DISTRIBUTION, ABUNDANCE, AND ACOUSTIC CHARACTERISTICS OF KOHALA
FOREST BIRDS
A THESIS SUBMITTED TO THE GRADUATE DIVISION OF THE UNIVERSITY OF
HAWAI‘I AT HILO IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF
MASTER OF SCIENCE
IN
TROPICAL CONSERVATION BIOLOGY AND ENVIRONMENTAL SCIENCE
JUNE 2020
By
Keith Burnett
Thesis Committee:
Pat Hart, Chairperson
Eben Paxton
Jonathan Price
Esther Sebastian-Gonzalez
Keywords: Abundance, Bioacoustics, Density, Forest birds, Hawai‘i
Preface
In the late 1800’s, Kohala was host to a wide variety of native bird species (Rothschild
1893; Wilson and Evans, 1974), yet there has been little ornithological work in the region, and
many species have since disappeared (van Riper 1982; Scott et al. 1986). Pigs have removed
much of the understory vegetation and suppressed the regeneration of canopy species, negatively
affecting the native forest bird population in Kohala, along with other pressures such as
introduced predators, anthropogenic habitat loss, and disease (van Riper and Scott 2001). The
introduction of avian malaria (Plasmodium relictum), has restricted native bird populations to
high elevation forests (LaPointe et al. 2010, van Riper et al. 1986), leaving areas like Kohala –
with a maximum elevation of 1670 m – with little suitable habitat for native species. Other native
bird communities that are limited to low elevation mountains, like those on Kauai, are already
exhibiting rapid rates of decline as climate change exacerbates the impacts of avian malaria,
further shrinking suitable habitat and impacting most if not all of their local range (Paxton et al.
2016). It is of utmost importance to monitor communities, such as Kohala, which harbor
endangered species and further understand how climate change and the encroachment of avian
malaria into higher elevation refuges impacts native Hawaiian forest birds (Paxton et al. 2018).
There are many reasons why there is a scarcity of ornithological work in Kohala, but
access and resource prioritization appear to be the two most influential. The native forest within
the state-owned Pu’u ‘O ‘Umi Natural Area Reserve on Kohala mountain is abutted by private
land with many owners, including three prominent ranches. There is only one public access road
to the forest edge, so coordination for access on private dirt roads is necessary for access to the
entire region without expensive helicopter operations, often stymied by the clouds and rain
consistently enveloping the forest. While the private landowners are generally amenable to
granting access to researchers, the coordination effort and necessity to minimize impacts on
ranch operations has been prohibitive. Prioritization of state and federal resources on Hawai’i
Island has also been influential on the lack of ornithological work in Kohala. Specifically,
funding and manpower has been directed in areas with documented extant endangered species,
such as Hakalau Forest National Wildlife Refuge, the kipuka region on Saddle Road, Hawai’i
Volcanoes National Park, Kamehameha Schools land in Keauhou, and Palila (Loxioides bailleui)
habitat on Maunakea. While endangered species have been documented in Kohala as recently as
the early 1970’s (van Riper 1982), the recent listing of the ‘i’iwi (Drepanis coccinea) as
threatened under the Endangered Species Act in 2017 has renewed the call for ornithological
work in the Kohala region among the conservation community in Hawai’i. Here I present the
first monitoring effort in Kohala since 1979, and the first examination of forest bird song
structure in Kohala.
Literature Cited
LaPointe, D. A., M. L. Goff, and C. T. Atkinson. 2010. Thermal constraints to the sporogonic
development and altitudinal distribution of avian malaria Plasmodium relictum in Hawai'i.
Journal of Parasitology 96(2):318-325.
Paxton, E. H., R. J. Camp, P. M. Gorresen, L. H. Crampton, D. L. Leonard, and E. A.
VanderWerf. 2016. Collapsing avian community on a Hawaiian island. Science Advances
2(9):e1600029.
Paxton, E. H., M. Laut, J. P. Vetter, and S. J. Kendall. 2018. Research and management priorities
for Hawaiian forest birds. The Condor: Ornithological Applications 120(3):557-565.
Rothschild, Lionel Walter Rothschild Baron, and H. Palmer. 1893. The avifauna of Laysan and
the neighbouring islands: with a complete history to date of the birds of the Hawaiian
possessions. RH Porter, .
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. :.
van Riper III, C. 1982. Censuses and breeding observations of the birds on Kohala Mountain,
Hawaii. The Wilson Bulletin :463-476.
Van Riper III, C., S. G. Van Riper, M. L. Goff, and M. Laird. 1986. The epizootiology and
ecological significance of malaria in Hawaiian land birds. Ecological Monographs 56(4):327-
344.
van Riper, C., and J. M. Scott. 2001. Limiting factors affecting Hawaiian native birds. Studies in
Avian Biology 22:221-233.
Wilson, S. B., and A. H. Evans. 1974. Aves Hawaiienses: the birds of the Sandwich Islands.
Arno Press, .
Chapter 1: Current distribution and abundance of Kohala forest birds
Abstract
Kohala, the northernmost volcano out of the five that comprise Hawai’i Island, is home to
the most spatially isolated population of Hawaiian forest birds on the island. The avian
community on Kohala is one of the few native bird populations in the state that has not been
monitored since the landmark Hawai’i Forest Bird Survey (HFBS) in 1979. Here, I examine
changes in the distribution and abundance of forest birds on Kohala since 1979. I surveyed 143
stations across 13 transects in Pu’u ‘O ‘Umi Natural Area Reserve in Kohala from March
through May 2017 and incorporated the 1979 HFBS data from 80 stations across 3 transects that
occurred in our study site. I detected 2806 individuals of 15 different species across the 143
survey stations and observed changes in species densities ranging from -8.4% (C. sandwichensis)
to +714% (D. coccinea). While equivalence testing showed meaningful increases in population
densities for all but one species, changes in survey protocol standards may have limited our
ability to make direct comparisons with the HFBS. I also document here the introduction and
establishment of C. diphone in Kohala.
Introduction
Understanding population dynamics and the ability to obtain accurate estimates of
population size are critical to making informed management decisions regarding avifauna (Scott
et al. 1981). Accurate population estimates of density and structure are necessary for both the
monitoring of endangered species, as well as the creation of viable recovery plans (Johnson et al.
2006). In Hawai’i, declines in forest bird populations first became evident in the late 1800’s
(Munro 1944), with declining populations and decreasing ranges continuing to the present (Scott
and Kepler 1985, Paxton et al. 2016). Habitat loss and introductions of diseases, competitors, and
predators led to fragmented and isolated native bird populations, all contributing heavily to their
decline (van Riper and Scott 2001). Hawai’i has seen the highest modern extinction rate for
avifauna than any other state in the United States (Loope 1998), and this, along with extinctions
in the Mariana Islands, is why the United States experienced a larger number of bird extinctions
than any other country in the 25-year period from 1980-2005 (Wilcove 2005).
The first comprehensive, state-wide assessment of native Hawaiian forest birds was
conducted from 1976 to 1983 with the landmark Hawai’i Forest Bird Survey (HFBS) – a series
of surveys throughout the main Hawaiian Islands (Scott et al. 1986). Many of the original HFBS
transects have not been re-sampled due to a lack of financial resources, but there is a growing
consensus in the Hawaiian scientific community for the need to increase the current spatial
coverage of monitoring efforts and to re-sample more of the original HFBS transects (Pratt et al.
2009). Since the HFBS, surveys have been conducted by state, federal, and not-for-profit
agencies, primarily for the purposes of monitoring population sizes and changes, and in 2008, the
data acquired from these surveys was analyzed and compiled into a central repository (Camp et
al. 2009). However, the HFBS has been the only centralized, state-wide, assessment to this date.
While the National Park Service does sample some sites on Hawai’i Island and Maui annually,
surveys by the state were rotated around the islands on a five-year cycle until very recently,
leaving other agencies and local groups to fill in the gaps.
The avian community within the Pu’u ‘O ‘Umi Natural Area Reserve on Kohala, the
northernmost volcano out of the five that comprise Hawai’i Island, is one of the few native
bird populations in the state that has not been monitored at all since the HFBS. With the recent
addition of the ‘I’iwi (Drepanis coccinea) to the national endangered species list, the Pu’u ‘O
‘Umi NAR is designated as Federal Critical Habitat for Threatened and Endangered Species.
Avian malaria (Plasmodium relictum) has restricted native bird populations to high elevation
forests above approximately 1500m elevation, where colder temperatures constrain the densities
of disease vectors and reduce the prevalence of avian malaria (LaPointe et al. 2010). The high-
elevation refuge in which native Hawaiian birds can thrive is further shrinking as mosquitoes are
predicted to increase in elevation with global warming (Atkinson et al. 2009, Fortini et al. 2015).
Kohala has a maximum elevation of 1670m, and the avian community there is isolated from all
other indigenous bird populations due to the prevalence of avian malaria at lower elevations
surrounding the mountain, coupled with extensive deforestation due to almost two centuries of
cattle ranching. With a climate-induced increase in the prevalence of avian malaria encroaching
on the existing indigenous bird habitat in Kohala, it is critical to monitor this unique population.
We present here the first demographics and distribution survey results of the forest bird
community in Kohala since 1979. The primary goals of this survey were to: 1) determine if there
have been changes in the composition of bird community between the two surveys; 2) document
current population range of native and non-native species across the landscape; and 3) assess
whether any of the native species have declined since the most recent survey in the late 1970’s.
Methods
The Kohala mountain complex is the remnant of a shield volcano which forms the
northern-most promontory of Hawai’i Island. The study area encompasses 35.6 km2 within the
Pu’u ‘O ‘Umi Natural Area Reserve which is managed by the Department of Land and Natural
Resources under the Department of Forestry and Wildlife, with the southeastern edge abutting
private pastureland owned by Parker Ranch. The habitat is primarily a closed to open canopy
‘ohi’a-dominant (Metrosideros polymorpha) wet forest with frequent bogs and large, deep
valleys with steep sides. The windward slope of Kohala contains large patches of open canopy
‘ohi’a forest with a thick ‘uluhe (Dicranopteris linearis) understory. Within the study site, there
is some variability in understory density and the assemblages of native and non-native plants.
Average rainfall in the Kohala mountain complex is 2290 mm per year (Van Riper 1976).
We surveyed 143 stations across 13 transects in Pu’u ‘O ‘Umi Natural Area Reserve from
March through May 2017, during the typical breeding season months. We incorporated the 1979
HFBS data from 80 stations across 3 transects that occurred in our study site to create a global
detection function that would allow for comparing densities between survey years. The original
HFBS transects were inaccessible as the transects have been unused and are overgrown. Rather
than disturb pristine forest to re-establish the HFBS transects, we used existing Department of
Forestry and Wildlife weed transects which were installed in similar locations as the original
HFBS transects, following the data entry and collection protocols outlined by the Hawai’i Forest
Bird Interagency Database Project Avian Monitoring Entry Form (Camp et al. 2016).
Subsequent to the HFBS, consensus has been that breeding season is the best time to
conduct most single and multi-species surveys (Fancy 1996, Simon et al. 2002, Reynolds et al.
2003, Spiegel et al. 2006, Camp et al. 2009) as density estimates are generally higher during
breeding season (Simon et al. 2002). The HFBS Kohala surveys occurred in the last 3 days of
July, while our surveys ranged from February through April, with most survey days occurring in
March. Rather than match survey timing with the HFBS, we chose to use the currently
acceptable survey window so that future studies can use our study as a baseline.
We conducted Variable Circular-Plot counts following the methods for point transect
distance sampling described by Scott et al. (1986). We used eight-minute counts at each station,
where trained and calibrated observers recorded the species, detection type (seen, heard, or both),
and distance in meters from survey station center-point to birds detected. Time of sampling, local
vegetation information (canopy closure classification, canopy height, canopy cover percentage,
canopy flower abundance, dominant canopy species, dominant understory species), and weather
conditions (cloud cover, rain, wind, noise, and gust) were also recorded (Table 1). To maximize
detectability, counts were made only during the first 4 hours after dawn and were conducted only
during periods with low wind speed and low rainfall (Scott et al. 1981). Surveying was halted
when heavy rain or heavy gusts (>20 kph) were present, hindering the ability to detect birds.
Our analysis focused on seven species of forest birds (4 indigenous and 3 introduced,
Table 2, Figure 1): Hawai’i ‘amakihi (Chlorodrepanis virens), ‘Apapane (Himatione sanguinea),
‘I’iwi (Drepanis coccinea), Hawai’i ‘elepaio (Chasiempis sandwichensis), Red-billed leiothrix
(Leiothrix lutea), Japanese bush warbler (Cettia diphone), and Warbling white-eye (Zosterops
japonicus). While other species were detected, there were not enough detections for accurate
estimations of density and distribution for those species.
We used the program Distance 7.0 (Thomas et al. 2010) to calculate species-specific
densities (birds/km2) from the point transect data. We included the data from the HFBS for
stations that occurred within the study site, and post-stratified for each year using the global
detection function calculated across pooled strata. Observations from both survey years were
pooled together and analyzed as a single database to calculate global detection functions for each
species following the methods outlined by Buckland et al. (2001, 2004) and Thomas et al.
(2010). All data were modelled as exact measures. We restricted a priori model selection to half
normal and hazard-rate detection functions with two orders of expansions series and covariate
variables (canopy closure, canopy height, dominant canopy species, dominant understory
species, cloud cover, visual detections, auditory only detections, detection type, gust, noise, rain,
time, observation date, and observer), following methods described by Buckland et al. (2001,
2004), Burnham and Anderson (1998, 2002), and Thomas et al. (2010). Bootstrap methods were
used to determine variance, except with D. coccinea, where analytical methods were used as
there were not enough detections to use bootstrap methods. Simulation studies demonstrate that
both bootstrap and analytic methods are comparable, but that the uncertainty in the confidence
limits is better reflected in bootstrap methods (Buckland et al. 2001). To map densities, we used
the station-specific densities for each species and used a 1km radius circle neighborhood, the
smallest radius that would provide full coverage for the study site, with the spatial analyst point
density function with ArcGIS Desktop 10.7.1 (ESRI 2019).
We used an equivalence-testing approach to determine an ecologically meaningful trend
in change of species density (Manly 2008, Dixon and Pechmann 2005, Camp et al. 2008). We
chose to test conservative equivalence bounds equal to a 50% change in the population (Camp et
al. 2009) over 38 years. We defined changes in population density as increasing, decreasing,
stable, or inconclusive. A negligible trend (i.e. stable population) occurred when the slope lay
within the equivalence region, while an ecologically meaningful trend (i.e. increasing or
decreasing population) occurred when the slope lay outside the equivalence region. An
inconclusive result occurred when sample size did not allow for precise variation estimates
(Dixon and Pechmann 2005).
Results
We detected 2806 individuals of 15 different species across the 143 survey stations.
Seven species were detected enough times to allow for accurate density (birds/km2) and
abundance (total population) estimations. We calculated the densities and abundances for these
seven species within the study site for both the 1979 HFBS and current 2017 surveys (Table 3)
and observed changes in species densities ranging from -8.4% (C. sandwichensis) to +714% (D.
coccinea). The highest population density was H. sanguinea with 1857 birds/km2 on average,
and the lowest population density was C. diphone with 126 birds/km2 on average. With the
exception of C. sandwichensis, we observed at least a 279% increase in species density. We used
the density estimates for each station to create density maps of the number of birds per square
kilometer for each species (Figures 2-8). Equivalence testing showed meaningful increases in
population densities for all species present in both surveys, with the exception of C.
sandwichensis, which proved inconclusive. We also observed the widespread occurrence of C.
diphone, a species which was not present at all within the study site during the HFBS survey in
1979.
Discussion
Our study has identified large increases in the population densities of 6 out of the 7 bird
species studied, suggesting that the native bird community in Kohala is not in decline. Any
positive change, though, should be regarded with extreme caution. Most of our study species
exhibit a breeding season from December to May, except for C. sandwichensis which breeds
from March through August (Simon et al. 2002). Therefore, all our study species were surveyed
outside of breeding season during the HFBS, except for C. sandwichensis, which is notably the
only species which exhibited a negative density change.
There are other difficulties aside from a bias in the survey timing that limit our ability to
make visual comparisons between the two survey years. One such difficulty is that the scope of
the HFBS species density maps is greater than the current study, making it difficult to determine
whether species ranges in Kohala have expanded or contracted. Species ranges in the current
study are limited by area surveyed, rather than the actual extent of species ranges. There is also a
stark difference between the resolution of the current species density maps which are rastered at
a much finer pixel scale than the hand-drawn HFBS maps.
While it is difficult to make fine-scale comparisons between years, there are some stark
differences which are obvious, such as a new species introduction, and the retreat of D. coccinea
from lower elevations. While the current density maps show increases in density since 1979 for
all but one species surveyed, the overall density patterns remain spatially similar for most
species. There are, however, two species which exhibit apparent changes in density patterns over
time, D. coccinea and H. sanguinea. In our current survey, D. coccinea is conspicuously absent
from two locations in which it was found in 1979: the northwestern-most corner of the survey
area, and a large section of the windward slope to the northeast. Both locations are below 1350 m
where avian malaria is likely to be more prevalent and are comprised of habitat which includes
more non-native plant species than the upper-elevation forest. As well, both D. coccinea and H.
sanguinea exhibit much higher densities than found in 1979 above 1350 m in the central-western
section of the study site. This area incudes management by the Natural Area Reserve System and
Kohala Watershed Partnership to exclude wild pigs via fenced units, remove non-native and
invasive plant species, and plant native understory species. It may also be an area that is a refuge
from avian malaria due to temperature and elevation. More research is necessary to determine
why these two species are increasing density in this area.
We also observed the current prevalence of a species, C. diphone, which was previously
undetected anywhere on Hawai’i Island during the HFBS in 1979 (Scott et al. 1986). C. diphone
was first observed on the Island of Hawai’i in 1997 (Nelson and Vitts 1998). With a diet that
consists of both insects and nectar (Scott et al. 1986), C. diphone overlaps the diet of all native
species in Kohala to some degree. Therefore, it is important to monitor how this population
changes over time, and how it may influence the population dynamics of native species in the
region.
Our study provides a framework of survey stations and protocol for future surveys in the
region. The forest bird community in Kohala should continue to be surveyed and monitored,
especially with the recent addition of D. coccinea to the endangered species list. Another study
species, C. sandwichensis, is currently listed as vulnerable. The number of native forest bird
populations in Hawai’i which exhibit stable or increasing population densities are greatly
outweighed by those in decline (Camp et al. 2009). Therefore, it is especially important to
continue to repeat surveys in Kohala during breeding season to determine with more certainty
whether the native populations here are truly stable or increasing. It may also be of interest to
repeat the survey during the same monthly time period as the HFBS to allow for a more direct
comparison. While most of the native population in Kohala is protected in the Pu’u ‘O ‘Umi
Natural Area Reserve, future surveys could warrant the creation of more intensive recovery and
protection plans in the region.
Acknowledgements
We thank the Hawai’i State Department of Land and Natural Resources for permission to
work at the study areas, and for their logistical and manpower support in conducting surveys.
Surveys within the Pu’u ‘O ‘Umi Natural Area Reserve in Kohala were conducted under
Department of Forestry and Wildlife permit NARSFY17-54. We would also like to thank Parker
Ranch for allowing easement to the Pu’u ‘O ‘Umi Natural Area Reserve, and University of
Hawai’i undergraduate student E. Lough for aiding in preparing transects and acting as a
secondary on many surveys.
Table 1. A list of the measured habitat and atmospheric variables recorded with a description of
what was recorded and in what increments.
Covariate Description
Time of Sampling 24hr reading
Cloud Cover 0-100% in 10% increments
Rain 0-4; No rain - Heavy rain
Wind 0-4 on the Beaufort Scale
Gust 0-4 on the Beaufort Scale
Noise 0-4; No noise - Could potentially
miss bird song at 10m distance
Canopy Flower Abundance 0-100% of bloom in closest 10
trees, in 10% increments. Trace
amounts recorded as 0.1
Canopy Cover Percentage 0-100% of sky covered by canopy,
in 10% increments
Canopy Height Average of estimated height, in
meters, of the 3 apparent tallest
trees within the circle plot
Canopy Closure
Classification Closed or Open
Dominant Canopy Species Single dominant species, or mixed
two species
Dominant Understory
Species Single dominant species, or mixed
two species
Table 2. Species with enough detections to yield population estimates along with their
endemic status, diet proclivity, and conservation status.
Species Status Diet Conservation Status
Hawai’i ‘amakihi (Hemignathus virens) Endemic Generalist Least concern
‘Apapane (Himatione sanguinea) Endemic Nectarivore Least concern
‘I’iwi (Drepanis coccinea) Endemic Nectarivore Threatened
Hawai’i ‘elepaio (Chasiempis sandwichensis) Endemic Insectivore Vulnerable
Red-billed leiothrix (Leiothrix lutea) Introduced Frugivore Least concern
Japanese bush warbler (Cettia diphone) Introduced Insectivorous/Generalist Least concern
Warbling white-eye (Zosterops japonicus) Introduced Generalist Least concern
Table 3. Population density (birds/km2), abundance (population estimate), and 95% confidence interval estimates by time period.
Sampling effort (number of stations sampled) and number of birds used to estimate densities are presented.
Species Year Density
95% Confidence
Interval Abundance
95% Confidence
Interval
No.
Stations
No.
Birds
Hawai’i ‘amakihi (Hemignathus virens)
1979 533.54 388.29 733.12 19652 14302 27004 59 186
2017 1491 1036.8 2144.3 54920 38188 78984 134 638
‘Apapane (Himatione sanguinea)
1979 324.06 224.77 467.21 11936 8279 17209 55 126
2017 1857.6 1409.7 2447.8 68423 51924 90164 140 881
‘I’iwi (Drepanis coccinea)
1979 18.74 6.4 54.86 690 236 2021 7 8
2017 133.8 79.02 226.56 4928 2910 8345 32 58
Hawai’i ‘elepaio (Chasiempis
sandwichensis)
1979 357.86 214.6 596.76 13181 7904 21981 61 127
2017 327.97 227.62 472.55 12080 8384 17406 93 193
Red-billed leiothrix (Leiothrix lutea)
1979 181.9 108.28 305.58 6700 3989 11256 61 170
2017 569.06 447.73 723.26 20961 16492 26641 141 434
Japanese bush warbler (Cettia diphone)
1979 - - - - - - - -
2017 126.87 41.5 387.84 4673 1529 14286 69 116
Warbling white-eye (Zosterops
japonicus)
1979 265.26 116.54 603.74 9770 4293 22238 40 65
2017 1266.9 712.86 2251.6 46666 26257 82936 128 412
House finch (Haemorhous mexicanus)
1979 - - - - - - 3 3
2017 - - - - - - 2 2
Melodious laughing thrush (Garrulax
canorus)
1979 - - - - - - 30 46
2017 - - - - - - 14 16
Eurasian skylark (Alauda arvensis)
1979 - - - - - - - -
2017 - - - - - - 1 1
Hawaiian hawk (Buteo solitarius)
1979 - - - - - - - -
2017 - - - - - - 2 4
Erckel's rancolin (Pternistis erfckelii)
1979 - - - - - - - -
2017 - - - - - - 7 12
Ring-necked pheasant (Phasianus
colchicus)
1979 - - - - - - 1 1
2017 - - - - - - 12 15
Khalij pheasant (Lophura leucomelanos)
1979 - - - - - - - -
2017 - - - - - - 9 14
Figure 1. Images of the seven species with population estimates. Top (left to right): Japanese
bush warbler (Cettia diphone), Red-billed leiothrix (Leiothrix lutea), Warbling white-eye
(Zosterops japonicus); Bottom (left to right): ‘I’iwi (Drepanis coccinea), Hawai’i ‘elepaio
(Chasiempis sandwichensis), Hawai’i ‘amakihi (Hemignathus virens), and ‘Apapane (Himatione
sanguinea).
Figure 2. Population density of Hawai’i ‘amakihi (Hemignathus virens) within the Pu’u ‘O ‘Umi
NAR study site in Kohala. Red circles indicate 1979 HFBS survey stations, while green circles
indicate the 2017 survey stations. Red contour lines indicate 500-feet elevation gradients, with
the highest gradient being 5000ft elevation.
Figure 3. Population density of ‘Apapane (Himatione sanguinea) within the Pu’u ‘O ‘Umi NAR
study site in Kohala. Red circles indicate 1979 HFBS survey stations, while green circles
indicate the 2017 survey stations. Red contour lines indicate 500-feet elevation gradients, with
the highest gradient being 5000ft elevation.
Figure 4. Population density of ‘I’iwi (Drepanis coccinea) within the Pu’u ‘O ‘Umi NAR study
site in Kohala. Red circles indicate 1979 HFBS survey stations, while green circles indicate the
2017 survey stations. Red contour lines indicate 500-feet elevation gradients, with the highest
gradient being 5000ft elevation.
Figure 5. Population density of Hawai’i ‘elepaio (Chasiempis sandwichensis) within the Pu’u ‘O
‘Umi NAR study site in Kohala. Red circles indicate 1979 HFBS survey stations, while green
circles indicate the 2017 survey stations. Red contour lines indicate 500-feet elevation gradients,
with the highest gradient being 5000ft elevation.
Figure 6. Population density of Red-billed leiothrix (Leiothrix lutea) within the Pu’u ‘O ‘Umi
NAR study site in Kohala. Red circles indicate 1979 HFBS survey stations, while green circles
indicate the 2017 survey stations. Red contour lines indicate 500-feet elevation gradients, with
the highest gradient being 5000ft elevation.
Figure 7. Population density of Japanese bush warbler (Cettia diphone) within the Pu’u ‘O ‘Umi
NAR study site in Kohala. Red circles indicate 1979 HFBS survey stations, while green circles
indicate the 2017 survey stations. Red contour lines indicate 500-feet elevation gradients, with
the highest gradient being 5000ft elevation.
Figure 8. Population density of Warbling white-eye (Zosterops japonicus) within the Pu’u ‘O
‘Umi NAR study site in Kohala. Red circles indicate 1979 HFBS survey stations, while green
circles indicate the 2017 survey stations. Red contour lines indicate 500-feet elevation gradients,
with the highest gradient being 5000ft elevation.
Literature Cited
Anonymous 2019. ArcGIS Desktop. 10.7.1 edition. Environmental Systems Research Institute,
Redlands, CA.
Atkinson, C. T., and D. A. LaPointe. 2009. Introduced avian diseases, climate change, and the
future of Hawaiian honeycreepers. Journal of Avian Medicine and Surgery 23(1):53-63.
Buckland, S. T., D. R. Anderson, K. P. Burnham, J. L. Laake, D. L. Borchers, and L. Thomas.
2004. Advanced distance sampling. Oxford University Press Oxford, .
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. :.
Burnham, K. P., and D. R. Anderson. 1998. Model selection and inference: a practical
information-theoretical approach. New-York: Springel-Verlag :.
Burnham, K. dr AnderSon. 2002. Model selection and multimodel inference: a practical
information-theoretic approach. Ecological Modelling.Springer Science & Business Media, New
York, New York, USA :.
Camp, R. J., P. M. Gorresen, T. K. Pratt, and B. L. Woodworth. 2009. Population trends of
native Hawaiian forest birds, 1976–2008: the data and statistical analyses :.
Camp, R. J., N. E. Seavy, P. M. Gorresen, and M. H. Reynolds. 2008. A statistical test to show
negligible trend: comment. Ecology 89(5):1469-1472.
Camp, R., R. Lam, V. Matsui, A. Cramer, T. Pratt, B. Woodsworth, P. M. Gorressen, and S.
Nash. 2016. Avian monitoring entry form version 2.1 Hawaii forest bird inter-agency database
project. :.
Dixon, P. M., and J. H. Pechmann. 2005. A statistical test to show negligible trend. Ecology
86(7):1751-1756.
Fancy, S. G., S. A. Sandin, M. H. Reynolds, and J. D. Jacobi. 1996. Distribution and population
status of the endangered'Akiapola'au. :.
Fortini, L. B., A. E. Vorsino, F. A. Amidon, E. H. Paxton, and J. D. Jacobi. 2015. Large-Scale
Range Collapse of Hawaiian Forest Birds under Climate Change and the Need for 21st Century
Conservation Options [corrected. PloS one 10(10):e0140389.
Johnson, L., R. J. Camp, K. W. Brinck, and P. C. Banko. 2006. Long‐term population
monitoring: Lessons learned from an endangered passerine in Hawai'i. Wildlife Society Bulletin
34(4):1055-1063.
LaPointe, D. A., M. L. Goff, and C. T. Atkinson. 2010. Thermal constraints to the sporogonic
development and altitudinal distribution of avian malaria Plasmodium relictum in Hawai'i.
Journal of Parasitology 96(2):318-325.
Loope, L. L. 1998. Hawaii and the Pacific islands. Status and trends of the nation’s biological
resources 2:747-774.
Manly, B. F. 2008. Statistics for environmental science and management. Chapman and
Hall/CRC, .
MUNRO, G. 1944. Birds of Hawaii. Honolulu, Hawaii. :.
Nelson, J., and A. Vitz. 1998. First reported sighting of Japanese bush-warbler (Cettia diphone)
on the island of Hawaii. ‘Elepaio 58:1.
Paxton, E. H., R. J. Camp, P. M. Gorresen, L. H. Crampton, D. L. Leonard, and E. A.
VanderWerf. 2016. Collapsing avian community on a Hawaiian island. Science Advances
2(9):e1600029.
Pratt, T. K., C. T. Atkinson, and P. C. Banko. 2009. Conservation biology of Hawaiian forest
birds: implications for island avifauna. Yale University Press, .
Reynolds, M. H., R. J. Camp, B. M. Nielson, and J. D. Jacobi. 2003. Evidence of change in a
low-elevation forest bird community of Hawai'i since 1979. Bird Conservation International
13(3):175-187.
Scott, J. M., J. D. Jacobi, and F. L. Ramsey. 1981. Avian surveys of large geographical areas: a
systematic approach. Wildlife Society Bulletin :190-200.
Scott, J. M., and C. B. Kepler. 1985. Distribution and abundance of Hawaiian native birds: a
status report. :.
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. :.
Scott, J. M., and F. L. Ramsey. 1981. Length of count period as a possible source of bias in
estimating bird densities. Red 1(61):58.
Simon, J. C., T. K. Pratt, K. E. Berlin, J. R. Kowalsky, S. G. Fancy, and J. S. Hatfield. 2002.
Temporal variation in bird counts within a Hawaiian rainforest. The Condor 104(3):469-481.
Spiegel, C. S., P. J. Hart, B. L. Woodworth, E. J. Tweed, and J. J. LeBRUN. 2006. Distribution
and abundance of forest birds in low-altitude habitat on Hawai'i Island: evidence for range
expansion of native species. Bird Conservation International 16(2):175-185.
Thomas, L., S. T. Buckland, E. A. Rexstad, J. L. Laake, S. Strindberg, S. L. Hedley, J. R.
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(1):5-
14.
VAN, R., and V. R. C III. 1976. ASPECTS OF HOUSE FINCH BREEDING BIOLOGY IN
HAWAII. :.
Wilcove, D. S. 2005. Ecology. Rediscovery of the ivory-billed woodpecker. Science (New York,
N.Y.) 308(5727):1422-1423.
Chapter 2: The effects of isolation on the song diversity and acoustic characteristics of a
Hawaiian forest bird
Abstract
Populations that are isolated have a high potential to diverge acoustically. As forest birds
use song primarily to defend territory and attract mates, major changes in the acoustic properties
of song can reinforce reproductive isolation, leading to assortative mating and speciation. While
some native species like the Hawai’i ‘amakihi have demonstrated an evolved tolerance to avian
malaria and begun to repopulate low-elevation habitats and reconnect populations, it is unclear
the level to which song divergence has already reinforced reproductive isolation. To address this
question, I examined how long-term isolation has affected the song diversity and acoustic traits
of the Kohala population of the Hawai’i ‘amakihi compared to the geographically-nearest four
populations. I recorded and analyzed 7,183 ‘amakihi songs across all five populations from 2013
to 2017, with a total of 57 unique song types. I found significant variability in song types among
all five populations and relatively uniform variation within populations. Principal component
analysis of temporal and frequency measurements revealed that the Kohala population is unique
in its acoustic characteristics. Future behavioral assays can indicate whether this uniqueness has
reinforced reproductive isolation in Kohala ‘amakihi over larger time scales.
Introduction
Nearly half of all bird species - the sub-order oscine Passerines - learn and transmit vocal
signals between individuals (Kroodsma and Miller 1996, Catchpole and Slater 2008), creating a
flow of socially learned information, or memes, across geographical space (Lynch 1996). A
meme is the simplest unit of cultural transmission (Dawkins 1976), and in bird song, acoustic
memes can be defined as an individual syllable, a segment of linked syllables, or an entire song
(Lynch 1996). Variations in meme and song diversity are linked to regional meme generation
and mutation during the learning process (Lynch et al. 1989, Slabbekoorn and Smith 2002), and
can be exacerbated by geographic isolation (Laiolo 2008, Parker et al. 2012). Habitat loss and
fragmentation have a profound effect on the cultural transmission of songs, with fragmentation
and anthropogenic barriers resulting in both greater differences between non-neighbors as well as
greater mimicry between neighbors (Laiolo and Tella 2005, Hart et al. 2017).
Similar to the way gene pools diverge following isolation, populations that are separated
have high potential to diverge acoustically, largely due to random sampling error and differences
in a species’ mode of song learning (Lynch 1966). If individuals move among populations,
memes will spread and reduce between-population acoustic diversity (Lynch 1996). Therefore,
species-dependent characteristics such as mobility and home-range size can affect song diversity,
especially in small populations (Pavlova et al. 2012, Sebastián-González and Hart 2017). Song
divergence can both affect a male’s success in obtaining a mate, as well as a male’s success in
establishing and defending a territory (Ellers and Slabberkoon 2003), leading to assortative
mating and speciation (Grant and Grant 1997). This rate of speciation can increase with greater
differences in acoustic traits (Irwin et al. 2001).
Divergence in song structure among populations can also be influenced by differences in
environmental conditions. Sound energy can be reduced or redirected by terrain features,
substrates, and vegetation, and since sounds differ in how far they can travel in a given habitat,
selection will likely favor individuals that produce sounds that have the greatest broadcast area,
especially in taxa which use song and call vocalizations to create and maintain contact with
conspecifics that are not in visual contact (Acoustic Adaptation Hypothesis: Morton 1975, Wiley
and Richards 1978, Boncoraglio and Saino 2007). Habitat-induced selection pressures can lead
to parallel divergence in both song and fitness measures and may give females auditory cues to
which males are more fit for a given habitat (Slabberkoorn and Smith 2002). Conspecific
populations that occupy different habitats can evolve temporal and frequency characteristics that
maximize transmission in each habitat, increasing acoustic variation (Tobias et al. 2010), and
competition for acoustic space within a habitat can influence birds to modify the temporal or
frequency aspects of their songs (Cardoso and Price 2010, Hart et al. 2015, Slabbekoorn 2017).
While many studies focus solely on either sexual selection via acoustic memes or habitat-
induced selection pressures and ecological divergence, it is more likely that a combination of the
two facets are responsible for speciation in songbirds (Slabberkoorn and Smith 2002).
Hawaiian honeycreepers are a group of oscine Passerines (Fringillidae) endemic to
Hawai’i, which are of particular conservation interest due to widespread, rapid population
declines and high rates of extinction post-human contact (Paxton et al. 2016). Anthropogenic
habitat destruction from land-use changes and introduced habitat modifiers such as ungulates, as
well as pressures from introduced diseases, competitors, and predators have led to the
fragmentation and isolation of habitat for all native Hawaiian birds including the honeycreepers
(van Riper and Scott, 2001). Since its introduction in Hawai’i, avian malaria (Plasmodium
relictum) has restricted Hawaiian honeycreeper populations to high elevation forests where
colder temperatures limit the densities of disease vectors (LaPointe et al. 2010, van Riper et al.
1986). However, some species such as the Hawaii ‘amakihi (Chlorodrepanis virens) have begun
to repopulate low-elevation habitats (Woodsworth et al. 2005) and have demonstrated an evolved
tolerance to avian malaria (Atkinson et al. 2013). We are only beginning to comprehend how
isolation and fragmentation affects acoustic meme flow and song structure among populations of
honeycreepers (see Sebastian-Gonzalez and Hart 2017, Pang-Ching et al. 2017, Paxton et al.
2019), and it is unclear the level to which song divergence has reinforced reproductive isolation
in intra-island populations of these species.
In this study, we examine how long-term isolation has affected the meme diversity and
acoustic traits of the Kohala population of the Hawai’i ‘amakihi – the most spatially isolated
population on Hawai’i island – compared to the geographically-nearest four populations of
‘amakihi. We hypothesize that isolation in a unique habitat has led to the divergence of songs in
the Kohala population of ‘amakihi, and that one of the four populations, likely the last population
to have information transfer with Kohala, will have a more similar meme composition to the
Kohala population. Specifically, we addressed the following questions: 1) What is the degree of
divergence in both meme repertoire and qualitative acoustic properties among the five
populations? 2) What population(s) exhibit the most similar meme repertoires to the Kohala
population?
Methods
Study Species
The Hawai’i ‘amakihi (Figure 1) is a generalist honeycreeper that feeds primarily on
nectar and invertebrates (Pimm and Pimm 1982) and is abundant in forest habitats on the islands
of Hawai’i and Maui (Lindsey et al. 1998). As oscine Passerines, ‘Amakihi are assumed to
acquire acoustic traits such as memes via cultural transmission (Beecher and Brenowitz 2005).
Here, we focused on the primary song, a single meme consisting of a repeated syllable in an
undulating trill (Figure 2), used principally for courtship interactions and territorial defense (van
Riper, 1987).
Study Sites
We recorded ‘amakihi songs from five populations on Hawai’i island: 1) Kohala
(KOHA) within the Pu’u ‘O ‘Umi Natural Area Reserve on Kohala, 2) Pu’u La’au (PULA) on
the Eastern flank of Maunakea, 3) Hualalai (HUAL) within the Pu’u Wa’awa’a Forest Reserve
on the Northeastern flank of Hualalai, 4) Hakalau (HNWR) within the Hakalau Forest National
Wildlife Refuge on the Western flank of Maunakea, and 5) the Kipukas (KIPU), a network of
naturally fragmented forest on the Eastern flank of Mauna Loa (Figure 3). The HUAL and PULA
populations are located in dry forest habitats with mean annual rainfall between 50-64cm/yr (van
Riper 1976), while the KOHA, HNWR, and KIPU populations are located in wet forest habitats
with mean annual rainfall between 229-500cm/yr (van Riper 1976, Pang-Ching et al. 2017).
Within each population, we chose 8 spatially separate recording sites at least 100m apart for
subsampling. The sites PULA were in an open shrubland with a sparse canopy of māmane
(Sophora chrysophylla). The sites at the other 4 locations were in a closed-canopy forest
dominated by ‘ōhi’a (Metrosideros polymorpha) and/or koa (Acacia koa). Average canopy
height was similar among the closed-canopy forest sites, though there was some variability in the
assemblages of native and non-native plants and understory density.
Acoustic Recordings
Acoustic recording at KOHA and PULA took place from 2016 through 2017 from March
through May of each year, in accordance with breeding season when the birds are most vocal
(van Riper 1987). Songs from HUAL, HNWR, and KIPU were taken from existing recordings
from 2013 through 2015 during the same months. At each site we stationed a recording device
pre-programmed to record at regular intervals during the hours of higher acoustic activity. All
recording was done with a Wildlife Acoustics Songmeter SM2 with an omnidirectional
microphone with a sensitivity of -35dBV pa-1 and frequency response of 20-20,000 Hz.
Recordings were made in 24-bit wav-file format with a sampling rate of 44.1kHz. At the KOHA
and PULA sites we programmed the devices to record for 5 minutes every 10 minutes from 6:00-
11:00, and additionally from 15:00-17:00. At the HUAL, HNWR, and KIPU sites, existing
recordings were for 5 minutes every 20 minutes, from 6:00-11:00, and additionally from 15:00-
16:00. Recording devices were placed approximately 1.5m from the ground and were left in the
field for approximately 2 weeks. A maximum of 2 recorders were active at any given time within
a location to cover the maximum temporal variability in the meme diversity of a population and
minimize temporal bias.
Song Analysis and Classification
For each recording site, we randomly selected 4 days of recording to analyze (one at the
beginning of the recording time, two in the middle, and one at the end). Both the PULA and
KOHA sites are in active hunting areas, with PULA also being in close proximity to a military
training site. If there was audible artillery fire, heavy rain, or wind during any days we selected,
we used the recordings from the nearest day without noise. In the selected recordings, we
manually searched for ‘amakihi songs, an acoustic meme comprised of a repeated syllable as
described above. We selected all songs with a definitive beginning and ending (i.e. were not
obscured by other noises) for analysis and classification. ‘Amakihi songs were
spectrographically imaged and analyzed using Raven Pro 1.4 (Bioacoustics Research Program
2011). We classified memes based on frequency range, spectrogram shape (i.e. the physical
shape of syllables within the spectrogram), and aural characteristics. All classification was
completed by a single person (ESG) to ensure classification consistency. We saved a sample of
the spectrogram from each meme class to be able to compare different classes posteriori, and to
allow for the repetition of the classification if necessary.
To quantify and compare frequency and temporal song characteristics among the five
populations, we measured five parameters: song length (total length of song from beginning to
end, in seconds), number of syllables in a song, peak frequency (the level at which the most
energy is expelled within a song, in Hz), low frequency (the lowest frequency level measured
during the entire song, in Hz), and high frequency (the highest frequency level measured during
the entire song, in Hz).
Statistical Analysis
Meme Analysis
To compare differences in meme repertoires within and among populations, we created a
presence-absence matrix of each meme for the 20 sites (5 populations and 4 recording sites
within population) and used a Jaccard similarity index as a measure of distance between song
types. The Jaccard coefficient measures the dissimilarity in richness of song types recorded
between each pair of sites, with a value of 1 indicating no shared memes and a value of 0
indicating all memes were the same between sites. We performed a Mantel test to determine if
the Jaccard index and physical distance between sites were correlated. We used three
nonparametric multivariate techniques with the R package “vegan” (Oksanen et al. 2017). First,
to visualize differences in meme types among populations, we used a nonmetric
multidimensional scaling (NMDS) ordination plot with the metaMDS function. Second, we
tested for differences in meme repertoire using a permutational multivariate analysis of variance
(PERMANOVA) with the adonis function. Last, we compared differences in the variability of
meme repertoire within populations using a permutational analysis of multivariate dispersion
(PERDISP) with the betadispersion function. PERMANOVA tests for differences in the
centroids of multivariate groups (Anderson and Walsh 2013), while PERDISP tests for the
homogeneity of distances of observations from the centroids of multivariate groups (Anderson
2006). P-values for test statistics (pseudo-F) of the main effects test and posterior pairwise
comparisons were based on 9999 permutations.
Acoustic Property Analysis
We also examined the differences in song acoustic characteristics among populations. We
first tested each acoustic variable in individual linear mixed models by location with recording
device sites as a random variable, using R package “lme4” (Bates et al. 2014). We then created a
principal component analysis (PCA) based on the five measured acoustic variables for each
recorded song. We used the predicted values of the principal component scores for the first two
axes in linear mixed models by location with sites as a random variable to determine which
locations displayed significant differences in the PCA axes.
All statistical analyses were conducted using the statistical program R version 3.6.1 (R
Core Team 2019).
Results
We recorded and analyzed 7,183 ‘amakihi songs across all five populations with a total
of 57 unique song types. Two-thirds of song types (n=38) were recorded within only one
location. Physical distance between sites was correlated with Jaccard similarity among sites
(partial mantel test; p = 0.009, rs = 0.2).
Meme Analysis
Nonmetric multidimensional scaling ordination indicated differentiation in song-type
repertoires among all populations. There was little overlap in song repertoires among the
populations and no overlap in song repertoires between the Kohala population and other four
populations (Figure 4). The PERMANOVA analysis supported the ordination, showing
significant differences in meme repertoires among populations (F4,15 = 2.2, p < 0.001, R2 = 0.37).
Pairwise comparisons showed that differences among populations were driven relatively equally,
with the most similar sites being Hakalau Forest National Wildlife Refuge compared to
Hualalai (p = 0.05), and the least similar sites being Hualalai compared to Pu’u La’au (p = 0.02)
(Table 1). The PERDISP analysis indicated that there was variability in memes within
populations (F4,15 = 3.17, p = 0.03) (Figure 5). Pairwise comparisons showed that variability was
mostly uniform, with only the Pu’u La’au population having a greater variability than the
Hualalai population (p = 0.05) (Table 1).
Acoustic Property Analysis
We found variability in both spectral and temporal aspects of ‘amakihi song among
populations, most notably in the KOHA and PULA populations. ‘Amakihi songs in the KOHA
population contain approximately 78% more notes (t26.76 = 7.8, p <0.001) with no difference in
song length among populations. ‘Amakihi songs in the PULA population were 28% lower in
low-end frequency than the other four populations (t27.73 = -5.14, p <0.001). A PCA analysis of
the five measured acoustic variables showed that three principal component axes explained 77%
of the variation in acoustic characteristics (Table 2). Based on linear mixed models for the first
two principal component axes with each recording device as a random variable, we found
significant differences in acoustic characteristics of ‘amakihi songs among populations.
Specifically, the KOHA population (PC1: t24.52 = -2.87, p < 0.01; PC2: t26.10 = 2.32, p = 0.03) and
PULA population (PC1: t25.19 = 2.3, p = 0.03; PC2: t26.66 = 5.74, p < 0.001) were found to be
unique in their acoustic characteristics (Figure 6).
Discussion
We found significant differences in both meme repertoires and the acoustic
characteristics of ‘amakihi songs among populations. Differences in meme repertoires were
strongly associated with geographical distance, however, the closest population to the Kohala
community was not the most similar in meme repertoires. The Hualalai population and part of
the Hakalau NWR population exhibited the most similar meme repertoires to the Kohala
‘amakihi. It is important to note that while we are attempting to identify which location(s) have
the most similar meme composition to the Kohala population, we did find that all locations were
significantly different from each other, so any degree of similarity is small.
Decreases in pair-wise similarity of acoustic signals with increasing distance among
populations has been documented over relatively large spatial scales for several avian taxa
including the Dupont’s lark Chersophilus duponti (Laiolo and Tella 2006), grey shrike-thrush
Colluricincla harmonica (Pavolva et al. 2012), and ‘I’iwi Drepanis coccinea (Sebastian-
Gonzalez and Hart 2017). While we also found similar isolation-by-distance patterns in meme
composition among our study sites, there are other factors to consider besides distance.
Habitat connectivity may also play a part in the levels of similarity among populations.
Cultural transmission may be limited by the organisms’s ability to move between habitats
(Laiolo and Tella 2006). Kohala is separated from neighboring populations by large areas of
pasture and agricultural land. Historically, ‘amakihi inhabited more contiguous forest habitats on
Hawai’i island (Pratt and Jacobi 2009), and while they are mobile within their home-range
(Knowlton et al. 2017) of 0.1-1.1 ha (Lindsey et al. 1998), the distance between suitable habitats
in PULA and HUAL to Kohala is 37 and 42 km respectively. Habitat loss due to land use
changes can severely affect cultural transmission (Laiolo and Tella 2005), and the anthropogenic
change of habitat between Kohala and the neighboring populations at HUAL and PULA could
have been compounded by other isolating factors such as disease. The Kohala population is
separated from all other populations by large tracts of warm, lowland habitat that contain a high
prevalence of avian malaria (Lapointe et al. 2010).
The high degree of song-type variability within the PULA population may also be
impacting comparisons with other populations. It is unclear what might be influencing this
increased variability at PULA, but this may also be habitat-driven. The habitat at Pu’u La’au is a
much more open, dry shrubland with a sparse, open canopy of māmane (Sophora chrysophylla)
in fragmented stands. More research is needed to determine why there is such high song type
variability in ‘amakihi at Pu’u La’au.
We also found significant temporal and spectral differences in the songs of the Kohala
and Pu’u La’au populations that set themselves apart from the other communities we studied.
The low-end frequency of ‘amakihi songs in the Pu’u La’au population were 28% lower than the
other populations, and the PCA axis with the strongest significance was the axis that included
low frequency. This observation does support the predictions of the Acoustic Adaptation
Hypothesis on the influence of open habitat on frequency characteristics of bird song: in open
habitats, lower frequency sounds attenuate more slowly, thereby increasing broadcast range
(Morton 1975). ‘Amakihi have also been found to follow the peak frequency predictions of the
Acoustic Adaptation Hypothesis in two of the same study areas (Sebastian-Gonzalez et al. 2018).
‘Amakihi songs in the Kohala population had 78% more notes than the other populations with no
difference in song length, and the PCA axis with the strongest significance was the axis that
included the number of notes per song. This faster song does not follow the Acoustic Adaptation
Hypothesis which states that in forested habitats such as Kohala, rapid frequency modulations
are masked by an increase in reverberations off stationary objects such as vegetation, thereby
reducing broadcast area (Wiley and Richards 1978, Wiley 1991). In a meta-analysis,
Boncoraglio and Saino (2007) found that there were no significant effects of habitat on temporal
aspects of song. This suggests that other selective pressures are influencing ‘amakihi songs in
Kohala.
Isolation due to habitat loss and introduced disease has likely led the Kohala population
of ‘amakihi to diverge both in meme repertoires and acoustic characteristics, a pattern which is
well documented in song-learning passerines (Slabbekoorn and Smith 2002, Parker et al. 2012).
Differences in habitat alone do not seem to be prompting these changes, and sexual selection on
these song traits may be reinforcing the reproductive isolation of this population. Behavioral
experiments should be conducted to determine whether other populations of ‘amakihi on Hawai’i
Island have a territorial or mating response to the recently diverged songs of the Kohala
population, as song divergence may have already acted as a reproductive-isolating mechanism. If
the recent trend of ‘amakihi demonstrating tolerance to avian malaria continues, the Kohala
population may integrate with other populations of ‘amakihi. An inability for other populations
to recognize the unique Kohala songs may indicate that the Kohala ‘amakihi population would
continue to remain reproductively separate with the potential to either diverge into a subspecies
or to go extinct.
Acknowledgements
We thank the Hawai’i State Department of Land and Natural Resources for permission to
work at the study areas. Recordings within the Pu’u ‘O ‘Umi Natural Area Reserve in Kohala
were conducted under Department of Forestry and Wildlife permit NARSFY17-54. We would
also like to thank Parker Ranch for allowing easement to the Pu’u ‘O ‘Umi Natural Area
Reserve.
Figure 1. Photograph of an adult Hawai’i ‘Amakihi
Figure 2. Example of an acoustic meme for Hawai’i ‘amakihi. The square in red is selecting an
entire song, while the square in blue is selecting a syllable. The y-axis represents song
frequency (kHz) while the x-axis represents time (s).
Figure 3. Map of study sites. Site name abbreviations can be found in the study site paragraph of
the methods section. Lower right square indicates the area of interest on Hawai’i Island.
Figure 4. Nonmetric multidimensional scaling (NMDS) ordination of song types (stress = 0.1)
based on Jaccard similarity indices of song type presence/absence for 20 sites sampled across 5
populations. Each point represents a site, while the dashed ellipses denote 95% confidence
intervals for each population. Shorter distances between sites indicates greater similarity in the
occurrence of song types than those further apart. NMDS represents the relative (rank-order)
similarities among points on an arbitrary scale, therefore the axes lack labels.
Pairs F R2
P-value
(PERMANOVA)
P-value
(PERDISP)
HNWR vs HUAL 1.45 0.19 0.05 0.34
HNWR vs KIPU 1.56 0.21 0.03 0.97
HNWR vs KOHA 2.28 0.28 0.03 0.53
HNWR vs PULA 1.45 0.19 0.02 0.75
HUAL vs KIPU 1.94 0.24 0.03 0.67
HUAL vs KOHA 3.16 0.35 0.03 0.99
HUAL vs PULA 2.2 0.26 0.02 0.05
KIPU vs KOHA 2.71 0.31 0.03 0.86
KIPU vs PULA 1.52 0.2 0.02 0.41
KOHA vs PULA 1.76 0.23 0.03 0.09
Table 1. PERMANOVA (among-population song variability) and PERDISP (within-population
song variability) pair-wise comparisons among locations. F and R2 fit values are for
PERMANOVA test. P-values for PERDISP test are based on Tukey’s post-hoc analysis.
Figure 5. Within-population song-type variability (multivariate distance-to-centroid) is greatest
within the Pu’u La’au (PULA) region. Post hoc pairwise comparisons of 9999 permutations
found that variability within PULA was significantly greater than HUAL, but otherwise
variability was similar among locations.
Acoustic Variables PC1 PC2 PC3
Number of notes -0.65 0.10 -0.10
Peak frequency 0.01 -0.02 0.97
Song length -0.58 0.52 0.04
Low frequency -0.32 -0.80 -0.09
High frequency -0.38 -0.29 0.21
Eigenvalue 1.35 1.02 1.01
Proportion of
variance 0.36 0.21 0.20
Cumulative
proportion 0.36 0.57 0.77
Table 2. Principal component loading values of five acoustic variables measured from ‘amakihi
songs recorded across five populations. Three principal component axes had eigenvalues >1 and
explained 77% of the variation in measured acoustic variables. Acoustic characteristics with the
strongest PCA loadings (>0.35) are in bold.
Figure 6. Graph of first two principal component axes for ‘amakihi songs recorded across five
populations with 95% confidence ellipses.
Literature Cited
Anderson, M. J. 2006. Distance‐based tests for homogeneity of multivariate dispersions.
Biometrics 62(1):245-253.
Anderson, M. J., and D. C. Walsh. 2013. PERMANOVA, ANOSIM, and the Mantel test in the
face of heterogeneous dispersions: what null hypothesis are you testing? Ecological Monographs
83(4):557-574.
Atkinson, C. T., K. S. Saili, R. B. Utzurrum, and S. I. Jarvi. 2013. Experimental evidence for
evolved tolerance to avian malaria in a wild population of low elevation Hawai ‘i ‘Amakihi
(Hemignathus virens). EcoHealth 10(4):366-375.
Badyaev, A. V., and E. S. Leaf. 1997. Habitat associations of song characteristics in
Phylloscopus and Hippolais warblers. The Auk :40-46.
Bates, D., M. Mächler, B. Bolker, and S. Walker. 2014. Fitting linear mixed-effects models using
lme4. arXiv preprint arXiv:1406.5823 :.
Beecher, M. D., and E. A. Brenowitz. 2005. Functional aspects of song learning in songbirds.
Trends in ecology & evolution 20(3):143-149.
Boncoraglio, G., and N. Saino. 2007. Habitat structure and the evolution of bird song: a meta‐
analysis of the evidence for the acoustic adaptation hypothesis. Functional Ecology 21(1):134-
142.
Boncoraglio, G., and N. Saino. 2007. Habitat structure and the evolution of bird song: a meta‐
analysis of the evidence for the acoustic adaptation hypothesis. Functional Ecology 21(1):134-
142.
Cardoso, G. C., and T. D. Price. 2010. Community convergence in bird song. Evolutionary
Ecology 24(2):447-461.
Catchpole, C. K., and P. J. Slater. 2003. Bird song: biological themes and variations. Cambridge
university press, .
Dawkinss, R. 1976. The selfish gene. :.
Ellers, J., and H. Slabbekoorn. 2003. Song divergence and male dispersal among bird
populations: a spatially explicit model testing the role of vocal learning. Animal Behaviour
65(4):671-681.
Grant, P. R., and B. R. Grant. 1997. Hybridization, sexual imprinting, and mate choice. The
American Naturalist 149(1):1-28.
Hart, P. J., E. Sebastián-González, A. Tanimoto, A. Thompson, T. Speetjens, M. Hopkins, and
M. Atencio-Picado. 2018. Birdsong characteristics are related to fragment size in a Neotropical
forest. Animal Behaviour 137:45-52.
III, C. v. R. 1987. Breeding ecology of the Hawaii common amakihi. The Condor 89(1):85-102.
Irwin, D. E., S. Bensch, and T. D. Price. 2001. Speciation in a ring. Nature 409(6818):333.
Knowlton, J. L., D. J. Flaspohler, E. H. Paxton, T. Fukami, C. P. Giardina, D. S. Gruner, and E.
E. W. Rankin. 2017. Movements of four native Hawaiian birds across a naturally fragmented
landscape. Journal of Avian Biology 48(7):921-931.
Kroodsma, D. E., and E. H. Miller. 1996. Ecology and evolution of acoustic communication in
birds. Comstock Pub., .
Laiolo, P. 2008. Characterizing the spatial structure of songbird cultures. Ecological
Applications 18(7):1774-1780.
Laiolo, P., and J. L. Tella. 2005. Habitat fragmentation affects culture transmission: patterns of
song matching in Dupont's lark. Journal of Applied Ecology 42(6):1183-1193.
Laiolo, P., and J. L. Tella. 2006. Landscape bioacoustics allow detection of the effects of habitat
patchiness on population structure. Ecology 87(5):1203-1214.
LaPointe, D. A., M. L. Goff, and C. T. Atkinson. 2010. Thermal constraints to the sporogonic
development and altitudinal distribution of avian malaria Plasmodium relictum in Hawai'i.
Journal of Parasitology 96(2):318-325.
Lindsey, G., E. Vanderwerf, H. Baker, and P. Baker. 1998. Hawaii amakihi (Hemignathus
virens). The Birds of North America Online 10:73.
Lynch, A. 1996. The population memetics of birdsong. Ecology and evolution of acoustic
communication in birds :181-197.
Lynch, A., G. M. Plunkett, A. J. Baker, and P. F. Jenkins. 1989. A model of cultural evolution of
chaffinch song derived with the meme concept. The American Naturalist 133(5):634-653.
Morton, E. S. 1975. Ecological sources of selection on avian sounds. The American Naturalist
109(965):17-34.
Oksanen, J., F. G. Blanchet, R. Kindt, P. Legendre, R. O’hara, G. L. Simpson, P. Solymos, M. H.
H. Stevens, and H. Wagner. 2010. Vegan: community ecology package. R package version 1.17-
4. http://cran.r-project.org>.Acesso em 23:2010.
Pang‐Ching, J. M., K. L. Paxton, E. H. Paxton, A. A. Pack, and P. J. Hart. 2018. The effect of
isolation, fragmentation, and population bottlenecks on song structure of a Hawaiian
honeycreeper. Ecology and Evolution 8(4):2076-2087.
Parker, K. A., M. J. Anderson, P. F. Jenkins, and D. H. Brunton. 2012. The effects of
translocation‐induced isolation and fragmentation on the cultural evolution of bird song. Ecology
Letters 15(8):778-785.
Parker, K. A., M. J. Anderson, P. F. Jenkins, and D. H. Brunton. 2012. The effects of
translocation‐induced isolation and fragmentation on the cultural evolution of bird song. Ecology
Letters 15(8):778-785.
Pavlova, A., J. N. Amos, M. I. Goretskaia, I. R. Beme, K. L. Buchanan, N. Takeuchi, J. Q.
Radford, and P. Sunnucks. 2012. Genes and song: genetic and social connections in fragmented
habitat in a woodland bird with limited dispersal. Ecology 93(7):1717-1727.
Paxton, E. H., R. J. Camp, P. M. Gorresen, L. H. Crampton, D. L. Leonard, and E. A.
VanderWerf. 2016. Collapsing avian community on a Hawaiian island. Science Advances
2(9):e1600029.
Paxton, K. L., E. Sebastián-González, J. M. Hite, L. H. Crampton, D. Kuhn, and P. J. Hart. 2019.
Loss of cultural song diversity and the convergence of songs in a declining Hawaiian forest bird
community. Royal Society open science 6(8):190719.
Pimm, S. L., and J. W. Pimm. 1982. Resource use, competition, and resource availability in
Hawaiian honeycreepers. Ecology 63(5):1468-1480.
Pratt, L., and J. Jacobi. 2009. Loss, degradation, and persistence of habitats. Conservation
biology of Hawaiian forest birds, implications for island avifauna.Yale University Press, New
Haven :137-158.
Sebastián‐González, E., and P. J. Hart. 2017. Birdsong meme diversity in a habitat landscape
depends on landscape and species characteristics. Oikos 126(10):1511-1521.
Slabbekoorn, H. 2017. Animal Communication: Competition for Acoustic Space in Birds and
Fish. Pages 327-363 BIOCOMMUNICATION: Sign-Mediated Interactions between Cells and
Organisms. World Scientific, .
Slabbekoorn, H., and T. B. Smith. 2002. Bird song, ecology and speciation. Philosophical
Transactions of the Royal Society of London.Series B: Biological Sciences 357(1420):493-503.
Slabbekoorn, H., and T. B. Smith. 2002. Habitat‐dependent song divergence in the little
greenbul: An analysis of environmental selection pressures on acoustic signals. Evolution
56(9):1849-1858.
Tobias, J. A., J. Aben, R. T. Brumfield, E. P. Derryberry, W. Halfwerk, H. Slabbekoorn, and N.
Seddon. 2010. Song divergence by sensory drive in Amazonian birds. Evolution: International
Journal of Organic Evolution 64(10):2820-2839.
Van Riper III, C., S. G. Van Riper, M. L. Goff, and M. Laird. 1986. The epizootiology and
ecological significance of malaria in Hawaiian land birds. Ecological Monographs 56(4):327-
344.
van Riper, C., and J. M. Scott. 2001. Limiting factors affecting Hawaiian native birds. Studies in
Avian Biology 22:221-233.
VAN, R., and V. R. C III. 1976. ASPECTS OF HOUSE FINCH BREEDING BIOLOGY IN
HAWAII. :.
Wiley, R. H. 1991. Associations of song properties with habitats for territorial oscine birds of
eastern North America. The American Naturalist 138(4):973-993.
Wiley, R. H., and D. G. Richards. 1978. Physical constraints on acoustic communication in the
atmosphere: implications for the evolution of animal vocalizations. Behavioral Ecology and
Sociobiology 3(1):69-94.
Woodworth, B. L., C. T. Atkinson, D. A. Lapointe, P. J. Hart, C. S. Spiegel, E. J. Tweed, C.
Henneman, J. Lebrun, T. Denette, R. Demots, K. L. Kozar, D. Triglia, D. Lease, A. Gregor, T.
Smith, and D. Duffy. 2005. Host population persistence in the face of introduced vector-borne
diseases: Hawaii amakihi and avian malaria. Proceedings of the National Academy of Sciences
of the United States of America 102(5):1531-1536.