By P.M. Jørgensen, S. Sheth, T. Consiglio & I. Jiménez

The most urgent concerns in conservation biology is species becoming extinct. The World Conservation Union (IUCN) has only assessed the probability of becoming extinct for 4% of all described plant species (IUCN 2006). Consequently, a standard procedure to streamline the use of herbarium specimen data to infer threat would be a valuable tool for plant conservation evaluation. A pilot project is underway at Missouri Botanical Garden’s Center for Conservation and Sustainable Development using a dataset of Neotropical Bignoniae and its results are promising (Sheth et al. in prep.). Passiflora will be included in the next group of organisms to be tested and evaluated and we outline here the methods we intend to use.

Herbarium specimens represent primary occurrence data that is used to approximate the geographical distribution of a species which in turn is used to estimate its threat status (Schatz et al. 2000; Valencia et al. 2000; Randrianasolo et al. 2002; Schatz 2002; Golding & Hurter 2003; Willis et al. 2003). The use of herbarium data in this context is associated with two potential sampling biases that may prevent the documented points of occurrence for a given species from being representative of its actual geographic distribution. First, plant collections may be spatially aggregated and major parts of a species geographic range may be inadequately sampled. If so, estimates of extinction risk may be an artifact of the degree to which the geographic range of a species is sampled. Second, it is possible that some species are rarely collected because they are difficult to detect or are hard to collect such that their geographic range sizes are underestimated. In this case, extinction risk may be an artifact of species detectability.

We will first use data from herbarium specimens to assess the extinction risk of all species currently recognized in Passiflora subgenus Decaloba. Subsequent to the identification of threatened species, we will test the hypothesis that extinction risk as assessed from herbarium data is an artifact of the detectability of species. In particular, after controlling for phylogeny, we will examine the effects of species detectability on estimates of geographic range size, and consequently, level of threat. If extinction risk is an artifact of species detectability, we predict a positive relationship between detectability and geographic range size and a negative relationship between detectability and level of threat.

The IUCN Red List Categories have several criteria that include measurements of population sizes (table 1 and 2). Most practical assessments have however, used criterion B, which focuses on species with restricted distributions that are also severely fragmented or represented by few locations, experiencing continuing decline, and/or undergoing extreme fluctuations (IUCN 2006). This criterion uses area of occupancy (AOO), extent of occurrence (EOO), number of locations, and continuing decline; all values that can be obtained or approximated from collection data, whereas population size assessments are not generally available.

We will use ArcView 3.2 and ArcMap 9.1 Geographic Information System (GIS) software ESRI 1999 & 2006) to create species distribution maps from occurrence points and calculate: (a) geographic range size, both as EOO, defined as the minimal convex polygon which encompasses all point localities where a taxon occurs (Willis et al. 2003), or AOO, calculated as the area of the number of 3 × 3 km grid cells where the species occurs; (b) number of locations, denoted as a “geographically or ecologically distinct area in which a single threatening event can rapidly affect all individuals of the taxon present;” and (c) continuing decline, which is a “recent, current or projected future decline which is liable to continue unless remedial measures are taken” (IUCN 2005). Scripts that automatically calculate both EOO and AOO from occurrence points have been developed for use in ArcView (Fay 2002; Consiglio 2006). By superimposing the occurrence points of the species, a map of protected areas (WDPAC 2006), and a human footprint map (values 10 or less, Sanderson et al. 2002); we determine the proportion of distribution that lies outside protected or pristine areas. The use of wilderness areas to estimate the chances of future decline is based on two main assumptions. The first is that, over the next few decades, few suitable habitats will persist outside these areas due to habitat destruction. Second, protected areas will remain relatively intact over time (Schatz et al. 2000). In order for a species to be assigned to a particular IUCN category of threat, it has to meet thresholds in all three criteria evaluated. For example, a species categorized as Critically Endangered may have an EOO < 100 km2, exist in only one location, and exhibit a continuing decline in AOO (see for a complete listing of criteria IUCN 2001)

We will use the same four indices of detectability as developed by Sheth et al (in prep). The first index is defined as the number of collection dates at which a species was collected in a grid cell (10×10 km) relative to the sum of collection dates of all species across grid cells where the target species occurs. The second index is a permutation of the first index; it considers the relative value of collector days (i.e. unique combination of collector and date). The first two indices are developed following MacKenzie et al. (2005) whereas the third and fourth indices are based on Cam et al. (2002). The third and fourth indices are measured as the probability that a plant collector gather the target species from a given spatial unit during a time unit provided that the species is present in the area. The detectability of a species can therefore be measured as the change in the number of collections of the target species associated with the change in collection dates across grid cells. For each species, we will count the number of collections per grid cell at certain time intervals and subsequently compute the slope resulting from a linear regression through the origin of the number of collection dates (independent variable) and the number of collections (dependent variable) across grid cells. The fourth index is expressed as the slope of a linear regression through the origin of the number of collector days and the total number of collections.

Next we wish to test if threat is independent of the phylogeny or if certain clades are more prone to extinction than others (Felsenstein 1985). We do this for two reasons, first we find it informative if extinction risk is inherited, and second we need to know if we can use the detectability indices in standard regression analysis. We plan to test phylogenetic independence of EOO, AOO, extinction risk (can be coded in various ways), and the four indices of detectability using the methods outlined by Abouheif (1999). We can use the program Phylogenetic Independence (Reeve & Abouheif 2003) to generate a null distribution for comparison to the observed values and test for phylogenetic correlation using phylogenetically independent contrasts (PIC) (Felsenstein 1985, Midford et al. 2005, Maddison & Maddison 2006). Incorporating branch length and topology from our phylogenetic tree will give us weighted differences between sister species that can be used in traditional statistical analysis (Felsenstein 1985).

Herbarium data has in the past been used to assess threat status of many species (e.g. Valencia et al. 2000), but these evaluations were performed without automated calculations of extent of occurrence, area of occupancy and without an evaluation of potential bias in data due to inter-specific variation in detectability. We have above outlined methods that will help us to achieve a better and less subjective evaluation of the species, and an evaluation that takes phylogeny into account. Preliminary assessments of all accepted species in subgenus Decaloba can be found in the linked tables.