Indonesia's threatened birds: over 500 million years of evolutionary heritage at risk

Using published IUCN red-lists for the threatened bird species of Indonesia (Stattersfield & Capper, 2000), we document that the species at risk in Indonesia are not a taxonomically random sample, consistent with patterns seen at the global scale (Purvis et al., 2000). This implies that a greater than random proportion of Indonesia's and the world's (through the loss of endemics) phylogenetic diversity (PD) is at risk (Von Euler, 2001). Using cytochrome b (Johns & Avise, 1998) and a distance-based taxonomy, we attempt to quantify this projected loss of PD in millions of years and conclude with a brief discussion of the conservation implications for Indonesia, and of the use of 'evolutionary heritage' as a measure for conservation at the geopolitical level. INTRODUCTION Using published IUCN red-lists for the threatened bird species of Indonesia (Stattersfield & Capper, 2000), we When a species becomes extinct, the evolutionary history investigate whether a significantly higher proportion of it embodied is lost with it (Nee & May, 1997). The more Indonesia's, and the world's (through the loss of distantly related a species is to its nearest relatives, the endemics) evolutionary history is at risk than if extinction more history is lost. If one important metric of risk were random with respect to taxonomic position (Von conservation value is evolutionary distinctness (Wilson, Euler, 2001). The amount of PD a particular country (or 1992), then measuring the loss of evolutionary history other geopolitical unit) stewards might be considered that (also called phylogenetic diversity or PD) through country's 'evolutionary heritage'. anthropogenic-mediated extinction is an important componentof conservation biology (Faith, 1992; Crozier, METHODS 1997; see also Morowitz, 1991). Theoretical work (Vane-Wright Humphries & Data collection Williams, 1991; Nee & May, 1997; see also Vazquez & Gittleman, 1998; Heard & Mooers, 2000) has quantified The Indonesian species list is from Andrew (1992), the expected pattern of evolutionary loss, and several augmented with data from Stattersfield &Capper (2000), groups have projected the worldwide amount of loss we Coates, Bishop & Gardner (1997), MacKinnon & might expect under reasonable scenarios of extinction for Phillipps (1993) and Beehler, Pratt & Zimmerman several taxonomic groups (McKinney, 1998; Russell et (1986). Taxonomy is based on Sibley & Monroe (1990). a!., 1998; Purvis et a!., 2000; Von Euler, 2001, Sechrest Conservation status data are from Stattersfield & et a!., 2002). However, conservation decisions are Capper (2000). Where taxonomy data differed between normally made at much smaller scales, and here we authors, Stattersfield & Capper (2000) was preferred present the first analysis of projected loss of evolutionary (e.g. species versus subspecies rank: see 'speciesnotes' at history at the scale of a geopolitical unit, using the birds www.sfu.ca/-amooers/evolheritage for taxonomic of Indonesia as a test case. Two-thirds of the world's bird decisions). Breeding and non-breeding species are species are threatened by habitat loss and degradation included in the analysis. We considered Indonesia to (Beissinger, 2000)and deforestation is the major threat in include East Timor and Irian Jaya (Papua) and its Indonesia; at current rates, most of Indonesia's speciessatellites, but not Sarawak, Sabah or Brunei. Exclusion rich lowland forests will have disappeared completely by of these areas did not affect the species list since there are the year 2010 (Jepson et a!., 2001). nospecies endemic to Sarawak, Sabah or Brunei. We classified each species two ways, as follows. (1) All correspondence to: Dr Arne 0. Mooers. Tel: +1 604-291-3979; endemic status: found only in Indonesia (endemic) versus Fax:1 604-291-3496; E-mail: [email protected] found in at least one other country (nonendemic); 184 A. 0. MOOERS AND R. A. ATKINS (2) threat status: not threatened (S), near threatened (nt), Evolution history at risk vulnerable (VU), endangered (EN), critically endangered (CR). Species lacking information on threat status (dataGiven that no full tree of the birds yet exists, we drew on deficient: DD) were excluded from further analyses. Sibley & Monroe's (1990) genetic distance-based According to Purvis et al. (2000), most data-deficient taxonomy, a compilation of cyt. b data from Johns & species are likely to have a high extinction risk, making Avise (1998; G. Johns, pers. comm.), and the standard 2% our approach conservative. We then constructed three data per million years 01 divergence for avian cyt. b (Moore sets by classifying our species following Purvis et al. & DeFilippis, 1997; Johns & Avise, 1998 and references (2000), producing sets of increasing conservation concern. therein) to make rough but very conservative calculations The most inclusive data set (NTplus) considered as of the amount of Indonesia's evolutionary history at risk. threatened' those species listed as nt, VU, EN and CR Preliminary analyses confirmed that estimates using the (Table 1). The other two data sets were designated VUplus entire data set were statistically indistinguishable from (and so lumping nt species with the not threatened those made using only genera with representatives in category) and ENplus (considering those species listed as Indonesia (results not shown). We therefore used the nt and VU as not threatened). entire data set and calculated the evolutionary history as follows: Analysis Species. We took the minimum genetic distance between sister species for 51 genera from 33 families and 11 orders Patterns ofrisk and calculated a taxonomically corrected (Harvey & Each of our three data sets corresponds to an extinction Mace, 1982) average as a conservative estimate of a regime. We compared the observed taxonomic patterns of species' age (T5). threat with patterns created by randomly sampling species repeatedly 2000 times across the taxonomy, using a Whole genera. Using data from 88 genera, 42 families and simple macro written in Microsoft Excel (available from 12 orders, we calculated genus ages using a Model I the authors). Each extinction regime specified a different regression of maximum intrageneric distance on (Log) proportion of threatened species to be sampled (Table 1) number of species sampled for cyt. b. The maximum We evaluated patterns of loss at the level of country distance is most likely to incorporate the first split in the (Indonesia) by comparing observed numbers of all higher genus (see Fig. 1). We then assumed that genus trees could taxa at risk with simulated values created by allocating the be approximated by a pure birth process. The appendix by same proportion of species at risk randomly to the species Sean Nee explicates how this assumption leads to the list. We evaluated patterns of loss to the world by expression for their PD: repeating this but recording only the number of endemic T(N-1 taxa affected. Actual patterns from Indonesia were PD = considered remarkable if they were larger than the most N extreme 5% of the simulations. 1 -2 with T being the age of a genus with N species (see Fig. I a). Random extinction in an expanding genus will lead to fewer deep branches than this model, meaning this Table 1. Projected loss of avian evolutionary history to Indonesia and expression will overestimate total PD. However, if there worldwide has been an adaptive radiation in the genus (much Taxa at risk' Projected loss speciation early in its history), branches will be longer, (MY)5 making this expression conservative. It is also Region, Riskd Species Genera Families Orders ERM Comb conservative if the genus has been at a stable size for a long time (see Nee & May, 1997 for a discussion of this Indonesia EN+ 44 5'° 0 0 78 85 last model). VU+ 114 23** 1 0 228 263 As an alternative, more ignorant approach, we also (I) NT+ 306 56*5* 5*** 1** 674 751 World EN+ 36 3** 0 0 62 66 took a taxonomically weighted average genus age T' VU+ 75 8*5* 0 0 134 146 and (2) assumed the genus form presented in Fig. 1(b), NT+ 158 1l* 0 0 270 288 leading to PD=2*T'4+(N_2)*T5. We report values from Asterisks denote significant deviations from taxonomically random patterns of risk: both methods. "P <0.1; '°.P <0.01; ",P <0.0005.) 'Millions of years (MY) of evolutionary history at risk. ERM assumes a log-linear Entire Families. Using data from 36 families in 12 orders, relationship between genus size and age and random (Equal Rates Markov) diversification; Comb assumes every genus is 4.54 MY old and has the shape in we calculated a taxonomically corrected average Fig. 1(b). See text for further details. Species and higher taSa at risk of being lost to Indonesia, or, being wholly endemic maximum intergeneric cyt. b distance and used this to to Indonesia, at risk of being lost worldwide, calculate the amount of added history lost if an entire 5 Nested sets of risk: 'NT' includes all neat threatened, vulnerable, endangered, or family goes extinct. Because the families involved were critically endangered; 'vu' excludes nt species; 'EN-v' includes only endangered and critically endangered species. Data Deficient species are excluded See text for very small (<3 spp.) this could be done without recourse more detail, to a pure birth model; this model is likely to be invalid at Indonesia's evolutionary heritage at risk 185 deeper levels in the avian tree (Von Euler, 2001). We A. Domestic genera simply added a single branch of the requisite length to the 60P1) estimated in the genera within that family. 50 Higher taxa. We treated PD for higher taxa on a case-by. 40 case basis (see results).