Abstract
Endemic Hawaiian forest birds have experienced dramatic population declines. The Big Island National Wildlife Refuge Complex (Refuge Complex) was established for the conservation of endangered forest birds and their habitats. Surveys have been conducted at two units of the Refuge Complex to monitor forest bird populations and their response to management actions. We analyzed survey data from 1987 to 2019 at the Hakalau Forest Unit (HFU) and from 1995 to 2019 at the Kona Forest Unit (KFU). We analyzed three strata at HFU: open-forest, closed-forest, and afforested-pasture, and two strata at KFU: upper (>1,524 m elevation) and lower (<1,524 m). In all years, ‘i‘iwi Vestiaria coccinea, ‘apapane Himatione sanguinea, and Hawai‘i ‘amakihi Chlorodrepanis virens virens were the most abundant species at HFU. Three endangered forest bird species, Hawai‘i ‘ākepa Loxops coccineus, ‘alawī Loxops mana (also known as Hawai‘i creeper) and ‘akiapōlā‘au Hemignathus wilsoni, had much lower densities. The most abundant species at KFU was ‘apapane, followed by Hawai‘i ‘amakihi at much lower densities. We found a continuation of several trends observed in previous analyses at HFU up to 2012, with most species’ trends upward in afforested-pasture stratum, stable in the open-forest stratum, and downward in the closed-forest stratum. However, more species were showing downward trends in all three strata during the most recent decade. Results were mixed at KFU, with most species’ trends downward in the upper stratum and upward in the lower stratum. Populations of endangered species were either locally extirpated at KFU or in numbers too low to reliably estimate population abundance. The Refuge Complex is important for conservation of forest birds on Hawai‘i Island. Our results show that HFU supports the majority of three endangered forest bird species. Threats to forest birds at the Refuge Complex appear to be having a negative impact. These threats include habitat loss, disease, feral ungulates, and nonnative predators. Continuing and enhancing management actions, such as forest restoration and removal of invasive species, could help mitigate these impacts and allow the Refuge Complex to remain a key site for forest bird conservation in Hawai‘i.
Native Hawaiian birds have experienced substantial population declines and many species extinctions (BirdLife International 2000; Scott et al. 2001). This is especially true for most species of forest birds (Banko and Banko 2009; Gorresen et al. 2009). Scott et al. (1986), realizing that there was little information on Hawaiian forest birds and their habitats, organized comprehensive surveys from 1976 to 1983—the Hawai‘i Forest Bird Survey or HFBS. These surveys showed that some of the highest quality forest bird habitat in Hawai‘i was located at upper elevation areas of the Hāmākua coast on Hawai‘i Island. The Hakalau Forest Unit (HFU) of the Big Island National Wildlife Refuge Complex (Refuge Complex) was established in this region in 1985 to conserve and restore endangered Hawaiian forest birds and their habitats. Habitat for three endangered forest bird species (US Endangered Species Act [ESA 1973, as amended]; Hawai‘i ‘ākepa Loxops coccineus, ‘alawī Loxops mana [also known as Hawai‘i creeper], and ‘akiapōlā‘au Hemignathus wilsoni) and one threatened species (‘i‘iwi Vestiaria coccinea) is found at HFU. The first three species are listed as endangered and ‘i‘iwi is listed as vulnerable under the International Union for Conservation of Nature Red List of Threatened Species (BirdLife International 2000). Four additional endemic forest bird species occur at HFU, Hawai‘i ‘elepaio Chasiempis sandwichensis, ‘ōma‘o Myadestes obscurus, Hawai‘i ‘amakihi Chlorodrepanis virens virens, and ‘apapane Himatione sanguinea. In 1997, the U.S. Fish and Wildlife Service acquired the Kona Forest Unit (KFU) with a similar purpose of providing protection and habitat for threatened and endangered species, including forest birds. Small but declining populations of two endangered forest bird species, Hawai‘i ‘ākepa and ‘alawī, as well as the threatened ‘i‘iwi, are found at KFU. This Unit supported the last wild population of ‘alalā Corvus hawaiiensis, now extinct in the wild (DLNR 2020, 2021; Star Advertiser 2020).
The Refuge Complex covers areas from 600 m to 2,400 m elevation, with most management efforts occurring between 1,500 m and 2,000 m. Higher elevation areas, above 1,500 m elevation, throughout Hawai‘i are essential refugia from disease-bearing, nonnative mosquitoes (LaPointe et al. 2010). However, as elsewhere in Hawai‘i, the upper elevation forests in the Refuge Complex have been degraded by decades of domestic and feral cattle Bos taurus grazing (Scott et al. 1986). Cattle graze and browse on understory plants and tree seedlings, resulting in forests converted to grasslands or understory species mostly eliminated in intensively grazed areas (Scott et al. 1986; Pratt and Jacobi 2009). Cattle can also damage mature trees via bark stripping, which can kill trees outright or make them vulnerable to disease (Perroy et al. 2021). Feral pigs Sus scrofa and invasive plants further degraded forest bird habitat. Feral pigs browse native plants, damage soils, inhibit native plant regeneration, alter nutrient cycling, and disperse invasive plants (Hess et al. 2006). Pig wallows and cavities created when pigs forage on hāpu‘u ferns Cibotium menziesii form mosquito breeding habitat (Atkinson et al. 1995; Aruch et al. 2007; LaPointe et al. 2009). Invasive plants compete with native plants and can change forest ecosystems (Pratt and Jacobi 2009). Management actions at HFU have included eradicating feral cattle and pigs, controlling invasive plants, and planting native trees and shrubs (USFWS 2010). Similar management is planned for KFU but has yet to be implemented. These actions and the low rate of occurrence of mosquito-borne disease have resulted in HFU being one of the most important areas for native forest birds on Hawai‘i Island, and the only area on the Island with historically positive or stable long-term trends for forest birds (Gorresen et al. 2009; Camp et al. 2010). The widespread decline of native birds in most other Hawaiian forests indicate continued pressures on forest bird populations (Paxton et al. 2016; Judge et al. 2021).
Surveys of forest bird populations have been conducted annually at HFU since 1987 and periodically at KFU since 1995. Data from these surveys were used to assess the status and trends of forest birds and evaluate whether management actions have been effective or if changes in stressors, such as the prevalence of mosquitos, have had an impact. Camp et al. (2010, 2016) completed comprehensive analyses of the status and trends for forest birds at HFU from 1987 to 2012. However, forest bird abundances and trends have not been estimated since the 2012 survey. Surveys conducted at KFU in 1995 and 1999–2001 were analyzed in Camp et al. (2009). Data from surveys conducted at KFU after 2001 have not been analyzed. We determined abundances and trends of forest bird populations at HFU and KFU using survey data collected through 2019.
Methods
Study site
The 15,390-ha HFU on the windward slope of Mauna Kea Volcano was divided into three strata for this analysis (Figure 1). The open-forest stratum at an elevational range of 1,400–1,920-m (area = 3,199 ha) was once intensively grazed forest but the forest canopy remained intact. The closed-forest stratum (area = 1,998 ha) at 1,450–1,750-m elevation was relatively unmodified by grazing. Vegetation in the open and closed-forest strata was dominated by native ‘ōhi‘a lehua Metrosideros polymorpha and koa Acacia koa/‘ōhi‘a forest (Jacobi 2017). The afforested-pasture stratum at 1,650–2,000-m elevation (area = 1,314 ha) was logged and grazed for decades prior to Refuge Complex establishment. It was mostly open grassland at the beginning of the study that has since been substantially reforested. By the end of the time series in 2019, it included planted stands of koa trees up to 30 y old, with some native trees and shrubs in the understory. Nonnative plant species occurred in all three strata, with the most injurious species being various pasture grasses, gorse Ulex europaeus, Florida blackberry Rubus argutus, banana poka Passiflora tripartite mollissima, English holly Ilex aquifolium, and photinia Stranvaesia davidiana davidiana (USFWS 2010). Feral cattle were eradicated from HFU. Numbers of feral pigs at HFU have fluctuated over the survey period in response to control efforts (Leopold et al. 2015) and were sighted in all strata since 2010 (S.J.K., U.S. Fish and Wildlife Service, Big Island National Wildlife Refuge Complex, pers. obs., 2021).
Location of the Big Island National Wildlife Refuge Complex Kona Forest Unit (KFU; open circle in islands map) and Hakalau Forest Unit (HFU; closed circle in islands map (Panel A), Hawai‘i Island, Hawai‘i, USA. The two strata at KFU are upper (dark gray) and lower (light gray; Panel B). The three strata at HFU are closed-forest (dark gray), open-forest (medium gray), and afforested-pasture (light gray; Panel C). Survey stations are represented with black dots. Surveys were conducted between 1987 and 2019 at HFU, and between 1999 and 2019 at KFU.
Location of the Big Island National Wildlife Refuge Complex Kona Forest Unit (KFU; open circle in islands map) and Hakalau Forest Unit (HFU; closed circle in islands map (Panel A), Hawai‘i Island, Hawai‘i, USA. The two strata at KFU are upper (dark gray) and lower (light gray; Panel B). The three strata at HFU are closed-forest (dark gray), open-forest (medium gray), and afforested-pasture (light gray; Panel C). Survey stations are represented with black dots. Surveys were conducted between 1987 and 2019 at HFU, and between 1999 and 2019 at KFU.
The 2,145-ha KFU of the Refuge Complex was established in 1997 and located on the leeward slope of Mauna Loa on the south Kona coast, Hawai‘i Island (Figure 1). An additional 4,105 ha were added to KFU in phases starting in 2017. No forest bird surveys have been conducted in these new additions. We did not make inference to bird populations in the new acquisitions. We split KFU analyses into two strata: the upper stratum from 1,524 to 1,829 m elevation (701 ha) and the lower stratum from 1,067 to 1,524 m (819 ha). We chose 1,524 m as the division between strata because that is the elevation below which populations are likely to be affected by mosquito-borne disease (Atkinson and LaPointe 2009). The upper stratum consisted of dry, subalpine ‘ōhi‘a forest, mesic koa/ōhi‘a forest, and mesic ‘ōhi‘a forest (Jacobi 1989). The lower stratum was dominated by wet, montane ‘ōhi‘a forest with areas of mesic ‘ōhi‘a forest at upper elevations and koa/‘ōhi‘a forest at lower elevations (Jacobi 1989). Nonnative plant species occurred throughout KFU, were most abundant at the lowest elevations, and included clidemia Clidemia hirta, strawberry guava Psidium cattleianum, and Christmas berry Schinus terebinthifolius. There were wild cattle and feral pigs in both strata of KFU.
Bird species
Eight and seven native passerine species persist in HFU and KFU, respectively. The Hawai‘i ‘elepaio is a monarch flycatcher (Monarchidae), and the ‘ōma‘o is a thrush (Turdidae). ‘Ōma‘o was extirpated from the leeward side of Hawai‘i Island (Judge et al. 2012; Wakelee and Fancy 2020) prior to Refuge Complex establishment and not detected at KFU during our study. Six Hawaiian honeycreepers (Fringillidae: Carduelinae) include the ‘akiapōlā‘au, Hawai‘i ‘amakihi, ‘alawī, Hawai‘i ‘ākepa, ‘i‘iwi, and ‘apapane. Hawai‘i ‘elepaio, ‘ōma‘o, ‘akiapōlā‘au, ‘alawī, and Hawai‘i ‘ākepa are endemic to Hawai‘i Island. Point-transect distance sampling survey methods (see below) were not suitable for estimating densities of ‘io or Hawaiian hawk Buteo solitarius, pueo or short-eared owl Asio flammeus, and nēnē or Hawaiian goose Branta sandvicensis. We do not present data for those species. In addition, 26 species of introduced birds have been observed at the Refuge Complex, but only two species, red-billed leiothrix Leiothrix lutea and warbling white-eye Zosterops japonicus, existed in densities large enough to estimate abundance and trends.
Bird sampling
The HFBS sampled the area that is now HFU in 1977 along three transects spaced about 3 km apart with 95 stations at 134-m intervals (Scott et al. 1986). In 1987, a new, overlapping, but separate series of 14 transects were established (Figure 1). The transect ranged in length from 1,750 to 5,000 m with 232 stations (10 to 29 stations/transect). Distances among transects were either 500 or 1,000 m and the distance among stations varied between 134 and 250 m. In 1999, 9 of the 14 transects were expanded to include lower elevation areas, in the closed-forest stratum, with an additional 111 stations. Transect 64 from the HFBS on southwest Hawai’i Island was sampled in 1978 and passed through the middle of KFU. In 1995, four new transects were established at KFU. These transects occurred in proximity to transect 64 but did not include that transect. Each transect was approximately 7,820 m long with 52 to 53 stations (215 total stations), spaced about 150 m apart (Figure 1). Stations at elevations lower than 1,067 m were periodically surveyed, but not frequent enough to be included in our analysis.
All surveys followed the point-transect distance sampling procedures implemented by Scott et al. (1986). Observers traversed transects to stop at predetermined points (also called stations) to conduct bird counts for an 8-min period. Observers received presurvey training to calibrate for distance estimation and learn bird vocalizations to minimize variability among observers and standardize for local conditions (Kepler and Scott 1981; Scott et al. 1986; Verner and Milne 1989). Observers recorded the detection type (heard, seen, or both) and horizontal distance from the station center point to individual birds detected. Observers also recorded cloud cover, rain intensity, wind strength on the Beaufort scale, gust strength on the Beaufort scale, and time of day at each station. Sampling was typically conducted between dawn and 1100 hours and halted when heavy rain or strong winds or gusts (>20 km/h) made observations difficult.
The HFBS was conducted in July, after most forest birds had finished breeding and were less vocal with many birds having dispersed in search of seasonal food sources (Scott et al. 1986; Ralph and Fancy 1994,; 1995; Smetzer et al. 2021). Surveys since HFBS (i.e., starting in 1987) were conducted mainly during March and April to correspond with peak breeding season and vocalization rates. To minimize bias associated with varying detectability throughout the year caused by the disparity in the months sampled and great difference in number and coverage of stations sampled, we excluded the HFBS data from our analyses. Data used in this study are available in Supplemental Material (Data S1–S3).
Density estimation
We used distance sampling to correct abundance estimates for individuals that go undetected following methods described in Buckland et al. (2001, 2004, 2015). We estimated a detection probability by fitting a detection function to the distance measurements, which we then used to compute density of birds (birds/ha). Distance sampling relies on the assumptions that all birds are detected with certainty at the station center point, birds are detected prior to any movement, and distances are measured without error. Buckland et al. (2001, 2004, 2015) describe distance sampling procedures and analyses in detail.
We estimated the total annual abundances by multiplying the strata-specific density by the area of each stratum. Species- and strata-specific densities were estimated using the Program R (R Core Team 2021) package Distance (version 1.0.3; Miller et al. 2019) in a stepwise process. We used conventional distance sampling methods to select the appropriate truncation, key detection function model, and fit adjustment terms. We truncated data at a distance where detection probability, using a preliminary half-normal detection function model, was about 0.1. This procedure facilitates modeling by deleting outliers and reducing the number of adjustment parameters needed to modify the detection function. Candidate key detection function models were limited to half-normal and hazard-rate functions. The selected key model had the lowest Akaike’s Information Criterion (AIC) value (Buckland et al. 2015). We applied adjustment terms to the selected key model, limited to expansion series of order two (Buckland et al. 2001:361, 365), where cosine and Hermite polynomial adjustment series were paired with the half-normal model, and cosine and simple polynomial adjustment terms were paired with the hazard-rate model.
To improve estimator precision, we incorporated sampling covariates to the selected key detection function model in the multiple covariate distance sampling (MCDS) engine of package Distance. We modeled covariates only for species with >50 strata-specific detections (Thomas et al. 2010). Covariates included cloud cover (in two categories: <50% cloud cover and ≥50% cloud cover), rain intensity (0–3 by integers), wind strength on the Beaufort scale (0–4 by integers), observer (see below), time of detection, type of detection (heard only, seen, or heard first then seen), and year (see below). We also included combinations of detection type covariates where birds that were heard first and then seen were pooled with birds that were (a) only seen or (b) birds only heard. We fit each detectability model in the candidate set to each species by strata.
A concern when monitoring birds over long periods is that bird detectability may differ among surveys as a result of turnover among observers, and environmental and climatological differences. Participation from many observers over the long time series led us to evaluate whether observers had an effect on the detection probability by modeling observer as a covariate. Many observers participated in relatively few surveys, so we combined those observers with <100 species-specific detections in a stratum at HFU; at KFU, we combined observers with <25 species-specific detections in a stratum. We grouped observers with similar coefficients and modeled the observer-pooled covariate in multiple covariate distance sampling. To capture environmental and climatological differences in annual detection probabilities, we developed a time-specific detectability covariate by combining years with similar coefficients and modeled the year-pooled covariate in multiple covariate distance sampling.
We estimated species- and strata-specific densities, and derived abundances from the model with the lowest AIC value. We evaluated the selected model by visual inspection of diagnostic plots and checking goodness of fit using an unweighted Cramér-von Mises test (Buckland et al. 2015). We used bootstrap methods to calculate species- and strata-specific 95% confidence interval (95% CI), standard error (SE), and coefficient of variation (CV). We used the delta method described in Thomas et al. (2010) to compute annual unit-wide density and variance estimates by combining area-weighted strata and unit estimates. These total annual population estimates were calculated only for years when all strata within a unit were surveyed. In addition, we computed the 2019 Refuge Complex-wide density and variance estimates using the same approach.
Trend detection
We evaluated trends in densities by regressing through the posterior distribution of densities estimated using a state-space model. State-space models consist of a state model and a measurement model. The state model describes the change of the state parameter, the true population size, from the previous time to the current time and reflects the stochastic subprocesses driving population change. The measurement model describes the relationship between the observations (i.e., annual abundance estimate corrected for detection probability and the state parameter). State-space models improve trend detection compared with simple regression models by differentiating between state process variation and measurement error. Process variation is the naturally occurring variation in abundances due to changing demographic and environmental conditions. Measurement error results from changing sampling conditions, availability of birds to be sampled, and stochastic measurement error. By partitioning the error into its component parts of process variation and measurement error, more precise abundance estimates can be produced and biologically implausible fluctuations eliminated (Camp 2021). We applied these time-varying rate of population change state-space model to the time series from 1987–2019 and 2010–2019 for HFU and 1999–2019 time series for KFU. At HFU, we analyzed the past decade of surveys (2010–2019) to describe recent short-term trajectories in the long-term trend in abundances. We felt this was important because environmental change and declines in Hawaiian forest bird populations have accelerated in the past two decades (Paxton et al. 2022). This time period also coincides with decadal management cycles and assists the Refuge Complex staff in developing and making resource-management strategies and decisions (USFWS 2010).











We fit a log-linear regression to each of the state-space model posterior annual density estimate iterations and computed the proportion of slopes in an equivalency framework (Camp et al. 2008). We used equivalence tests to distinguish between cases of a negligible trend from the inability to statistically detect a trend. We set equivalence bounds to identify a 25% change in the population over 25 y, or a −0.0119 and 0.0093 annual rate of change. We interpret the likelihood of a trend based on the evidence of the posterior odds as weak (P < 0.5), moderate (0.5 ≤ P < 0.7), strong (0.7 ≤ P < 0.9), and very strong (P ≥ 0.9). We defined trends as upward (>0.0093), downward (<−0.0119), stable, or inconclusive. A stable trend occurred when the slope was within the equivalence region. An inconclusive result occurred when P < 0.5 in all three upward, downward, and stable trend categories, which was a result of imprecise densities.
Results
Surveys were conducted at HFU each year between 1987 and 2019 with only one year (2009) not surveyed (Table S1, Supplemental Material). The number of stations surveyed over the study period ranged from 194 to 343 per year, with a total of 8,368 stations surveyed (Table S1, Supplemental Material). The open-forest stratum was surveyed the longest, since 1987, with 136 to 204 stations surveyed annually. The closed-forest stratum was surveyed since 1999 with 61 to 113 stations annually. The afforested-pasture stratum was surveyed intermittently since 1987 with 34 to 54 stations surveyed annually. We did not use survey data from 1992 to 1995 in the afforested-pasture stratum in our analyses because too few stations (n = 2) were sampled in those years. Surveys at HFU were primarily conducted from mid-March to early April (Table S1, Supplemental Material).
Surveys were conducted intermittently at KFU since 1995 (Table S1, Supplemental Material). Sampling in the lower stratum occurred in 1995, 1999, 2000, 2009, 2012, 2013, 2015, 2018, and 2019. The upper stratum was surveyed in the same years, plus 2001, 2005, and 2006. We used survey data from 1995 to estimate population abundance but not trend because the large gap between the 1995 and 1999 surveys means the 1995 estimates would strongly influence trends. There were 71 stations in the upper stratum and 78 stations in the lower stratum. A total of 1,545 stations was surveyed at KFU, primarily in late March to mid-April.
In 2019 at HFU, Hawai‘i ‘amakihi was the most frequently detected species, occurring on 95% of the stations with a relative abundance of 4.06 birds/station (Table S2, Supplemental Material). Other species frequently detected included ‘i‘iwi (92% of stations, 6.24 birds/station), ‘apapane (91% of stations, 5.15 birds/station), ‘ōma‘o (85% of stations, 2.06 birds/station), red-billed leiothrix (82% of stations, 1.79 birds/station), warbling white-eye (74% of stations, 1.80 birds/station), and Hawai‘i ‘elepaio (59% of stations, 0.91 birds/station). The frequency of detections for the three endangered species was ‘alawī on 34% of stations, 0.57 birds/station; Hawai‘i ‘ākepa on 26% of stations, 0.40 birds/station; and ‘akiapōlā‘au on 10% of stations, 0.12 birds/station. Northern cardinal Cardinalis cardinalis occurred on 23% of stations with 0.32 birds/station, and Japanese bush-warbler Horornis diphone occurred on 12% of the stations with 0.16 birds/station. Japanese bush-warbler was not detected at HFU until 2010; since 2015, ≥25 birds were detected in several years. During the entire study period, 69 ‘akiapōlā‘au were detected in the afforested-pasture stratum at HFU, fewer than the >75–100 detections recommended for distance sampling by Buckland et al. (2001). We included ‘akiapōlā‘au in the analyses because it is one of the endangered species found at HFU and close to the recommended number of detections to reliably estimate densities. There were only 12 detections of Hawai‘i ‘ākepa in the afforested-pasture; we did not estimate their densities in this stratum.
In 2019 at KFU, ‘apapane (99% of stations, 17.16 birds/station) was the most frequently detected species. In all other years, except 2018 and 2019, it occurred at 100% of the stations (Table S2, Supplemental Material). Other species frequently detected included Hawai‘i ‘amakihi (86% of stations, 3.59 birds/station), warbling white-eye (71% of stations, 2.06 birds/station), ‘i‘iwi (68% of stations, 2.10 birds/station), Hawai‘i ‘elepaio (38% of stations, 0.55 birds/station), red-billed leiothrix (38% of stations, 0.56 birds/station), and northern cardinal (36% of stations, 0.46 bird/station). During the entire study period, ‘akiapōlā‘au (n = 1), ‘alawī (n = 68), and Hawai‘i ‘ākepa (n = 60) were observed at KFU, but received too few detections per stratum to reliably estimate abundances. ‘Ōma‘o was not detected at KFU.
Detection function modeling
The best-fit detection function models selected were either the key model or key model with covariates; however, no key models including adjustment terms were selected to estimate forest bird densities at either Refuge Complex unit (Tables S3–S6, Figures S1–S2, Supplemental Material). All but one model at KFU and all but five of the models at HFU had detection type or detection types grouped as a covariate. Two models at HFU had observer as a covariate: ‘ōma‘o and red-billed leiothrix in the afforested-pasture stratum. Three models at HFU and one at KFU included only the key detection function: ‘akiapōlā‘au in the afforested-pasture and closed-forest strata at HFU, ‘alawī in the afforested-pasture stratum at HFU, and red-billed leiothrix at KFU.
Population estimates
Population estimates at HFU varied by species and through time. In 2019, ‘i‘iwi was the most abundant species with a density of 16.9 birds/ha (95% CI = 13.4–21.1) and 110,028 individuals (95% CI = 87,360–137,712; Tables S7 and S8, Supplemental Material). In 2019, densities of the three endangered species (‘alawī, Hawai‘i ‘ākepa, and ‘akiapōlā‘au) were <1.5 birds/ha, with fewer than 10,000 individuals of each species.
Density and population abundance also varied by strata at HFU. In 2019, ‘i‘iwi was the most abundant species in the open-forest and closed-forest strata, with 27.5 birds/ha (95%CI = 25.3–30.0) or 88,109 individuals (95% CI = 80,851–95,962) in the open forest and 14.9 birds/ha (95% CI = 12.8–17.4) or 29,811 individuals (95% CI = 25,569–34,656) in the closed forest (Tables S7, S8, Supplemental Material). In the afforested-pasture stratum, Hawai‘i ‘amakihi was the most abundant species with a density of 7.8 birds/ha (95% CI = 6.3–9.6) or 10,305 individuals (95% CI = 8,267–12,651; Tables S7, S8, Supplemental Material).
Bird abundance was greater in the open-forest stratum, the stratum with the largest area, than in either the closed-forest or the afforested-pasture strata (Table S8, Supplemental Material). Since 2008, the open-forest had the largest overall forest bird densities (Table S7, Supplemental Material). In all years, >50% of the population abundance of all species combined were in the open-forest stratum. The proportion of the population in this stratum increased through time as population abundance in the closed-forest stratum declined. In 2019, 64% of HFU forest birds were in the open-forest stratum, while 28% of the population occurred in the closed-forest stratum. Throughout the study period, only small proportions (<10% in most years) of populations were estimated in the afforested-pasture stratum. This stratum had lower densities compared with other strata (Table S7, Supplemental Material). However, abundance in this stratum increased through time, from 3% in 1999 to 8% in 2019.
Population abundance and densities were considerably different at KFU compared with HFU. In all years, ‘apapane was the most abundant species at KFU. In 2019, the total density of ‘apapane was 71.8 birds/ha (95% CI = 61.6–84.1) or 109,121 individuals (95% CI = 93,672–127,755; Tables S9, S10, Supplemental Material). ‘Apapane density in the upper stratum was 52.4 birds/ha (95% CI = 31.0–83.5) with 36,720 individuals (95% CI = 21,698–58,506) and 93.3 birds/ha (95% CI = 82.7–104.9) with 76,384 individuals (95% CI = 67,713–85,924) in the lower stratum.
At KFU, populations and densities were greater in the lower stratum than the upper stratum (Tables S9, S10, Supplemental Material). Across years, 54% to 62% of the population of all species were in the lower stratum. In 2019, 62% of KFU forest birds were in the lower stratum and 38% of the population occurred in the upper stratum.
Population trends
Most of the eight native and two nonnative species at HFU had stable, long-term (1987–2019) population trends in the open-forest stratum, with only Hawai‘i ‘ākepa showing a downward trend; ‘apapane and warbling white-eye had upward trends (Figure 2; Table 1; Table S11, Supplemental Material). However, in the closed-forest stratum all but three species had downward population trends. Of the three species, Hawai‘i ‘ākepa had an inconclusive trend, ‘apapane had a stable trend, and warbling white-eye had an upward trend (Figure 2; Table 1; Table S11, Supplemental Material). Although Hawai‘i ‘ākepa trends were inconclusive in the closed-forest stratum, they showed weak evidence for a downward trend (0.49) and nearly reached the 0.50 cutoff (Table S11, Supplemental Material). In the afforested-pasture, all species had upward long-term population trends.
Population trends, 1987–2019, for forest birds at Big Island National Wildlife Refuge Complex, Hakalau Forest Unit, Hawai‘i Island, Hawai‘i, USA. Black dots are the detection-corrected abundances estimated from program Distance with bootstrap-derived 95% confidence intervals (CIs) of the estimates (whisker bars). State-space population estimates are represented by the blue dots, and the gray ribbons are the 95%CIs. The blue lines indicate changes in the population. X-axes are stratum and panel specific.
Population trends, 1987–2019, for forest birds at Big Island National Wildlife Refuge Complex, Hakalau Forest Unit, Hawai‘i Island, Hawai‘i, USA. Black dots are the detection-corrected abundances estimated from program Distance with bootstrap-derived 95% confidence intervals (CIs) of the estimates (whisker bars). State-space population estimates are represented by the blue dots, and the gray ribbons are the 95%CIs. The blue lines indicate changes in the population. X-axes are stratum and panel specific.
Summary of trends in detection-corrected abundances of forest birds at Hakalau Forest Unit in open-forest, closed-forest, afforested-pasture, and combined (total) strata of the Big Island National Wildlife Refuge Complex, Hawai‘i Island, Hawai‘i, USA. Time period is 1987–2019 in the open-forest and afforested-pasture strata, and 1999–2019 in the closed-forest and total strata. The ecological relevance of a trend was based on a 25% change in relative abundance over 25 y.

Results were mixed for the combined strata estimates (Figure 2; Table 1; Table S11, Supplemental Material). Three of the native species, Hawai‘i ‘elepaio, ‘ōma‘o, and ‘i‘iwi, had stable populations. The posterior probabilities for Hawai‘i ‘amakihi were nearly equally divided between stable (0.48) and upward (0.50). Two of the endangered species, ‘alawī and Hawai‘i ‘ākepa, had downward population trends, and results were inconclusive for the third endangered species, ‘akiapōlā‘au. Warbling white-eye population trends were upward. Red-billed leiothrix trends were inconclusive.
Population trends were somewhat different for the most recent decade (2010–2019) compared with the long-term trend (1987–2019) at HFU. Three of the native species (‘alawī, Hawai‘i ‘ākepa, and ‘apapane) and the nonnative warbling white-eye had downward population trends in the open-forest stratum (Figure 2; Table 2; Table S12, Supplemental Material). Five species had stable populations; results were inconclusive for ‘akiapōlā‘au. In the recent decade, nearly all species, native and nonnative, had downward trends in the closed-forest stratum with only the ‘ōma‘o showing an upward trend. Three of the more common species at the Refuge Complex, Hawai‘i ‘amakihi, ‘apapane, and warbling white-eye, had downward population trends in the afforested-pasture stratum in the recent decade. Results for ‘i‘iwi were inconclusive, and five species had upward population trends in this stratum. Results were mixed for the combined strata estimates. ‘Akiapōlā‘au and ‘ōma‘o had upward trends, but ‘alawī, Hawai‘i ‘ākepa, ‘apapane, and warbling white-eye had downward trends. Results were inconclusive for Hawai‘i ‘elepaio, Hawai‘i ‘amakihi, ‘i‘iwi, and red-billed leiothrix. However, the evidence of the overall ‘i‘iwi population trend (0.49) was borderline for classification as downward.
Summary of short-term trends in detection-corrected abundances of forest birds at Hakalau Forest Unit in open-forest, closed-forest, pasture, and combined (total) areas, 2010–2019, of the Big Island National Wildlife Refuge Complex, Hawai‘i Island, Hawai‘i, USA. The ecological relevance of a trend was based on a 25% change in relative abundance over 25 y.

At KFU, Hawai‘i ‘elepaio had downward population trends (1999–2019) in the upper stratum and upward trends in the lower stratum (Figure 3; Table 3; Table S13, Supplemental Material). Trends were inconclusive for Hawai‘i ‘elepaio combining the two strata, but was borderline for classification as upward (0.49). Hawai‘i ‘amakihi were stable in both strata and overall. ‘I‘iwi trends were downward in the upper stratum, inconclusive in the lower stratum, and downward overall. ‘Apapane trends were upward in both strata and overall. Red-billed leiothrix trends were downward overall, upward in the upper stratum, and downward in the lower stratum. Warbling white-eye trends were inconclusive in the upper stratum and overall, but stable in the lower stratum. Irregular and infrequent sampling at KFU resulted in standard errors that were higher with more inconclusive or weak evidence for trends than at HFU where surveys were conducted consistently and frequently.
Population trends, 1999–2019, for forest birds at Big Island National Wildlife Refuge Complex, Kona Forest Unit, Hawai‘i Island, Hawai‘i, USA. Black dots are the detection-corrected abundances estimated from program Distance with bootstrap-derived 95% confidence intervals (CIs) of the estimates (whisker bars). State-space population estimates are represented by the blue dots, and the gray ribbons are the 95%CIs. The blue lines indicate changes in the population.
Population trends, 1999–2019, for forest birds at Big Island National Wildlife Refuge Complex, Kona Forest Unit, Hawai‘i Island, Hawai‘i, USA. Black dots are the detection-corrected abundances estimated from program Distance with bootstrap-derived 95% confidence intervals (CIs) of the estimates (whisker bars). State-space population estimates are represented by the blue dots, and the gray ribbons are the 95%CIs. The blue lines indicate changes in the population.
Summary of trends in detection-corrected abundances of forest birds at Kona Forest Unit in the upper stratum (<1,524 m), lower stratum (<1,524 m), and combined (total), 1999–2019, of the Big Island National Wildlife Refuge Complex, Hawai‘i Island, Hawai‘i, USA. The ecological relevance of a trend was based on a 25% change in relative abundance over 25 y.

Discussion
Forest bird population trends across both units of the Refuge Complex are showing considerable variation among strata and for both long-term and decadal timeframes. Some forest bird species continue to show upward or stable population trends; these trends were estimated at both Refuge Complex units previously by Gorresen et al. (2009) and Camp et al. (2010, 2016). We observed stable or upward, long-term (1987–2019) population trends for the combined strata estimates for Hawai‘i ‘elepaio, ‘ōma‘o, ‘i‘iwi, ‘apapane, and warbling white-eye at HFU (Table 1), and Hawai‘i ‘amakihi and ‘apapane at KFU (Table 3). However, our analyses found that populations of several species had more negative trends, particularly populations at lower elevations, during the entire study period, and in each stratum for the most recent decade at HFU. At HFU, we found that five species (‘ōma‘o, Hawai‘i ‘amakihi, ‘alawī, ‘i‘iwi, and red-billed leiothrix) went from having inconclusive population trends in the time period up to 2012 in the closed-forest stratum to having downward population trends with the addition of data through 2019. Two other species, which had upward population trends through 2012, changed to inconclusive (Hawai‘i ‘ākepa) or stable (‘apapane) trends in our analysis. Much of this change appears to be driven by population trends in the recent decade when all species except ‘ōma‘o had downward trends in the closed-forest stratum.
Trends also changed from the previous analyses in the open-forest stratum. Most species in the current analyses showed stable population trends in this stratum, but many of the current trends were less positive than trends observed between 1987 and 2012 (Camp et al. 2016), including for three of the listed species. ‘Alawī and ‘akiapōlā‘au went from having an upward population trend to having stable trends. In the recent decade (2010–2019), the trend for ‘alawī was downward and ‘akiapōlā‘au was inconclusive. Hawai‘i ‘ākepa went from an inconclusive trend between 1987 and 2012 to a downward trend in both the long-term and recent decade periods.
Camp et al. (2010) documented upward trends for Hawai‘i ‘amakihi, ‘i‘iwi, ‘apapane, warbling white-eye, and red-billed leiothrix in the afforested-pasture stratum between 1987 and 2008. Paxton et al. (2017) found all species, except ‘ākepa, had upward trends from 1987 to 2012 in the afforested-pasture stratum. Our results extended the time series through 2019 and corroborated these upward trends.
At KFU, we found Hawai‘i ‘elepaio trends to be inconclusive but trended upward (posterior probability of 0.49). In the lower stratum, the Hawai‘i ‘elepaio trend was upward during our study. In contrast, trends were previously inconclusive but trended downward (Camp et al. 2009). In the upper stratum of KFU, Hawai‘i ‘elepaio continued the downward trend previously observed. ‘Elepaio populations elsewhere on Hawai‘i Island are declining, particularly populations at higher elevations (Gorresen et al. 2009). ‘Elepaio populations on Kaua‘i Island are declining at the periphery of its range and its range is contracting upslope (Paxton et al. 2016, 2020). Camp et al. (2009) found Hawai‘i ‘amakihi population trends above and below 1,500 m to be inconclusive at KFU, whereas with additional data, we found that ‘amakihi populations were stable. ‘I‘iwi previously had stable populations in the upper stratum of KFU, but we found this population was trending downward. Our results were inconclusive for ‘i‘iwi in the lower stratum, which were declining through 2001 (Camp et al. 2009). Numbers of endangered forest birds at KFU remained very low with only a few individuals detected and only in the upper stratum. ‘Apapane trends were upward in both KFU strata. Camp et al. (2009) documented stable populations in both Pu‘u Wa‘awa‘a Forest Bird Sanctuary and South Kona areas since the HFBS, but densities at KFU increased nearly 10-fold from the 1978 survey to 1995–2001. Densities of ‘apapane at KFU continued to increase and in 2019 were >25 times those estimated in 1978. ‘Apapane trends on Kaua‘i and Maui were mixed. On Kaua‘i, trends were downward between 2000 and 2012, but densities in 2018 had increased and were similar to estimates in the 1990s and 2000s (Paxton et al. 2016, 2020). On Maui, densities between 2012 and 2017 were similar and differences in densities were statistically inconclusive (Judge et al. 2019).
Island-wide populations of the listed passerines, ‘akiapōlā‘au, ‘alawī, Hawai‘i ‘ākepa, and ‘i‘iwi, are showing downward trends, similar to those found at the Refuge Complex (Camp et al. 2009, 2010). The endangered ‘akiapōlā‘au densities vary widely among years and sites, trends were mixed, and population ranges were contracting (Camp et al. 2009). At KFU, we have not detected ‘akiapōlā‘au since one bird was observed in 1995. The species was not detected during surveys of other westside or southside sites (Camp et al. 2009; Pratt et al. 2010; Judge et al. 2017; A.W., Division of Forestry and Wildlife, pers. comm. 2022). Taken together, ‘akiapōlā‘au is likely extirpated from the Kona region. In the only other area where ‘akiapōlā‘au have recently been documented, Kulani–Keauhou Forests, densities were lower than at HFU in 2019, 0.095 to 0.103 birds/ha compared with 0.18 birds/ha, respectively (Camp et al. 2009). Trends were stable at Kulani–Keahhou Forests, while trends were inconclusive as a result of mixed results among the strata at HFU.
Few ‘alawī remain in the Kona district. We detected between 0 and 24 birds during KFU surveys. Similar numbers of birds were detected during north Kona surveys (0–22; Pratt et al. 2010) but 0 to 1 ‘alawī were detected during surveys in south Kona (Camp et al. 2009; Judge et al. 2017). Judge et al. (2017) detected 26 ‘alawī in forests on the southside of the Island. Densities have increased in southside forests recently, but the species has declined since HFBS and now are restricted to forests and woodlands >1,500-m elevation. Densities of ‘alawī in HFU in 2019 averaged 1.52 birds/ha compared with 0.16 birds/ha on the southside of the Island. While overall, long-term trends were downward at HFU, apparently driven by declines in the closed-forest stratum, populations in the open-forest stratum were stable and upward in the afforested-pasture stratum.
The endangered Hawai‘i ‘ākepa population has generally been stable since first surveyed by Scott et al. (1986; Judge et al. 2018). Historically, HFU has been a stronghold for this species, supporting some of the largest island-wide densities (Judge et al. 2018). Scott et al. (1986) estimated ‘ākepa densities approaching 3 birds/ha in open woodlands at Hawai‘i Volcanoes National Park, similar to the largest densities we estimated in the closed-forest stratum at HFU (Table S7, Supplemental Material). Despite downward trends in ‘ākepa populations at HFU, this unit had greater ‘ākepa densities than were reported in the other areas of the Island (Judge et al. 2018). We estimated the most recent density as 1.11 birds/ha, which is more than twice the densities of the other subpopulations (range 0.07 to 0.52 birds/ha; Judge et al. 2018). Previous densities of ‘ākepa at KFU ranged from 0 to 0.42 birds/ha (Camp et al. 2009); however, we had too few detections at KFU to estimate densities in our analysis.
A substantial portion (90%) of the threatened ‘i‘iwi global population occurs on Hawai‘i Island; the species is mostly declining or extirpated on the other islands (Paxton et al. 2013, 2016, 2020; Brinck 2020). The largest densities of ‘i‘iwi in their range continue to occur in HFU (Camp et al. 2009; Paxton et al. 2013), where we estimated stable long-term population trends. Reported densities elsewhere on the Island range from <1 bird/ha to <10 birds/ha (Camp et al. 2009; Paxton et al. 2013; Judge et al. 2018; Burnett et al. 2021). Population trends elsewhere on Hawai‘i Island were generally declining (Paxton et al. 2013). In our analysis, ‘i‘iwi populations were declining at KFU, but our estimated densities of 5–10 birds/ha exceed those found at most other areas on the Island (Camp et al. 2009; Paxton et al. 2013; Judge et al. 2018; Burnett et al. 2021). Only in the Kulani–Keauhou Forests were the ‘i‘iwi densities comparable to those at KFU (Camp et al. 2009; Paxton et al. 2013).
In 2019, ‘apapane densities at KFU were 71.79 birds/ha, much greater densities than observed in this area in 1978 (2.58–2.79 birds/ha) or 1999 (29.85–30.80 birds/ha; Camp et al. 2009). Previously the greatest densities of ‘apapane recorded were 33.35 birds/ha in the Kilauea summit in 1979 (Ralph and Fancy 1995). When the HFBS was conducted in 1978, ‘i‘iwi densities at KFU were greater than ‘apapane (Scott et al. 1986). While ‘apapane densities have dramatically increased at KFU, ‘i‘iwi densities have changed very little. ‘I‘iwi compete with ‘apapane and limit their access to resources (Fancy and Ralph 2020a; 2020b). The relatively low density of ‘i‘iwi at KFU may have allowed for the large increase of the ‘apapane populations (Smith et al. 1995). We did not observe a similar dramatic increase of ‘apapane populations at HFU, where ‘i‘iwi densities were two to three times greater than KFU. A similar pattern of high ‘apapane densities relative to ‘i‘iwi densities observed at KFU also occurred in the Ka‘ū forest, but this pattern has not occurred elsewhere on Hawai‘i Island (Camp et al. 2009; Burnett et al. 2021).
It is unclear what is driving the trends we observed for bird populations supported by the Refuge Complex. Many forest bird populations are declining throughout Hawai‘i and this trend is expected to continue or accelerate in a warming climate (Benning et al. 2002; Fortini et al. 2015; Liao et al. 2015). There is insufficient evidence to attribute recent declines to increased disease in HFU (LaPointe et al. 2016; D.L., U.S. Geological Survey, pers. comm. 2023; S.M., University of Hawai‘i at Hilo, pers. comm. 2023), but disease is expected to continue to be a threat to forest bird populations at the Refuge Complex. It could be that we observed natural fluctuations or oscillations in bird populations, but this seems unlikely because species with different life histories and foraging strategies experienced declines and there were differing trends among species with similar life histories and foraging strategies. For example, short-term trends in the closed-forest strata at HFU varied for the two frugivores, ‘ōma‘o had an upward trend but red-billed leiothrix had a downward trend. However, all insectivores and nectivores had downward trends. Conducting demographic studies along with continued mosquito and disease prevalence monitoring could identify whether disease or environmental change are affecting populations.
Conclusions
The Refuge Complex remains one of the more important conservation areas for forest birds in Hawai‘i (Gorresen et al. 2009; USFWS 2010). Comparing the 2019 Refuge Complex populations estimates to those presented in Gorresen et al. (2009), Paxton et al. (2013) and Judge et al. (2018), we found that HFU supports large proportions of the global populations of four listed forest bird species (Table 4). Management actions implemented thus far have been fundamental to the historic positive trends (Camp et al. 2010, 2016; Paxton et al. 2017), but it appears that additional stressors are having negative influences on forest bird populations resulting in the recent downward trends. Considering the importance of the Refuge Complex for conservation of forest birds in Hawai‘i, monitoring populations frequently and consistently may reveal trajectories and trends more quickly, and permit detecting smaller changes in population sizes. Monitoring abundances is retrospective, and effective conservation may be improved by more closely linking forest bird monitoring with surveys of habitat, invasive species, and other threats. In Supplemental Materials (Text S1) we discuss stressors on forest bird populations and management actions to address these stressors. Data from ongoing monitoring can help managers prioritize management actions to the benefit of forest birds at the Refuge Complex.
Population estimates (and 95% confidence intervals) for forest birds at the Big Island National Wildlife Refuge Complex, Hawai‘i Island, Hawai‘i, USA, in 2019; global and Hawai‘i Island population estimates for each species; and the proportion of the global population and Hawai‘i Island population at the Refuge Complex. Global and statewide estimates for Hawai‘i ‘ākepa Loxops coccineus are from Judge et al. (2018) and for ‘i‘iwi Vestiaria coccinea are from Paxton et al. (2013). Hawai‘i Island estimates for ‘apapane Himatione sanguinea are from Scott et al. (1986). Population estimates for the other species are from Gorresen et al. (2009).

Supplemental Material
Please note: The Journal of Fish and Wildlife Management is not responsible for the content or functionality of any supplemental material. Queries should be directed to the corresponding author for the article.
Text S1. Implemented and ongoing management actions to alleviate stressors on forest bird populations at the Hakalau and Kona Forest units of the Big Island National Wildlife Refuge Complex, Hawai‘i, 1987–2019.
Available: https://doi.org/10.3996/JFWM-22-035.S1 (960 KB DOCX)
Data S1. Metadata describing data from forest bird surveys at Hakalau and Kona Forest Units of the Big Island National Wildlife Refuge Complex, Hawai‘i, 1987–2019 used to estimate forest bird abundance and trends.
Available: https://doi.org/10.3996/JFWM-22-035.S2 (37 KB XML)
Data S2. Point-transect distance sampling data from forest bird surveys at Hakalau Forest Unit of the Big Island National Wildlife Refuge Complex, Hawai‘i, 1987 to 2019 used to estimate forest bird abundance and trends.
Available: https://doi.org/10.3996/JFWM-22-035.S3 (17.324 MB DOCX)
Data S3. Point-transect distance sampling data from Kona Forest Unit of the Big Island National Wildlife Refuge Complex, Hawai‘i from 1995 to 2019 forest bird surveys used to estimate forest bird abundance and trends.
Available: https://doi.org/10.3996/JFWM-22-035.S4 (3.127 MB XLSX)
Table S1. Forest bird point-count survey effort by year at the Hakalau and Kona Forest Units of the Big Island National Wildlife Refuge Complex, Hawai‘i, 1987–2019. NA = no surveys were conducted.
Available: https://doi.org/10.3996/JFWM-22-035.S5 (18 KB DOCX)
Table S2. Number of birds detected, indices of bird occurrence (occur; %), and relative abundance (birds/station) during point count surveys of forest birds at the Hakalau and Kona Forest units of the Big Island National Wildlife Refuge Complex, Hawai‘i, 1987 to 2019.
Available: https://doi.org/10.3996/JFWM-22-035.S6 (53 KB XLSX)
Table S3. Detection function parameters used in Program Distance to derive population densities of forest bird species calculated from forest bird point-count survey data from Hakalau Forest Unit of the Big Island National Wildlife Refuge Complex, Hawai‘i, 1987–2019. NA = insufficient data.
Available: https://doi.org/10.3996/JFWM-22-035.S7 (20 KB DOCX)
Table S4. Detection function parameters used in Program Distance to derive population densities of forest bird species calculated from forest bird point-count survey data from Kona Forest Unit of the Big Island National Wildlife Refuge Complex, Hawai‘i, 1995–2019.
Available: https://doi.org/10.3996/JFWM-22-035.S8 (17 KB DOCX)
Table S5. Models, number of parameters (param), Akaike’s Information Criterion (AIC), and change in AIC (ΔAIC) values used in Program Distance to estimate forest bird populations at Hakalau Forest Unit of the Big Island National Wildlife Refuge Complex, Hawai‘i, 1987–2019. Models that failed to converge, and models with covariates that had <50 observations for some variables were not included. Observations with missing covariates were not included in the analysis.
Available: https://doi.org/10.3996/JFWM-22-035.S9 (59 KB DOCX)
Table S6. Models, number of parameters (param), Akaike’s Information Criterion (AIC), and change in AIC (ΔAIC) values used in Program Distance to estimate forest bird populations at Kona Forest Unit of the Big Island National Wildlife Refuge Complex, Hawai‘i, 1987–2019. Models that failed to converge and models with covariates that had <50 observations for some variables were not included. Observations with missing covariates were not included in the analysis.
Available: https://doi.org/10.3996/JFWM-22-035.S10 (38 KB DOCX)
Table S7. Densities (birds/ha), standard error (SE), coefficient of variation (CV), and 95% confidence intervals (lower, upper) calculated with Program Distance for forest birds at Hakalau Forest Unit of the Big Island National Wildlife Refuge Complex, Hawai‘i, 1987–2019. Estimates are by strata and from state-space models.
Available: https://doi.org/10.3996/JFWM-22-035.S11 (142 KB DOCX)
Table S8. Population estimates, standard error (SE), coefficient of variation (CV), and 95% confidence intervals (lower, upper) calculated with Program Distance for forest birds at Hakalau Forest Unit of the Big Island National Wildlife Refuge Complex, Hawai‘i, 1987–2019. Estimates are by strata and from Distance and state-space models. - = surveys were not conducted.
Available: https://doi.org/10.3996/JFWM-22-035.S12 (270 KB DOCX)
Table S9. Densities (birds/ha), standard error (SE), coefficient of variation (CV), and 95% confidence intervals (lower, upper) calculated with Program Distance for forest birds at Kona Forest Unit of the Big Island National Wildlife Refuge Complex, Hawai‘i, 1995–2019. Estimates are by strata and from Distance for 1995 and state-space models for other years. - = surveys were not conducted.
Available: https://doi.org/10.3996/JFWM-22-035.S13 (91 KB DOCX)
Table S10. Population estimates, standard error (SE), coefficient of variation (CV), and 95% confidence intervals (lower, upper) calculated with Program Distance for forest birds at Kona Forest Unit of the Big Island National Wildlife Refuge Complex, Hawai‘i, 1995–2019. Estimates are by strata and from Distance and state-space models. - = surveys were not conducted.
Available: https://doi.org/10.3996/JFWM-22-035.S14 (126 KB DOCX)
Table S11. State-space model trends (β) with lower and upper 95% confidence intervals (CI) calculated from forest bird detection-corrected counts at Hakalau Forest Unit of the Big Island National Wildlife Refuge Complex, Hawai‘i, in open-forest (open), closed-forest (closed), afforested-pasture (pasture), and combined (total) strata. Posterior probability is the likelihood of a trend based on the evidence of the posterior odds. Time period is 1987–2019 in open-forest and afforested-pasture strata, and 1999–2019 in the closed-forest and total strata. The ecological relevance of a trend was based on a 25% change in relative abundance over 25 Years. Trend was interpreted as upward = ↑, stable = ↔, downward = ↓, or inconclusive = Inc. NA = insufficient data.
Available: https://doi.org/10.3996/JFWM-22-035.S15 (26 KB DOCX)
Table S12. State-space model short-term trends (β) with lower and upper 95% confidence intervals (CI) calculated from forest bird detection-corrected counts at Hakalau Forest Unit of the Big Island National Wildlife Refuge Complex, Hawai‘i, 2010–2019 in open-forest (open), closed-forest (closed), afforested-pasture (pasture), and combined (total) strata. Posterior probability is the likelihood of a trend based on the evidence of the posterior odds. The ecological relevance of a trend was based on a 25% change in relative abundance over 25 Years. Trend was interpreted as upward = ↑, stable = ↔, downward = ↓, or inconclusive = Inc. NA = insufficient data.
Available: https://doi.org/10.3996/JFWM-22-035.S16 (27 KB DOCX)
Table S13. State-space model trends (β) with lower and upper 95% confidence intervals (CI) calculated from detection-corrected counts of forest birds at Kona Forest Unit of the Big Island National Wildlife Refuge Complex, Hawai‘i, in upper (<1,524 m elevation), lower (<1,524 m elevation), and combined (total) strata, 1999–2019. Posterior probability is the likelihood of a trend based on the evidence of the posterior odds. The ecological relevance of a trend was based on a 25% change in relative abundance over 25 y. Trend was interpreted as upward = ↑, stable = ↔, downward = ↓, or inconclusive = Inc.
Available: https://doi.org/10.3996/JFWM-22-035.S17 (39 KB DOCX)
Figure S1. Detection probability and probability density from Program Distance of models used to estimate forest bird density at Hakalau Forest Unit of the Big Island National Wildlife Refuge Complex, Hawai‘i, 1987–2019. Solid lines are the detection function of the fitted model and dotted lines represent the detection function of each covariate factor.
Available: https://doi.org/10.3996/JFWM-22-035.S18 (730 KB DOCX)
Figure S2. Detection probability and probability density from Program Distance of models used to estimate forest bird density at Kona Forest Unit of the Big Island National Wildlife Refuge Complex, Hawai‘i, 1995–2019. Solid lines are the detection function of the fitted model and dotted lines represent the detection function of each covariate factor.
Available: https://doi.org/10.3996/JFWM-22-035.S19 (368 KB DOCX)
Reference S1. Banko PC, Peck R, Munstermann M, Jaenecke K. 2022. Host plant associations of Lepidoptera and implications for forest bird management at Hakalau Forest National Wildlife Refuge. Hawai‘i Cooperative Studies Unit Technical Report HCSU-104. Hilo: University of Hawai‘i at Hilo.
Available: https://doi.org/10.3996/JFWM-22-035.S20 (1.936 MB PDF) and http://hdl.handle.net/10790/5387
Reference S2. Brinck KW. 2020. Forest bird population trends within Haleakalā National Park. Hawai‘i Cooperative Studies Unit Technical Report HCSU-097. Hilo: University of Hawai‘i at Hilo.
Available: https://doi.org/10.3996/JFWM-22-035.S21 (5.371 MB PDF) and http://hdl.handle.net/10790/5382
Reference S3. Camp RJ, Gorresen PM, Pratt TK, Woodworth BL. 2009. Population trends of native Hawaiian forest birds, 1976–2008: the data and statistical analyses. Hawai‘i Cooperative Studies Unit Technical Report HCSU-012. Hilo: University of Hawai‘i at Hilo.
Available: https://doi.org/10.3996/JFWM-22-035.S22 (8.317 MB PDF) and http://hdl.handle.net/10790/2692
Reference S4. Cuddihy LW, Stone CP. 1990. Alteration of native Hawaiian vegetation; effects of humans, their activities and introductions. Hilo: Cooperative National Park Resources Studies Unit, University of Hawaii.
Available: https://doi.org/10.3996/JFWM-22-035.S23 (1.932 MB PDF) and https://www.semanticscholar.org/paper/Alteration-of-Native-Hawaiian-Vegetation%3A-Effects-Cuddihy-Stone/056ae7d1c489263e2cecfbafc2c73b24074a4991
Reference S5. [DHHL] Department of Hawaiian Homelands. 2009. ‘Aina Mauna Legacy Program. Kailua: Ho‘okuleana LLC.
Available: https://doi.org/10.3996/JFWM-22-035.S24 (21.431 MB PDF) and https://dhhl.hawaii.gov/wp-content/uploads/2011/05/Aina_Mauna_Legacy_Program_FINAL.pdf
Reference S6. [DOFAW] Department of Land and Natural Resources Division of Forestry and Wildlife. 2016. Laupāhoehoe Natural Area Reserve management plan.
Available: https://doi.org/10.3996/JFWM-22-035.S25 (7.276 MB PDF) and https://dlnr.hawaii.gov/ecosystems/files/2013/07/Laupahoehoe_mngt_plan_10192016_final-with-signature.small_.pdf
Reference S7. [DLNR] Department of Land and Natural Resources. 2020. Adaptation is key to overcoming challenges faced in ‘Alalā Recovery Program. Posted online Oct 5, 2020.
Available: https://doi.org/10.3996/JFWM-22-035.S26 (276 KB PDF) and https://dlnr.hawaii.gov/blog/2020/10/05/nr20-156/
Reference S8. [DLNR] Department of Land and Natural Resources. 2021. Next steps in ‘alalā recovery include Maui Nui & ‘io research. Posted on Mar 31, 2021.
Available: https://doi.org/10.3996/JFWM-22-035.S27 (254 KB PDF) and https://dlnr.hawaii.gov/blog/2021/03/31/nr21-061/
Reference S9. Jacobi JD. 1989. Vegetation maps of the upland plant communities on the Islands of Hawai‘i, Maui, Moloka‘i, and Lāna‘i. Technical Report 68. Honolulu: Cooperative National Park Resources Studies Unit, Department of Botany, University of Hawai‘i at Mānoa.
Available: https://doi.org/10.3996/JFWM-22-035.S28 (548 KB PDF) and http://hdl.handle.net/10125/4192
Reference S10. Jacobi JD. 2017. Vegetation map for the Hakalau Forest Unit of the Big Island National Wildlife Refuge Complex on the Island of Hawai‘i. Hawai‘i Cooperative Studies Unit Technical Report 84. Hilo: University of Hawai‘i at Hilo.
Available: https://doi.org/10.3996/JFWM-22-035.S29 (10.517 MB PDF) http://hdl.handle.net/10790/3300
Reference S11. Judge SW, Camp RJ, Sedgwick DE, Squibb CL, Hart PJ. 2017. Pacific Island landbird monitoring report, Hawai‘i Volcanoes National Park, 2015–2016: tract groups 1 and 2. Natural Resource Report NPS/PACN/NRR—2017/1407. Fort Collins, Colorado: National Park Service.
Available: https://doi.org/10.3996/JFWM-22-035.S30 (6.885 MB PDF) and https://pubs.er.usgs.gov/publication/70040383
Reference S12. Judge SW, Camp RJ, Warren CC, Berthold LK, Mounce HL, Hart PJ, Monello RJ. 2019. Pacific island landbird monitoring annual report, Haleakalā National Park and East Maui Island, 2017. Natural Resource Report NPS/PACN/NRR—2019/1949. Fort Collins, Colorado: National Park Service.
Available: https://doi.org/10.3996/JFWM-22-035.S31 (22.847 MB PDF) and https://doi.org/10.13140/RG.2.2.18347.80168
Reference S13. LaPointe DA, Gaudioso-Levita JM, Atkinson CT, Egan A, Hayes K. 2016. Changes in the prevalence of avian disease and mosquito vectors at Hakalau Forest National Wildlife Refuge: a 14-year perspective and assessment of future risk. Hawai‘i Cooperative Studies Unit Technical Report 73. Hilo: University of Hawai‘i at Hilo.
Available: https://doi.org/10.3996/JFWM-22-035.S32 (2.244 MB PDF) and http://hdl.handle.net/10790/2670
Reference S14. Leopold CR, Hess SC, Kendall SJ. 2015. Vegetation and non-native ungulate monitoring at the Big Island National Wildlife Refuge Complex 2010–2014. Technical Report 62. Hilo: Hawai‘i Cooperative Studies Unit, University of Hawai‘i at Hilo.
Available: https://doi.org/10.3996/JFWM-22-035.S33 (2.999 MB PDF) and http://hdl.handle.net/10790/2588
Reference S15. Paxton EH, Brinck KW, Crampton LH, Hite J, Costantini M. 2020. 2018 Kauai forest bird population estimates and trends. Hawai‘i Cooperative Studies Unit Technical Report HCSU-098. Hilo: University of Hawai‘i at Hilo.
Available: https://doi.org/10.3996/JFWM-22-035.S34 (1.917 MB PDF) and http://hdl.handle.net/10790/5507
Reference S16. Paxton EH, Gorresen PM, Camp RJ. 2013. Abundance, distribution, and population trends of the iconic Hawaiian honeycreeper, the ‘i‘iwi (Vestiaria coccinea) throughout the Hawaiian Islands. U.S. Geological Survey Open-File Report 2013-1150. Reston, Virginia: U.S. Geological Survey.
Available: https://doi.org/10.3996/JFWM-22-035.S35 (3.080 MB PDF) and https://pubs.usgs.gov/of/2013/1150/
Reference S17. Paxton EH, Laut M, Enomoto S, Bogardus M. 2022. Hawaiian forest bird conservation strategies for minimizing the risk of extinction: biological and biocultural considerations. Hawai‘i Cooperative Studies Unit Technical Report HCSU-103. Hilo: University of Hawai‘i at Hilo.
Available: https://doi.org/10.3996/JFWM-22-035.S36 (2.439 MB PDF) and http://hdl.handle.net/10790/5386
Reference S18. Peck RW, Banko PC, Stelmach M. 2014. Arthropod community structure on bark of koa (Acacia koa) and ‘o-hi‘a (Metrosideros polymorpha) at Hakalau Forest National Wildlife Refuge, Hawai‘i Island, Hawai‘i. Hawai‘i Cooperative Studies Unit Technical Report HCSU-050. Hilo: University of Hawai‘i at Hilo.
Available: https://doi.org/10.3996/JFWM-22-035.S37 (1.276 MB PDF) and http://hdl.handle.net/10790/2614
Reference S19. Spurr EB, Foote D, Lindsey GD, Perry CF. 2013. Efficacy of hand-broadcast application of diphacinone bait for rodent control in Hawaiian montane forests. Hawaii Cooperative Studies Unit Technical Report HCSU-043. Hilo: University of Hawai‘i at Hilo.
Available: https://doi.org/10.3996/JFWM-22-035.S38 (461 KB PDF) and http://hdl.handle.net/10790/2621
Acknowledgments
We thank the many trained bird counters who collected data over the years. We especially thank Jack Jeffrey, former Refuge Complex biologist, who led and organized many of the surveys. Thanks are extended to Kevin Brinck (Hawai‘i Cooperative Studies Unit, University of Hawai‘i at Hilo), Richard Camp, and Ayesha Genz for providing training in R-Distance and the scripts used as the basis for conducting our analyses. Funding for surveys was provided by U.S. Fish and Wildlife Service (USFWS)–Big Island National Wildlife Refuge Complex. Funding was provided by USFWS–Refuges Inventory and Monitoring Program to Steven Kendall and Rachel Rounds. Funding for Ayesha Genz was provided under Cooperative Agreement G21AC10163 for the U.S. Geological Survey Pacific Island Ecosystems Research Center, and by U.S. Geological Survey Ecosystems Mission Area Appropriation to Richard Camp for this analysis. We thank Sarah Nash, Eben Paxton, and Alex Wang for comments that improved our manuscript.
Any use of trade, product, website, or firm names in this publication is for descriptive purposes only and does not imply endorsement by the U.S. Government.
References
The findings and conclusions in this article are those of the author(s) and do not necessarily represent the views of the U.S. Fish and Wildlife Service.
Author notes
Citation: Kendall SJ, Rounds RA, Camp RJ, Genz AS, Cady T, Ball DL. 2023. Forest bird populations at the Big Island National Wildlife Refuge Complex, Hawai‘i. Journal of Fish and Wildlife Management 14(2):410–432; e1944-687X. https://doi.org/10.3996/JFWM-22-035