May 2013 - Predicting birdsong from space

A new study published online in Evolutionary Applications reports on a remarkable tool to use remote-sensing data to predict animal behaviour (birdsong), across vast spatial scales.


The collaborative effort was spearheaded by Prof. Tom Smith from the Center for Tropical Research and the Institute of the Environment and Sustainability at the University of California Los Angeles. The team exploited space expertise from the NASA Jet Propulsion Laboratory and included Dr. Hans Slabbekoorn from the IBL.

The researchers used remotely sensed data to predict the song characteristics of the little greenbul (Andropadus virens), a widely distributed African passerine, found across secondary and mature rainforest habitats and the rainforest-savanna ecotone. Dr. Hans Slabbekoorn started to record songs of this species back in 1999 on invitation by Prof. Tom Smith. The recordings served in a comparative study of acoustics, morphology, and genetics. The data of this earlier study revealed that ecology was more important than geographic distance in explaining song variation across 12 sites in Cameroon. This was regarded as an important finding and in line with a role for song in ecological speciation given that birds at the recorded sites were all of the same species and that they also showed habitat-dependent divergence in morphology despite evidence of on-going gene flow (Slabbekoorn & Smith, Evolution 2002).

Prof. Tom Smith and Dr. Hans Slabbekoorn in Cameroon in search for little greenbuls, back in 1999

Dr. Alex Kirschel (now at the University of Cyprus, Nicosia) followed up on collecting recordings of this same species in Cameroon and Uganda. He found out by playing back recordings that rainforest greenbuls in Cameroon and Uganda respond more strongly to songs from rainforest birds than to songs from ecotone birds, even if the rainforest recordings are made far away in the other country and the ecotone forest songs are from nearby birds within the same country. This pattern was not confirmed for the response of ecotone birds, but suggests that especially ecological song divergence and associated response divergence can affect gene flow among populations and supports the hypothesis that ecological gradients are more important to avian speciation and explaining rainforest diversity than geographical isolation in glacial refuges (Kirschel et al., Evolution 2011).

The current study concerns a follow-up combining all available song data, which were all re-analysed for this purpose by Dr. Selvino de Kort (who defended his PhD-thesis at the IBL in 2002 and who is now at the University of Manchester). Habitat characteristics were inferred using three spaceborne sensors and captured various aspects such as canopy greenness, structure and moisture, and tree cover. Predictions of the song characteristics of little greenbul songs were made across a large unsampled region, including Southern Cameroon and parts of the neighboring countries of Equatorial Guinea, Gabon, Republic of Congo, and the Central African Republic. The ability to model and predict a complex trait such as vocal behaviour, from remotely sensed data related to key structural variation in habitat, suggests the utility of satellite-based layers for identifying variation in other complex behavioral and phenotypic characteristics of any other species (Smith et al., Evolutionary Applications 2013).


Predicting bird song from space - paper
United Academics newsitem
Website of Tom Smith
Website of Alex Kirschel
Website of Selvino de Kort
Website of Hans Slabbekoorn
NASA Jet Propulsion Laboratory

Webcommunication Science - Published: 20 May 2013