Ecological monitoring provides critical information regarding species, community, and ecosystem trends, including those associated with human disturbance (Haughland, et al., 2010). However, false conclusions can be drawn from such monitoring if sampling is not carefully designed and conducted. This can have implications if results do not represent the real-world situation and may lead to inadequate and inappropriate conservation actions (Kéry, 2002). This is further complicated for species that are infrequently detected (e.g., cryptic < 30% detection; Spetch et al., 2017) or are rare in nature (< 30% occupancy probability; Spetch et al., 2017) since detection rates are typically low over large spatial scales and therefore characterized by a high frequency of zero observations during monitoring despite the species being truly present (reported as false negatives) (MacKenzie et al, 2006). Ecological monitoring should ideally report detection probabilities and confidence levels. Only by doing this will it be possible to quantitatively understand imperfect detection for some rare or cryptic species. By inclusion of these metrics, improvements in sampling design should be considered to ensure monitoring efforts maximize the potential for positive conservation outcomes.
Imperfect detection occurs when individuals, populations or species are not found during surveys despite their presence (Martin et al., 2005). This may lead to estimates of abundance, distribution and species richness being negatively biased (Kéry & Schmidt, 2008), and spatial or temporal patterns in species abundance, distribution and biodiversity being obscured (Link & Sauer, 1998; Pollock, et al., 2002). It is imperative that monitoring programs quantify imperfect detection by considering detection probabilities, thus allowing the removal of bias from estimates (Tanadini & Schmidt, 2011). Occupancy modelling is one such tool that can estimate occurrence while accounting for imperfect detection. An advantage of estimating detection probability for ecological monitoring programs is that conservation practitioners can optimize survey timing to ensure high detection rates (McGrath et al., 2015).
Improvements in technology may allow for more efficient and effective monitoring strategies to increase detectability, particularly for declining and/or cryptic species (Taylor et al., 2014; Diggins et al., 2020). Passive acoustic recorders (PARs) are one such technology that enable capture of animal vocalisations (Ross et al., 2023). By programming PARs to record their acoustic environment remotely for extended periods and at any time of the day, rich datasets can be gathered that are otherwise not feasible with in-field survey techniques (Celis-murillo, Deppe & Ward 2012). This can be coupled with in-field data loggers that measure abiotic variables to gather additional temporal and spatial data to identify covariates that might influence activity (Willacy et al. 2015; Sugai, et al., 2018). Additionally, presence/absence data obtained from PARs can be paired with traditional count surveys to improve estimates of abundance (Doser et al. 2021). These tools can also enable the construction of phenological models, enhancing ecological monitoring (Hoffmann & Mitchell, 2022).
Some amphibians are cryptic and rare, and their populations can have considerable natural fluctuations (Green, 2003; Beranek et al., 2022), which makes long-term monitoring and the detection of threat-related declines difficult (Barata, Griffiths & Ridout, 2017). Accurate monitoring of target species can be impacted by a range of ecological, environmental, economic, and logistical challenges (Bakker et al., 2010), including imperfect detection. These constraints have led to amphibians being under-represented in distribution and abundance studies, resulting in scarce historic data for many species (Recher & Lim 1990).
Effective conservation of threatened species relies on adequate information regarding their life history traits, such as those associated with reproduction, as it is the interaction between these traits that influence fitness (Roznik et al., 2015). For example, identifying breeding strategies offers an insight into why populations might vary in size and structure (Petranka & Thomas, 1995; Gould et al., 2022a), how frequent and for how long breeding events occur (Díaz-Paniagua, 1992; Paracuellos et al., 2022), as well as the timing of these events (Dixon & Heyer, 1968). The Wells (1977) explosive versus prolonged breeding paradigm presents a continuum of breeding strategies that varies among frogs and populations. Determining breeding strategies selected by species and populations can provide critical information for determining distribution and abundance and thus conservation assessment. While our understanding of the life history strategies of numerous anuran amphibians remains limited, deploying PARs at breeding habitats is useful because most anuran amphibians engage in vocalisation during their reproductive processes (Gerhardt and Huber, 2002; Köhler et al., 2017).
We combined acoustic monitoring with occupancy modelling to increase our understanding of spatial and temporal breeding patterns of a threatened and cryptic burrowing species, the giant burrowing frog, Heleioporus australiacus australiacus (HAA). We present this as a case study to demonstrate a process where rich data generated through acoustic surveys performed by PARs can be used to estimate detection probability with high precision to allow sufficient statistical power to model the probability of occupancy of cryptic species that vocalise.
We made several predictions based on the current understanding of spatial and temporal breeding patterns for this species (see Appendix Section A1 for a summary of HAA). We predicted that this species will have a low probability of detection (< 30%; Spetch et al., 2017) and a low probability of occupancy (< 30%; Spetch et al., 2017) driven by environmental and habitat specialisation for breeding (Stauber, 2006; Penman et al., 2005a; Penman, Lemckert, & Mahony, 2006a). Additionally, we hypothesized that breeding detectability would be positively correlated with temperature and rainfall as HAA is considered a late summer breeding species and has been shown to move to breeding habitats following rainfall (Penman et al., 2005b; Penman, Lemckert & Mahony, 2006a; Penman, Lemckert & Mahony, 2008a). Lastly, we predicted that breeding site occupancy would occur more often in ephemeral water bodies compared to permanent sites (Stauber, 2006) and positively correlated with the percentage of sandy soil profiles within the immediate landscape of the breeding site, as previous surveys indicate that HAA breeding habitats occur on sandy soils and, except for one population in Ulladulla, are not found on clay-based soils (Daly, 2019).