relatively untouched when compared with the lowland forest that surrounds them, it is clear that the scale of deforestation is such that they are under increasing threat. This treat is, perhaps, most acute on the Malay Peninsula where the development of hill areas for human settlement is increasing and there are still large tracts of undisturbed and unsearched montane forests. Information on the location and extent of populations of these species is needed if strategic planning is to be adopted on behalf of these species. Existing conservation areas, such as Kerinci-Seblat on Sumatra and the Sungei Kayan-Mentarang region on Borneo (McGowan et al., 1995) could be assessed rst. On Java, in contrast, populations of the montane endemic partridge Arborophila javanica have been mapped leaving A. orientalis as the only montane species in need of additional survey work. Conservation e ort at known sites is required on this landmass. There is similar lack of information on the partridge and pheasant species inhabiting the Lesser Sundas. Historical records name localities that have not been traced or simply state that a species occurs on an island. Given the small size of habitat patches on these islands, there is a need for data to be collected at a ner resolution. It should be noted that no partridge and pheasant species is known to be restricted to these islands where it appears that they have been replaced by the related megapodes (Olson, 1980; Dekker, 1989). (iii) Assessing potentially available habitat relies on sound information on species habitat use and accurate vegetation maps. The assessment made here uses the data that are currently available, but two problems are clear. The rst is the di erentiation of vegetation into biologically meaningful classes and the second is the assignation of partridge and pheasant species to suitable habitat use classes. Whilst the distinction between lowland and montane forest is acceptable initially, it is evident that this is insu cient. The boundary between forest on hill slopes and that on level ground is considered to be important in revealing avian distributions in the subregion (Wells, 1988). Thus, species may in fact be restricted to lowland forest on hillsides, whereas current vegetation maps include forest on both hills and level land as one forest type. Attention to undisturbed non-forest areas is also needed to assess habitat availability for non-forest specialists, such as Coturnix chinensis.

Partridge and pheasant status in South East Asia 1333 Acknowledgements We thank the World Conservation Monitoring Centre for access to forest cover data, BirdLife International for restricted range bird species information, and the World Pheasant Association for some contemporary distribution data collected for the Action Plans. Edward Dickinson and F.E. Warr at the Natural History Museum in Tring kindly allowed access to references in their libraries. Simon Blyth helped with access to the habitat data held by the World Conservation Monitoring Centre and Mike Dodd assisted with GIS analyses. The work was funded by a Leverhulme Trust grant to The Open University. References Blasius, W. (1883) VoÈ gel von Borneo, im SuÈ dosten der Insel gessammelt von Herrn F.J. Grabowsky. Verh. Zoolbot. Gessels. Wien Jahr. 1883, 3±92. BuÈ ttikofer, J. (1900) Zoological results of the Dutch Scienti c Expedition to Central Borneo. The birds. Notes from the Leyden Museum 21, 145±276. Chasen, F. N. and Kloss, C. B. (1932) On some birds from Pontianak, Dutch West Borneo. Treubia 14, 11±18. Collar, N.J., Crosby, M.J. and Statters eld, A.J. (1994) Birds to Watch 2: The World List of Threatened Birds. Cambridge, UK: BirdLife International. Collins, N.M., Sayer, J.A. and Whitmore, T.C. (eds) (1991) The Conservation Atlas of Tropical Forests: Asia and the Paci c. London and Basingstoke, UK: Macmillan Press. Cranbrook, Earl of (ed.) (1988) Mammals: distribution and ecology. In Key Environments: Malaysia pp. 146±66. Pergamon Press. Dekker, R.W.R.J. (1989) Predation and the western limits of megapode distribution (Megapodiidae; Aves). J. Biogeog. 16, 317±21. Delacour, J. (1977) The Pheasants of the World. 2nd edition. Hindhead, UK: Spur Publications. Evans, T.D. and Timmins, R.J. (1995) The status of the green peafowl Pavo muticus in Laos. Forktail 11, 11±32. Gaston, K.J. (1994) Measuring geographic range sizes. Ecography 17, 198±205. Holmes, D.A. (1973) Bird notes from southernmost Thailand, 1972. The Natural History Bulletin of the Siam Society 25, 39±66. Holmes, D.A. and Wells, D.R. (1975) Further observations on the birds of south Thailand. The Natural History Bulletin of the Siam Society 26, 61±78. Johnsgard, P.A. (1986) The Pheasants of the World. Oxford University Press. Johnsgard, P.A. (1988) The Quails, Partridges and Francolins of the World. Oxford University Press. McGowan, P.J.K. (1994) Display dispersion and micro-habitat use by the Malaysian peacock pheasant Polyplectron malacense in Peninsular Malaysia. J. Trop. Ecol. 10, 229±44. McGowan, P.J.K., Dowell, S.D., Carroll, J.P. and Aebischer, N.J. (1995) Partridges, quails, francolins, snowcocks and guineafowl: status survey and action plan 1995±1999. Gland, Switzerland: IUCN. McGowan, P.J.K. and Garson, P.J. (1995) Pheasants: status survey and action plan 1995±1999. Gland, Switzerland: IUCN. Medway, Lord and Wells, D.R. (1976) Birds of the Malay Peninsula. Volume 5. London, UK: Witherby. Myers, N. (1988) Threatened biotas: ``Hot spots'' in tropical forests. The Environmentalist, 8, 187± 208. Olson, S.L. (1980) The signi cance of the distribution of the Megapodiidae. Emu 80, 21±4. Richards, P.W. (1952) The Tropical Rain Forest: An Ecological Study. Cambridge, UK: Cambridge University Press.

1334 McGowan and Gillman Robson, C.R., Eames, J.C., Wolstencroft, J.A., Nguyen Cu and Truong Van La (1989) Recent records of birds from Viet Nam. Forktail 5, 71±97. Robson, C.R., Eames, J.C., Nguyen Cu and Truong Van La (1993a) Birds recorded during the third BirdLife/Forest Birds Working Group expedition in Viet Nam. Forktail 9, 89±119. Robson, C.R., Eames, J.C., Nguyen Cu and Truong Van La (1993b) Further recent records of birds from Viet Nam. Forktail 8, 25±52. Round, P.D. (1988) Resident forest birds in Thailand: their status and conservation. Monograph no. 2. Cambridge, UK: International Council for Bird Preservation. Sharpe, R.B. (1877) Contributions to the ornithology of Borneo. Ibis Series 4 1, 1±25. Species Survival Commission (1994) IUCN Red List categories. Gland, Switzerland: IUCN. Treherne, J. and Cranbrook, Earl of (1988) Preface. In Key Environments: Malaysia (Earl of Cranbrook, ed.) pp. vii±viii. Pergamon Press. van Balen, B. (1992) Distribution, status and conservation of the forest partridges in the Greater Sundas (Indonesia) with special reference to the chestnut-bellied partridge (Arborophila javanica). Gibier Faune Sauvage 9, 561±9. van Balen, S., Prawiradilaga, D.M. and Indrawan, M. (1995) The distribution and status of green peafowl Paco muticus in Java. Biol. Conserv. 71, 289±97. Walker, D. (1982) Speculations on the origin and evolution of Sahul-Sunda rain forests. In Biological Diversi cation in the Tropics (G. Prance, ed.) pp. 554±75. New York, USA: Columbia University Press. Wells, D.R. (1985) The forest avifauna of western Malesia and its conservation. In Conservation of Tropical Forest Birds (A.W. Diamond and T. E. Lovejoy, eds) pp. 213±32. Cambridge, UK: International Council for Bird Preservation. Wells, D.R. (1988) Birds. In Key Environments: Malaysia (Earl of Cranbrook, ed.) pp. 167±95. Pergamon Press. Whitmore, T.C. (1984a) Tropical Rain Forests of the Far East. 2nd Edition. Oxford University Press. Whitmore, T.C. (1984b) A vegetation map of Malesia at scale 1:5 million. J. Biogeog. 11, 461±71.

Partridge and pheasant status in South East Asia 1335 Appendix. Number of localities from which each species has been recorded in South East Asia both prior to 1970 and from 1970 Borneo Sumatra Java Lesser Sundas Malay Peninsula All Sunda subregion Period Habitat <1970 ³ 1970 all <1970 ³ 1970 all <1970 ³ 1970 all <1970 ³ 1970 all <1970 ³ 1970 all <1970 ³ 1970 all Number of localities sampled 92 41 114 70 33 93 63 57 100 18 8 26 84 72 133 327 211 466 Species Arborophila charltonii (n) Arborophila hyperythra (e) Arborophila javanica (e) Arborophila orientalis (e) Arborophila rubrirostris (e) Argusianus argus (n) Caloperdix oculea (n) Coturnix chinensis Coturnix ypsilophorus L 11 2 12 3 0 3 9 4 13 23 6 28 M 7 4 9 7 4 9 M 7 11 13 7 11 13 M 10 2 11 2 2 4 7 5 10 19 9 25 M 11 3 13 11 3 13 L 35 24 58 8 16 23 15 40 51 58 80 132 L/M 4 1 5 10 3 13 9 8 14 23 12 32 N 15 6 20 7 0 7 10 5 14 4 4 7 8 2 10 44 17 58 N 7 0 7 7 0 7 Gallus gallus N 7 6 13 17 12 26 3 2 5 16 23 35 43 43 70 Gallus varius (e) N 23 15 32 10 4 14 33 19 46 Haematortyx sanguiniceps (e) Lophura bulweri (e) M 12 5 15 12 5 15 L 24 6 29 24 6 29

1336 McGowan and Gillman Appendix. (Continued ) Borneo Sumatra Java Lesser Sundas Malay Peninsula All Sunda subregion Period Habitat <1970 ³ 1970 all <1970 ³ 1970 all <1970 ³ 1970 all <1970 ³ 1970 all <1970 ³ 1970 all <1970 ³ 1970 all Number of localities sampled 92 41 114 70 33 93 63 57 100 18 8 26 84 72 133 327 211 466 Species Lophura erythrophthalma (e) Lophura ignita (n) Lophura inornata (e) Melanoperdix nigra (e) L 11 7 18 4 1 5 6 13 17 21 21 40 L 26 13 38 6 4 10 8 8 15 40 25 63 M 16 3 16 16 3 16 L 8 4 12 2 1 3 7 3 10 17 8 25 Pavo muticus N 40 31 55 10 0 10 50 31 65 M 23 8 29 23 8 29 Polyplectron chalcurum (e) Polyplectron inopinatum (e) Polyplectron malacense (n) Polyplectron schleiermacheri (e) Rheinardia ocellata Rhizothera longirostris (n) M 11 6 14 11 6 14 L 13 6 19 13 6 19 L 7 1 8 7 1 8 L 1 2 2 1 2 2 L 9 2 11 5 1 6 8 12 18 22 15 35

Partridge and pheasant status in South East Asia 1337 Appendix. (Continued ) Borneo Sumatra Java Lesser Sundas Malay Peninsula All Sunda subregion Period Habitat <1970 ³ 1970 all <1970 ³ 1970 all <1970 ³ 1970 all <1970 ³ 1970 all <1970 ³ 1970 all <1970 ³ 1970 all Number of localities sampled 92 41 114 70 33 93 63 57 100 18 8 26 84 72 133 327 211 466 Species Rollulus rouloul (n) L 22 13 34 12 3 15 16 23 38 50 39 87 `All' gives the total number of localities for each species and allows for some localities having been sampled both before and after 1970. Habitat classes follow Wells (1985) and are denoted by L (lowland forest specialist), M (montane forest specialist) and N (not a forest specialist). Caloperdix oculea is considered a lowland forest specialist on Sumatra and Borneo and a montane forest specialist on the Malay Peninsula. After the species name (n) indicates a near subregional endemic (see text) and (e) a subregional endemic. The distribution of four species in the subregion is based on fewer than ten records and two of these species are subregional endemics that are restricted to Borneo: Arborophila hyperythra and Polyplectron schleiermacheri. Another four species that are endemic to montane forest on one landmass only, have been recorded from 15 or fewer sites: Arborophila javanica, Arborophila rubrirostris, Haematortyx sanguiniceps, Polyplectron inopinatum. On all landmasses apart from Java, there are fewer than 20 localities recorded for at least 70% of species. The lack of information is most acute on Sumatra, where 12 out of 14 species (85%) are known from fewer than 20 localities.