Study Area

Field studies were conducted at the remote Halepaakai field site in the central mountains of Kauai (Figure 1). The 100 ha study area lies between 1,205 and 1,400 m in elevation and consists of largely intact ‘ōhi‘a-dominant wet forest. Halepaakai is located ∼5 km west of Mount Waialeale, where mean annual precipitation is ∼10 m (Giambelluca et al. 2013). This section of forest is encompassed by the Alakai Wilderness and is considered the best remaining habitat for indigenous forest bird species on Kauai.

FIGURE 1.

Halepaakai study area, depicting subplot boundaries, 100-foot contours, and major stream courses.

Nest Searching and Monitoring

Nest searching and monitoring were conducted from March through June of 2012 and 2013, which is the peak of the breeding season for both species based on observations made in 2011 (R. L. Hammond personal observation). Halepaakai was divided into 4 subplots, and observers searched for nests following methods in Martin and Geupel (1993). Nests were monitored using a spotting scope located 10–25 m from nests to minimize disturbance and to improve observations of these high-canopy nesters. Nests were monitored every 2–3 days until termination, except when monitored near hatching or fledging, in which case they were typically monitored every 1–2 days. Nests were defined as successful when they fledged at least one chick (Martin and Geupel 1993) and failed when no chicks fledged. Even with spotting scopes, precise estimates of fledge dates were difficult to obtain. To avoid incorrectly assigning fates to nests, nests that terminated within 5 days of the projected fledge date with no obvious evidence of predation (e.g., chick remains in nest, disheveled nest) and those for which fledglings could not be located were characterized as unknown fate because chicks younger than 5 days from fledging were easily characterized as too young to fledge.

Fledglings were sought for up to 1 week after a nest was found empty and presumed to have fledged. A 70 m radius around nests was searched because fledglings of multiple honeycreeper species at Halepaakai (including ‘Akikiki) are capable of such movements on the first day of fledging (R. L. Hammond personal observation). Family groups from successful nests were monitored after fledging for the remainder of the season to determine juvenile dependency length.

Cause of failure was investigated using an SM-202T (Teklink Security, Ontario, CA, USA) wireless camera mounted on a telescoping 9 m pole, and nests were collected to further investigate failures when video was inconclusive. Presence of rodent feces, eggshell fragments, chick remains, and disturbance of nest material were used to identify predation events. To determine if weather directly caused nest failures, observers searched for eggs and nestlings that might have been ejected from nests when failures occurred within 2 days of a storm event (i.e. >40 mm of precipitation; Kuntz 2008). After nests terminated, nest height was measured using a clinometer.

Clutch size was not measured because of difficulty accessing nests and the need to minimize disturbance, but brood size was measured based on the number of chicks in nests that survived to the late nestling stage. Brood sizes and nest contents (i.e. eggs found in nests after a sole chick fledged) were used to estimate egg hatching rate (i.e. the proportion of total eggs laid that hatched). Hatching rate was estimated assuming that every nest initially contained 2 eggs (because brood sizes for both species were typically 2; see Results). A second, more conservative, estimate was also calculated by assuming that inaccessible nests that failed to hatch, or contained only one chick, originally contained one egg.

For each species, we calculated mean nest height ±SD; mean length of nest construction, incubation, and nestling periods ±SD; and brood size. Nests found in the building stage that did not progress to incubation were not used in these calculations. Nests were used to estimate incubation and nestling period lengths only when they were monitored for the entire period. Incubation was defined as the length in days from the date egg laying began to the earliest date on which hatching evidence (e.g., female popping on nest, feeding young, removing eggshells or fecal sacs) was recorded. Both species probably lay a 2-egg clutch at a rate of one egg daily (Eddinger 1972a, 1972b, VanderWerf and Roberts 2008). In 2012, hatching dates were estimated for nests found before incubation based on data from closely related species to target critical monitoring periods: Hawaii Creeper (Loxops mana; Lepson and Woodworth 2002) for ‘Akikiki, and ‘Ākepa or Hawaiian Akepa (Loxops coccineus; Lepson and Freed 1997) for ‘Akeke‘e. In 2013, hatching dates were projected from the start of incubation using the mean estimate for incubation period length derived from the 2012 breeding season. Nestling period lengths were defined as the length in days from the estimated hatching date to the first date of fledging. Fledging dates were projected from estimated hatching dates by using nestling period lengths of closely related species in 2012 or by using mean nestling period lengths from 2012 for nests in 2013 to target critical monitoring periods for catching fledgling events. Range estimates for incubation and nestling periods may be high or low by 1–3 days; exact hatching and fledging dates were rarely known because nests were not checked every day, the inside of nests were not visible to confirm hatching, and fledglings were rarely seen leaving the nest. Observations recorded during monitoring were compiled to draw inferences about male and female roles in the nesting cycle (e.g., nest building, incubation, provisioning of female and young). ArcMap 10.0 (ESRI, Inc.) was used to estimate nearest-neighbor nest distances (±SD) when home ranges of 2 simultaneously breeding pairs abutted.

‘Akikiki at Halepaakai were captured and color-banded opportunistically or as part of a different study to estimate occupancy and habitat use since 2008 (L. A. Benhke personal communication). As a result, 2 pairs consisting of one banded member and 3 pairs with both members marked were monitored during this study. These color-banded birds were used to determine if pairs exhibited inter- and intra-annual mate fidelity and to document double-brooding and renesting attempts.

Statistical Analysis

Only nests that progressed to the incubation period and had known fates were used for statistical analysis. Nests' monitoring data were compiled for each species and used to derive estimates of nesting success ±SE using program MARK (White and Burnham 1999), which uses a maximum likelihood method of estimating a daily survival rate (DSR) for nests based on exposure days (Mayfield 1961). The generalized linear modeling framework provided in program MARK was used to build competing models of DSR that incorporated individual covariates.

The variables of nest height, nest age, and time across the breeding season (i.e. a linear or quadratic time trend as the breeding season progresses) were selected a priori and used to build 6 models for each species. Nest height might be an important predictor of nesting success if lower nests are more susceptible to rodent predation (Woodworth and Pratt 2009), as has been shown in the O‘ahu ‘Elepaio or Oahu Elepaio (Chasiempis ibidis; VanderWerf 2012b). Alternatively, nests high in the canopy may have lower success if weather is a primary cause of nest failure (Simon et al. 2000, Kuntz 2008; Maui Forest Bird Recovery Project personal communication) because they are more exposed to severe weather events. Nest age was considered for examination to determine if nest failure was more likely to occur early or late in the nesting cycle. The linear (time) and quadratic (time²) models of nesting success across the breeding season were considered to assess temporal effects potentially correlated with factors such as weather, food availability, and predators. Small sample sizes prevented the modeling of inter-annual effects on DSR. A global model was included to evaluate the additive effects of all variables modeled. The null model, having no additive effects to DSR, was used to derive estimates of constant DSR for each species, and these estimates were used to calculate cumulative estimates of nesting success for the entire nesting period by exponentiating the respective DSR to the mean nest period estimates from this study (see Results). For each species, a cumulative estimate of nesting success was calculated for each year and cumulatively for both years, and variances were calculated using the delta method (Powell 2007b). Akaike's Information Criteria with a small sample size correction (AIC_c) was used for model selection, and the model with the lowest AIC_c was considered best. Because no model in either model set received substantial support over other models in the set (i.e. ≥90%), model uncertainty was addressed by reporting parameter estimates for variables in each confidence set (Burnham and Anderson 2002).

‘Akikiki

We located 20 ‘Akikiki nests during the study (Table 1). Nest construction began in early March, with most pairs showing signs of building by mid-March, and the last evidence of construction was observed in mid-May. All 20 ‘Akikiki nests were placed in ‘ōhi‘a trees; 16 nests were placed in small branches just beneath the canopy, 3 were placed on a horizontal branch, and 1 was built next to the trunk. ‘Akikiki nest heights ranged from 4.5–16.6 m, with a mean of 9.2 ± 2.6 m (n = 17). Seven active nests were found when a neighboring pair's nest had already been located and was concurrently active. One of these nests was the nearest-neighbor for 2 nests, with the 2 different nearest-neighbor nests being active at different periods of time during its nest cycle. Nearest-neighbor nest distance was estimated at 268 ± 85 m (n = 4, range 194–363 m).

TABLE 1.

‘Akikiki and ‘Akeke‘e nest fates at Halepaakai (A = abandoned, S = successful, F = failed, D = depredated, O = failure not attributed to predation, C = nest failed with unknown cause, U = unknown fate). Daily survival rate (DSR) ± SE and cumulative estimates of nesting success ± SE derived from the null model (i.e. constant DSR with no additive effects) and listed with confidence intervals in parentheses.

Nest construction lasted from 13 to >20 days based on 4 nests found near construction initiation. On one occasion a male was confirmed helping with nest construction. Incubation ranged from 13 to 20 days with a mean of 16.5 ± 2.7 days (n = 7), and the nestling period ranged from 16 to 23 days with a mean of 19.0 ± 2.3 days (n = 7). Males were never observed helping with incubation. Putative males often fed putative females at or near the nest during nest construction, incubation, and early in the nestling period. Both parents fed nestlings, but putative females received food from the putative male at the nest and transferred it to the chicks on the day of and day after hatch, after which time males fed nestlings directly.

Brood size was 2 chicks for 11 nests and 1 chick for 4 nests. Of the 1-chick nests, 2 nests contained 1 unhatched egg, and the other nests could not be accessed to determine clutch size. One of the unhatched eggs was candled and confirmed to be infertile. Fifteen nests could be used to estimate hatching rate. Excluding a nest that failed early in incubation and assuming that 2 eggs were laid in every nest, hatching rate of eggs was as low as 87% but may have been as high as 93% if the 2 inaccessible nests containing only 1 chick originally had only 1 egg.

Monitoring of 3 pairs in which both members were banded suggested that ‘Akikiki were monogamous. The only color-banded pair that was monitored both years appeared to be monogamous throughout the study. Double-brooding was witnessed twice. In 2012, a color-banded pair successfully fledged a single chick that likely died within a week of fledging. This pair's second brood of 2 chicks progressed to the late nestling stage and may have fledged, but fledging was not confirmed because the field crews left at the end of June, prior to the projected fledge date. In 2013, a different color-banded pair was observed fledging 2 broods, and frequent observations of the male foraging with 2 hatch-year (HY) birds suggest males largely assume responsibility for the first brood when pairs double-brood. ‘Akikiki HYs were observed remaining with the family group until field crews left the field site each year, suggesting a juvenile dependency period of more than 3 months. One color-banded family group was observed with a second-year (SY) offspring at Halepaakai in January 2012. The pair's young separated from the parents in March. Based on the similarity of facial patterns of some of the breeding birds to those of HY birds, ‘Akikiki may nest in their second year, and one color-banded SY bird was observed attempting to build a nest without a mate in 2012. A renest may have been attempted for one pair after their first nest failed, but this could not be confirmed because the birds were not color-banded.

Sixteen ‘Akikiki nests had sufficient data for use in candidate modeling, spanning 415 exposure days. Three of these nests failed during the course of the study. One of the nest failures was apparently depredated by a rodent in the late nestling stage, one failed in the incubation stage but could not be accessed to determine a cause of failure, and the third failed in the late nestling stage but was seemingly undisturbed and contained one egg.

The model of quadratic time was the best model (Table 2), suggesting a decrease in nesting success at the middle of the breeding season, and differed from the second-best model by a ΔAIC_c of ≥2. The model of quadratic time received 82% of the support in the model set, but 95% confidence intervals (CI) around the parameter estimates included zero for this variable (β_time = −2.97 ± 3.42, CI = −9.80 to 3.87 and β_time² = 0.05 ± 0.06, CI = −0.07–0.16) and all variables in the confidence set (β_{nest height} = −0.25 ± 0.24, CI = −0.73–0.22; β_time 0.02 ± 0.03, CI = −0.04–0.08; β_{nest age} = −0.001 ± 0.054, CI = −0.110–0.108), suggesting that a real effect of any variable is uncertain. The cumulative estimate of nesting success for the constant DSR was 0.77 ± 0.12 (Table 1).

TABLE 2.

Confidence model sets of daily survival rate (DSR) for nests of ‘Akikiki and Akeke‘e monitored in 2012 and 2013.

‘Akeke‘e

Eight ‘Akeke‘e nests were found during the study (Table 1). Nest construction probably begins in early March for some pairs based on one nest found in the late building stage on March 16, 2013, but most females in the study (n = 6) began incubating in late March or early April, suggesting that nest construction typically begins in mid to late March. All ‘Akeke‘e nests were located in the small branches in the canopy of ‘ōhi‘a (n = 8), and mean nest height was 11.1 ± 2.3 m (range 8.7–14.5 m, n = 8). ‘Akeke‘e were difficult to follow, and nests were located for only a small proportion of breeding pairs in the study area.

Construction duration was not measured because no nests were found earlier than the late building stage. The mean length of the incubation period was 16.9 ± 2.0 days (range 15–19 days, n = 4), and the mean nestling period was 18.0 ± 2.3 days (range 16–22 days, n = 5). Only females were observed constructing and incubating the nest. Males often provisioned females at or near the nest during construction, incubation, and early in the nestling period. Males fed only females at the nest during hatching day and the day after hatch while females were solely responsible for feeding the hatchlings, but afterward both adults fed the nestlings.

Brood size was 2 chicks in 5 of 6 successful nests; the other successful nest had only 1 chick and it could not be accessed to determine clutch size. Both nest failures terminated during incubation; one was attributed to hatching failure because the female apparently incubated for 1 week after the projected hatching date before abandoning. The other failure was attributed to poor nest attendance by the female and was therefore excluded from hatching rate estimation. Based on 7 nests, hatching rate was 79% if 2 eggs were assumed to have been laid in each nest, but 93% if 1 egg was assumed to have been laid in the nest that failed to hatch and in the nest that fledged only 1 chick.

Breeding behavior and nesting were challenging to fully document for ‘Akeke‘e, and several aspects of their biology remain incompletely known. One pair may have raised a second brood, where the assumed male of a brooding female fed 2 HY birds multiple times at 20–100 m from the nest. Parentage was uncertain, however, because these birds were not color-banded and the female was never seen directly interacting with the family group. It is unclear if ‘Akeke‘e breed in their second year because no birds were color-banded, and they could not be discriminated by age. Fledglings were attended by both adults after departing the nest. Juvenile dependency was possibly as short as 2 months because HY birds were seen traveling and foraging in pairs late in the study period for 20–30 min without an adult nearby. These birds also showed signs of molting juvenile plumage and made only adult calls.

All 8 nests found had known fates and were used to build candidate models based on 207 exposure days. Of 2 total nest failures, one was attributed to poor nest attendance by the female, and one was apparently due to hatching failure.

All models were included in the 95% confidence set, and the model of nest height was the best model but received only 30% of the support in the model set (Table 2). The ΔAIC_c for the nest height model differed from the second-best model (quadratic time) and the third-best model (null) by 0.63 and 0.73, respectively, suggesting that all 3 models have some explanatory value. Additionally, 95% confidence intervals around all parameter estimates included zero (β_{nest height} = −0.67 ± 0.51, CI = −1.68 to 0.35; β_time = −0.70 ± 0.90, CI = −2.49 to 1.10 and β_time² = 0.007 ± 0.010, CI = −0.013 to 0.027; β_{nest age} = −0.02 ± 0.07, CI = −0.17 to 0.12; β_time = 0.004 ± 0.026, CI = −0.048 to 0.055), suggesting that all variables modeled may not be good predictors of DSR. The cumulative estimate of nesting success for constant DSR was 0.71 ± 0.17 (Table 1).