Globally, seabirds can be important drivers of nutrient cycling via allochthonous deposition of nutrients such as nitrogen (N ) and phosphorus (P) that influence all trophic levels in the ecosystems where they breed and nest (Anderson and Polis 1999, Mulder and Keall 2001, Whelan et al. 2008, Towns et al. 2009, Grant-Hoffman et al. 2010). Guano naturally contains N and P, nutrients that commonly limit ecosystem processes in forest ecosystems globally (Hutchinson 1950, Elser et al. 2007). Guano composition for Whitecapped Noddies (Anous minutus) and Wedgetailed Shearwaters (Puffinus pacificus) is approximately 7.3% N and 1.5% P (Smith and Johnson 1995), but N content of guano increases to 14.9%–28.6% for pelagic seabirds such as Northern Fulmar (Fulmarus glacialis) and Manx Shearwater (Puffinus puffinus) (Burger et al. 1978, Bird et al. 2008). As such, seabird defecation can increase available pools of N and P in terrestrial ecosystems by 100 and 400 times, respectively (Mulder et al. 2011). In turn, marine-derived nutrient subsidies affect a suite of processes including primary productivity (Mulder and Keall 2001), plant community composition (Anderson et al. 2008, Mizota 2009), and the population size of top predators (Rose and Polis 1998).

The role of seabird guano in the transport of nutrients from marine sources to land has been studied in many coastal ecosystems, but tropical montane ecosystems where many burrowing seabirds nest have been poorly studied (Mulder et al. 2011). Nutrient levels in highly weathered tropical ecosystems can be low, particularly for P, and with high rainfall and steep slopes N and P can be readily lost through leaching and erosion (Ehleringer et al. 1986, Posada and Schuur 2011). In these nutrient-poor ecosystems, the addition of guano rich in N and P may be vital to maintaining or enhancing plant communities and ecosystem process rates (Mizota 2009).

Seabirds face many challenges including habitat loss, introduction of predators, collision with man-made structures, light pollution, toxins, change in prey availability, and poisoning (Millenium Ecosystem Assessment 2005, Hebshi et al. 2008, Duffy 2010, Griesemer and Holmes 2011, Loss et al. 2012, 2015, Wiley et al. 2013). These challenges have typically led to severe population declines and in some cases extinctions, resulting in reduced nutrient inputs to the terrestrial habitats that seabirds traditionally occupied. In Hawai‘i, seabirds that nest in montane forest ecosystems have experienced severe population declines, with some seabird species extirpated from the Hawaiian Islands and others remaining in greatly reduced ranges and numbers (Olson and James 1994, Burney et al. 2001, Hearty et al. 2005). The decline of nutrient flux from marine to terrestrial ecosystems due to the reduction of seabirds on the main Hawaiian Islands may have serious effects on plant community dynamics and ecosystem processes. Thus understanding how native plants utilize nutrient subsidies is important to inform the conservation and restoration of native habitats.

Two seabird species that have experienced drastic population declines in the montane regions of Hawai‘i historically, as well as in the last 25 yr, are the Newell's Shearwater (Puffinus newelli ) and Hawaiian Petrel (Pterodroma sandwichensis), which are federally listed as threatened and endangered, respectively (James 1991, Burney et al. 2001, Griesemer and Holmes 2011). Formerly numerous and widespread, these two species are currently limited to remote colonies in hard-to-access locations due to loss of habitat and an increase in introduced predators. We sought to determine whether Newell's Shearwaters and Hawaiian Petrels in low numbers still influence soil nutrient availability and plant nutrient uptake. Specifically, we sought to answer three primary research questions. First, do low numbers of seabirds increase the availability of macronutrients and micronutrients in wet montane forest soils? Second, do the dominant plants in wet montane ecosystems utilize marine-derived N, and if so to what extent? Third, does avian nutrient subsidy influence plant species composition? We hypothesized that (1) soil micronutrient and macronutrient availability would be higher around the seabird colonies than in areas without seabirds but only minimally given the greatly reduced seabird populations (Wainright et al. 1998, Liu et al. 2006); (2) δ¹⁵N values would be higher in soil and foliage at seabird nesting sites, reflecting a marinederivednutrient subsidy (Caut et al. 2012, Kazama et al. 2013); and (3) the plant community composition in seabird plots would be biased toward nitrophilic plants adapted to high nutrient levels (Vitousek and Farrington 1997, Martinelli et al. 1999, Bond et al. 2010).

materials and methods

Study Site

Though greatly reduced from historic levels, the island of Kaua‘i is home to the densest populations of montane nesting seabirds in the Hawaiian archipelago. Study sites were located in the montane forests of Upper Limahuli Preserve and Hono O Nā Pali, Kaua‘i. We considered two treatment types in each of these areas: active seabird colonies and nonseabird areas (areas without current seabird colonies and with no evidence of recent nesting) (Figure 1). Notably, it is likely that seabirds historically nested in most, if not all, montane areas in the past, but the control sites contained no burrows (new or old), bird sign (e.g., feathers or guano), or records of bird activity since 2006, when the Kaua‘i Endangered Seabird Restoration Project began working in the area. Furthermore, the density of seabirds in the most heavily used areas is only 0.04 burrows m^-2; this is low for colonial nesting seabirds such as shearwaters and petrels, which are known to nest up to 0.76 burrows m^-2 for Gray-faced Petrel (Pterodroma gouldi ) (Whitehead et al. 2014) and 0.08 burrows m^-2 for Cook's Petrel (Pterodroma cookii ) (Rayner et al. 2007).

We established 24 plots on ridgetops: nine seabird and four nonseabird plots in Upper Limahuli Preserve and eight seabird and three nonseabird plots in Hono O Nā Pali (Figure 1). On the geologically older Hawaiian island of Kaua‘i, ecosystem processes are typically limited by P availability (Crews et al. 1995), which is expected to apply to our study sites as well. Seabird and nonseabird plots were selected opportunistically in areas with and without seabirds, respectively. Steep slopes, lack of helicopter landing locations, and low seabird numbers made random or uniform plot selection unrealistic. Sample size for isotopic comparisons was based on an a priori power analysis (Faul 2009). Effect size was set at 1.27 based on published soil and Metrosideros polymorpha δ15N values (Vitousek et al. 1989); error probability (α) was set at .05, power (1 - β error probability) was .95, and total sample size was calculated as n = 24. Each plot was 5 m in diameter and established in different seabird subcolony clumps. The Upper Limahuli Preserve ranges in elevation from 750 to 980 m, with plots located above 800 m. Hono O Nā Pali is located above 1,200 m, and plots were situated between 1,210 and 1,287 m.

Figure1.

Location of plots in Upper Limahuli Preserve (ULP) and Hono O Nā Pali (Pihea), Kaua‘i, Hawai‘i. Seabird plots are represented with a triangle, and nonseabird plots are represented with a circle. Image courtesy of National Geographic, Esri, DeLorme, HERE, UNEP-WCMC, USGS, NASA, ESA, METI, NRCAN, GEBCO, NOAA, iPC.

All soils in the two study areas were surveyed by reconnaissance survey. The soils in Upper Limahuli are classified as Alakai mucky peat and Waialeale mucky silty clay loam, and the soils in Hono O Nā Pali are classified as rough mountainous land (Soil Report for Island of Kaua‘i, Hawai‘i 2014). Alakai mucky peat taxonomic classification is clayey, ferrihumic, dysic, isomesic Terric Haplosaprists with pH values typically less than 4.0. Waialeale mucky silty clay loam is classified as veryfine, isotic, isothermic Typic Epiaquods with pH values commonly less than 4.4. Based on similar topography and rainfall, it is likely that the rough mountainous land in Hono O Nā Pali has the same classification as the soils in Upper Limahuli.

All plots were located in wet montane forest with the majority of vegetation being native and dominated by Metrosideros polymorpha (‘öhi‘a) in the canopy and the false staghorn fern Dicranopteris linearis (uluhe) in the understory. No nitrogen-fixing plants were present, although there is evidence of N-fixation in the litter of D. linearis (Russell and Vitousek 1997). Mean annual precipitation at both sites ranges between 2,500 and 3,000 mm with rain occurring throughout the year (Giambelluca et al. 2013). Mean annual temperature is 13°C, with warmest temperatures occurring in August and September (Juvik and Juvik 1998). Both study sites are remote with relatively intact forests exposed to introduced pigs (Sus scrofa), cats (Felis catus), mice (Mus musculus), and two rat species (Rattus exulans and Rattus rattus). In recent years, control and restoration measures have been implemented, including ungulate-proof fencing and pig removal in Upper Limahuli, and trapping for invasive animals and nonnative plant removal in both sites (Jon-Carl Watson, Limahuli Preserve Operations Manager at National Tropical Botanical Garden, pers. comm.).

Isotopic Analysis

Nitrogen isotopic ratios were determined for the top 10 cm of mineral soil, sunlit foliage of both M. polymorpha and D. linearis in seabird and nonseabird plots at both sites as well as seabird guano. Samples were composited (five soil samples and five samples each of M. polymorpha and D. linearis foliage) in each plot. Soil samples were collected using a 1.27 cm diameter soil corer, and sunlit, live leaves were collected from the newest fully mature cohort. Hawaiian Petrel and Newell's Shearwater guano samples were opportunistically collected from field sites during the 2014 breeding season. Twelve relatively fresh guano samples were placed in plastic bags (Ziploc) and kept in a cooler on ice until they could be frozen. Organic materials were handpicked out of guano samples before analysis. Freshness could not be determined, but the high rainfall at these sites should preclude guano from remaining on the ground for more than 2 days of rain. The mean carbon:nitrogen ratio for the sampled seabird guano was 1.09 ± 0.21 (n = 8), and δ¹⁵N = 8.23‰ ± 1.68‰ (n = 8).

Soil and foliar samples were dried at 70°C, sieved through a 2 mm mesh, homogenized, and powdered in a ball mill (Carter and Gregorich 2006). Guano samples were freezedried and ground using a mortar and pestle. The isotopic composition of all samples was analyzed at the University of Hawai‘i at Mānoa Biogeochemical Stable Isotope Facility using a continuous flow mass spectrometer (ThermoFinnigan Deltaplus XP) coupled with an elemental analyzer (Costech ECS 4010) via a Conflo IV interface. Nitrogen isotopic compositions are expressed as δ¹⁵N values in ‰ relative to air:

where R_sample is the isotopic ratio (¹⁵N/¹⁴N ) of the sample and Rstandard is the ¹⁵N/¹⁴N of air. Accuracy and precision of δ¹⁵N values were <0.2‰, as determined from reference materials analyzed every 10 samples (glycine and a tuna muscle homogenate with δ¹⁵N values of 11.25 ± 0.04‰ and 12.97 ± 0.06‰, respectively, which were determined by extensive characterization using National Institute of Standards and Technology certified reference materials, and their δ¹⁵N values were verified independently in other laboratories). Percentage N (%N ) was determined for soil and foliage samples from the results of isotopic analysis.

A two-component N isotope mass balance mixing model was used to determine the proportion of marine-derived nutrients in the top 10 cm of soil and the foliage of the two dominant plants in the plots (Phillips et al. 2005, West et al. 2006). As such, we used the mixing model equation from Dawson et al. (2002):

where δ T is the total sampled isotopic value, δA and δ B are the two source values, and ∫A is the portion of the total value that is provided by source A (Dawson et al. 2002). The nonseabird plot data were used to establish terrestrial δ 15N end member, and the δ 15N value of seabird guano was used as the marine end member. Solutions from the mixing model provide the percentage of total N in soil and foliage of M. polymorpha and D. linearis in the seabird plots that was marinederived(Fry 2006, Phillips 2012, Dawson et al. 2002, Boecklen et al. 2011). Uncertainty in the marine-derived fraction of total N was determined by propagation of error using the analytical solution of the partial differentiation of general Taylor series approximation determined using the two-component stable isotope mixing model (see Gelwicks and Hayes 1990, Phillips and Gregg 2001).

Statistical Analyses

Statistical analyses were performed in SPSS 22 (SPSS Inc. 2007). Levine's test was used to assess homogeneity of variance in soil nutrient concentration as well as the plant community composition. Of all nutrients analyzed (%N, total inorganic N, NO₃^-, NH₄⁺, SO₄^2-, PO₄^3-, and Ca₊²), only NH₄⁺ did not pass Levine's test and these data were log10 transformed for analysis. All plant community composition data passed Levine's test. One-way analysis of variance (ANOVA), with significance set at α = .05, was used to test for differences in available soil nutrients as well as relative differences in δ¹⁵N values in soil and M. polymorpha and D. linearis foliage between treatments. All results are presented as means ± SE, unless otherwise noted. We used t tests to determine differences between treatments for %N in M. polymorpha and D. linearis. For plant community composition, we analyzed canopy cover, average vegetation height, total Metrosideros cover, and total number of species present to look for differences between treatments, also using t tests.

TABLE 1

Mean and Standard Error (SE) Inorganic Soil Nutrient Availability for Seabird and Nonseabird Plots (µg/10 cm^-2/4 weeks)

results

Isotopic Analysis

Foliage of M. polymorpha had significantly higher δ¹⁵N values in the seabird than in the nonseabird plots (F = 5.07; df = 1, 21; P = .036) (Table 2). Although δ¹⁵N values in soil and D. linearis leaves between the two treatments were not statistically different (F = 2.78; df = 1, 21; P = .11; F = 2.82; df = 1, 21; P = .11, respectively), there was a positive trend toward increasing δ¹⁵N values in the seabird plots compared with the nonseabird plots (Table 2). Results from the mixing model indicated that 32% of the total soil N was derived from seabirds in the seabird plots. Foliar N of M. polymorpha in seabird plots was 27.9% (8% SE) from a marine source, and D. linearis foliage contained 16.9% (0.08 SE) N from a marine source (Table 3). However, %N did not differ between seabird and nonseabird plots (Table 4) for soil (t = 0.81, df = 32, P = .43), M. polymorpha foliage (t = 0.17, df = 20, P = .26), or D. linearis foliage (t = 0.92, df = 21, P = .37).

TABLE 2

Mean and Standard Error (SE) of δ¹⁵N Values of Soil and Two Dominant Plant Species in Seabird and Nonseabird Plots on Kaua‘i

TABLE 3

Mean Proportions and Propagated Error for a Two-Source, One-Isotope (δ¹⁵N ) Model for Soil and Two Dominant Plant Species in Seabird and Nonseabird Plots P = .26), or D. linearis foliage (t = 0.92, df = 21, P = .37).

TABLE 4

Mean %N in Soil, M. polymorpha, and D. linearis

discussion

We found support for the first hypothesis because more inorganic N in the form of NH₄₊ was found where seabirds were present. In support of the second hypothesis, 32% of soil nitrogen was from a marine source, and that marine-derived N accounted for 17%–28% of foliar N in the two dominant native plants in the study system. However, we did not find evidence to support the third hypothesis: that plant community composition would differ with and without allochthonous input of nutrients by seabirds.

Microbial processes can affect the nitrogen isotopic composition of plant-soil systems (Amundson 2003, Szpak 2014). Nitrogen in guano is deposited primarily as uric acid (C₅H₄O₃N₄) (Bird et al. 2008), and microbes mineralize this organic N into NH₄⁺ and NO₃ -. Inorganic nitrogen as NH₄+ and NO₃^- is available to most plants in soil solution, making this a potentially important addition to the ecosystem. Ammonium can be lost from the system through oxidation (the first step in soil nitrification), which can lead to N loss as nitrates are leached from the soil or through ammonia volatilization. Nitrogen transformations such as ammonia volatilization and denitrification can affect the δ¹⁵N values of plants and soil because there are large nitrogen isotope fractionations associated with these processes. We do not, however, believe that nitrogen loss in the systems studied was the main cause of 15N enrichment in seabird sites. The soil types in these areas are acidic (pH less than 4.5), and under these acidic conditions ammonia would be protonated so that the dominant form of reduced N would be ammonium. In Hawaiian rain forests in regions with mean annual rainfall exceeding ∼2,500 mm, soil microbial denitrification completely consumes nitrate in local soil environments, preventing expression of the isotope effect associated with denitrification (Houlton et al. 2006). Under these conditions δ¹⁵N values of soils converge on the δ¹⁵N values of the N input. We anticipate that these potential losses of N may make seabird subsidies more critical to the ecosystem than currently understood.

The impacts of seabirds on soil and plant characteristics vary across systems and depend at least partially on the life history of the seabirds in question. Durrett et al. (2014) found that trees and shrubs differed in their response to the addition of marine nutrients, with trees increasing slowly in foliar %N and δ¹⁵N values with increasing population density while shrubs showed a strong positive response at low densities and negative responses at higher densities. In ecosystems that are N and /or P limited, the addition of guano may increase primary productivity and select for fastgrowingplants that can take advantage of periodic resource subsidies, or seabirds that nest in high density may make the soils toxic to plants with the excessive addition of nutrients (Wainright et al. 1998, Anderson and Polis 1999, Kolb et al. 2010). In similar studies based on coastal seabird colonies, nutrient pulses caused changes that were observed through the food web and even back into the marine environment. These changes included increased plant productivity, increased plant predation, increased arthropod and lizard density, and increased coastal nutrient influx into the nearshore environment leading to increased plankton growth (Barrett et al. 2005, Spiller et al. 2010, McCauley et al. 2012). On Kaua‘i, Newell's Shearwater and Hawaiian Petrel nest colonially but in low densities in burrows. Because burrow-nesting birds are not observed to reach densities that cause nutrient toxicity to plants, seabirds likely played a larger supporting role historically in nutrient cycling in this ecosystem when population numbers were higher.

We estimated the quantity of nutrients potentially added to our study sites by guano input. The two sites total about 200 ha, 160 ha in Upper Limahuli and 40 ha in Hono O Nā Pali, including nonseabird areas as well as areas where the seabirds are nesting. We have no official estimates of seabird population numbers in these areas, but using Griesemer's island-wide estimates for Newell's Shearwater as well as estimates from the field, we estimated 500 pairs (± 250) of Newell's Shear water and Hawaiian Petrel combined between the two sites, with proportionally more Newell's Shearwater in Upper Limahuli and proportionally more Hawaiian Petrels in Hono O Nā Pali (Griesemer and Holmes 2011). No studies have been conducted to indicate how much Newell's Shearwater or Hawaiian Petrel eat or excrete per day. However, the wandering albatross weighs approximately 10 kg and was found to consume 2 kg of food per day (Salamolard and Weimerskirch 1993). The average weight for Newell's Shearwater and Hawaiian Petrel is 0.4 kg (Ainley et al. 1997, Judge et al. 2014). Therefore, because small organisms require more food per unit body mass than larger organisms, we used the allometric relationship:

Isotopic Analysis

Seabirds have been shown to increase N levels in soil and surrounding organisms via marine-derived N (Wainright et al. 1998, Wait et al. 2005, Mizota 2009). However, it was previously unknown if this is also the case in wet tropical montane regions characterized by high rainfall, warm temperatures, and low current population densities of seabirds (Martinelli et al. 1999, Garcia et al. 2002). Although marine N was clearly higher in M. polymorpha, the effect of seabird-added nitrogen may be masked in D. linearis by nitrogen fixation that may occur in the litter (Russell and Vitousek 1997). It should also be kept in mind that although the nonseabird sites had no current evidence for nesting, they were likely colonized in the past and may have a legacy of high δ¹⁵N values in soil. Thus, using these nonseabird sites likely resulted in an overestimation of terrestrial δ¹⁵N end member values in the isotope mass balance mixing model and thus systematically underestimated the proportion of marine-derived N available in the soil and incorporated into the foliage of both studied species.

Vitousek et al. (1989) and Martinelli et al. (1999) measured δ¹⁵N of M. polymorpha foliage in nonseabird areas in Hawai‘i across multiple islands and found mean δ¹⁵N values of -3.3 (± 2.3‰ SD) and -2.8 (± 2.6‰ SD), respectively. A comparison of their values with this study is complicated by difference in substrate age between sampling sites in these prior studies and ours on Kaua‘i. Martinelli et al. (1999) also measured a δ¹⁵N value of -0.5‰ from a single M. polymorpha on Kaua‘i from a nonseabird area. It is unknown if the δ¹⁵N value of -0.5‰ is an outlier or representative of M. polymorpha on Kaua‘i. However, the archipelago averages of -3.3‰ and -2.8‰ in nonseabird areas are slightly lower than the nonseabird values that we measured for M. polymorpha (-2.3‰), and this also indicates a potential underestimate of the importance of current marine-derived N as presented here.

Results of the isotopic mixing models indicate that dominant plants in this ecosystem utilize at least some N derived from a marine source, and soil and foliage of both plant species showed marine influence. Although standard error was high, the amount of marinesourced N was higher across all sampled substrates in seabird plots, and δ¹⁵N values were significantly higher in M. polymorpha foliage in seabird plots compared with control plots.

Plant Community Composition

None of the plant species composition measures indicated differences between seabird and nonseabird plots. This is likely due not only to the limited amount of nitrogen being added to the seabird sites, but also to the species-depauperate nature of the Islands and isolation of the study sites. There are a limited number of native species on the Islands to populate these areas, and the isolated nature of the study sites means less influence of the nonnative and invasive plant species that occur in high densities in more disturbed areas in Hawai‘i. Invasive plant species were actively managed in Upper Limahuli and pulled opportunistically in Hono O Nā Pali, though the density of invasive plants is low in both sites due to the remote locations. Another reason for the lack of differences in some of the measured variables is that other drivers besides the addition of nutrients by seabirds may be more influential, particularly in lowdensity seabird sites such as ours. Other researchers have found that on high- and medium-density seabird islands, seabirds drove ecosystem properties such as δ¹⁵N values (soil and leaf ), soil and leaf N, NH₄⁺, and NO₃^-. However, in low-density colonies other ecosystem processes drove these ecosystem properties more than the seabirds (Durrett et al. 2014).

conclusions

Despite being at historically low population densities, seabirds contribute to the ecosystems where they still nest in montane Hawai‘i via marine-derived nutrient deposition. These study sites in Kaua‘i contain some of the last relatively intact tropical montane ecosystems with native seabirds, yet very little research has occurred there. Studies in the arctic and in coastal systems are abundant (Polis and Hurd 1996, Mulder et al. 2011, Gagnon et al. 2013) and generally show seabirds to increase nutrient availability and biodiversity in the arctic (Keatley et al. 2009, Zmudczyńska et al. 2012) and to fertilize or even create toxic conditions in coastal ecosystems depending on seabird density (Kolb et al. 2010, VanderWerf et al. 2014). However, comprehensive studies about how seabirds and their nutrient subsidies impact tropical montane ecosystems are notably lacking (Hawke and Holdaway 2009).

Our control plots may have a historical legacy of seabirds and thus may still contain nutrients from seabirds. Therefore, our estimate of allochthonous nutrient input by seabirds is likely conservative. The influx of N and P may be more important to the resiliency of these ecosystems, especially in the face of climate change and other stressors, than is currently understood (Perry et al. 2010, Doughty et al. 2015). Historically seabirds in the Hawaiian Islands may have played a major role in controlling soil fertility in areas where they nested. The montane forest seabirds in the Hawaiian Islands have been greatly reduced both in number and in range, yet even in vastly reduced numbers these birds impact soil and vegetation nutrient content. The decrease in nutrient flow and its impacts on the ecosystem are vital for restoration project managers to understand as they attempt to rebuild the ecosystem and restore endemic plants and wildlife. We know that seabird declines have sweeping effects throughout the food web in warm temperate New Zealand (Fukami et al. 2006), but the consequences of seabird declines on tropical vegetation dynamics and ecosystem processes are largely unknown.