Where are all the hummingbirds? Data fusion can tell us

Submitted by editor on 23 February 2021. Get the paper!
Figure 1. Two chestnut-breasted coronet (Boissonneaua matthewsii) having a discussion about the recent Ecography paper by Zurell et al. 2020 on ‘A standard protocol for reporting species distribution models’. Photo courtersy of Adriana Soto.

 

By Diego Ellis-Soto (diego [dot] ellissoto [at] yale [dot] edu), on behalf of all co-authors

 

Biodiversity worldwide, particularly montane biodiversity, is increasingly impacted by human activities and climate change (Elsen et al. 2020). To better manage and conserve biodiversity, we need comprehensive estimates of species distributions. In this study in particular, we introduce a new data fusion approach which allows combining multiple sources of biodiversity information to improve our predictions on hummingbird species across the Americas.

 

Hummingbirds have accompanied me since I was a young; from the glittering-bellied emerald (Chlorostilbon lucidus) feeding on red lilies at my grandparents’ house in Uruguay, to a pair of rofous-tailed hummingbirds (Amazilia tzacatl) building their nest in my family’s garden in San Jose Costa Rica, up to majestic chestnut-breasted coronet (Boissonneaua matthewsii) swirling around in the Ecuadorian Cloud forest of Mindo (Fig. 1), to the migratory ruby-throated hummingbird (Archilochus colubris) greeting me on my way to my office at Yale University, in Connecticut, USA. Each of these species have in common that biodiversity information in form of point locations, species range maps, and species-specific elevation limits are available. With a group of international collaborators, we made use of these data combinations through a Poisson point process modeling approach (Warton and Shepherd 2010), to estimate the continental scale distribution of hummingbirds across the Americas at 1 km2 resolution. To assess whether our philosophy of borrowing strengths across data types improves predictions on modeling biodiversity, we compared our Hummingbird distribution models with inventory datasets across the Andean region of South America (publicly downloadable through our publication). Results reveal that our newly proposed method outperforms traditional methods more than 92% of the times, with greatest gains in model performance for data deficient species (Ellis-Soto et al. 2021).

 

Feeling encouraged by these results, we stacked predictions of single hummingbirds, to create detailed species richness maps of hummingbirds across the Americas and identify the hotspot for these birds to be located in southern Colombia (Fig. 2). We are currently implementing this workflow at the Yale Center for Biodiversity and Global Change (bgc.yale.edu) with the Map of Life (mol.org) to predict the distributions of thousands of vertebrate species at fine spatial-resolutions to more accurately model species distributions in a rapidly changing world to improve our understanding of biodiversity for conservation planners.

Figure 2. Hummingbird species richness based on the stacked predictions of 276 hummingbird species distribution models (Fig. 5a of Ellis-Soto et al. 2021).

Read more at the Half Earth Project.

References

 

Ellis‐Soto, D., Merow, C., Amatulli, G., Parra, J. L. and Jetz, W. 2021. Continental‐scale 1 km hummingbird diversity derived from fusing point records with lateral and elevational expert information. Ecography, in press, <https://doi.org/10.1111/ecog.05119>.

 

Elsen, P.R. et al. 2020. Topography and human pressure in mountain ranges alter expected species responses to climate change. – Nat. Comm. 11: 1974.

 

Warton D.I. and Shepherd, L. C. 2010. Poisson point process models solve the ‘pseudo‐absence problem' for presence‐only data in ecology. – Ann. Appl. Stat.4: 1383–1402.

 

Zurell, D. et al. 2020. A standard protocol for reporting species distribution models. – Ecography 43: 1261-1277.

 

Categories: 
General

Comments