Accounting for the indirect area effect in stacked species distribution models to map species richness in a montane biodiversity hotspot. Issue 11 (10th September 2015)

• Record Type: Journal Article
• Authors: Pouteau, Robin; Bayle, Élise; Blanchard, Élodie; Birnbaum, Philippe; Cassan, Jean‐Jérôme; Hequet, Vanessa; Ibanez, Thomas; Vandrot, Hervé; Robertson, Mark
• Is Part Of: Diversity & distributions. Volume 21:Issue 11(2015:Nov.)
• Page Start: 1329
• Page End: 1338
• Publication Date: 2015-09-10
• Subjects: Biodiversity -- Periodicals; Biodiversity conservation -- Periodicals; 577
• DOI: 10.1111/ddi.12374
• Languages: English
• ISSNs: 1366-9516

Abstract

<abstract abstract-type="main" id="ddi12374-abs-0001"> <title>Abstract</title> <sec id="ddi12374-sec-0001" sec-type="section"> <title>Aim</title> <p>Understanding how species richness is distributed is a critical prerequisite to implement efficient conservation strategies in biodiversity hotspots. Stacked species distribution models (S‐SDMs) provide new opportunities to map species richness, but the accuracy of this method may decline with elevation. Here, we test whether variation in model accuracy arises from increasingly unpredictable environments or the decreasing availability of area with increasing elevation, which might affect the regional pool of species and, as a result, the local species richness (the indirect area effect).</p> </sec> <sec id="ddi12374-sec-0002" sec-type="section"> <title>Location</title> <p>The New Caledonian biodiversity hotspot (south‐west Pacific Ocean).</p> </sec> <sec id="ddi12374-sec-0003" sec-type="section"> <title>Methods</title> <p>An individual MAXENT model was built for 562 tree species by combining eight 100‐m resolution environmental variables with <italic>c</italic>. 10, 000 occurrence records. For each species, a map was produced at a one‐hectare scale indicating the estimated habitat suitability. All models were then summed, and the resulting estimates were compared with richness measured in 11 independent one‐hectare inventories. To account for the indirect area effect, S‐SDM estimates were adjusted through the Arrhenius' equation linking the number of species hosted by a habitat with its surface area.</p> </sec> <sec id="ddi12374-sec-0004" sec-type="section"> <title>Results</title> <p>The S‐SDM predictions ranged from 95 to 251 species (mean = 153) while field inventories were lower, ranging from 39 to 131 species (mean = 90). Overall, the S‐SDM increasingly overestimated richness as elevation increased. Taking into account the indirect area effect de‐correlated residuals from elevation and induced a significant correlation between modelled and measured species richness.</p> </sec> <sec id="ddi12374-sec-0005" sec-type="section"> <title>Conclusion</title> <p>The decreasing accuracy of the S‐SDM with elevation was explained by the decreasing availability of habitat influencing regional diversity found in each elevational band. Despite remaining difficulties to predict species richness when addressing the indirect area effect, our findings represent a significant step forward towards improved S‐SDM designing.</p> </sec> </abstract>