We have documented that the introduced Japanese white-eye (Zosterops japonicus) has outcompeted an entire community of
native birds at Hakalau Forest National Wildlife Refuge. We estimate that the refuge has lost an average of 31.5% of individuals
of its eight native passerine species in a 3373 ha area to competition, and up to 50% of one endangered species. While we believe that the white-eye needs to be controlled, the USGS-BRD still asserts that native bird populations are stable. We document the extent of the decline of native species in this case report and show that data quality ultimately dictates appropriate action, even when statistical analysis is performed correctly.
Introduced species have contributed to human caused global change [1,2]. They are homogenizing communities by adding new species to established communities around the world , and Hawaii is no exception where birds were intentionally introduced . When introduced species invade native habitats [5,6], they may cause enormous conservation problems to native and endangered species [7-11]. Great expense is incurred in controlling them [12,13]. Insect pests and weeds in agricultural settings are also frequently introduced species with costly management [14,15].
We present a case report of an introduced bird, the Japanese white-eye (Zosterops japonicus) and its impacts on native Hawaiian forest birds. One objective is to show that this introduced bird needs to be reduced in numbers while there is still time for native species to recover. The second objective is to highlight the role that misleading trend analysis as a statistical tool to visualize annual census data has played in preventing their control, stoking a particularly serious problem that managers need to solve. We will present the case documenting why failure to understand the limitation of trend analysis generates downstream impacts with fatal consequences for an entire Hawaiian forest bird community.
The historic and prehistoric loss of birds in Hawaii has been well-documented . Attempts were made in the 1920s to introduce birds from other continents so people could enjoy them in their yards, to control insects in agricultural districts, or as species beneficial for ecosystem service . The white-eye was introduced to the Hawaiian Islands in 1929 to control insects. During the Hawaii Forest Bird Survey in the mid-1970s, it was found on all main islands in every forested habitat , and competed with many species of native birds . However, there was a large pocket of low white-eye density on the Island of Hawaii where the native bird community was most complete, and much of this parcel of land eventually became Hakalau Forest National Wildlife Refuge (HFNWR) in 1985 (Fig. 1). This included about half of the 3373 ha open forest area and most of the 1998 ha closed forest area covering a large portion of the windward slope of Mauna Kea.
Figure 1. Density of Japanese white-eyes in 1977, on land that eventually became Hakalau Forest National Wildlife Refuge, from Scott et al. . Lighter areas indicate low density. The “H”, which represents the refuge, lies in a pocket of low white-eye density. Figure is from Freed and Cann  who showed that the low density persisted until 1999 from diffuse competition.
Diffuse competition kept the white-eye at low density from 1977 through 1999 , and describes the stable state where each native species overlapped multiple foraging substrates with the white-eye (Fig. 2). This prevented the white-eye from expanding in numbers. The foraging substrate overlap occurred in a forest with 87.7% Metrosideros polymorpha and 2.9% Acacia koa , so this forest approached a monoculture, which may have facilitated diffuse competition. Generally, reasons for low density for long periods of time before increase in numbers are largely unknown [22,23]. Diffuse competition to keep the introduced bird at low density is the first well-documented case of this process .
Beginning in 1989, the refuge planted hundreds of thousands of seedlings of A. koa in the 1314 ha pasture area. Although rare in the old-growth forest, this tree was planted because it is fast-growing and as a nitrogen-fixing legume would enrich the pasture soil for other native plant species. The ultimate reason for restoration was to create high elevation refugia for Hawaiian birds with little prior exposure to introduced avian malaria (Plasmodium relictum). We previously showed that the air temperature in the old-growth forest was too cool for the pathogen to develop in the mosquito vector, and that resident birds testing positive likely stemmed from infectious mosquitoes that were carried by wind to upper elevations . Wildlife specialists hoped that this restoration strategy would eventually provide critical high elevation habitat that would allow native birds more time to evolve tolerance to the disease. Annual census estimates to keep track of general trends were initiated in 1987 to help evaluate the utility of various management practices.
Figure 2. Foraging substrate overlap between the Japanese white-eye and native species, from Freed et al.  and Freed and Cann . The white-eye was maintained at low density by diffuse competition from the native community. We approached all who had performed research on the refuge and the refuge staff to construct the figure.
Why did habitat restoration become problematic?
The white-eye quickly colonized this restoration area from intermittent streams with trees in the area, and when A. koa first developed a canopy, the white-eye population began a cycle of exponential growth . Beginning in 2000, we documented their surge into the old-growth forest adjacent to and lower in elevation to the restoration area [20,25]. By 2002, in our long-term study site at 1900 m in the pocket of former whiteeye low density, the invaders had increased 6.6-fold compared with 1988, years with comparable numbers of net-hours . Even more white-eyes were captured per net hr through 2006 in the 1900 m, 1770 m, and 1585 m sites in the pocket. The 1900 m and 1585 m sites bracketed the restoration area and were 8 km apart, indicating that the white-eye increased in areas of the open forest area contiguous with the restoration area. A delay of several years was involved in the increased density in the closed forest area , probably because juvenile white-eyes were displaced by higher density of adults in the open forest sites . All of this information has been summarized
for managers and population biologists alike.
With higher density in 2002, the white-eye began to outcompete each species of native birds. We understand that many biologists remain unconvinced of competitive effects regulating established communities, but a coherent picture has now emerged, given our long-term studies on breeding biology and the metrics associated with the annual cycles of these birds. Starting that year all species of native birds had many changes in molt, reflecting imbalance in the metabolic cost of replacing worn feathers . Both young and adult birds took up to twice as long to complete their molt, a process described as extended molt. Early molt of adults of each species likely followed breeding failure. Molt-breeding overlap, previously very rare, occurred in females of most species. Asymmetric molt of primary flight feathers also occurred throughout the native community, requiring birds to expend extra energy to effect a turn in flight. These changes were also reflected in our banding data with lower recapture rates of native species compared to control years in the 1990s .
In 2004, young of all species measured (seven of eight) had stunted growth, for the first time in 18 years, evidenced by lower fledgling mass and shorter bills after growth had stopped . Juvenile survival was lower in the majority of native species with lower mass, typical of food limitation [28,29]. Shorter bills indicate severe food limitation. Birds with shorter bills had lower second-year survival in the majority of species. Even older adults of all species that survived their second-year with shorter bills were recaptured at lower frequency throughout the community than similar aged birds with normal bills. About half a million years of bill size evolution, the age of the Island of Hawaii, was altered. The lower survival indicates normalizing selection, not evolutionary change.
Changes in growth were associated with differences in whiteeye captures in mist-nets. During 2004-2005 there was a 5-fold difference in white-eyes between the combined 1900 and 1770 m sites in the open-forest area and the 1645 m site in the closed-forest area, while normal growth occurred in the 1645 site during those years . This comparison provides indirect evidence of white-eye competition.
Changes in molt suggested additional evidence of competition . Non-normal molt increased over background levels in 2002, concurrent with the white-eye increase. A reduced level of non-normal molt occurred in 2004 when white-eye numbers were lowest compared with contiguous years, associated with a temporary drop in propagule pressure . Highest prevalence of non-normal molt occurred in 2006 in the 1645 m site associated with an increase in white-eye density in the closed forest area. The highest prevalence of asymmetric molt occurred in 2003, when the white-eye had highest density on the 1900 m site. The combination of stunted grow and changes in molt caused by differences in white-eye density is strong evidence of competition.
There is no comparable data set reflecting a long-term intensive study of a banded community of forest birds from other Hawaiian islands contemporaneous with ours, so it is difficult to know what kinds of generalizations could be made in the absence of replication. Do our observations reflect long-term climatic shifts affecting phenology? Do they document increased mortality associated with emerging unknown/undetected pathogens? Predators and habitat loss are general features of the degraded Hawaiian ecosystem on all islands. The assembly of introduced bird species varies island by island, as do the ages, elevations, and size of the remaining forest habitat. Yet HFNWR represents one of the last, best hopes for preservation of the remaining avifauna, and the stakes could not be higher if we fail to heed these warnings. What follows now is a description of our efforts to alert decision makers about our findings and their reactions.
How scientists miss clues to an ecosystem in distress while evaluating “total evidence”
We summarized these community-wide effects and formally documented a switch from diffuse competition to reverse diffuse competition . The Japanese white-eye thus outcompeted the entire native passerine community. The competition was strictly exploitative competition because we never saw agonistic behavior initiated from the white-eye to any native species. Managers still conducted annual census counts, saw birds, and walked transects, unaware of the dynamics underway.
How did a significant increase in density result in an introduced bird outcompeting an entire native community? There are two hypotheses. First, reconsider Figure 2 which showed the basis for diffuse competition. We used Pianka’s  niche overlap from foraging substrates method to show that the white-eye overlapped more substrates with each native species than a native species overlapped on average with the rest of the native community. We tested three different models, one with all foraging substrates as if they were equally abundant, one based on relative abundance of M. polymorpha and A. koa and estimated relative use by the different species, and one based on the lesser used foraging substrates . The last two models are shown in Figure 3. All three models showed greater overlap by the white-eye for each native species than native species had with each other . This hypothesis explains why native species were outcompeted by increased numbers of white-eyes.
Figure 3. Niche overlap between the white-eye and native species. A. Overlap based on forest structure and estimated use of foraging substrates. The white-eye had more overlap with each native species than a native species overlapped with the rest of the community. B. Overlap based on greater use of lowest used foraging substrates. This figure from Freed and Cann  indicates that the white-eye is an extreme generalist.
Second, consider the phylogeny of the white-eye, which is in the bird family Zosteropidae, famous for range expansion and morphological niche diversification . Scott et al.  showed when the congener Zosterops lateralis colonized an area in the South Pacific, the bird had a very broad foraging niche. Close inspection indicated that birds with different length bills had narrower niches. The Japanese white-eye had stunted bill growth like the native Hawaiian species. However, instead of lower survival it had higher survival . It is possible that Japanese white-eyes with different bill lengths also collectively had very broad foraging niches. This would contribute to higher survival and the higher survival would spread the competitive effect in the native community (Table 1). Further research is important to determine if this change reflects an unappreciated phylogenetic plasticity in the Japanese white-eye. This hypothesis supplements the first to account for why one species of introduced bird outcompeted an entire native community.
Table 1. Decline of native birds averaged between 1996-2005 and 2006-2007.
Latin binomials of species in order listed: Loxops coccineus coccineus, Oreomystis mana, Hemignathus monroi, Hemignathus virens virens, Vestiaria coccinea, Himatione sanguinea, Chasiempis sandwichensis sandwichensis, Myadestes obscurus. The Hawaii akepa, Hawaii creeper, and akiapolaau are endangered species.
No other introduced biotic threat to native birds matched the changes in condition of native birds in time and space. These threats include avian malaria, chewing lice (Phthiraptera , yellow-jacket wasps (Vespula pensylvatica), and diverse mammalian and avian predators . A yellow-jacket irruption in 1995 had similar effects on white-eyes and native birds , but afterwards these wasps were monitored and controlled. These biotic factors do not match the changes in condition of native birds in time or space .
However, an abiotic change affected native birds subject to white-eye competition. Slight but significant climate cooling began in 2007 and extended to 2012 . The mechanism of cooling at 1900 m elevation resulted from movement of the inversion layer. Descent of the layer brings the cooler air above the warm air to elevations like 1900 m. Evidence of movement is indicated by frost on the ground on clear nights and lower clouds the next morning. The greatest change in density of native birds occurred between 2006 and 2007 (Fig. 4).
We modeled the decline of each native species over the entire 3373 ha open forest site from survey data between 2002 and 2007 using multimodel inference . We used two competition models (cumulative and discounted) and three climate cooling models dealing with cooler temperatures at different times . The three endangered species declined just from a cumulative competition model. Four other native species had support from both a competition model and a cooling model. Food limitation in small birds that are struggling to maintain body temperature with a narrow physiological range further stresses some species.
Demography ultimately revealed the devastating changes to us all
The status of the native community in 2006-2007 compared with 1996-2005 is shown in Table 1. We selected 1996-2007 because those years had lower coefficients of variation of density values than years 1987-1995 . The declines were substantial, particularly for the endangered species. Adults with shorter bills were recaptured less frequently , and from this we infer that the shorter-billed birds suffered higher mortality. Seven of eight native species had support from a competition model . The white-eye has reduced carrying capacity considerably for native species and is a driver of ecological change . It is the most serious case of interspecific competition in birds [37, 38], and is a tragedy that its surge in population occurred in old-growth forest from exponential growth in the restoration area. Most heretofore recognized problems with habitat restoration involve weedy plants within the restoration site .
We also used years 1997-2007 in the open forest area to model piecewise regression with an estimated breakpoint. Code for R and Splus is available in Freed and Cann . The breakpoint was estimated at year 2004 for most species, and the second slopes were all negative for all species . We also estimated breakpoints in the closed forest area, and combining all species on the same hypothesis, there were significant declines in that area as well . In fact, endangered species were seen at lower elevations than during the Hawaii Forest Bird Survey. They moved to where white-eyes were at lower density and where infectious mosquitoes were more prevalent.
Misleading trend analysis
We have advised the refuge since 2004 that white-eyes need to be controlled. By 2007, almost one in ten birds in the 3373 ha open forest area was a white-eye . Both Table 1 and the fact that the refuge is becoming an “introduced bird refuge” are reasons why white-eyes need to be controlled. But the refuge has ignored all research by Freed and Cann in the references here. These references represent peer-reviewed formal studies.
The reason for the declines is a flawed statistical analysis of annual census data that was used to justify management inaction, based on the assertion that the bird populations on the refuge were stable or increasing. It has been published by the USGS-BRD scientists and taken as a valid depiction of the status of birds on the refuge [40-42]. But we showed that years 1987-1995 had significantly higher coefficients of variation of census data than years 1996-2007 for each native species . This could result from some surveyors missing detections, thus contributing to the higher variance and the low density values. This situation is entirely understandable given that none had much experience in the old-growth forest at Hakalau during those years. Surveyors became better with time and may have been joined by more experienced personnel, generating the lower coefficient of variation during later years.
Figure 4 reveals another fundamental issue with trend analysis. The fluctuations between years 1987-1989 represent an impossible situation for native birds that are not replicated later in the series. The year 1987 starts out with a high native community-wide density. There was a drop of 56.6% from 1987 to 1988. This was not reflected on our study site where adult survival was constant between 1987 and 2000 . In addition, the year 1987 had a higher than average capture rate of young birds in the series 1987-2005 (Figure 5). Thus survival and recruitment on our study site cannot account for the drop in density in 1988.
Figure 4. Native community-wide density from years 1987 through 2007, from Freed and Cann . The years 1987 through 1995 had higher coefficients of variation than years 1996-2007. That is one reason that Camp et al. [40-42] had a misleading trend analysis. The years 1987-1989 represent a biologically impossible response of these birds. See Table 2.
Figure 5. Captures of young birds per net hour from our 1900 m site, indicating that recruitment was not responsible for the apparently increased density. Freed and Cann  also show that no change in adult survival occurred.
We can extend this from our study site to the entire 3373 open forest area. Three of the native species can have two successful broods per year with two eggs per clutch. Table 2 shows differences in density between 1987 through 1989 over the
entire 3373 open forest area. The three species had comparable declines between 1987 and 1988 ranging from 66 to 73%. But the increases between 1988 and 1989 are impossible, with each of these species more than doubling its density. With juvenile survival documented in Freed and Cann  in Table 2, it would be impossible for populations to experience more than a doubling in numbers in a single year without a change in adult survival, which did not occur. We know of no case in birds where a population at the level of these three species doubles in a single year.
Table 2. Density of three honeycreepers which can have up to two broods per year.
1from Camp et al. 
2from Freed and Cann 
There is a lesson in statistics in this: The quality of data is just as important as the appropriate statistical model. This is especially true of survey data where there is no control over climate, bird behavior, or breeding status which can vary among years, or among observers on different transects. The statistical analysis may be valid, but if the data include outliers and years with higher coefficients of variation, the analysis is in vain. Reviewers and editors of manuscripts focus on the statistical analysis and usually do not question the quality of the data except as they pertain to the statistical model used. Unfortunately Camp et al. [40-42] ignored coefficients of variation that were published in Freed and Cann . Reviewers and editors of Camp et al.  apparently did likewise.
Additional evidence of misleading trend analysis exists. The years 1988 and 2007 have similar low community-wide densities. Yet in 1988 the endangered Hawaii akepa population was viable, but after 2000 was not viable [20,21]. Specific life-history adaptations were dismantled by white-eye competition leading to lower survival [21,43,44]. There was no indication of recovery as we had 67 color-banded birds in 2005 in our 1900 m study site, of which 36 were detected in 2006 [20,21]. Only 31 akepa were reported in 2013. This is consistent with the fact that akepa had declined in the 3373 ha open forest area by over 40% (Table 1).
The endangered Hawaii creeper fared even worse. It declined in the open forest area by 50%. Even this decline is overly optimistic, because females experienced higher mortality from molt-breeding overlap and the remaining 50% of the birds had an adult sex ratio of 28% females .
Moreover, in March 2008, we inspected a study site in the high density area for 2 hours. We did not detect a single native bird of any species. Only one pair of akepa was breeding when we returned in May. This decline of native birds had never occurred in 20 years of research.
The sum total of information indicates a significant decline of native species. As unimaginable it might be that a single introduced bird outcompeted an entire native community, it is based on niche overlap which is the basis of diffuse competition , which became reversed. Tilman  asserted that if a species had no life-history tradeoffs and was not more susceptible to predators and parasites, it would outcompete the rest of the community. The higher survival of the white-eye indicates no life-history tradeoffs and no more susceptibility to predators and parasites than native species.
The quality of data problem might extend to other regions of the world where censuses are conducted. There are many papers published about trend analyses of census data [35, 48], as problems occur in modeling populations with widespread ranges, with serious population dynamics, or where the transects change over time. These references do not question the quality of the survey data as we have, which raises more fundamental issues.
Czech  studied the capacity of the national wildlife refuge system to conserve threatened and endangered species compared to similar habitat and species outside refuges. He found that national wildlife refuges conserved such species generally well as long as commercial activities such as logging and mining were constrained. The restoration area was solely above the old-growth forest of HFNWR, implying that habitat and birds outside the refuge would be less impacted by the whiteeye.
One reason for the steep declines in the refuge would be the consequence of ignoring peer-reviewed research in favor of Camp et al. [40-42]. If independent, peer-reviewed research is ignored, we predict these declines will continue, as in the case of the endangered akepa and creeper. This should send a chill through the conservation community, as researchers chase fewer funding sources and agencies hold the permitting powers. Bad news means the messengers get shot.
Our Proposed Solution to the Problem
When we approached the refuge in 2004 about controlling white-eyes the response we received is what Simberloff  identified as xenophobic. Then there was an official meeting in 2008 for the new 15 year conservation management plan for the refuge. We provided extensive written comments about this plan which said little about controlling introduced avian competitors. There was no support for controlling white-eyes because the USGS-BRD asserted that all native bird populations were stable. There was followed by a meeting in October 2013 about forest birds where the USGS-BRD continued to assert stable populations. When demographic changes documented in Freed and Cann  were pointed out, dialogue ended with federal biologists insisting that there was no demographic information to evaluate their analyses against.
Control of the Japanese white-eye is both essential and complicated, and is supported by studies regarding control of other introduced bird species [51,52]. Control must occur in both the restoration area and old-growth forest. Control in the restoration area will reduce density of the white-eye and permit native species populations to expand their range in the area. Propagule pressure might even be reduced from the restoration area to the forest. Control of white-eyes must occur simultaneously in the forest so native birds can recover. Success in control will be indicated by diffuse competition occurring in the restoration area and re-occurring in the old-growth forest .
We raise this alarm because we feel it is not just a problem at HFNWR. Native birds are declining elsewhere in the state. Camp et al.  showed that native birds were declining in the former Kau Forest Reserve on Mauna Loa and the Alakai Swamp on the Island of Kauai. In the latter area the Kauai akepa (Loxops caeruleirostris) and Kauai creeper (Oreomystis bairdi) have been drastically declining, as have their relatives at Hakalau. Given this case report at Hakalau, competition with white-eyes should be the first hypothesis to be considered and critically examined for preventing further declines of the native Hawaiian avifauna.
We appreciate advice from T.A.H. Smith. Our funding agencies were the John D. and Catherine T. MacArthur Foundation World Environment and Resources Program (8900287), the National Science Foundation (DBI 96-02547), and the National Center for Environmental Research (Science to Achieve Results, Environmental Protection Agency R82-9093). This research was enabled by the University of Hawaii Institutional Animal Care and Use Committee protocol 00-005-15. We thank the many students, interns, and volunteers who provided the demographic data over the years.