Predicting the potential distribution of an endangered cryptic subterranean mammal from few occurrence records
Introduction
Knowledge of geographical distributions and habitat preferences are central to the conservation and population management of threatened species (Guisan et al., 2006, Thomaes et al., 2008). Ecological niche models have been widely applied in conservation biology to predict the potential distributions of species using correlative models that exploit a species–environment relationship (Anderson et al., 2009, Chefaoui et al., 2005, De Siqueira et al., 2009, Peterson, 2001, Raxworthy et al., 2007, Rushton et al., 2004, Titeux et al., 2007). Ecological niche models require occurrence records and a set of environmental predictor variables in order to predict a species’ potential distribution (Gaubert et al., 2006, Peterson, 2001). Many species of conservation concern are cryptic or range-restricted and often very few occurrence records are available for these species (Pearson et al., 2007). Recently success has been achieved in calibrating ecological niche models with fewer than ten occurrence records (De Siqueira et al., 2009, Pearson et al., 2007). In view of this, ecological niche models have the potential to make a contribution to conservation in two important ways, especially for species with few occurrence records. First, they can be used to direct field surveys that are designed to search for new populations of rare and poorly known species by identifying potentially suitable areas (Bourg et al., 2005, De Siqueira et al., 2009, Guisan et al., 2006). Second, they can be used in conservation area selection for systematic biodiversity plans (Cowling et al., 2003, Loiselle et al., 2003, Wilson et al., 2005). This planning process identifies parcels of land that have high biodiversity value using the principles of systematic biodiversity planning (Driver et al., 2003, Margules and Pressey, 2000). These biodiversity plans require a set of targets to be defined, which are usually based on several elements of biodiversity (Driver et al., 2003) but can include species distributions and particularly threatened or range-restricted species (Loiselle et al., 2003, Wilson et al., 2005). In this way species of conservation concern will gain a level of protection from threatening processes by being part of a network of conservation areas, whilst the biodiversity of the region will simultaneously be conserved.
The endangered Juliana's golden mole (Neamblysomus julianae; Chrysochloridae) is a small (35 g), habitat-specific mammal that is particularly difficult to study due to its subterranean lifestyle (Jackson et al., 2008, Jackson et al., 2009). Despite occurring less than 15 km from the centre of South Africa's capital city, Pretoria, this species was only described in 1972 (Bronner & Bennett, 2005). Although recent studies have investigated certain aspects of its biology (Jackson et al., 2007, Jackson et al., 2008, Jackson et al., 2009), it is still relatively poorly known. The distribution range of the species is highly restricted. It is known to occur on the Bronberg Ridge (BR) in eastern Pretoria, 120 km north in the Nylsvley Nature Reserve (NNR), and 350 km east in the south western Kruger National Park (KNP) (Bronner & Bennett, 2005). Based on dental characteristics (Bronner, 1995) and recent genetic evidence (Maree, pers. comm.), animals from the KNP population differ significantly from the other two populations and evidence suggests that the KNP population could be treated as a separate species. Juliana's golden mole thus appears to be restricted to the BR and NNR areas.
The BR is less than 15 km in length and the suitable sandy soils are on average less than 1 km in width (Bosch, 2004). The NNR population is described from the reserve where approximately 800 ha constitutes suitable habitat (Jackson, 2007). Based on the current understanding of its distribution, the geographic areas occupied by these two populations are very small and not adequate to ensure the long-term persistence of the species. The small size and fossorial lifestyle means that this species has poor dispersal capabilities. Unlike mole-rats that are powerful diggers, excavating soil with their large teeth, golden moles use only their forelimbs and have far weaker tunnelling capabilities that restricts them to softer soil types. They are thus extremely susceptible to habitat fragmentation, as even relatively minor obstacles such as roads can constitute impenetrable barriers. These factors in combination with small geographical ranges have resulted in ten of the 21 golden mole species appearing on the IUCN red list, with an additional two species listed as data deficient. Of these species, five are listed amongst the ten most endangered mammal species in South Africa (IUCN, 2004).
Recently a thorough ecological assessment that specifically documented the species’ habitat requirements was undertaken (Jackson et al., 2008). Using this information we aimed firstly to make use of ecological niche modelling to produce a potential distribution map for Juliana's golden mole that could be used to search for new populations. Secondly, we aimed to produce a model that could be used for area selection in systematic biodiversity planning.
Section snippets
Study region
The study region was defined as four of South Africa's provinces: Gauteng; North West; Mpumalanga; and Limpopo (Fig. 1). This region is dominated by the savanna biome (Mucina & Rutherford, 2006). This area was chosen since the species is only known from savanna vegetation types (Jackson et al., 2008). Rainfall in the savanna biome is highly seasonal, resulting in wet summers and dry winters. The region is generally warm, with a sub-tropical to tropical thermal regime and is generally free of
Results
The first model, calibrated with only four occurrence records, indicated a small part of the map region as being suitable for the species (Fig. 2). Large areas of unsuitable habitat (<0.4) separate the areas of high suitability (>0.8). The jackknife analysis indicated that the models were successful at predicting known presences as suitable (proportion correctly predicted: 0.5; p = 0.00054). All four of the occurrence records were located in areas that were predicted to be highly suitable for the
The models
The models that we have presented here have been successful at identifying regions that are suitable for the species. The first model that was calibrated with only four occurrence records was useful as it allowed us to identify further sites that were occupied by the species. The jackknife analysis indicated that this model successfully predicted known presences as suitable 50% of the time. However, this proportion implies that the model was unsuccessful and omitted known presences 50% of the
Acknowledgements
Trine H. Setsaas and Simon Byron are thanked for assisting with field surveys. The WWF Green Trust, a partnership between WWF South Africa and Nedbank (South Africa), kindly funded part of this study.
References (41)
- et al.
Using distribution models to test alternative hypotheses about a species’ environmental limits and recovery prospects
Biological Conservation
(2009) - et al.
Modeling species’ geographic distributions for preliminary conservation assessments: An implementation with the spiny pocket mice (Heteromys) of Ecuador
Biological Conservation
(2004) - et al.
Potential distribution modelling, niche characterization and conservation status assessment using GIS tools: A case study of Iberian Copris species
Biological Conservation
(2005) - et al.
A conservation plan for a global biodiversity hotspot—the Cape Floristic Region, South Africa
Biological Conservation
(2003) - et al.
Something from nothing: Using landscape similarity and ecological niche modeling to find rare plant species
Journal for Nature Conservation
(2009) - et al.
Natural history collections and the conservation of poorly known taxa: Ecological niche modelling in central African rainforest genets (Genetta spp.)
Biological Conservation
(2006) - et al.
Predictive modelling and ground validation of the spatial distribution of the New Zealand long-tailed bat (Chalinolobus tuberculatus)
Biological Conservation
(2006) - et al.
Maximum entropy modeling of species geographic distributions
Ecological Modelling
(2006) - et al.
Applying species distribution modelling for the conservation of the threatened saproxylic Stag Beetle (Lucanus cervus)
Biological Conservation
(2008) - et al.
Fitness-related parameters improve presence-only distribution modelling for conservation practice: The case of the red-backed shrike
Biological Conservation
(2007)
Sensitivity of conservation planning to different approaches to using predicted species distribution data
Biological Conservation
Overview of the agricultural natural resources of South Africa. ARC-ISCW report no. GW/A/2004/13
Different climatic envelopes among invasive populations may lead to underestimations of current and future biological invasions
Diversity and Distributions
Supplementary discussion of the results of the geological investigations above Willow Quarries on Zwavelpoort 373 JR and impact on the Bronberg Mountain
Putting a cart before the search: Successful habitat prediction for a rare forest herb
Ecology
Order Afrosoricida
Planning for living landscapes: Perspectives and lessons from South Africa
Novel methods improve prediction of species’ distributions from occurrence data
Ecography
A review of methods for the assessment of prediction errors in conservation presence/absence models
Environmental Conservation
Cited by (55)
Spatial modelling of Acropora muricata and Porites lutea distribution using environmental descriptors across Lakshadweep–Chagos Archipelago
2021, Regional Studies in Marine ScienceCitation Excerpt :The present attempt is the first of its kind to predict hard coral distribution using modelling techniques in Lakshadweep Chagos Archipelago. MaxEnt method employed in the present study has successfully predicted a variety of species’ occurrence (Jackson and Robertson, 2010; Papes and Gaubert, 2007; Saatchi et al., 2008; Mingyang et al., 2008; Suárez-Seoane et al., 2008; Carroll, 2010). As the coral reefs are very sensitive as well as an important source of ecosystem services, the distribution data will have remarkable applications.
Prioritising search effort to locate previously unknown populations of endangered marine reptiles
2020, Global Ecology and ConservationNew insights from RADseq data on differentiation in the Hottentot golden mole species complex from South Africa
2020, Molecular Phylogenetics and EvolutionFuture climate and habitat distribution of Himalayan Musk Deer (Moschus chrysogaster)
2018, Ecological InformaticsCitation Excerpt :It is one of the most widely used software package for environmental niche modelling and can achieve high predictive accuracies even with low presence only data (Pacifici et al., 2015; Phillips and Dudík, 2008). MaxEnt has been used for either single or multiple mammalian species in Nepal (Aryal et al., 2016) as well as other geographic regions including Denmark (Flojgaard et al., 2009), Spain (Morueta-Holme et al., 2010), South Africa (Jackson and Robertson, 2011), South America (Marino et al., 2011), Northern Europe (Hof et al., 2012), Malaysia (Nazeri et al., 2012), Bangladesh (Alamgir et al., 2015), and China (Hu and Jiang, 2010; Su et al., 2015). A total of 80 (from a total of 192, after removing multiple presences within a 1 km2 raster grid cell) distribution points of HMD within the territory of Nepal were used for processing in MaxEnt software, sourced from different literature (Aryal et al., 2010; Aryal and Subedi, 2011; Chalise, 2012; Subedi et al., 2012).
Distribution of Wildlife and Illegal Human Activities in the Lampi Marine National Park (Myanmar)
2019, Environmental Conservation