Judge et at. • REOCCURRENCE OF ‘OMAO ON HAWAI I ISLAND 
677 
Mauna Loa 
n_n_n 
m 
gt. 
V s—- 
3 000 n. 
000 
'Oma'o detections 
0 incidental observations 
# 2010 HAVO survey 
A 1978 HFBS 
•••••••other surveys with detections 
.other surveys with no detections 
Land cover 
j ] Mesic/wet forest 
[ | Dry shrubland/woodland 
| 1 Alpine scrub or unvegetated 
] Sub-alpine/alpine life zone 
FIG. 1. Detections of ‘Oma'o during 10 forest bird surveys on Hawai'i Island. The species’ range is nearly continuous 
in forests on the windward side of the island, while there have only been scarce detections on the leeward side. The species 
was not seen nor heard for >30 years in leeward forests and woodlands until it was detected during the HAVO land bird 
surveys in 2010. Tract areas are II Ilonomalino. P = Papa, and K = Northwest Kahuku. The graphic also depicts sub- 
alpine and alpine incidental observations ol the species; as well as potential habitats that have not been formally surveyed. 
2005). Covuriates included cloud cover, rain, 
wind, gusts, observer, time of detection, canopy 
cover, canopy height, panel member, detection type, 
and tract. All covariates were trealed as a factor, 
except lime of detection, which was treated as both a 
factor and a continuous covariate. Assessing time of 
detection as a continuous covariate helped ascertain 
if the detection rate varied during the morning. Each 
detection rate model in the candidate set was lit to 
data pooled across tracts, including 269 detections in 
windward tracts (Judge et aJ. 2011); the model 
selected was that with the lowest second-order 
Akaike s Inlomiation Criterion corrected for small 
sample sizes (AICc) (Table I) (Buckland et al. 
2001, Burnham and Anderson 2002). Data were 
truncated at a distance where detection probability 
was -10%. This procedure facilitated modeling by 
deleting outliers and reducing the number of 
parameters needed to modify the detection function. 
Tract-specific densities were estimated from the 
