Lormee et al. • POPULATION SURVEY OF LEACH'S STORM-PETREL 
249 
TABLE 2. Detection probability of Leach's Storm- 
Petrel burrows on Grand Colombier Island. Saint Pierre and 
Miquelon Archipelago. Dev = deviance, Np = number of 
parameters, Gof = goodness of fit P-valuc The selected 
model is in bold. 
Model 
ATCr 
AAlCr 
Dev 
Np 
Col 
Observer 
23.85 
_ 
19.84 
2 
0.99 
Constant 
24.37 
0.52 
22.36 
1 
0.77 
Habitat 
26.35 
2.5 
22.36 
2 
0.64 
Observer X Habitat 
27.80 
3.95 
19.80 
4 
1 
Calculation of Leach's Storm-Petrel Breeding 
Population Size .—The mean Leach's Storm-Petrel 
breeding population size (AO was calculated as: 
Ox Dx A 
N= 
where O is the mean burrow 
px(\-B)' 
occupancy probability. D is the mean burrow 
density estimated from plots, A is the surface area 
of the island, p is the burrow detection probability, 
and B is the breeding failure probability. The 
variance of N was calculated using the delta 
method following Sober (1982). All values are 
means ± SE, unless otherwise staled. 
RESULTS 
Burrow Occupancy Probability — Occupancy 
probability was neither affected by slope angle 
nor by vegetation type (slope: P\ = 2.13, P = 
0.14; habitat: F\ - 0.50, P = 0.47) and was 
considered similar for every sector (0,546 ± 
0.029). 
Pail lire Probability. —Breeding failure was 
estimated from 73 burrows (32 of the 105 active 
burrows initially chosen could not be found 
during the second visit because of fern growth). 
Failure probability was 0.068 ± 0.029 on 10 July, 
just before we started the survey. Only 3% of the 
active nests contained a recently hatched chick by 
this date. 
Detection Probability. —The total number of 
burrows used to estimate detection probability of 
all observers in = 5) was 513 (219 in fern and 294 
in herbaceous habitats). Burrow densities within 
the plots used to estimate detection probability 
were comparable to those in the entire set of plots 
i plots sampled on plateau: t- test = 1.679, df = 57. 
P - 0.098: plots sampled on island slope: /-test = 
1.745. dt = 114. P = 0.084). Forty-three burrows 
were detected by primary observers only. 59 by 
secondary observers only, and 411 by both 
observers. The detection probability of each 
observer was estimated for 185 ± 3.3 burrows. 
All tested models had a good fit to the data. Both 
constant and observer effect models received 
relatively similar support (AAICY < 2; Table 2). 
Detection probabilities based on the observer 
effect model, ranged from 0.787 ± 0.040 to 
0.913 ± 0.018. We chose the most parsimonious 
model to estimate detection probabilities, i.e.. the 
constant model (detection probability of a burrow 
of Leach Storm-Petrel was neither affected by 
observer nor by habitat type). Detection probabil¬ 
ity for a single observer was obtained from the 
constant model and estimated at 0.89 ± 0.01. 
Relationships Among Burrow Density , Slope, 
and Habitat Type .—Burrow density was positive¬ 
ly affected by slope (Z = 7.016, P < 0.0001). We 
separated the island into four sectors depending 
upon the importance of slope angle: plateau, 
southern and northern sides, and steep area. 
Burrow density was estimated specifically for 
each sector (Table 2). This stratification w as used 
to estimate the number of breeding pairs. 
Leach's Storm-Petrel Breeding Population .— 
We estimated the Leach's Storm-Petrel breeding 
population size at 363,787 ± 19,991 (95% Cl = 
295,502-432,072; Table 3) pairs on Grand Co¬ 
lombier Island in 2008 considering burrow 
occupancy, nest failure, detection probability, 
and specific burrow density for each sector. 
DISCUSSION 
Survey Method .—The estimates of detection 
probability suggested an individual observer may 
miss up to 11% of burrows. We strongly 
encourage systematic estimation of detection 
probability during surveys to increase their 
accuracy and the power to detect subtle temporal 
changes in population size. Detection probability 
estimation would be particularly relevant if 
observers are expected to change throughout 
successive surveys. 
Failure at the egg stage appears to be the main 
factor affecting breeding success for Leach’s 
Storm-Petrel (Bicfcnell et al. 2009). Breeding 
failure preceding the survey was low at our study 
site (—7%) in comparison with reported values in 
the literature (hatching success = 77.9 ± 5.1%, 
min = 66%, max = 86%) (Huntington et al. 
1996). We may have slightly underestimated 
incubation failure rate since the laying period 
starts in early June and some breeding pairs may 
have tailed in the 2 weeks preceding monitoring 
of burrows. Additional failures also occur near 
