Abstract
The koala, Phascolarctos cinereus, is an iconic Australian wildlife species, but faces rapid decline in South-East Queensland (SEQLD). For conservation planning, estimating koala populations is crucial. Systematic surveys are the most common approach to estimate koala populations, but such surveys are restricted to small geographic areas, they are costly and conducted infrequently. Public interest and participation in the collection of koala sightings is increasing in popularity, but such data is generally not used for population estimation. We used incidental sightings of koalas reported by members of the public from 1997-2013 in SEQLD to estimate the yearly spatio-temporal koala sightings density. For this, a spatio-temporal point process model was developed accounting for observed koala density, spatio-temporal detection bias and clustering. The density of koalas varied throughout the study period due to the heterogeneous nature of koala habitat in SEQLD, with density estimates ranging between 0.005 to 8.9 koalas per km2.
The percentage of land areas with very low sightings densities (0-0.25 koalas per km2) remained similar throughout the study period representing in average (SD) 68.3% (0.06) of the total study area. However, land areas with more koalas per km2showed larger annual variations, with koala mean (SD) densities of 0.25-0.5, 0.5-1, 1-2, 2-5 and > 5 koalas per km2 representing 16.8% (0.21), 13.8% (0.25), 0.7% (0.20), 0.3% (0.13), and 0.2% (0.1) of the study area in South-East Queensland, respectively.
We did find that clustering of koala sightings was not prominently different between the mating and non-mating seasons of koalas. While acknowledging the limitations associated sightings data, we developed a statistical model that addressed the spatio-temporal bias associated with observed koala sightings and provided long-term relative koala density estimates for one of the largest koala populations of Australia. In future research, the proposed model proposed here could be used for systematic survey data and ultimately for combining koala survey data with koala sightings data and remove the spatial bias more reliably.
Key words: Koala, modelling, population, citizen science, Queensland

Introduction

Direct observations and counting of koalas in the field faces many challenges, because koalas are difficult to detect in their natural habitat and they are widely dispersed (Ellis, et al. 2013, Masters, et al. 2004, McGregor, et al. 2013). Various methods are used to count koalas and estimate koala population density: systematic field surveys (David S. Dique, et al. 2003), distance sampling (D. S. Dique, et al. 2003), counting the number of faecal pellets under trees (Seabrook, et al. 2011, Sullivan, et al. 2002, Sullivan, et al. 2003), capture-mark-recapture methods (Masters, Duka, Berris and Moss 2004) and community surveys (Hollow 2015). Counts of vocalisations heard and spotlight surveys are also sometimes used (Smith and Andrews 1997).
Systematic surveys of strip transects (Dique, et al. 2004) and distance sampling using line transects are the most common methods to estimate koala density, but distance sampling techniques are only suitable for small areas because they are labour intensive and therefore expensive (Kjeldsen, et al. 2015). Research has described the relationship between koala’s tree preference and the presence of koala scat (Ellis, FitzGibbon, Melzer, Wilson, Johnston, Bercovitch, Dique and Carrick 2013) and koala scat prevalence has been shown to correlate well with koala density (Ellis, et al. 1998, Lunney, et al. 2009). Scat surveys are less expensive than systematic surveys and have been therefore been used to estimate koala population density (McAlpine, et al. 2006, Rhodes, et al. 2008). Postal surveys of targeted communities and incidental sightings of koalas by members of the public have also been used to estimate population counts (Cork, et al. 2000, Lunney, Crowther, Shannon and Bryant 2009, Lunney, et al. 2016, Predavec, et al. 2016, Sequeira, et al. 2014). These methods are suitable for smaller geographical areas with varying success and require statistical analysis to estimate population counts based on reported koala numbers (Santika, et al. 2014).
To review koala density estimates for conservation purposes, it is important to generate long-term datasets of koala populations over large geographical areas rather than to generate population counts at infrequent intervals (Ellis, Sullivan, Lisle and Carrick 1998, Lunney, et al. 2014, Lunney, Predavec, Miller, Shannon, Fisher, Moon, Matthews, Turbill and Rhodes 2016). Long-term survey data have been used to predict koala populations. A recent study in South-East Queensland (Santika, et al. 2015) attempted to estimate the geographic distribution of koala populations across a wide geographical area using spatial modelling techniques informed by long term (Rhodes, et al. 2015) line and strip transect survey data (1996 and 2015) collected by distance sampling. Ecological modelling techniques can also provide an alternative to active, longitudinal data collection (Schmolke, et al. 2010), although their validity is questionable in the absence of standardised methods for estimating wildlife density and distribution (Dique, Preece, Thompson and Villiers 2004, McGregor, Kerr and Krockenberger 2013).
With the advancement of communication technologies and the widespread availability of dedicated mobile applications, public participation in the collection of wildlife data is increasing in popularity. Attempts have been made to estimate wildlife populations using incidental sighting data alone, and/or in combination with survey data (Dorazio 2014, Sequeira, Roetman, Daniels, Baker and Bradshaw 2014). For instance, member of the public were invited to take part in collecting data on koala sightings as part of a program titled the ‘Great Koala Count’ in the Australian states of New South Wales and South Australia in 2012 (Sequeira, Roetman, Daniels, Baker and Bradshaw 2014). The ‘Great Koala Count’ has generated large number of incidental koala sighting using specific guidelines for data collection in pre-identified geographical areas in those two states. In South-East Queensland incidental koala sightings are collected since 1997, although no formal field protocols are provided to members of the public for the data collection (Dissanayake, et al. 2019). While these data have been used to describe koala population trends and to describe spatial biases identified (Dissanayake, Stevenson, Allavena and Henning 2019), it has not yet been used to estimate long-term koala density. In this study, we have developed a modelling approach to estimate koala density from observed sightings data over a period of 17 years, while addressing spatio-temporal detection biases and potentially clustering of observations.

Materials and methods