Intense trapping of the critically endangered Yellow-crested Cockatoo Cacatua sulphurea  for the International pet trade has devastated its populations across Indonesia such that populations >100 individuals remain at only a handful of sites. We combined distance sampling with density surface modeling (DSM) to predict local densities and estimate total population size for one of these areas, Komodo Island, part of Komodo National Park (KNP) in Indonesia. We modeled local density based on topography (topographic wetness index) and habitat types (percentage of palm savanna and deciduous monsoon forest). Our population estimate of 1,113 (95% CI: 587–2,109) individuals on Komodo Island was considerably larger than previous conservative estimates. Our density surface maps showed cockatoos to be absent over much of the island, but present at high densities in wooded valleys. Coincidence between our DSM and a set of independent cockatoo observations was high (93%).
Standardized annual counts by KNP staff in selected areas of the island showed increases in cockatoo records from <400 in 2011 to ~650 in 2017. Taken together, our results indicate that KNP, alongside and indeed because of preserving its iconic Komodo Dragons Varanus komodoensis, is succeeding in protecting a significant population of Indonesia’s rarest cockatoo species. To our knowledge this is the first time DSM has been applied to a Critically Endangered species. Our findings highlight the potential of DSM for locating abundance hotspots, identifying habitat associations, and estimating global population size in a range of threatened taxa, especially if independent datasets can be used to validate model predictions.

The dataset was collected by point count distance sampling and analysed with density surface modeling in R.  

To avoid informing illegal trapping of the critically endangered target species, all location information has been deleted from the version published here. This was requested by the journal but makes it impossible to re-produce our results with the dataset provided.