Leveraging big citizen science data to understand urban tolerance of bird species

Submitted by editor on 16 January 2020. Get the paper!

Fig. 1. Three example species, showing their theoretical grouping based on response to urbanization – and their values of VIIRS night-time lights at a continental scale.

By Corey Callaghan

Urbanization is negatively impacting biodiversity. But urbanization processes differentially affect species: there are species-specific responses to urbanization. Traditionally, urban ecologists have measured a species’ response to urbanization categorically – i.e. urban avoiders, adapters, or exploiters. And sometimes even as being ‘urban’ or ‘non-urban’. Yet all species fall along a continuum of urban tolerance. Unfortunately, collecting the necessary amounts of data to know precisely where a species falls along this continuum has been traditionally costly and time-consuming.


Currently, citizen science data are rapidly shaping the future of ecology and conservation (see here, here, or here for some relevant commentaries). Massive datasets like eBird and iNaturalist are collecting millions of observations per year. In May of 2019, eBird collected 7.5 observations per second – throughout the entire month! But how can these data be used to better understand biodiversity responses to urbanization?


We previously used a novel approach to assign an ‘urban tolerance score’ to species based on continental-scale data (see here for an interactive phylogenetic tree for Australian birds). This urban tolerance score is a measure of a species’ urbanness. It is calculated by (1) taking all citizen science observations for a species from eBird, (2) assigning a value of night-time lights (i.e. radiance observed from space) to every observation, and (3) taking the median of all these observations. See Fig. 1 for an example.

After this process, we are left with every species falling along a continuum of their urban tolerance. We also looked at how these species-specific scores can be transformed into community-level scores. Importantly, though, the question remained: how well do these continental-scale measures of urbanness – based on big citizen science data – correlate with local-scale responses to urbanization?

In this paper, we tested this question. Using a year-long intensive sampling of birds along an urbanization gradient in the Blue Mountains World Heritage region, we modelled the local-scale response of urbanization for 49 species. We then investigated the relationship between the continental-scale urban scores and these local-scale modelled responses. The measure of urbanization was at both a different spatial extent and spatial grain between the two methods. For the continental data, we used VIIRS night-time lights within a 5 km buffer around every observations whereas for the local-scale data we estimated the percent impervious surface using Google Earth imager within a 250-meter buffer surrounding each sampling location. We found evidence for a strong relationship between these two seemingly disparate measurements of urban tolerance scores (Fig. 2). We also resampled these and found that ~250 citizen science observations were necessary to produce reliable measures of continental-scale urbanness.

Figure 2. a) the relationship between continental and local-scale urbanness was strongly related (see here for an interactive version), and b) about 250 observations were needed before the R2 values levelled off providing reliable relationship between continental and local-scale urbanness.

Ultimately, this method is very novel, but with the continued increase of citizen science data (in our case eBird data), combined with the increased accessibility of remotely-sensed layers, we believe this method will prove valuable in measuring the urbanness of birds and other taxa in the future, providing a mechanism to track how species, populations, and taxa are responding to increasing urbanization.


Read the paper here.


Corey T. Callaghan on behalf of all co-authors.