Lessons from the Tōhoku tsunami: A model for island avifauna conservation prioritization

Abstract Earthquake‐generated tsunamis threaten coastal areas and low‐lying islands with sudden flooding. Although human hazards and infrastructure damage have been well documented for tsunamis in recent decades, the effects on wildlife communities rarely have been quantified. We describe a tsunami that hit the world's largest remaining tropical seabird rookery and estimate the effects of sudden flooding on 23 bird species nesting on Pacific islands more than 3,800 km from the epicenter. We used global positioning systems, tide gauge data, and satellite imagery to quantify characteristics of the Tōhoku earthquake‐generated tsunami (11 March 2011) and its inundation extent across four Hawaiian Islands. We estimated short‐term effects of sudden flooding to bird communities using spatially explicit data from Midway Atoll and Laysan Island, Hawai'i. We describe variation in species vulnerability based on breeding phenology, nesting habitat, and life history traits. The tsunami inundated 21%–100% of each island's area at Midway Atoll and Laysan Island. Procellariformes (albatrosses and petrels) chick and egg losses exceeded 258,500 at Midway Atoll while albatross chick losses at Laysan Island exceeded 21,400. The tsunami struck at night and during the peak of nesting for 14 colonial seabird species. Strongly philopatric Procellariformes were vulnerable to the tsunami. Nonmigratory, endemic, endangered Laysan Teal (Anas laysanensis) were sensitive to ecosystem effects such as habitat changes and carcass‐initiated epizootics of avian botulism, and its populations declined approximately 40% on both atolls post‐tsunami. Catastrophic flooding of Pacific islands occurs periodically not only from tsunamis, but also from storm surge and rainfall; with sea‐level rise, the frequency of sudden flooding events will likely increase. As invasive predators occupy habitat on higher elevation Hawaiian Islands and globally important avian populations are concentrated on low‐lying islands, additional conservation strategies may be warranted to increase resilience of island biodiversity encountering tsunamis and rising sea levels.


| INTRODUCTION
Seismic events have produced hazardous tsunamis that affect coastlines, damage infrastructure, and cause human causalities approximately once per year globally (Bernard, 2001). The Hawaiian Archipelago is vulnerable to tsunamis generated in most parts of the Pacific Basin (NOAA 2017). Depending on water depth, tsunamis can propagate from the epicenter at speeds of greater than 1,000 km/hr (NOAA, 2017) refracting near islands and bays, so their impacts vary greatly between locations. Little is known about tsunamis on remote islands, and the effects of sudden flooding on island biodiversity are not well studied. After the 26 December 2004 Indian Ocean tsunami following the Sumatra-Andaman earthquake (Mw = 9.1; max water height near source 50.9 m; NOAA, 2015b), researchers documented effects to wildlife and ecosystems including King Penguins (Aptenodytes patagonicus) in the Crozet Archipelago (Viera, Le Bohec, Cote, & Groscolas, 2006); Nicobar Megapodes (Megapodius nicobariensis; Sivakumar, 2009), and mangrove, coral reef, and forest ecosystems (Ramachandran et al., 2005) of the Nicobar Islands; and plant communities of Phuket Island (Hayasaka, Goka, Thawatchai, & Fujiwara, 2012).

| Tide station data
To characterize the water levels around the Hawaiian archipelago, we extracted unprocessed data from the tide gauge at Sand Island and used 1-min intervals from 11 to 13 March 2011 (Station 1619910; NOAA 2014). We compared the 1-minute water levels with 15-second data to verify that a smaller time increment was not needed. We also extracted data from Kahului Harbor, Maui (Station 1615680) and across the Pacific at Crescent City, California (Station 9419750; NOAA 2015).

| Habitat mapping
We estimated nesting habitat of breeding bird communities by generalizing the land cover classification from satellite imagery into four nest substrates or sites: (1) in trees and shrubs, (2) underneath the trees, shrubs, and dense grasses, (3) on the ground, or (4) below ground in subterranean burrows and crevices (Table 1). We classified the land cover as described in Reynolds et al. (2012) using
Other species are not monitored for abundance, but their breeding phenology and habitat use is described to infer populations' vulnerability (Table 1; Figure 3).

| Uncertainty
Flooding extent uncertainty is primarily a function of the accuracy of the mapped inundation line (<5 m error) and species nest distributions. Nest sites were mapped as points (e.g., Masked and Brown Booby) while colony boundaries were mapped as linear features (e.g., Red-footed Booby and Great Frigatebird), both accurate to within 5-10 m. For this model, we assumed the data reflected typical longterm colony, distribution, and habitat for each species. Nest losses due to flooding were projected from empirical data of nesting distributions (Table 2) and were assumed to represent the nesting pattern at the time of the tsunami.

| Inundation characteristics
The tide station at Sand Island, Midway Atoll recorded eight events with water heights of at least 30 cm from 1952 through March 2017 (Table 3; (Table 4). The maximum wave run-up distance from the coastline was approximately 500 m at Sand Island, 500-800 m at Eastern Island, and 300 m at Laysan Island; Spit Island experienced complete overwash (approximately 300 m; Numerous wildlife and vegetation impacts 10,11 Data sources: 1: NOAA (2017)

| Flooding of seabird nesting areas
The Laysan albatross nests at Eastern Island experienced 75%-79% inundation and the only Short-tailed Albatross nest was flooded (Table 5).
On all islands, Black-footed albatross nests were concentrated in coastal areas (Figure 6), while Laysan Albatross and Bonin Petrel nests were distributed across the atoll (Figures 7 and 8; also see Reynolds, Courtot, Berkowitz et al., 2015).
At Laysan Island, the tsunami flooding covered 26% of Blackfooted and 17% of Laysan Albatross nesting habitat (Figure 9). Ten percent of Masked and 31% of Brown Booby nests that were mapped were flooded (Table 5). Eleven percent of the nesting area used by Red-footed Boobies and Great Frigatebirds was inundated (Table 6).
At Midway Atoll, 24% of Casuarina equisetifolia trees and 45% of the other tree/shrub habitat (e.g., Scaevola taccada, Tournefortia argentea) was flooded (Table 7). At Laysan, most of the suitable nesting habitat had <10% inundation except for bare ground, which experienced 31% inundation ( Table 7). The peak of the Masked Booby (Sula dactylatra) nesting season coincided with tsunami inundation (Figure 3), and of their potential ground-nesting habitat, 44% was inundated at Midway Atoll and 17% at Laysan Island (Tables 1 and 7). Similarly, tree and shrub canopy habitat used by White Terns (Gygis alba) was inundated at 33% at Midway Atoll and 5% at Laysan Island (Tables 1 and 7).

| Historical tsunamis and data limitations
We found accounts but with few details for 10 tsunamis reaching the Northwestern Hawaiian Islands (Table 3). Reports from Midway Atoll describe a 1952 tsunami that deposited sand and debris on runways and moved buildings on Sand Island (Lander & Lockridge, 1989;NOAA, 2017 (Pararas-Carayannis & Calebaugh, 1977;Rice, 1959). Although damage to infrastructure at Midway Atoll from the Tōhoku tsunami was reported as "limited" by NOAA (2017), we estimated with our spatially explicit models that more than 40% of the atoll was inundated T A B L E 6 Projected habitat inundation at Midway Atoll and Laysan Island, Hawai'i during the March 2011 Tōhoku tsunami for species with distribution or habitat use data Reynolds, Courtot, Berkowitz et al., 2015)  T A B L E 7 Land cover at Midway Atoll and Laysan Island, Hawai'i classified and quantified from WorldView-2 satellite imagery and area inundated by the March 2011 Tōhoku tsunami. Nesting habitat classified as: tree and shrub canopy (C), underneath trees, shrubs and dense bunch grasses on the ground (U), on the ground with vegetation or bare ground (G), and subterranean burrows and crevices (S). For more detailed descriptions of land cover classes, including species information, see Reynolds et al. (2012) and Reynolds, Courtot, Berkowitz et al. (2015) Land cover class

| Other studies of tsunami impacts on wildlife
Few studies have documented the impacts of tsunamis on wildlife populations. After the December 2004 Indian Ocean tsunami following the Sumatra-Andaman earthquake, short-term effects (e.g., Sivakumar, 2009;Viera et al., 2006)

| Coastal vegetation impact
The estimated speed of waves washing over land at Midway Atoll was 27.7 km/hr (R. Weiss, Virginia Tech, personal communication).
Coastal vegetation was physically uprooted by waves (Figure 13), and trees and shrubs in flooded areas died due to salt water exposure. Vegetation cover, structure, and composition were affected,  (Collen, Garton, & Gardner, 2009;Sundaresan, 1993) and indeed, at Midway Atoll during this event, dense stands of Scaevola taccada were associated with protective dunes and appeared less affected than other areas (JLK, personal observations).

| Vulnerability of bird communities
Using the Tōhoku tsunami as a case study, we provided a model to show the variation in vulnerability and impact to island biodiversity from sudden inundation events. This study illustrated how vulnerable low-lying islands are to sudden flooding. Population vulnerability to catastrophic events varies with differences in spatial and temporal exposure and sensitivity to the effects of the event (e.g., Gardali, Seavy, DiGaudio, & Comrack, 2012;Reynolds, Courtot, Berkowitz et al., 2015). Exposure to inundation is greater for species that concentrate near the coastline (e.g., Brown Booby and Black-footed Albatross), compared to species that typically nest farther inland (e.g., Red-footed Booby, Great Frigatebird), or species with nest distributions across the island (e.g., Laysan Finch).
Ground-nesting species such as albatrosses and Gray-backed and Sooty terns may suffer direct mortality from sudden flooding, or chicks displaced from their nests may not survive to independence.
Tree-and shrub-nesting species (e.g., White Tern) may be less ex- boobies, and frigatebirds are long-lived seabirds with deferred maturity, low fecundity, and high adult survival (Nelson, 1978;Warham, 1996;Weimerskirch, 2001). In contrast, species that lay replacement eggs or have asynchronous or aseasonal breeding (e.g., Black Noddy, White Tern) will likely be less sensitive to sudden flooding during their nesting season compared with species limited to a single nesting attempt or nest synchronously (e.g., albatrosses and petrels; Reynolds, Courtot, Berkowitz et al., 2015).
As sea level rises, insular wildlife will be increasingly exposed to flooding (Baker, Littnan, & Johnston, 2006;Bellard, Leclerc, & Courchamp, 2014;Reynolds et al., 2012). Higher sea levels in the coming decades potentially will cause greater inundation extents, increased flooding depths and durations, and increased flooding frequency (Eversole & Andrews, 2014), all of which have the potential to cause declines in bird populations and colony collapses.

| Conservation conclusions
Range-restricted and island species are vulnerable to extinction (Jenkins, Van Houtan, Pimm, & Sexton, 2015), and our spatially explicit models of tsunami inundation highlight the exposure of species nesting on Pacific islands. On the four islands of our study, a range of up to 6-10 million birds may rely on a combined area of about 9.3 km 2 , with a mean elevation of less than 3.5 m . Midway Atoll and Laysan Island support the largest colonies of Black-footed and Laysan albatrosses globally and, in total, more than 95% of the global populations of these strongly philopatric albatrosses nest on the low-lying Northwestern Hawaiian Islands (Arata, Sievert, & Naughton, 2009;Fefer, Harrison, & Naughton, 1984). These islands also support the global populations of two resident land bird spe-  (Suryan et al., 2006(Suryan et al., , 2008, warrants review in light of new scenarios of sea-level rise (Reynolds, Courtot, Berkowitz et al., 2015) and tsunami forecasts at Midway Atoll (Gica, 2015).
Reintroduction of other endangered taxa (e.g., Laysan Finch, Laysan Teal, Nihoa Finch (Telespiza ultima)), endangered plants, or Hawaii green turtles to additional islands could be considered a useful short-term or intermediate step in a conservation strategy to reduce extinction risks, or as an alternative to captivity, until a longer-term strategy to restore larger and higher elevation sites for recolonization (via translocation, social attraction, or natural immigration) can be implemented.

ACKNOWLEDGMENTS
We thank P. Leary and E. Flint of USFWS for technical assistance and logistical support with field data collection. We also thank C. Krause and G. Nielsen for assistance with the collection of Black-footed Albatross nest location points and data processing. Albatross census data were collected by numerous USFWS volunteers. We thank J.

CONFLICT OF INTEREST
None declared.

AUTHOR CONTRIBUTIONS
M.H.R and K.N.C. conceived and designed the study; J.L.K and K.N.C collected field data; P.B. and K.N.C. analyzed data; all authors contributed to writing the final manuscript.