A comparison of methods for estimating abundance of unmarked Hochstetter's frogs
- Wildlife Ecology Group, School of Natural Sciences, Massey University, Private Bag 11 222, Te Papa-i-Oea Palmerston North, 4442, Aotearoa New Zealand
- Wildland Consultants Ltd, PO Box 50539, Porirua 5240, Aotearoa New Zealand
- Current address: Centre for Biodiversity and Restoration Ecology, School of Biological Sciences, Victoria University of Wellington, PO Box 600, Te Whanganui-a-Tara Wellington 6140, Aotearoa New Zealand
- Windy Hill-Rosalie Bay Catchment Trust, 429 Rosalie Bay Road, Aotea Great Barrier Island 0991, Aotearoa New Zealand
The Hochstetter’s frog (Leiopelma hochstetteri) is a nationally At Risk – Declining species, but management decisions for this species are limited by the lack of established monitoring protocols and analytical methods. We compared methods for inferring spatial and temporal patterns in abundance on Aotea (Great Barrier Island) using count data collected from fifteen 100 m stream transects in 2012, 2015 and 2021. Each transect was surveyed 2–3 times on the same day each year. Frogs were not marked, but individuals were identified in 2021 based on their body sizes and locations to facilitate the use of closed-population capture-mark-recapture (CMR) methods. We compared patterns in abundance estimates derived from Bayesian formulations of CMR (2021 only), N-mixture, Poisson regression of single counts, and occupancy models. Abundance estimates from CMR and N-mixture models were realistic and reasonably precise if detection probability (p) was assumed constant among transects. N-mixture estimates were 17% lower than CMR estimates but closely correlated with them. Relaxing the assumption of constant p among transects made little difference to CMR estimates but greatly reduced the precision of N-mixture estimates. Assuming constant p among transects, the N-mixture abundance estimates for the 15 transects were consistent among years. The 95% credible interval for the change in abundance from 2012–2021 ranged from a 24% decrease to a 10% increase. Mean first counts were 32% as high as N-mixture estimates, reflecting the estimated detection probability for first surveys. However, the spatial and temporal patterns inferred from single counts were consistent with those from N-mixture, and the change over time was estimated with only slightly lower precision. Estimated occupancy probabilities were correlated with N-mixture estimates but could not distinguish among transects with greater than 50 frogs and could not be used to infer changes over time.