Mobile Menu Open Mobile Menu Close

Distribution of great spotted kiwi (Apteryx maxima), 2012-2021

  • Publication Type

    Journal Article

  • Publication Year


  • Author(s)

    Toy, R., Toy, S., MacKenzie, D., Simister, K., Yong, S.

  • Journal Name


  • Volume, Issue

    69, 1

  • Pagination


  • Article Type



Great spotted kiwi, roroa, Apteryx maxima, Apteryx haastii, distribution, acoustic recorder, call rate, occupancy modelling, detection probability

Distribution of great spotted kiwi (Apteryx maxima), 2012-2021

Notornis, 69 (1), 1-18

Toy, R., Toy, S., MacKenzie, D., Simister, K., Yong, S. (2022)

Article Type: Paper



Conservation management requires knowledge of the distribution of species and how this changes over time. Great spotted kiwi (roroa, Apteryx maxima) is classified as globally threatened, ‘Vulnerable’ by the IUCN. It occurs only in the northwest of the South Island of New Zealand, is nocturnal and occurs at low density in mainly remote, mountainous terrain. To determine its distribution, we deployed acoustic recorders at 1,215 locations across 1,400,000 ha between 2012 and 2021. We analysed 3,356 nights of recordings to determine presence and call rates at each location. Roroa were distributed across 848,000 ha, but we identified a core area in northwest Nelson representing just 12% of the distribution (101,000 ha). Within the core, call rates exceeded 3 calls/h at many locations. Call rates provide only a relative indication of abundance but, outside the core, call rates fewer than 0.3 calls/h are common, suggesting that roroa are relatively sparse over much of their distribution. We used a static occupancy model with climatic, topographic and land-cover class variables to better understand the distribution. Eighty percent of recorder-nights had a detection probability exceeding 50%. At this probability, 73% of 5 x 5 km cells surveyed were sampled sufficiently to exceed 90% probability of detection if roroa were present. Annual rainfall and land-cover class appear most important for modelling occupancy. However, comparison of probability of occupancy and actual distribution suggests that variables not included in the modelling, which might include predation, also affect the distribution.