healthmedicinet

Study species

Laysan and Black-footed albatrosses breed sympatrically on atolls throughout the NWHI
and show periodic declines in their reproductive success. The breeding cycle of both
Laysan and Black-footed albatrosses is well defined; eggs are laid in November and
December, and the incubation period lasts for approximately two months. Adult albatrosses
then brood chicks for several weeks (late January through mid to late February), guard
and provision the chicks for approximately one month and then return to the breeding
site to feed them regularly before the chicks fledge in June or July 41], 42]. Approximately 4300 Black-footed albatrosses and 3200 Laysan albatrosses nest at
Tern Island, representing 6.9 and 0.5 % of the total population in the North Pacific,
respectively 43]. The albatross colony at Midway Atoll is much larger, comprising approximately 408,000
Laysan and 22,000 Black-footed albatrosses 43]. However, standardized surveys of the colony at Midway Atoll were initiated much
later than those at Tern Island (1990s on Midway vs. 1980 on Tern Island) and the
resulting time series is considerably shorter.

Albatross telemetry and reproductive success data

We examined the foraging behavior and reproductive success of Laysan and Black-footed
albatrosses on Tern Island, French Frigate Shoals, NWHI (23.87°N, 166.28°W). Data
on albatross reproductive success at Tern Island (defined as the number of chicks
fledged per eggs laid) were obtained from the United States Fish and Wildlife Service
(USFWS). Standardized surveys of breeding birds and active nests have been conducted
at Tern Island since 1980; we used data from 1981-2012 (November-February 1981/82-2011/12)
to be consistent with available SST data. Reproductive success data are referred to
by the January/February calendar year (e.g., data from the November–February 1981/1982
breeding season are referred to as data from 1982). Albatross nests on Tern Island
were assigned to survey plots and were numbered and monitored by USFWS personnel throughout
each breeding season. Chicks were banded, and each nest was monitored weekly for hatching
chicks, chicks present within 30 m of the nest site, and dead chicks; chicks were
assumed to be dead if not found within 30 m of the nest site for three successive
observations.

Nest desertion by adults is the primary cause of reproductive failure in Laysan and
Black-footed albatrosses, and can be associated with poor foraging conditions, death
of the adults, or inexperienced breeders. Short-term climatic events such as flooding
and storms can also influence reproductive success over short time periods 44], 45], USFWS unpublished data]. However, given the protracted breeding season of Laysan
and Black-footed albatrosses (from November to June), such short-term events are unlikely
to influence reproductive success of the entire colony. At Tern Island, reproductive
success of these species appears to show periodic declines during which reproductive
success remains depressed for 2–3 years. We suggest that this pattern of reproductive
success is consistent with long-term oceanographic variability, and we assess this
hypothesis below.

We recorded albatross movements during the incubation and brooding periods from 2002–2006
and from 2008–2012 using satellite platform terminal transmitters (30 g Pico-100,
Microwave Telemetry, Columbia, MD, 42 g SPOT4 and SPOT5, Wildlife Computers, Redmond,
WA) and GPS data loggers (40 g Technosmart GPS, 35 g TechnoSmart GiPSy, 32 g EO Technologies,
and 30 g igotU, GT-120, Mobile Action Technology Inc, Taiwan). Tags were attached
with Tesa adhesive tape (Tesa, Hamburg, Germany) to dorsal feathers. All tag weights
represent water-proofed packaged tags and are well below the maximum mass threshold
recommended for albatrosses 46]. Only complete trips in which tracks covered the entire trip (leaving from and retuning
to Tern Island) were included in the analysis; a total of 93 trips for Laysan and
97 trips for Black-footed albatrosses were included in the analysis (Table 1), and only one trip per individual was tracked and included in the analysis. Trip
duration decreased towards the end of the incubation period for both species; we therefore
included only incubating trips during the first two months of the incubation period
(November and December), when trip duration remained consistent (Pearson’s correlation
coefficient between trip duration and day of breeding season 0.15, p value??0.4 for both species). Trip duration was consistent throughout the brooding
period (late January through late February) for both species (Pearson’s correlation
coefficient between trip duration and day of breeding season??0.15, p value??0.3 for both species). GPS and PTT tag data were resampled to a six-hour
time scale in order to provide sufficiently detailed spatial information at a time
scale that was appropriate for both tag types. Resampling was conducted using the
Minimum Covariance Determinant (MCD) in the MASS library (version 7.3-31) of the R
statistical package (version 3.0.2) in order to provide a robust estimate of location
at each time step that is not strongly influenced by outliers occurring due to the
spatial resolution of telemetry data. When fewer than four locations were available
within a time window, MCD cannot be computed and the coordinate-wise median was used
47].

Table 1. Tracks used in analyses by species, breeding stage, tag type and year

For each albatross foraging trip, we assessed cumulative trip distance, maximum distance
travelled from Tern Island, and the duration of each trip using the ArgosFilter (version
0.63) and MASS libraries in R. To examine albatross movement in relation to the location
of the TZCF, we calculated distances to TZCF for each location on each track using
daily rasters of distance to TZCF (see below) and assessed which tagged birds spent
time north of the TZCF. We produced kernel density distributions for both species
during incubating and brooding periods. Kernel densities were calculated with ArcGIS
10.2.2 Spatial Analyst using a fixed radius of 100 km for incubating trips and 50 km
for brooding trips. We then delineated habitat use as the 95, 50 and 25 % isopleths
of kernel density distributions 29], 33].

Transition Zone Chlorophyll Front (TZCF)

The TZCF is a basin-wide front, spanning more than 8000 km from west to east across
the North Pacific. Representing a zone of convergence, the TZCF separates the cool,
well-mixed, nutrient-rich waters of the subarctic gyre from the warmer, stratified,
nutrient-poor waters of the subtropical gyre 35]. The TZCF is characterized by a sharp chlorophyll gradient, and is defined by a surface
chlorophyll value of 0.2 mg/m
³35]. The 18 °C isotherm in SST has been demonstrated to provide a proxy for the location
of the TZCF 37].

During the breeding season of albatrosses in the North Pacific (November–June), there
is considerable interannual variability in the location of the TZCF 37] but the front is generally closest to Tern Island in January and February, coinciding
with the brooding period (Fig. 2). Given the importance of the TZCF as feeding grounds for breeding albatrosses, the
proximity of the front to the NWHI likely has important implications for both Laysan
and Black-footed albatrosses. Historical records of SST are available at a finer temporal
and spatial resolution compared to chlorophyll and can thus be more useful in historical
models of habitat use 37]. We localized the TZCF on a daily time scale using daily Group for High Resolution
SST (GRHSST) images with a 5 km resolution. We created shapefiles of TZCF location
for each day of the incubating and brooding periods from 1982–2012 (November-February
1981/1982-2011/2012) and generated rasters of distance to TZCF for each day in order
to examine variability in frontal location relative to albatross reproductive success
and foraging distribution. Within the albatross brooding period, the TZCF can range
from a minimum of approximately 400 km to a maximum of approximately 1150 km from
Tern Island. All spatial analyses were conducted in ArcGIS 10.2.2 using the Spatial
Analyst extension and using the raster (version 2.3-34), geosphere (version 1.3-8)
and Imap (version 1.32) R Statistical packages 48].

Large-scale Oceanographic and Climatic Indices

We obtained time series of the North Pacific Gyre Oscillation index (NPGO) and the
Multivariate El Niño Southern Oscillation (ENSO) Index (MEI) to represent large-scale
climatic and oceanographic variability in the central North Pacific (MEI and NPGO
data obtained from http://www.esrl.noaa.gov/psd/enso/mei/table.html and http://www.o3d.org/npgo/npgo.php, respectively). NPGO represents the strength of the North Pacific Gyre 49], 50], where high/low values indicate expansion/contraction of the gyre, respectively.
MEI identifies El Niño events by incorporating variability in six oceanographic and
climatic variables over the tropical Pacific 51], with positive values of MEI representing El Niño conditions, and negative values
representing La Niña conditions. El Niño events have been linked with southern migrations
in the TZCF 37] and the front migrates in association with expansion/contraction of the North Pacific
gyre, making MEI and NPGO important variables to evaluate in our models of albatross
behavior and reproductive success.

We used GRHSST data (see above) to examine SST on an annual scale relative to reproductive
success, and at a daily scale relative to albatross trip metrics. The location of
the TZCF represents the boundary between the subarctic and subtropical gyres and therefore
reflects broad-scale patterns in SST, but variability in SST at a finer spatial scale
may influence search effort of foraging albatrosses 33] and was therefore examined separately. We produced rasters of mean SST for each year
during the albatross incubating and brooding periods from 1981–2012 (November–February
1981/82–2011/12), as well as the 31-year mean for this time period. We then produced
rasters of SST anomalies (SSTa) using the formula SSTa
_i
?=?SST
_i
– SST
_mean
, where i is the year, SST
_i
is the mean SST from November–February for year i and SST
_mean
is 31-year SST mean from the November–February time period. As with reproductive success
data, SST data are referred to by the calendar year of the January/February period
(i.e., SST data from November–February 1981/1982 are referred to as 1982 SST data).
We assessed SST within the 95 % kernel density isopleths for Laysan and Black-footed
albatrosses during the brooding and incubating stages, respectively. There are considerable
differences in habitat use between species and breeding stages, and therefore it was
important to assess changes in SST in each region independently 52].

Statistical Analyses

In order to examine the effects of oceanographic and climatic indices on albatross
reproductive success, we examined average, minimum and maximum values of SST, MEI
and NPGO annually during the albatross breeding season (November–February). We examined
time-lagged effects on albatross reproductive success using cross-correlation functions
(CCFs). CCFs are useful for identifying predictor variables (x
_t
) that might have lagged effects on a dependent variable (y
_t
), and examine correlations between the dependent variable and predictor variables
at different time lags (correlations between x
_t-h
and y
_t
for different time lags represented by h?=?0, 1, 2, 3, etc.). Here we identified variables
that had time-lagged effects on albatross reproductive success for time lags of 0
to 5 years. Variables with significant CCFs were time-lagged and included as lagged
variables for further analyses. However, after applying model selection (described
below), time lagged variables were not included in final models. The final models
of albatross reproductive success included the following variables: minimum distance
to TZCF, minimum NPGO, minimum MEI and mean SST in brooding habitat. Albatross trips
were examined over the ten-year period, and were evaluated relative to both broad-scale
(monthly MEI and NPGO) and finer-scale (daily SST and TZCF) oceanographic and climatic
variables. Though MEI and NPGO are broad-scale metrics, these variables vary considerably
over a period of several months and were therefore sampled at the midpoint of albatross
foraging trips along with finer-scale variables. To quantify the effects of these
variables on albatross trip metrics, we assessed minimum, maximum and mean values
of distance to TZCF, NPGO, MEI and SST in brooding or incubating habitat. After model
selection, minimum SST was used in trip metrics for Laysan albatrosses, while mean
SST was used in models for Black-footed albatrosses.

We examined relationships between La Niña events, gyre expansion, TZCF location and
SST on albatross trip metrics and reproductive success using several analytical approaches.
We used Pearson’s correlation coefficients and Wilcoxon signed rank tests to examine
relationships among different oceanographic variables and between trip metrics and
reproductive success. We applied Principal Component Analysis (PCA) combined with
Generalized Linear Models (GLMs) to examine the overall effects of multiple climatic
indices. PCA provides a means of summarizing the variability across a number of correlated
variables (MEI, NPGO, SST, proximity of TZCF to Tern Island) into fewer independent,
orthogonal axes. GLMs are statistical regression models 53] that allow for different types of predictors (continuous, binary, ordinal) to be
evaluated [e.g., 54]. Here we constructed separate PCAs to summarize environmental variation and to evaluate
their effects on Laysan and Black-footed albatrosses, respectively, at two scales:
at an annual level to evaluate effects on reproductive success, and at the trip level
to assess effects on albatross trip metrics. We then used GLMs to evaluate how reproductive
success and trip metrics were influenced by the environment (PC axes), and included
PC axes with greater than 15 % of variance explained as predictor variables. We examined
the two species’ brooding and incubating periods separately due to the differences
in the timing, spatial habitat use of albatrosses, and associated differences in environmental
variables. This allowed us to resolve how relationships with predictor variables differed
between the brooding and incubating periods, thus indicating how metrics of albatross
foraging differed during the most constraining period of breeding (brooding). For
all GLMs, we used Akaike Information Criterion (AIC) 55] to select variables for the most parsimonious model 56]. PCAs and GLMs were performed using the stats (version 3.0.2) and mgcv (version 1.7-29)
R Statistical packages, respectively.

To further examine relationships between albatross trip metrics and proximate oceanographic
variables, we compared trip metrics with MEI, NPGO, SST, and distance to TZCF (assessed
as values above/below the mean values) using Wilcoxon signed rank tests. We also compared
trip metrics below/above mean reproductive success using Wilcoxon signed rank tests.

We used spatial correlations to illustrate how the relationship between SSTa and albatross
reproductive success varied spatially. Using the 31-year time series for both reproductive
success and SSTa (November–February average for each year from 1981–2012) described
above, annual reproductive success was correlated with each SSTa grid cell using Pearson’s
correlation coefficients, producing a spatial correlation between these two datasets
for each grid cell.