Abstract
We present a framework to assess the ex-situ conservation status of plant species, by estimating the fraction of extant genetic diversity that has been conserved as seeds or other living material outside their natural habitat. We use a species distribution model to predict the area that is suitable for a species, and use inclusion and exclusion buffers to estimate its range. The range is divided into a number of zones that is proportional to the square root of its size. The geographic and environmental distance between zones is combined into a single distance metric that is used as to set weights on the links of a network connecting neighboring zones. An adjusted network is created by setting the distance between zones with seed samples to zero and halving the distance to neighboring zones for which there is no seed sample. The ex-situ conservation score (XC, between 0 and 1) is then computed as the reduction in the sum of the weights of the shortest paths between the nodes in the adjusted network relative to the sum the weights of the shortest paths in the original network. We adjust XC for small seed sample sizes and for the number of seed samples with unknown geographic origin. We illustrate our framework for 61 wild Vigna species in Africa. Twenty-three species were not conserved ex-situ (XC = 0), and 42 species had very low conservation scores (XC < 0.2). Range sizes were very different from the suitable area sizes predicted with the species distribution model (r = 0.09). The geographic and environmental distances were weakly correlated (r = 0.29), illustrating the importance of considering both. The Pearson correlation coefficient between XC and seed sample size for each species was 0.94, suggesting that seed sample size can be useful for quickly evaluating ex-situ conservation status.