Research Paper

INTRODUCTION

Palila (Loxioides bailleui) are critically endangered, finch-billed Hawaiian honeycreepers (family Fringillidae), and the last of the seed specialists in this adaptive radiation remaining within the main Hawaiian Islands. Approximately 2.8 million years ago, Palila diverged from the ancestors of the Laysan Finch (Telespiza cantans) and Nihoa Finch (Telespiza ultima; Lerner et al. 2011), and colonized the dry forests of at least Kauaʻi, Oʻahu, and the Island of Hawaiʻi, based on subfossil and historical evidence (Olson and James 1982, Burney et al. 2001, Banko et al. 2020). They evolved to feed almost exclusively upon the seeds of endemic māmane trees (Sophora chrysophylla; family Fabaceae), which contain compounds likely to be toxic to most other species (Banko et al. 2002a). However, their current dependence on māmane seeds is markedly increased from a century ago, based on reconstructed diets showing greater exploitation of caterpillars (Van Houtan et al. 2024). Palila also occasionally take naio fruits (Myoporum sandwicense; family Scrophulariaceae) when māmane is scarce (Hess et al. 2014). Within historical times, likely due to conversion of dry forests to agriculture, Palila persisted only on the island of Hawaiʻi, whereas the Lesser Koa Finch (Rhodoacanthis flaviceps), Greater Koa Finch (Rhodoacanthis palmeri), and Kona Grosbeak (Chloridops kona), other seedeaters, quickly disappeared (Banko and Banko 2009). Today, Palila are restricted to subalpine dry forest (2000- to 3000-m elevation; Banko et al. 2020) on the southwest slope of Mauna Kea, where habitat conditions are rapidly changing because of interactions among climate change, fire occurrence, invasive grasses, and browsing by feral ungulates (Thaxton and Jacobi 2009, Banko et al. 2013, 2014).

By the late 18th century, Europeans introduced cattle (Bos taurus), domestic sheep (Ovis aries), and goats (Capra hircus), which browsed and degraded much of the remaining māmane and naio forest (Scowcroft 1983, Hess et al. 1999). Palila subsequently vanished from the slopes of Hualālai and Mauna Loa. A nearly successful attempt to eradicate feral livestock from Mauna Kea in the 1930s and 1940s evolved into a sustained game management plan by the 1950s (Juvik and Juvik 1984). Feral sheep were allowed to proliferate and were even augmented by mouflon (O. aries musimon), introduced in the 1960s (Juvik and Juvik 1984), which reversed any recent forest recovery (Reddy et al. 2012). This precipitated a series of lawsuits that eventually ruled that the State of Hawaii must remove feral sheep from Palila Critical Habitat (Juvik and Juvik 1984).

Despite the court’s orders, sheep remain in substantial numbers in Palila Critical Habitat (Banko et al. 2014), whereas the Palila population has declined by an estimated 89% between 1998 and 2021 (Genz et al. 2022). Reducing feral ungulate populations after 1980 promoted a moderate level of māmane regeneration (Hess et al. 1999, Reddy et al. 2012). However, Palila numbers oscillated throughout the 1980s and 1990s, likely tracking with droughts impacting the annual māmane seed-pod output (Jacobi et al. 1996, Lindsey et al. 1997, Gray et al. 1999, Hess et al. 2001). Palila then started to markedly decline around the turn of the 21st century, steadily from 2005 until 2010, and more slowly afterward, with the population estimate reaching an all-time low of 678 birds (95% Confidence Interval: 452-940) in 2021 (Leonard et al. 2008, Banko et al. 2009, Gorresen et al. 2009, Banko et al. 2013, Genz et al. 2022). A myriad of threats, including drought, fire, invasive mammalian predators, especially feral cats (Felis catus), invasive weeds, māmane fungal pathogens, and introduced insects that reduce food resources continue to have an impact on Palila survival and reproduction (Hughes et al. 1991, D’Antonio and Vitousek 1992, Jacobi et al. 1996, Hess et al. 1999, Gardner and Trujillo 2001, Hess et al. 2004, Oboyski et al. 2004, USFWS 2006, Banko et al. 2009, 2013, 2020).

These declines have occurred despite decades of ongoing management efforts. Currently, the State of Hawaii Department of Land and Natural Resources, Division of Forestry and Wildlife (DLNR DOFAW), manages Palila Critical Habitat through sheep eradication efforts, critical habitat fencing, forest restoration, predator control, and fire prevention (https://dlnr.hawaii.gov/restoremaunakea/management/). Public hunting and aerial shooting are the primary strategies to eliminate ungulates. As of 2024 60 miles (96 km) of 6 ft high (1.8 m) fencing have replaced the fence surrounding Palila Critical Habitat originally built by the Civilian Conservation Corps in 1937. Mauna Kea Forest Restoration Project (MKFRP; https://dlnr.hawaii.gov/restoremaunakea/management/forest-restoration/), a partnership between the DLNR and Pacific Cooperative Studies Unit, University of Hawaiʻi at Mānoa, oversee outplanting of māmane and other dry forest plants in the Kaʻohe (Figure 1) and Puʻu Mali Restoration Areas, as well as other sites throughout Palila Critical Habitat, and monitor natural regeneration and survival of māmane. MKFRP has planted over 40,000 trees and restored over 500 acres of forest since 2017. They also maintain a predator control grid targeted at feral cats. To mitigate fire risk, the DLNR maintains roads and firebreaks, whereas MKFRP removes fountain grass (Cenchrus setaceus) from restoration areas and roadsides. In 2016 DLNR installed a 40,000-gallon diptank within Palila Critical Habitat for fire control.

In addition to management, DLNR and MKFRP partner with other agencies, including the Three Mountain Alliance, U.S. Geological Survey, U.S. Fish and Wildlife Service, American Bird Conservancy, and National Park Service to conduct annual and, as of 2022, also quarterly, Palila surveys. This long-term monitoring has allowed for detection and response to changes in the Palila population. Our study provides the most recent update on the status and trend of the Palila population since 1998 to inform conservation and management decisions, based on new annual survey data from 2022, 2023, and 2024. We hypothesized that continuing, decades-long efforts to suppress feral sheep populations, increase and maintain fencing, outplant māmane, remove invasive predators, and mitigate fire threats would have restored habitat and reduced threats sufficiently to allow the Palila population trajectory to change since the all-time low in 2021 (Banko et al. 2014, 2020).

METHODS

Study area & bird sampling

Since 1980, 95% of the Palila population has inhabited a 64.4 km² area on the southwestern slope of Mauna Kea volcano, representing only about 5% of its historical range (Banko et al. 2013), within the federally designated Palila Critical Habitat (USFWS 1977, Scott et al. 1984, Banko et al. 2013, Camp et al. 2014; Figure 1). We refer to this area hereafter as the core survey area. Palila follow the seasonal abundance of māmane pods along an elevational gradient within this core survey area (Jacobi et al. 1996, Hess et al. 2001). Whereas the Palila population distribution shifts with its primary food source, its seasonal ranges overlap extensively. Therefore, its winter range area, which is surveyed annually in January or February, provides a stable and representative basis for evaluating population abundance and trends.

In 1980 transects spanning the core survey area and extending from the tree line (approximately 3000 m) downslope were established based on the historical Palila range and information from a 1975 survey (Van Riper III 1978, Scott et al. 1984). Additional transects were added in 1998 to improve population-estimation precision and provide a more comprehensive coverage of Palila distribution during the survey period (Figure 1; Johnson et al. 2006). Consequently, this analysis is restricted to the timespan of the annual survey effort from 1998 through 2024.

During the three most recent surveys, 25 January to 17 February 2022, 2 to 13 February 2023, and 30 January to 15 February 2024, 13 bird-survey transects inside the core survey area (transects 101–108 and 122–126; Figure 1) were surveyed one or more times. In addition to the core survey area, supplemental transects to the east of the core range were surveyed in all three years to look for possible range expansion. These included transects 109–121 in 2022, 109 in 2023, and 109 and 110 in 2024. Within the Kaʻohe mitigation area (Banko et al. 2009; Figure 1), surveys were conducted on supplemental sampling points, also called stations, to the lower portions of transects 102, 124, and 125 in 2022, and the lower portions of transects 101, 102, 124, and 125 in 2023 and 2024.

Within the core survey area, the 2022 survey consisted of 417 counts at 417 sampling points, the 2023 survey consisted of 415 counts at 408 points, and the 2024 survey consisted of 396 counts at 391 points (Figure 1; Table S1, Appendix 1). The adjacent, supplemental points outside of the core survey area were each counted once in 2022 at 325 counts, in 2023 at 78 counts, and in 2024 at 89 counts.

Since the introduction of transects 122–126 in 1998, Palila abundance estimates were produced by U.S. Geological Survey (USGS) and partners (Camp and Banko 2012, Camp et al. 2014 and 2016, Genz et al. 2018 and 2022). Prior to 2008 surveys were conducted mountain-wide, but the lack of detections outside the southwestern slope led to focusing effort on the core survey area with the intent to survey the entire mountain every five years starting in 2012 (D. L. Leonard Jr., State of Hawaii, Department of Land and Natural Resources Division of Forestry and Wildlife, 2012, personal communication).

Surveys were conducted using point-transect distance sampling to estimate Palila abundance and range. Most forest bird surveys in the Hawaiian Islands last eight minutes (Camp et al. 2009). However, Palila counts last six minutes because their woodland habitat is more open than mesic and wet-forest habitats, allowing for easier and more rapid detection. Prior to surveys, observers were trained in aural and visual bird identification and calibrated in their distance estimation and weather assessment to minimize bias and standardize data collection of local environmental conditions, following the protocols of Kepler and Scott (1981), Scott et al. (1986), and Verner and Milne (1989), updated and expanded by Camp et al. (2011). When available, laser rangefinders were distributed to aid in distance calibration and for verification during surveys.

Counts commenced at sunrise and continued up to four hours, concluding at approximately 11:00 HST. During each count, trained and calibrated observers recorded the species, detection type, i.e., heard, seen, or both, and horizontal distance of each bird from the observer, recorded as the exact distance rounded to the nearest meter. Time of sampling and weather conditions, i.e., cloud cover, rain intensity, wind strength, and wind gust strength (hereafter gust strength), were also recorded, and surveying was postponed when conditions hindered the ability to detect birds (wind and gust strengths > 20 km/h or heavy rain).

Abundance estimation

Distance analysis fits a detection function to estimate the probability of detecting a bird at a given distance from the observer, with the probability of detection decreasing with increasing distance. Previous analysis of Palila surveys demonstrated that the probability of detection was affected not only by distance but also by covariates such as detection type, sampling conditions, time of day, and survey year (Genz et al. 2022). A detection function was therefore fitted to models with and without these covariates as sample size post-stratification by year allowed (Thomas et al. 2010). Modeling covariates gives better precision, assuming the same key function for all factors (Buckland et al. 2015). Pooling covariates that have a similarly shaped detection function, based on similar coefficient values, can improve model fit, substantially reduce computation time, avoid overfitting, and further increase precision. Detection type was evaluated as heard only (detection type = 1) versus seen first (detection type = 2) or heard first and then seen (detection type = 4), where seen-first detections were pooled with heard-then-seen detections (detection type 1 versus 2 and 4). For weather conditions, wind strength was pooled into counts with light wind (Beaufort scale 0 to 1) or heavy wind (2 or greater). Gust strength was similarly pooled into counts with light gusts or heavy gusts using the same scale. We also included time since the earliest survey start, because time of day may affect bird activity levels and therefore detectability. Survey years were included both as discrete factors and as five blocks of pooled years, based on similar parameter coefficients (Figure S1, Appendix 1). With each additional year of data, estimates of these effects become more precise and the improved detection function may cause population estimates of previous years to change slightly from previous analyses.

Density estimates (birds per km²) were calculated from point-transect distance sampling data using the R package (R Core Team 2024) Distance, Version 1.0.9 (Miller et al. 2019). Statistically, our sample units were the random transects, because survey points were dependent upon the location of previous points within transects. For several years, sampling points were visited more than once within a sampling period to reduce estimator variance, i.e., reduce the amount of within-site variability. This sampling scheme, however, does not fit with the constraints imposed on bootstrap processing of Distance in R while keeping transect as the sample unit. Therefore, we chose survey points as our sampling unit, which could have underestimated the variances around estimates. Population abundance estimates were the product of the density estimate times the area of the core survey area (64.4 km²). The 2022 to 2024 data were pooled with detections from previous surveys since 1998. Candidate models were limited to half-normal and hazard-rate detection functions with expansion series of order two, following the standard protocol of Buckland et al. (2001) for point-transect distance sampling, where half-normal is paired with cosine and Hermite polynomial adjustments, and hazard rate is paired with cosine and simple polynomial adjustments. Survey effort in a given year was adjusted by the number of times the station was counted in that year. To improve model precision, potential sampling covariates were incorporated in the multiple covariate distance sampling engine of Distance. Covariates included the weather conditions, time of sampling, type of detection, and year of survey, and were only fit with the best-fit key function (Buckland et al. 2001). The right-tail truncation for the 1998 to 2018 data was 87.5 m (Genz et al. 2018), the distance where the detection probability was 10% in the non-truncated model. The 10% truncation distance in this analysis was 90 m, which is a so-called heaping distance, i.e., rounding of distances to favored values. Because Buckland et al. (2015) state that a heaping distance should not be selected as a truncation distance, we chose the previous truncation distance (Genz et al. 2022) of 87.5 m. Truncation facilitates modeling by deleting outliers and reducing the number of adjustment parameters needed to modify the detection function. The detection probability model selected was the one having the lowest Akaike’s information criterion (AIC; Buckland et al. 2001, Burnham and Anderson 2002). Visual inspection of diagnostic plots was conducted, and model fit was evaluated with a Cramér-von Mises test (Buckland et al. 2015). Annual population densities for each survey were calculated using the global detection function, i.e., the most parsimonious model, and the pooled data were post-stratified by year and location, i.e., inside/outside core survey areas. The 95% confidence intervals for the annual density estimates were derived from the 2.5th and 97.5th percentiles using bootstrap methods in Distance for 1000 iterations (Buckland et al. 2001, Thomas et al. 2010).

Trend detection

We assessed the trend in Palila abundance in two different ways. First, we used the bootstrap sample estimates generated to evaluate uncertainty in population abundance, to approximate the long-term population trend, i.e., 1998 to 2024, with a log-linear regression model. We fitted a log-linear regression to each of the 999 bootstrap estimates of the population and computed regression intercepts, slopes, and 95% confidence intervals using 2.5 and 97.5 percentiles in R. We evaluated the trend over four time periods to determine whether the direction and/or strength of the trend differed between shorter versus longer timeframes: the full modern time span of 26 years, the most recent 15 years, the most recent 10 years, and the most recent five years. The evidence of a trend was derived from the bootstrap distribution of slopes following Camp et al. (2015), where a meaningful trend in the population was considered a 25% change over 25 years (log values -0.119 and 0.0093). Diagnostics demonstrated that the log-linear regressions of trends met model assumptions according to visual inspection of residual plots. Shapiro-Wilk normality test (w = 0.90, P = 0.011) for the complete 26-year time series indicated that the abundance estimates were not normally distributed; however the 15-, 10-, and five-year time series estimates did have normal distributions. Temporal autocorrelation was evident across all four of the time series.

Because of the temporal autocorrelation, we also used a Bayesian state-space model on the log-scale, which provided an inherently auto-correlated alternative-trends assessment. A state-space model partitions model uncertainty into portions attributable to observation error, due to random noise in the environment affecting detectability and measurement, and process error, due to stochastic fluctuations in population outside of the overall trend. Such a state-space model can be interpreted as a biologically informed smoother and provides annual estimates consistent with the observed inter-annual noise. We used diffuse priors for the model parameters: a normal distribution with mean zero and standard deviation of 10 for the annual mean population change and exponential priors with mean 0.1 as priors for the standard deviation of slope and the observation error.

The state-space model was fit using the rjags package, Version 4-16 (Plummer 2003), which uses a Markov chain Monte Carlo algorithm. The trends were centered on the mean bootstrap abundance estimate across the time series to improve model convergence. The model parameters were estimated from 1,010,000 iterations for each of six chains, i.e., model runs, after first discarding 10,000 iterations as a warm-up period. The six chains were pooled, i.e., 6,000,000 total iterations, to calculate the posterior distribution. Gelman-Rubin convergence statistics for all estimated parameters were well below 1.01, which is less than the 1.1 threshold that indicates convergence (Gelman and Rubin 1992).

Both methods were assessed in an equivalence-testing approach (Camp et al. 2008) using the observed distribution of slopes of the bootstrap distribution and the posterior distribution of the slopes from the Bayesian state-space model. We chose biologically meaningful thresholds for the overall population trend as a 25% change in the population over a 25-year period, with annual rate of change equal to -0.0119 and 0.0093 (for decreasing and increasing) on the log-scale. A biologically meaningful trend occurs when the posterior probability distribution of the slope lies outside the equivalence region, whereas a negligible trend occurs when the slope is within the equivalence region. An inconclusive result occurs when small sample size and high variation in estimates results in the posterior distribution of the slope providing weak evidence in the three outcomes, i.e., increasing, stable, and decreasing. The strength of evidence for a trend was based on the distribution defined as weak, moderate, strong, or very strong based on the percentage of bootstrap slopes in each category: weak if < 50%, moderate if between 50% and 70%, strong if between 70% and 90%, and very strong if > 90%.

Finally, to test our hypothesis of whether the Palila population changed after 2021, we conducted a Bayesian change-point analysis on the estimates produced from distance sampling using the bcp package in R, Version 1.8.4 (Erdman and Emerson 2007). Change-point analysis helps to identify differences in mean abundances based on the posterior probability of a change in value at each point in the time series.

Sampling condition evaluation

The sampling conditions, particularly weather that could hinder or prohibit detecting birds, can adversely affect population estimates. We compared the distributions and probability densities of weather covariates by survey year using violin plots. Violin plots are similar to box plots in that they depict the summary statistics of mean, median, and interquartile ranges, but also display the full distribution of the data and include a kernel density to show structures in the data. Assessments were evaluated visually.

RESULTS

Bird sampling

Within the 64.4-km² core survey area of the southwestern flank of Mauna Kea, the number of Palila detected decreased by 21.8% from 101 in 2021 to 79 in 2022. There was a 38.0% decrease in detections between 2022 and 2023, with 79 in 2022 and 49 in 2023, and a slight increase of 8.2% between 2023 and 2024, with 49 in 2023 and 53 in 2024 (Figure 2). No Palila were detected outside of the core survey area during annual counts between 2020 and 2024 (Table 1).

Abundance estimation

The best model, the hazard-rate detection function model with year block as the covariate, had the lowest AIC by more than 29 units (Table S2, Appendix 1). Inspection of the diagnostic plots, i.e., probability detection function and probability density, indicated that the model adequately fit the data (Figure S2, Appendix 1). The Cramér-von Mises test was non-significant at the alpha = 0.05 level (test statistic = 0.43, P = 0.060, α = 0.05) indicating that the detection function did not statistically differ from the distance histogram.

In 2022 the Palila population in the core survey area was estimated within a confidence interval (CI) of 367 to 742, with a point estimate of 545 (Table 1, Figure 3). In 2023 the Palila population in the core survey area was estimated within a CI of 374 to 842 birds, with a point estimate of 596. In 2024 the Palila population in the core survey area was estimated within a CI of 412 to 970 birds, with a point estimate of 666. Estimates for previous years were within the 95% CIs of a previous analysis (Genz et al. 2022).

Trend detection

Between 1998 and 2005 Palila numbers fluctuated moderately between 4000 and 6800 except for an unusually low estimate in 2000 (Table S1, Appendix 1; Figure 3). A steep decline began in 2006 and continued steadily through 2010. For the next decade, the population decline appears to have leveled off as estimates fluctuated moderately (average CV = 0.14) around 1000 birds, with a local peak in 2019. The population then sharply declined again to 679 birds in 2021 and fluctuated around 600 birds through 2024. The observed mean decline during 1998 to 2024 has been 205 birds per year, with an overall decline of 90%. The bootstrap log-linear regression model showed very strong evidence (posterior probability P = 1.0) of a downward trend in Palila abundance for all four time series: the last five, 10, 15, and 26 years. The log-linear state-space model (Figure 4) also showed very strong evidence of a downward trend with 100% posterior probability of population decline for all four time series (Bayesian R² = 0.81).

Because we suspected the unusually low estimate in 2000 to be observation error rather than a true population decline, we input the mean of the 1999 and 2001 abundance estimates for the year 2000 in the change-point analysis. We found a very strong (P ≥ 0.9) Bayesian probability of a change-point (0.99) to occur in 2005, preceding an approximate halving in the population, and a strong (0.7 ≤ P < 0.9) probability (0.73) occurring in 2001 (Figure 5). All other change point probabilities were weak (P < 0.5) with no evidence of a change-point in 2021 or thereafter.

Sampling condition evaluation

The 2021 abundance estimates are about half that of 2020 estimates. Similar to 2021, abundance estimates declined more than 35% in 2000, 2006, 2010, and 2015, immediately after which estimates increased, except in the case of 2006. Because of the lower numbers of detections in 2022, 2023, and 2024, fewer covariates could be reliably modeled with sufficient sample size (at least five per level) post-stratification by year, resulting in a smaller set of candidate models (Table S2, Appendix 1). This limited our ability to detect any effects from weather covariates or observers on detectability. For this reason, we inspected the distribution of weather covariate values to identify any clear outliers or patterns across survey years. Weather conditions during these surveys did not contribute to low estimates because they did not differ markedly from conditions during other surveys (Figure S3, Appendix 1). This is because surveys were halted when sampling conditions impeded bird detection, e.g., heavy rains or strong winds/gusts.

DISCUSSION

We rejected our hypothesis that the Palila population trajectory would change from 2021 to 2024 and concluded that decades-long management efforts were insufficient to promote recovery, given that (1) the new abundance estimates did not differ significantly from that of the 2021 survey; and (2) they did not introduce any new change points in the time series. In fact, the estimated abundance of Palila in 2022 at 367 to 742 birds (95% CI; point estimate: 545) was the lowest estimate since regular surveys began in 1980 (Scott et al. 1984), even correcting for differences in sampling frame prior to 1998 when the survey area was expanded (Johnson et al. 2006). The population apparently increased slightly in 2023 to 374 to 842 birds (95% CI; point estimate: 596), and further in 2024 to 412 to 970 birds (95% CI; point estimate: 666); however, both point estimates were within the 95% CIs of the 2021 and 2022 estimates. Whereas 2023 and 2024 estimates appear to suggest that the population stabilized after 2021, no new changepoints indicated a halt in the decline, and estimates remain well below pre-2021 abundances, which exceeded 1000 birds. The estimates from the last four years hover around what Franklin (1980) had called the minimum viable population to prevent genetic drift, i.e., 500 individuals.

We produced four state-space trend assessments to examine how different time periods might affect our conclusions, and all showed a 100% certainty of decline, even the most optimistic time series for the most recent five years. The mean decline during 1998 to 2024 was 205 birds per year, resulting in a 90% decline in the population over the 26-year monitoring period (Table S1, Appendix 1; Figures 3 and 4). Our change-point analysis identified a significant shift in Palila population estimates in 2001, preceding a roughly 25% decline, and an even stronger shift in 2005, preceding a roughly halving of the population (Figure 5). However, evidence of change points outside of these years was weak (P < 0.5). Posterior probabilities of change close to zero between 2010 and 2020 suggest that the Palila decline was temporarily arrested in these years. A small peak in 2020 followed by posterior probabilities close to zero suggest that the decline observed in 2020 to 2021 has remained unchanged. Further surveys and future evaluation of critical slowing down, which is often associated with changes in recruitment or survival, could provide additional insights of population dynamics pertinent to management response (Rozek et al. 2017).

Weather conditions in 2022, the year of all-time low Palila abundance, were not markedly different from those of previous surveys, nor were the conditions different during other years when estimates were low, e.g., 2000, 2006, 2010, 2015, and 2021, and thus are not likely to contribute to the low estimates (Figure S3, Appendix 1). This is most likely because surveys are halted when weather conditions hinder detection. However, the decrease by over 50% from 1999 to 2000, followed by a more than 100% increase in 2001 (Table S1, Appendix 1; Figures 3 and 4), is not biologically realistic given the constraints of Palila reproductive output, and likely represent observer error. In contrast, the Palila did not recover from a 45% decrease between 2005 and 2006, which likely represents a true population crash. Future incorporation of a time-to-detection model into the interface of the web-app used to collect Palila point count data could help to reduce observation bias in abundance estimates (Amundson et al. 2014).

Whereas abnormal estimates are unlikely explained by weather conditions affecting detectability, climate change likely played a broader role in the population dynamics reflected in our 1998 to 2024 time series. The 2005 to 2006 decline and further dips in the population, e.g., 49% decline from 2009 to 2010, and 38% decline from 2014 to 2015, coincide with a significant browning event from 2006 to 2013, based on satellite imagery from the Moderate Resolution Imaging Spectroradiometer sensor and the Normalized Difference Vegetation Index (Gallerani et al. 2025). This appears to be associated with intermittent drought, which reduces māmane seed-pod output and lowers Palila reproduction and survival (Lindsey et al. 1997, Gray et al. 1999, Leonard et al. 2008, Banko et al. 2013). Drought on Mauna Kea often closely followed El Niño Southern Oscillation events (ENSO), such as those occurring in 1985, from 1991 to 1992, from 2004 to 2006, and in 2009 (Jacobi et al. 1996, Lindsey et al. 1997, Gray et al. 1999, Leonard 2008, Banko et al. 2013; https://www.cpc.ncep.noaa.gov/data/indices/). However, Banko et al. (2013) reported drought occurring in 74% of months between 2000 and 2011, suggesting that drought events are not isolated to large ENSO events. Whereas the low Palila years do not time perfectly with ENSO and drought, there are likely delayed effects from drought interacting with continued ungulate browsing, which effectively lower the carrying capacity of the māmane forest (Banko et al. 2013, Brinck and Banko 2014). Although the last major ENSO event in 2016 (> 2.0 on the Southern Oscillation Index) did not coincide with any drought events based on a 12-month comparison of standard precipitation index (SPI-12), it did coincide with a brief severe drought based on a three-month comparison of standard precipitation index (SPI-3), and preceded a slight decline in the Palila population from 2016 to 2019, before the population temporarily recovered in 2020 (Table S1, Appendix 1; Figures 3 and 4; Longman et al. 2023).

All-time low Palila numbers in 2021 and 2022, not coinciding with any major ENSO events or severe droughts, suggest that other pre-existing stressors continue to pose existential threats to Palila, despite frequent and long-term efforts to research and mitigate them. The greatest known driver of Palila declines has been extensive habitat loss from deforestation by humans and browsing by ungulates (Scowcroft 1983, Scott et al. 1984, Hess et al. 1999, Banko et al. 2014). Although public hunting and routine aerial shoots conducted by DLNR DOFAW have removed over 20,000 sheep between 1998 and 2023 (Ian Cole, East Hawaiʻi Wildlife Manager, DLNR DOFAW, 2025, personal communication), a substantial population remains within the Palila Critical Habitat. The highest cull years between 2007 to 2013, with a peak of nearly 3000 sheep, partially coincide with a decline in Palila from 2007 to 2010 (Figure 3; Banko et al. 2014), which could suggest that high cull rates reflect sheep population increases. Similarly, a relatively constant take of a few hundred culls per year from 2015 to 2019, followed by an uptick to over 500 culls in 2020 and 2022, could indicate a stable-to-increasing sheep population; however, differences in control efforts between years have not been accounted for. Censusing feral sheep and quantifying cull effort could help to assess their ongoing impact on Palila habitat and the effort required to fully eradicate them on Mauna Kea, as ordered by the U.S. District Court, District of Hawaii. Previous research has revealed that feral sheep browsing lowers māmane seed output by reducing seedling recruitment and overall forest cover, and by thinning the lower branches and foliage of individual trees (Banko et al. 2013). This cumulatively reduces the carrying capacity of the dry forest by lowering māmane seed-pod availability, having a strong impact on Palila reproduction and survival (Lindsey et al. 1995, Banko and Farmer 2014, Banko et al. 2020). Browsing is also suspected to have largely contributed to the extirpation of ʻakiapōlāʻau (Hemignathus wilsoni) from Mauna Kea’s subalpine zone (Scowcroft 1983, Banko et al. 2014). However, the direct links between ungulates and bird survival and reproduction have not been investigated. Likewise, studies of māmane seed-pod output on the survival of different demographic groups, i.e., sex and age, have not, to our knowledge, been conducted, including population viability analysis. Future work incorporating capture-mark-recapture and radio telemetry could clarify this relationship.

Whereas some have proposed that ungulates could locally reduce fuel loads of invasive weeds, especially fountain grass, that increase fire occurrence, introduced ungulates apparently do not reduce fire occurrence and severity at landscape scales in similar ecosystems (D’Antonio and Vitousek 1992, Blackmore and Vitousek 2000, Williams et al. 2006, Thaxton and Jacobi 2009). Quantifying expansion or reduction of invasive plants associated with fire risk could help to identify priority areas for fire breaks and restoration. Invasive plants also degrade Palila habitat because they suppress māmane regeneration, and possibly host insects that threaten Cydia plicada moths, an important food source for Palila nestlings (Hughes et al. 1991, Hess et al. 1999, Banko et al. 2002b, 2020). Understanding the impact of hymenopterans, including introduced ants, predatory wasps, and parasitoid wasps, on C. plicada caterpillar availability (Oboyski et al. 2004, Banko et al. 2020) could clarify whether they have contributed to the Palila’s decline. Other insects, such as naio thrips (Klambothrips myopori; order Thysanoptera) threaten a secondary food source of Palila, and currently lack a biological control agent in the State of Hawaii (Banko et al. 2014, 2020). Similarly, introduced fungal pathogens, such as Armillaria mellea, have increased māmane mortality (Gardner and Trujillo 2001), but their wider ecological impact and treatment methods remain mostly unstudied.

Palila predators include feral cats, black rats (Rattus rattus), and pueo (Asio flammeus sandwichensis), which consume Palila eggs, nestlings, and adults (Jacobi et al. 1996, Banko et al. 2020). One study found cats alone had depredated 11% of Palila nests annually (Hess et al. 2004). Since 2008, roughly 900 feral cats have been removed from the Palila Critical Habitat by MKFRP (Joe Kern, MKFRP, 2025, personal communication). However, quantifiable information relating trap effort to invasive predator abundance and take remains unavailable and could help to assess the effort required to keep predator numbers at optimal levels for Palila recovery.

In addition to, and partly exacerbated by, these threats, Palila have a slow reproductive rate, with most pairs producing only two eggs and an average of one to two fledglings per nest per breeding season, with mean annual survivorship in their first years estimated at 0.36 ± 0.08 (Lindsey et al. 1995, Banko et al. 2009, 2020). Effectively, Palila are not replacing the previous generation of breeding adults each year, which greatly limits their rate of recruitment. For this reason, we find the doubling in abundance estimates between 2000 and 2001 to be biologically unrealistic.

Finally, the status of Palila habitat and its relation to their current distribution remain largely unknown. A recent analysis from quarterly surveys suggests that the area of highest Palila density has contracted and the density within this region has also decreased (Camp et al. 2025). The 2017 mountain-wide survey also found no Palila along the southeastern and eastern slopes of Mauna Kea, where they have occurred historically. There were two detections on the northern slope, where wild birds were translocated between 1997 to 1998 and 2004 to 2006, and where captive-reared birds were released from 2003 to 2005, in 2009, and in 2019 (Banko and Farmer 2014, https://dlnr.hawaii.gov/blog/2019/05/21/nr19-101). However, Palila have not been detected on annual surveys outside the core survey area, at lower elevation, since 2019, corroborating the conclusion from the quarterly surveys. Their overall range remains relatively constant, at 5% of its historical extent (Figure 1), since annual surveys began, despite efforts from DOFAW and MKFRP to outplant māmane and exclude grazing ungulates from Puʻu Mali and Kaʻohe Restoration Areas. Although diseases such as avian malaria (Plasmodium relictum) limit the potential for restoration at lower elevations to be effective (Banko et al. 2009), the absence of Palila from apparently suitable habitat at higher elevations is not fully understood. Quantifiable information on the changes in māmane forest structure and composition coinciding with restoration efforts could help to provide an estimate of how much habitat is available to Palila and the effort required to expand habitat for a self-sustaining population.

CONCLUSION

In summary, management actions for Palila since at least 2003 have included public hunting and aerial shooting of feral ungulates, fence repair and construction, habitat restoration, fire suppression, invasive weed removal, predator control, captive breeding, and translocation (https://dlnr.hawaii.gov/restoremaunakea/management/forest-restoration/). Decades of research on the demography, breeding ecology, habitat use, food ecology, and vegetation ecology associated with Palila have also been conducted (Banko and Farmer 2014). However, these substantial endeavors have not reversed a 100% certain decline of the Palila since 1998, even if they temporarily arrested it between 2010 to 2020. If the greatest threats to Palila are already known, our results suggest these are outpacing decades-long efforts to mitigate their impact on the population. Perhaps the hardest threat to combat is climate change, because Palila population declines are strongly associated with severe drought events that suppress māmane reproduction (Banko et al. 2014). At roughly 666 individuals, the current Palila population, having barely departed from its all-time low, and with a severely contracted range at 95%, is likely at an inflection point in its future. Jacobi et al. (1996:368) warned that “the risk of extinction to a population from fire or some other catastrophe is greater when the populations is concentrated in a small area.” The authors were referring to a subpopulation of Palila in the southeastern portion of Palila Critical Habitat, which has since been extirpated. Continued long-term monitoring can help to assess the efficacy of existing and future adaptive management decisions for Palila persistence and recovery.

LITERATURE CITED

Amundson, C. L., J. A. Royle, and C. M. Handel. 2014. A hierarchical model combining distance sampling and time removal to estimate detection probability during avian point counts. Auk 131:476-494. https://doi.org/10.1642/AUK-14-11.1

Banko, P. C., K. W. Brinck, C. Farmer, and S. C. Hess. 2009. Recovery programs: Palila. Pages 513-529 in T. K. Pratt, C. T. Atkinson, P. C. Banko, J. D. Jacobi, and B. L. Woodworth, editors. Conservation biology of Hawaiian forest birds: implications for island avifauna. Yale University Press, New Haven, Connecticut, USA.

Banko, P. C., R. J. Camp, C. Farmer, K. W. Brinck, D. L. Leonard, and R. M. Stephens. 2013. Response of Palila and other subalpine Hawaiian forest bird species to prolonged drought and habitat degradation by feral ungulates. Biological Conservation 157:70-77. https://doi.org/10.1016/j.biocon.2012.07.013

Banko, P. C., M. L. Cipollini, G. W. Breton, E. Paulk, M. Wink and I. Izhaki. 2002a. Seed chemistry of Sophora chrysophylla (Mâmane) in relation to the diet of the specialist seed predator Loxioides bailleui (Palila) in Hawaiʻi. Journal of Chemical Ecology 28:1389-1406. https://doi.org/10.1023/A:1016248502927

Banko, P. C., and C. Farmer. 2014. Palila restoration research, 1996–2012. Technical Report HCSU-046A, Hawaiʻi Cooperative Studies Unit, University of Hawaiʻi at Hilo, Hawaiʻi, USA. http://hdl.handle.net/10790/2618

Banko, P. C., S. C. Hess, P. G. Scowcroft, C. Farmer, J. D. Jacobi, R. M. Stephens, R. J. Camp, D. L. Leonard, K. W. Brinck, J. O. Juvik, et al. 2014. Evaluating the long-term management of introduced ungulates to protect the Palila, an endangered bird, and its critical habitat in subalpine forest of Mauna Kea, Hawaiʻi. Arctic, Antarctic, and Alpine Research 46:871-889. https://doi.org/10.1657/1938-4246-46.4.871

Banko, P. C., L. Johnson, G. D. Lindsey, S. G. Fancy, T. K. Pratt, J. D. Jacobi, and W. E. Banko. 2020. Palila (Loxioides bailleui), version 1.0. In A. F. Poole and F. B. Gill, editors. Birds of the world. Cornell Lab of Ornithology, Ithaca, New York, USA. https://doi.org/10.2173/bow.palila.01

Banko, P. C., P. T. Oboyski, J. W. Slotterback, S. J. Dougill, D. M. Goltz, L. Johnson, M. E. Laut, and C. Murray. 2002b. Availability of food resources, distribution of invasive species, and conservation of a Hawaiian bird along a gradient of elevation. Journal of Biogeography 29:789-808. https://doi.org/10.1046/j.1365-2699.2002.00724.x

Banko, W. E., and P. C. Banko. 2009. Historic decline and extinction. Pages 25-58 in T. K. Pratt, C. T. Atkinson, P. C. Banko, J. D. Jacobi, and B. L. Woodworth, editors. Conservation biology of Hawaiian forest birds: implications for island avifauna. Yale University Press, New Haven, Connecticut, USA.

Blackmore, M., and P. M. Vitousek. 2000. Cattle grazing, forest loss, and fuel loading in a dry forest ecosystem at Puʻu WaʻaWaʻa Ranch, Hawaiʻi. Biotropica 32:625-632. https://doi.org/10.1111/j.1744-7429.2000.tb00509.x

Brinck, K., and P. C. Banko. 2014. Changes in Mauna Kea dry forest structure 2000–2014. Technical Report HCSU-054, Hawaiʻi Cooperative Studies Unit, University of Hawaiʻi at Hilo, Hawaiʻi, USA. http://hdl.handle.net/10790/2610

Buckland, S. T., D. R. Anderson, K. P. Burnham, J. L. Laake, D. L. Borchers, and L. Thomas. 2001. Introduction to distance sampling: estimating abundance of biological populations. Oxford University Press, Oxford, UK.

Buckland, S. T., E. A. Rexstad, T. A. Marques, and C. S. Oedekoven. 2015. Distance sampling: methods and applications. Springer International, Cham, Switzerland. https://doi.org/10.1007/978-3-319-19219-2

Burney, D. A., H. F. James, L. P. Burney, S. L. Olson, W. Kikuchi, W. L. Wagner, M. Burney, D. McCloskey, D. Kikuchi, F. V. Grady, et al. 2001. Fossil evidence for a diverse biota from Kauaʻi and its transformation since human arrival. Ecological Monographs 71:615-641. https://doi.org/10.1890/0012-9615(2001)071[0615:FEFADB]2.0.CO;2

Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodel inference: a practical information-theoretic approach. Second edition. Springer-Verlag, New York, New York, USA.

Camp, R. J., C. K. Asing, N. Hunt, A. Wang, C. Farmer, L. Neitmann, and P. C. Banko. 2025. Fine-grained temporal population monitoring of a declining, critically endangered Hawaiian honeycreeper. Frontiers in Conservation Science 6:1564661. https://doi.org/10.3389/fcosc.2025.1564661

Camp, R. J., and P. C. Banko. 2012. Palila abundance estimates and trend. Technical Report HCSU-033, Hawaiʻi Cooperative Studies Unit, University of Hawaiʻi at Hilo, Hawaiʻi, USA. http://hdl.handle.net/10790/2630

Camp, R. J., K. W. Brinck, and P. C. Banko. 2014. Palila abundance estimates and trend. Technical Report HCSU-053, Hawaiʻi Cooperative Studies Unit, University of Hawaiʻi at Hilo, Hawaiʻi, USA. http://hdl.handle.net/10790/2611

Camp, R. J., K. W. Brinck, and P. C. Banko. 2016. 2015–2016 Palila abundance estimates. Technical Report HCSU-076, Hawaiʻi Cooperative Studies Unit, University of Hawaiʻi at Hilo, Hawaiʻi, USA. http://hdl.handle.net/10790/2750

Camp, R. J., K. W. Brinck, P. M. Gorresen, F. A. Amidon, P. M. Radley, S. P. Berkowitz, and P. C. Banko. 2015. Current land bird distribution and trends in population abundance between 1982 and 2012 on Rota, Mariana Islands. Journal of Fish and Wildlife Management 6:511-540. https://doi.org/10.3996/112014-JFWM-085

Camp, R. J., M. Gorresen, T. K. Pratt, and B. L. Woodworth. 2009. Population trends of native Hawaiian forest birds: 1976–2008. Technical Report HCSU-012, Hawaiʻi Cooperative Studies Unit, University of Hawaiʻi at Hilo, Hawaiʻi, USA. http://hdl.handle.net/10790/2692

Camp, R. J., T. K. Pratt, C. Bailey, and D. Hu. 2011. Landbirds vital sign monitoring protocol — Pacific Island Network. Version 1.0. Natural Resources Report NPS/PACN/NRR—2011/402, National Park Service, Fort Collins, Colorado, USA. https://www.govinfo.gov/content/pkg/GOVPUB-I29-PURL-gpo132229/pdf/GOVPUB-I29-PURL-gpo132229.pdf

Camp, R. J., N. E. Seavy, P. M. Gorresen, and M. H. Reynolds. 2008. A statistical test to show negligible trend: comment. Ecology 89:1469-1472. https://doi.org/10.1890/07-0462.1

D’Antonio, C. M., and P. M. Vitousek. 1992. Biological invasions by exotic grasses, the grass/fire cycle, and global change. Annual Review of Ecology, Evolution, and Systematics 23:63-87. https://doi.org/10.1146/annurev.es.23.110192.000431

Erdman, C., and J. W. Emerson. 2007. bcp: an R package for performing a Bayesian analysis of change point problems. Journal of Statistical Software 23:1-13. https://doi.org/10.18637/jss.v023.i03

Franklin, I. R. 1980. Evolutionary change in small populations. Pages 135-149 in M. E. Soulé and B. A. Wilcox, editors. Conservation biology: an evolutionary-ecological perspective. Sinauer Associates, Sunderland, Massachusetts, USA.

Gallerani, E. M., R. J. Camp, P. C. Banko, A. Madson, C. Dong, L. Berio Fortini, Z. Ma, and T. W. Gillespie. 2025. Breaking down Palila decline: assessing the role of drought and vegetation health in the population loss of an endangered Hawaiian Honeycreeper. Global Ecology and Conservation 62:e03831. https://doi.org/10.1016/j.gecco.2025.e03831

Gardner, D. E., and E. E. Trujillo. 2001. Association of Armillaria mellea with māmane decline at Puʻu Laʻau. Newsletter of the Hawaiian Botanical Society 40:33-34.

Gelman, A., and D. B. Rubin. 1992. Inference from iterative simulation using multiple sequences. Statistical Science 7:457-511. https://doi.org/10.1214/ss/1177011136

Genz, A. S., K. W. Brinck, C. K. Asing, L. Berry, R. J. Camp, and P. C. Banko. 2022. 2019–2021 Palila abundance estimates and trend. Technical Report HCSU-101, Hawaiʻi Cooperative Studies Unit, University of Hawaiʻi at Hilo, Hawaiʻi, USA. http://hdl.handle.net/10790/6858

Genz, A. S., K. W. Brinck, R. J. Camp, and P. C. Banko. 2018. 2017–2018 Palila abundance estimates and trend. Technical Report HCSU-086. Hawaiʻi Cooperative Studies Unit, University of Hawaiʻi at Hilo, Hawaiʻi, USA. http://hdl.handle.net/10790/4431

Gorresen, P. M., R. J. Camp, M. H. Reynolds, T. K. Pratt, and B. L. Woodworth. 2009. Status and trends of native Hawaiian songbirds. Pages 108-136 in T. K. Pratt, C. T. Atkinson, P. C. Banko, J. D. Jacobi, and B. L. Woodworth, editors. Conservation biology of Hawaiian forest birds: implications for island avifauna. Yale University Press, New Haven, Connecticut, USA.

Gray, E. M., P. C. Banko, S. J. Dougill, D. Goltz, L. Johnson, M. E. Laut, J. D. Semones, and M. R. Wiley. 1999. Breeding and nonbreeding censuses of the 1998 Palila population on Mauna Kea, Hawaiʻi. ʻElepaio 59:33-39.

Hess, S. C., P. C. Banko, G. J. Brenner, and J. D. Jacobi. 1999. Factors related to the recovery of subalpine woodland on Mauna Kea, Hawaiʻi. Biotropica 31:212-219. https://doi.org/10.1111/j.1744-7429.1999.tb00133.x

Hess, S. C., P. C. Banko, D. M. Goltz, R. M. Danner, and K. W. Brinck. 2004. Strategies for reducing feral cat threats to endangered Hawaiian birds. Pages 21-26 in Proceedings of the Vertebrate Pest Conference (Visalia, 2004). University of California, Davis, California, USA.

Hess, S. C., P. C. Banko, L. J. Miller, and L. P. Laniawe. 2014. Habitat and food preferences of the endangered Palila (Loxioides bailleui) on Mauna Kea, Hawaiʻi. Wilson Journal of Ornithology 126:728-738. https://doi.org/10.1676/13-220.1

Hess, S. C., P. C. Banko, M. H. Reynolds, G. J. Brenner, and L. P. Laniawe. 2001. Drepanidine movements in relation to food availability in subalpine woodland on Mauna Kea, Hawaiʻi. Studies in Avian Biology 22:154-163.

Hughes, F., P. M. Vitousek, and T. Tunison. 1991. Alien grass invasion and fire in the seasonal submontane zone of Hawaiʻi. Ecology 72:743-746. https://doi.org/10.2307/2937215

Jacobi, J. D., S. G. Fancy, J. G. Giffin, and J. M. Scott. 1996. Long-term population variability in the Palila, an endangered Hawaiian honeycreeper. Pacific Science 50:363-370.

Johnson, L., R. J. Camp, K. W. Brinck, and P. C. Banko. 2006. Long-term population monitoring: lessons learned from an endangered passerine in Hawaiʻi. Wildlife Society Bulletin 34:1055-1063. https://doi.org/10.2193/0091-7648(2006)34[1055:LPMLLF]2.0.CO;2

Juvik, J. O., and S. P. Juvik. 1984. Mauna Kea and the myth of multiple use. Mountain Research and Development 4:191-202. https://doi.org/10.2307/3673140

Kepler, C. B., and J. M. Scott. 1981. Reducing bird count variability by training observers. Studies in Avian Biology 6:366-371.

Leonard, D. L., Jr., P. C. Banko, K. W. Brinck, C. Farmer, and R. J. Camp. 2008. Recent surveys indicate rapid decline of Palila population. ʻElepaio 68:27-30.

Lerner, H. R., M. Meyer, H. F. James, M. Hofreiter, and R. C. Fleischer. 2011. Multilocus resolution of phylogeny and timescale in the extant adaptive radiation of Hawaiian honeycreepers. Current Biology 21:1838-1844. https://doi.org/10.1016/j.cub.2011.09.039

Lindsey, G. D., S. G. Fancy, M. H. Reynolds, T. K. Pratt, K. A. Wilson, P. C. Banko, and J. D. Jacobi. 1995. Population structure and survival of Palila. Condor 97:528-535. https://doi.org/10.2307/1369038

Lindsey, G. D., T. K. Pratt, M. H. Reynolds, and J. D. Jacobi. 1997. Response of six species of Hawaiian forest birds to a 1991–1992 El Niño drought. Wilson Bulletin 109:339-343.

Longman, R. J., D. J. Ford, A. Anderson, A. G. Frazier, and C. P. Giardina. 2023. Climate change, climate variability, and drought portfolio leeward Mauna Kea. Pacific Drought Knowledge Exchange CCVD Series Version 5.4. East-West Center, Honolulu, Hawaiʻi.

Miller, D. L., E. Rexstad, L. Thomas, L. Marshall, and J. L. Laake. 2019. Distance sampling in R. Journal of Statistical Software 89:1-28. https://doi.org/10.18637/jss.v089.i01

Oboyski, P. T., J. W. Slotterback, and P. C. Banko. 2004. Differential parasitism of seed-feeding Cydia (Lepidoptera: Tortricidae) by native and alien wasp species relative to elevation in subalpine Sophora (Fabaceae) forests on Mauna Kea, Hawaiʻi. Journal of Insect Conservation 8:229-240. https://doi.org/10.1023/B:JICO.0000045821.71118.54

Olson, S. L., and H. F. James. 1982. Prodromus of the fossil avifauna of the Hawaiian Islands. Smithsonian Contributions to Zoology No. 365. Smithsonian Institution Press, Washington, D.C., USA. https://doi.org/10.5479/si.00810282.365

Plummer, M. 2003. JAGS: a program for analysis of Bayesian graphical models using Gibbs sampling. Pages 1-10 in Proceedings of the Third International Workshop on Distributed Statistical Computing (Vienna, 2003). Technische Universität Wien, Vienna, Austria.

R Core Team. 2024. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/

Reddy, E., D. H. Van Vuren, P. G. Scowcroft, J. B. Kauffman, and L. Perry. 2012. Long-term response of the mâmane forest to feral herbivore management on Mauna Kea, Hawaiʻi. Pacific Conservation Biology 18:123-132. https://doi.org/10.1071/PC120123

Rozek, J. C., R. J. Camp, and J. M. Reed. 2017. No evidence of critical slowing down in two endangered Hawaiian honeycreepers. PLoS ONE 12:e0187518. https://doi.org/10.1371/journal.pone.0187518

Scott, J.M., S. Mountainspring, F. L. Ramsey, and C. B. Kepler. 1986. Forest bird communities of the Hawaiian Islands: their dynamics, ecology, and conservation. Studies in Avian Biology 9:1-431

Scott, J. M., S. Mountainspring, C. van Riper III, C. B. Kepler, J. D. Jacobi, T. A. Burr, and J. G. Giffin. 1984. Annual variation in the distribution, abundance, and habitat response of the Palila (Loxioides bailleui). Auk 101:647-664. https://doi.org/10.2307/4086892

Scowcroft, P. G. 1983. Tree cover changes in māmane (Sophora chrysophylla) forests grazed by sheep and cattle. Pacific Science 37:109-119.

Thaxton, J. M., and J. D. Jacobi. 2009. Assessment of fuels, potential fire behavior, and management options in subalpine vegetation on Mauna Kea Volcano, Hawaiʻi. Technical Report HCSU-013. Hawaiʻi Cooperative Studies Unit, University of Hawaiʻi at Hilo, Hawaiʻi, USA. http://hdl.handle.net/10790/2693

Thomas, L., S. T. Buckland, E. A. Rextad, J. L. Laake, S. Strindberg, S. L. Hedley, J. R. B. Bishop, T. A. Marques, and K. P. Burnham. 2010. Distance software: design and analysis of distance sampling surveys for estimating population size. Journal of Applied Ecology 47:5-14. https://doi.org/10.1111/j.1365-2664.2009.01737.x

U.S. Fish and Wildlife Service (USFWS). 1977. Palila critical habitat, Federal Register Rule 42 FR 40689. USFWS, Washington, D.C., USA.

U.S. Fish and Wildlife Service (USFWS). 2006. Revised recovery plan for Hawaiian forest birds. Region 1, USFWS, Portland, Oregon, USA. https://www.fws.gov/program/recovery/recovery-plans

U.S. Geological Survey. 2018. National elevation dataset. Pages 83-110 in D. B. Gesch, G. A. Evans, M. J. Oimoen, and S. Arundel, editors. The national elevation dataset. American Society for Photogrammetry and Remote Sensing, Baton Rouge, Louisiana, USA. https://www.usgs.gov/publications/national-elevation-dataset

Van Houtan, K. S., T. O. Gagné, P. Banko, M. E. Hagemann, R. W. Peck, and C. T. Yarnes. 2024. Climatic drought and trophic disruption in an endemic subalpine Hawaiian forest bird. Biological Conservation 299:110823. https://doi.org/10.1016/j.biocon.2024.110823

Van Ravenzwaaij, D., P. Cassey, and S.D. Brown. 2018. A simple introduction to Markov Chain Monte-Carlo sampling. Psychonomic Bulletin and Review 25:143-154. https://doi.org/10.3758/s13423-016-1015-8

Van Riper III, C., 1978. The breeding ecology of the Amakihi (Loxops virens) and Palila (Psittirostra bailleui) on Mauna Kea, Hawaiʻi. Dissertation. University of Hawaiʻi at Mânoa, Honolulu, Hawaiʻi, USA.

Verner, J. and K. A. Milne. 1989. Coping with sources of variability when monitoring population trends. Annales Zoologici Fennici 26:191-199.

Williams, R. J., C.-H. Wahren, R. A. Bradstock, and W. J. Muller. 2006. Does alpine grazing reduce blazing? A landscape test of a widely held hypothesis. Austral Ecology 31:925-936. https://doi.org/10.1111/j.1442-9993.2006.01655.x