Spatio-temporal gap analysis of OBIS-SEAMAP project data: assessment and way forward.

Abstract

The OBIS-SEAMAP project has acquired and served high-quality marine mammal, seabird, and sea turtle data to the public since its inception in 2002. As data accumulated, spatial and temporal biases resulted and a comprehensive gap analysis was needed in order to assess coverage to direct data acquisition for the OBIS-SEAMAP project and for taxa researchers should true gaps in knowledge exist. All datasets published on OBIS-SEAMAP up to February 2009 were summarized spatially and temporally. Seabirds comprised the greatest number of records, compared to the other two taxa, and most records were from shipboard surveys, compared to the other three platforms. Many of the point observations and polyline tracklines were located in northern and central Atlantic and the northeastern and central-eastern Pacific. The Southern Hemisphere generally had the lowest representation of data, with the least number of records in the southern Atlantic and western Pacific regions. Temporally, records of observations for all taxa were the lowest in fall although the number of animals sighted was lowest in the winter. Oceanographic coverage of observations varied by platform for each taxa, which showed that using two or more platforms represented habitat ranges better than using only one alone. Accessible and published datasets not already incorporated do exist within spatial and temporal gaps identified. Other related open-source data portals also contain data that fill gaps, emphasizing the importance of dedicated data exchange. Temporal and spatial gaps were mostly a result of data acquisition effort, development of regional partnerships and collaborations, and ease of field data collection. Future directions should include fostering partnerships with researchers in the Southern Hemisphere while targeting datasets containing species with limited representation. These results can facilitate prioritizing datasets needed to be represented and for planning research for true gaps in space and time.

Full text

Page 1

Spatio-Temporal Gap Analysis of OBIS-SEAMAP Project
Data: Assessment and Way Forward
Connie Y. Kot1*, Ei Fujioka1, Lucie J. Hazen1, Benjamin D. Best1, Andrew J. Read2, Patrick N. Halpin1
1Marine Geospatial Ecology Lab, Nicholas School of the Environment, Duke University, Beaufort, North Carolina, United States of America, 2Nicholas School of the
Environment, Duke University, Beaufort, North Carolina, United States of America
Abstract
The OBIS-SEAMAP project has acquired and served high-quality marine mammal, seabird, and sea turtle data to the public
since its inception in 2002. As data accumulated, spatial and temporal biases resulted and a comprehensive gap analysis
was needed in order to assess coverage to direct data acquisition for the OBIS-SEAMAP project and for taxa researchers
should true gaps in knowledge exist. All datasets published on OBIS-SEAMAP up to February 2009 were summarized
spatially and temporally. Seabirds comprised the greatest number of records, compared to the other two taxa, and most
records were from shipboard surveys, compared to the other three platforms. Many of the point observations and polyline
tracklines were located in northern and central Atlantic and the northeastern and central-eastern Pacific. The Southern
Hemisphere generally had the lowest representation of data, with the least number of records in the southern Atlantic and
western Pacific regions. Temporally, records of observations for all taxa were the lowest in fall although the number of
animals sighted was lowest in the winter. Oceanographic coverage of observations varied by platform for each taxa, which
showed that using two or more platforms represented habitat ranges better than using only one alone. Accessible and
published datasets not already incorporated do exist within spatial and temporal gaps identified. Other related open-source
data portals also contain data that fill gaps, emphasizing the importance of dedicated data exchange. Temporal and spatial
gaps were mostly a result of data acquisition effort, development of regional partnerships and collaborations, and ease of
field data collection. Future directions should include fostering partnerships with researchers in the Southern Hemisphere
while targeting datasets containing species with limited representation. These results can facilitate prioritizing datasets
needed to be represented and for planning research for true gaps in space and time.
Citation: Kot CY, Fujioka E, Hazen LJ, Best BD, Read AJ, et al. (2010) Spatio-Temporal Gap Analysis of OBIS-SEAMAP Project Data: Assessment and Way
Forward. PLoS ONE 5(9): e12990. doi:10.1371/journal.pone.0012990
Editor: Yan Ropert-Coudert, Institut Pluridisciplinaire Hubert Curien, France
Received June 29, 2010; Accepted August 31, 2010; Published September 24, 2010
Copyright: ? 2010 Kot et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted
use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: Current funding for OBIS-SEAMAP comes from the National Science Foundation (http://www.nsf.gov) and the National Oceanic Partnership Program
(http://www.nopp.org) 3-year award number NSF-OCE-07-39199. The funders had no role in study design, data collection and analysis, decision to publish, or
preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: connie.kot@duke.edu
Introduction
The need for the conservation of marine mammals, seabirds,
and sea turtles, is increasing due to the on-going and long-term
negative effects of direct harvests/kills, indirect fisheries catch, and
habitat alteration and degradation. A detailed understanding of
the spatial and temporal patterns of species distribution and
diversity is critical for quantifying populations, the significance of
adverse events, and the potential for mitigation. Resources may be
limited for researchers to gather enough information over long
periods of time, across large regions, and on multiple species to
fully assess conservation requirements. Therefore, a world data
commons where multiple datasets are available can facilitate this
fundamental need.
Established by the Census of Marine Life program [1] in 2002,
Duke University is leading the OBIS-SEAMAP project (Ocean
Biogeographic Information System – Spatial Ecological Analysis of
Megavertebrate Populations project, herein called SEAMAP)
involving a consortium of organizations and individuals who
share a vision to make marine biogeographic data freely available
to the public [2]. SEAMAP is one of the participating network
data nodes of OBIS (http://www.iobis.org), which in turn, is a
member and data provider of the Global Biodiversity Information
Facility (GBIF; http://www.gbif.org) [3]. Data are aggregated by
specific marine taxa (SEAMAP), up to all marine biogeographic
data (OBIS), and finally to global (including terrestrial) biogeo-
graphic data (GBIF). Compared to GBIF and OBIS, SEAMAP
promotes the storage and publication of many more types of data
(i.e., effort, animal behavior, etc.) while providing additional
features and tools for both data providers and potential users
interested in marine megavertebrates [4].
The SEAMAP project’s objective is to compile any existing geo-
referenced data for marine mammals, seabirds, and sea turtles that
can be used to better understand the spatial and temporal patterns
of species distribution and diversity in the global ocean, such as at-
sea surveys (from a shipboard or aerial platform), land-based
counts (from shore), satellite telemetry data (from tagged animals),
and stranding data (from shore). These high quality data are
standardized with a minimum set of fields including the taxon
name, latitude/longitude position, and date/time for observations,
begin/end dates and locations for effort data, and individual
identification codes for satellite telemetry tag data [4]. Besides
direct contributions, data can be delivered to SEAMAP from other
data portals in an automated fashion. For instance, many of the
PLoS ONE | www.plosone.org1 September 2010 | Volume 5 | Issue 9 | e12990

Page 2

sea turtle and telemetry datasets were fed into SEAMAP from the
Satellite Tracking and Analysis Tool (STAT) available at
SeaTurtle.org [5]. This tool enables direct consumption of the
real-time ARGOS (Advanced Research and Global Observation
Satellite) data. STAT users can freely sign up for their satellite tag
data to be processed online with an option to upload and publish
the data onto SEAMAP with a simple tick of a box.
Data collected for the SEAMAP project are publicly available
on a web-based system (http://seamap.env.duke.edu/) that is
intended for educators, students, managers, and researchers and
provides advanced mapping tools, a variety of supplemental
products (i.e., species profiles and photos), and tabular data
summaries. The SEAMAP website also provides metadata for all
datasets, including information on the data providers and
collectors, description of techniques used to gather the data,
survey effort details (when available), and methods to process the
data for analysis. The interactive website features sophisticated
querying and mapping capabilities for displaying specific data
from multiple datasets across space and time. In addition, remotely
sensed oceanographic data such as bathymetry and sea surface
temperature can be viewed alongside observations. These
environmental data are associated with observations synchronous
in time and space, and available for download.
Currently, the SEAMAP database hosts over 2.3 million records
with over 280 datasets from individuals, government agencies,
non-profit organizations, and academic institutions. Ongoing
dataset contributions from each of the three marine megafauna
communities, both solicited and unsolicited, consistently increase
numbers of observations and species holdings while expanding the
temporal and geographic range of high quality data. However, it is
apparent that temporal and geographic biases and gaps exist
within the data holdings. A comprehensive gap analysis provides
information that can facilitate prioritization in targeting new
SEAMAP datasets to fill these gaps. In turn, results from the
analysis provide direct feedback to management agencies and the
research community for planning future surveys when a true gap
in knowledge exists.
While the term ‘‘gap analysis’’ is used here to broadly refer to
analyzing missing data, gap analysis has been referred to in the
past as the approach of identifying conservation gaps by overlaying
species distributions predicted from the environment with existing
protected areas per the U.S. NBII Gap Analysis Program (http://
gapanalysis.nbii.gov) [6]. This program’s more specific gap
analysis application with SEAMAP data would certainly add
great conservation value, but is beyond the scope of this paper.
Here, we more simply describe the gaps in the SEAMAP data
holdings, bound by taxa, space, time, and environment.
Results
As of February 12, 2009, SEAMAP has published over 2.2
million records from 240 datasets. About 20% of these include
records with spatial errors (missing latitude or longitude coordi-
nates or on land), missing observation dates, and observations that
were not marine mammals, seabirds or sea turtles. The remaining
published observations (1,839,510 records with 9,862,073 individ-
ual marine mammals, seabirds, and sea turtles observed) from 234
different datasets had all of the required data for the gap analysis.
Points that fell on land were excluded, except for records from four
shore-based surveys (988 records). The majority of records
consisted of seabird observations from 67 datasets (1,607,041
records; 7,534,777 individuals; 230 different taxa). Combined
observations of marine mammals from 181 datasets (154,485
records; 2,246,463 individuals; 116 different taxa) and sea turtles
from 104 datasets (77,984 records; 80,833 individuals; 8 different
taxa) made up less than half of the data and generally overlapped
spatially with seabird distribution, although sea turtle observations
were much lower in number in northern Europe. Since a large
proportion of observation data (over 87% of records, and over
76% of animals) consisted of seabirds, temporal and spatial
distribution results were broken down by taxa rather than lumping
all three taxa together.
Overall, most datasets and records were gathered from
shipboard surveys (n=115; 1,478,671 records), followed by aerial
surveys (n=66; 251,107 records), tag data (n=49; 100,527
records), and shore-based observations (n=4; 9,205 records).
The majority of marine mammals and seabird datasets and
records came from shipboard observations (63% and 86% of
records, respectively), but sea turtle data mostly came from tags
(86% of records). When comparing the number of individuals
(number of records * group size) from aerial, shipboard, and shore
surveys, the relationship between the number of records and
animals obtained varied by platform and taxa. Shore surveys gave
the greatest group sizes for marine mammals (mean=28) and
seabirds (mean=1,161). After shore surveys, shipboard surveys
had the next highest group size for marine mammals (mean=16)
though only slightly higher than data gathered from aerial surveys
(mean=14). For seabirds, aerial survey datasets had a slightly
higher mean group size over the shipboard datasets (mean=6 and
4, respectively). Sea turtle mean group sizes did not differ among
platforms (mean=1).
As of the February 2009 cut-off, three shore surveys for marine
mammals included data from whale strandings, seal haulouts, and
sea otter sightings while the one shore-based seabird dataset
included data from seabird colonies. Although SEAMAP currently
displays two shore-based datasets on sea turtles (i.e., sea turtle
nesting sites), no shore-based sea turtle data was included in the
gap analysis because data were unavailable for public download at
the time of this analysis.
Trackline data published on SEAMAP was associated with the
majority (.78%) of datasets examined in this analysis. Since only a
few survey trackline datasets flagged records as either ‘‘on’’ or
‘‘off’’ effort, it was assumed that all data were ‘‘on’’ or qualified as
a period or track length where observations were actively being
sought and recorded when a dataset did not have flags. A total of
184 observation datasets had associated trackline data (1,338,771
records within Food and Agriculture Organization [FAO] global
statistical fishing areas [7]), mainly from datasets collected by
shipboard surveys (n=85) and aerial surveys (n=54), followed by
satellite tag data (n=45). Globally, 3,956,372 km of tracklines are
represented on SEAMAP. These tracklines, or polylines, represent
the shortest distance connecting points reported as positions within
a survey track or within a satellite tag path. However, the total
length of effort calculated per region may be an underestimate
since survey effort and a tagged animal’s path are unlikely to be a
straight line between recorded locations.
Temporal distribution
For the 234 datasets published on SEAMAP and included in
this analysis, most contained data observed during the late 1990s,
with the maximum number of datasets contributing to any one
year occurring in 1999 (n=36; Figure 1). Marine mammal data
published on SEAMAP ranged from 1935–2009, sea turtle data
ranged from 1966–2009, and seabird data ranged from 1940–
2008 (Figure 2). For marine mammals, the peak number of records
was in 1992 (n=15,219) and the peak in observed animals was in
1993 (n=179,720; Figure 2a). For seabirds, SEAMAP had the
highest number of records occurring in 1981 (n=119,573) and the
OBIS-SEAMAP Gap Analysis
PLoS ONE | www.plosone.org2September 2010 | Volume 5 | Issue 9 | e12990

Page 3

highest number of seabirds observed in 1980 (n=541,156;
Figure 2b). Finally, sea turtle records and individuals were
extremely low for most years within the time range with a drastic
peak in records during 2007 (n=13,447) and animals in 2008
(n=13,273; Figure 2c). This recent jump in the number of records
and observations was due to five datasets, two of which were from
satellite tags contributing the majority of the data in 2007 and
2008 (12,375 and 12,149 records respectively).
Recorded observations and number of individuals were highest
in the summer for all three major taxa (Figure 3). This also
coincided with the length of tracklines published on SEAMAP per
season with summer having the greatest coverage, followed by fall,
spring, and winter. When observations were broken down by taxa,
the number of records and animals were next highest for marine
mammals and sea turtles in the fall, followed by spring and then
winter (Figures 3a and 3c). For seabirds, however, high records
were found in winter, followed by spring and then fall while the
number for animals observed was higher in the fall, followed by
spring and then winter (Figure 3b). The higher number of seabird
records in the winter was mostly due to the increased number of
aerial surveys compared to fall while the higher number of
seabirds observed in the fall was mostly due to more shipboard
surveys than in the winter. July and August were the peak summer
months while December and January had some of the lowest
numbers. However, April had the least number of records for
marine mammals and seabirds.
Spatial distribution
Aggregated SEAMAP data revealed large gaps in the Southern
Hemisphere, especially in the Pacific and Atlantic Oceans, while
there was a high concentration of records in the Pacific northeast
(NE), and northern Atlantic (Figure 4a; Table 1). The highest
number of datasets were found to contribute data in the Pacific NE
(n=72), followed closely by the Atlantic central-west (CW; n=69);
regions with the least number of datasets (n=3 for each region)
were the Arctic, Atlantic southwest (SW), Atlantic-Antarctic, and
Pacific-Antarctic. Marine mammal records were present in every
region, but were particularly low in the Pacific CW and northwest
(NW) as well as the Atlantic SW and southeast (SE; Figure 4b).
High concentrations of marine mammal records and animals were
in the Pacific central east (CE), northern Atlantic, and Mediter-
ranean. Seabird records and number of animals were low in most
of the southern regions, with major gaps in the Atlantic CE and
SE, Pacific SW and SE, with no records found in the Pacific CW
(Figure 4c). Compared to the other taxa, a higher proportion of
seabird records were found in the Pacific NE, Arctic, and Indian-
Antarctic. Sea turtles, however, had high numbers of records from
the Atlantic CW and CE regions, but gaps in the southern Pacific,
Arctic and Antarctic regions (Figure 4d). Absence of sea turtle
observations near the poles is expected as this ectothermic taxa
requires warmer habitats.
Within FAO regions, high densities of records for all taxa were
shown to be particularly biased towards the coastal areas with
decreasing densities moving further offshore (Figure 5). Coastal
biases were particularly apparent for marine mammals and
seabirds around the United States (US) and United Kingdom
(Figures 5b and 5c). Less clustered areas (low record densities)
further offshore were a result of satellite tag datasets that tracked
individual animal movements across large regions such as seabirds
in the Pacific NE and CE (Figure 5c) and sea turtles in the Pacific
CW (Figures 5d).
The relative distribution of tracklines among and within regions
sensibly coincides with observational records (Figures 5 and 6).
Most of the published tracklines were concentrated in the north
and central Atlantic regions (highest km of tracklines was found in
the Atlantic NW region). SEAMAP did not have any published
datasets with tracklines in the Pacific-Antarctic (Table 2). The area
covered in most regions was minimal and the presence of effort
within an area was less than 50% for all regions (Table 2), with
most tracklines concentrated near the coast. The amount of areal
coverage calculated per region is likely an overestimate since the
buffer around the trackline was larger than what is typical (survey/
animal swaths are usually ,100 km of the ship, plane, satellite
telemetry tag path).
When broken down by platform, densities of shipboard survey
tracks were the highest at around 7.4 km per square km, found
near Great Abaco Island, Bahamas (Figure 6a) where high
numbers of sea turtles (loggerhead Caretta caretta being the most
abundant species) and cetaceans (bottlenose dolphin Tursiops
truncatus being the most abundant species) were observed. Track-
line density for aerial surveys was highest around 7 km per square
km, found in Puget Sound, Washington, US (Figure 6b), where
high numbers of seabirds (northern fulmar Fulmarus glacialis and
common murre Uria aalge being the most abundant species) were
observed. Tracklines for satellite tag data were the least dense out
of the three platforms, with a maximum of around 3.7 km per
Figure 1. The number of datasets published on SEAMAP containing data records for each year of observations.
doi:10.1371/journal.pone.0012990.g001
OBIS-SEAMAP Gap Analysis
PLoS ONE | www.plosone.org3 September 2010 | Volume 5 | Issue 9 | e12990

Page 4

square km found near St. Petersburg, Florida, US (Figure 6c)
where high numbers of sea turtles (loggerheads being the most
abundant species) were tagged and tracked. Other hotspots for
satellite tag trackline density appeared near Grand Manan Island,
New Brunswick, Canada where many harbor porpoises (Phocoena
phocoena) were tracked and near Pamlico Sound, North Carolina,
US where loggerhead sea turtles were the most abundant species
tracked (Figure 6c).
Figure 2. The number of records and animals published on SEAMAP each year. Number of records (solid red line) and animals (dashed blue
line) for: (A) marine mammals; (B) seabirds; and (C) sea turtles. (Marine mammal, seabird, and sea turtle icon credit: Tracey Saxby, IAN Image Library,
ian.umces.edu/imagelibrary/).
doi:10.1371/journal.pone.0012990.g002
OBIS-SEAMAP Gap Analysis
PLoS ONE | www.plosone.org4 September 2010 | Volume 5 | Issue 9 | e12990

Page 5

Oceanographic coverage
It was not possible to associate environmental variables with
many of the dates and locations for observations because the
availability of such variables was restricted to more recent years
(Table 3) and limited by cloud cover. About 92% of observations
were associated with bathymetry because temporal restrictions did
Figure 3. The number of records and animals published on SEAMAP each month. Number of records (solid red line) and animals (dashed
blue line) for: (A) marine mammals; (B) seabirds; and (C) sea turtles. (Marine mammal, seabird, and sea turtle icon credit: Tracey Saxby, IAN Image
Library, ian.umces.edu/imagelibrary/).
doi:10.1371/journal.pone.0012990.g003
OBIS-SEAMAP Gap Analysis
PLoS ONE | www.plosone.org5 September 2010 | Volume 5 | Issue 9 | e12990

Page 6

not apply, though data were lacking in coastal areas. Sea surface
temperature (SST) data were available for over 20 years and could
be associated with about 57% of observations in time and space.
For sea surface height (SSH) and chlorophyll (CHL) data, only
about a quarter or less of the observations were matched in time
and space mostly due to the temporal restrictions of the
environmental data. Besides pre-dating the availability of satellite
coverage and cloud cover, environmental data may also be missing
due to calculation error associated with coastal cells (and
equatorial cells for SSH) or satellite instrument malfunctions.
For SST and CHL, the trends of using weekly/8 day, monthly,
and yearly values for each taxa were the same (i.e., yearly,
monthly, and weekly SST values were highest for sea turtles and
lowest for seabirds) and, similarly, weekly values of SSH followed
the same relationship with taxa. Therefore, only monthly values
were presented here to show seasonal averages, modes, and ranges
(see Tables 4 and 5).
The distribution of bathymetry, SST, CHL, and SSH by taxa
and by platform all deviated significantly from normal (Shapiro-
Wilk test, p,0.001). In addition, most environmental layers by
taxa and by platform significantly deviated from unimodality
(p,0.001). Exceptions included bathymetry sampled by sea turtle
sightings gathered by shipboard survey and by satellite tag, yearly
SSH sampled by sea turtle sightings gathered by satellite tag, and
SSH sampled by seabird sightings gathered by shipboard survey
(p.0.001).
Depth values associated with observations ranged from 1 to
9,294 m, but were heavily skewed towards shallow waters
(,350 m; median=82 m; mode=53 m). Average depths varied
significantly with taxa and platform (Kruskal-Wallis test, p,0.01;
Table 4). Sea turtles were observed over the deepest waters,
followed by marine mammals, and then seabirds; datasets
Figure 4. Relative abundance of records per square km published on SEAMAP in FAO statistical fishing areas. Abundance for: (A) all
marine mammals, seabirds, and sea turtles; (B) marine mammals; (C) seabirds; and (D) sea turtles. Each color represents one quintile (hot to cool
colors: red, orange, yellow, light blue, and dark blue), containing the number of records that fell into the .79th(high), 79-60th(medium-high), 59-40th
(medium), 39-20th(medium-low), and ,20th(low) percentiles, respectively. See Table 1 for FAO global statistical fishing area codes and regions.
doi:10.1371/journal.pone.0012990.g004
Table 1. Regional summary of all marine mammal, seabird,
and sea turtle observations data published on SEAMAP.
Region
(FAO Code)
Records
(n)
Animals
(n)
Records per
million km2
Arctic (18)8,364 78,263 1,145.75
Atlantic CE (34)9,906 15,409707.57
Atlantic CW (31)58,193 371,6413,958.71
Atlantic NE (27)1,146,130 3,585,44367,818.34
Atlantic NW (21)250,2572,844,695 48,126.35
Atlantic SE (47)2,0252,601 108.87
Atlantic SW (41)1,454 8,10082.61
Atlantic-Antarctic (48) 6,679 54,928543.01
Indian E (57) 5,48810,353 184.16
Indian W (51)4,922 42,663 162.98
Indian-Antarctic (58)14,435 115,103 1,145.63
Mediterranean (37)5,142 170,7391,714.00
Pacific CE (77)137,9241,611,324 2,820.53
Pacific CW (71)2,599 5,47978.28
Pacific NE (67) 173,318793,335 23,109.07
Pacific NW (61)3,2583,939 158.93
Pacific SE (87)2,9723,30999.07
Pacific SW (81)3,198134,807 112.61
Pacific-Antarctic (88) 3,2469,942312.12
doi:10.1371/journal.pone.0012990.t001
OBIS-SEAMAP Gap Analysis
PLoS ONE | www.plosone.org6September 2010 | Volume 5 | Issue 9 | e12990

Page 7

gathered by tag were over deeper waters than ships, followed by
aerial surveys and shore surveys (Table 4). For marine mammals,
the number of records for all four platforms peaked in shallow
waters; the most frequent bathymetry for observations collected by
ship, tag, and shore was at 1 m while data gathered by aerial
surveys were mostly at 12 m (Table 5). Kernel density distributions
for marine mammal depths differed greatly among platforms
(Figure 7a). For seabirds, all platforms had high numbers of
observations from shallow waters, while satellite tag data also had
an abundance of data over depths around 5,500 m (Figure 7b).
Sea turtles were also mostly observed in shallow waters, although
the satellite tags allowed for a greater number of observations over
deep water, with most data over waters around 5,600 m deep
(Figure 7c, Table 5). Not including the shore-based datasets, data
gathered by aerial surveys had the smallest range compared to
other platforms, with a range of only about 4,600 m (Table 5).
Monthly SST values associated with observations ranged from
22.92 to 35.17 degrees Celsius (uC), with most observations at
22.9uC although many temperatures were around the median
(11.10uC). Like bathymetry, average SST (weekly, monthly, and
yearly) varied significantly with taxa and platform (Kruskal-Wallis
test, p,0.01; Table 4). Sea turtles were observed in higher SSTs,
followed by marine mammals, and then seabirds; datasets
gathered by tag were in warmer waters than shipboard surveys,
followed by aerial surveys, and then shore surveys (Table 4). For
marine mammals, most shipboard observations were in the
warmest waters, followed by planes, tags and then shore
observations (Table 5). Kernel density distributions for marine
mammal SSTs varied greatly among the different platforms
(Figure 7d). For seabirds, aerial and shipboard surveys had high
numbers of observations around 7 and 10uC, respectively, while
satellite tag data included an abundance of data recorded in
warmer waters (Figure 7e, Table 5). Sea turtles were primarily
observed in warm waters, with modes between 17–31uC for each
platform (Figure 7f, Table 5).
Monthly SSH values associated with observations ranged from
216.5 to 314.8 cm (median=119.1 cm; mode=157 cm). SSH
(daily and monthly) varied significantly with taxa and platform
(Kruskal-Wallis test, p,0.01). Sea turtles were observed in higher
SSHs, followed by marine mammals, and then seabirds; datasets
gathered by tag were higher in SSH than aerial surveys, followed
by shipboard, and then shore-based surveys (Table 4). For marine
mammals, the number of records for all four platforms peaked in
waters around 120 cm, with shipboard and aerial observations
peaking again near 160 cm (Figure 7g; Table 5). For seabirds, all
platforms peaked at different SSHs, with shipboard observations in
lower waters than aerial observations, followed by satellite tags
(Figure 7h, Table 5). For sea turtles, all platforms also peaked at
different SSHs, with shipboard survey data having peaks slightly
overlapping with both aerial and tag data (Figure 7i). Compared
with other environmental variables, SSH modes seem to be most
divided among platforms per taxa.
Monthly CHL values associated with observations ranged from
0.01 to 44.15 mg/m3(median=0.86 mg/m3; mode=1.50 mg/
m3). CHL (weekly and monthly) varied significantly with taxa and
platform (Kruskal-Wallis test, p,0.01). Marine mammals were
observed in highest average CHLs, followed by seabirds, and then
sea turtles; datasets gathered from shore were higher in average
CHL than aerial surveys, followed by shipboard surveys, and then
tag data (Table 4). For all taxa and all platforms, the number of
records peaked in waters of low CHL concentration (Figures 7j–l,
Table 5). Shore surveys yielded the widest range of data for CHL,
while sea turtles observed by shipboard surveys had the smallest
range along with relatively low CHL concentrations (Table 5).
Figure 5. Relative abundance of records per square degree published on SEAMAP in 1 degree square cells. Abundance for: (A) all
marine mammals, seabirds, and sea turtles; (B) marine mammals; (C) seabirds; and (D) sea turtles. Each color represents one quintile (hot to cool
colors: red, orange, yellow, light blue, and dark blue), containing the number of records that fell into the .79th(high), 79-60th(medium-high), 59-40th
(medium), 39-20th(medium-low), and ,20th (low) percentiles, respectively.
doi:10.1371/journal.pone.0012990.g005
OBIS-SEAMAP Gap Analysis
PLoS ONE | www.plosone.org7September 2010 | Volume 5 | Issue 9 | e12990

Page 8

Identifying potential datasets
A total of 636 references were identified with many high quality
datasets within all regions, including regions in which the
SEAMAP project currently holds little or no data (Text S1).
There were no regions that showed an absence of seabird, marine
mammal, or sea turtle observations collected from either dedicated
surveys or satellite telemetry. Initially, the number of potential
datasets identified grew exponentially with time spent on searching
with the use of literature, networking, and websites due to the
snowball sampling effect [8]; the collaborative nature of many
Figure 6. Relative density of tracklines published on SEAMAP for global FAO regions. Trackline data collected by: (A) shipboard surveys;
(B) aerial surveys; and (C) satellite telemetry tag. Insets highlight the highest density per platform; FAO region borders are solid lines and US EEZ
borders are dashed lines. (Ship, plane, and satellite icon credit: Tracey Saxby, IAN Image Library, ian.umces.edu/imagelibrary/).
doi:10.1371/journal.pone.0012990.g006
OBIS-SEAMAP Gap Analysis
PLoS ONE | www.plosone.org8 September 2010 | Volume 5 | Issue 9 | e12990

Page 9

studies and information in the bibliography and acknowledgments
led to information on other organizations with similar datasets.
The first round of data requests, primarily via e-mail, to acquire
many of the identified datasets showed significant potential. As the
search and review of literature progresses, there usually comes a
point at which the rate of information gained with respect to effort
decreases [9]. For the scope of this gap analysis, the timeframe
allocated to searching for datasets had not yet reached this point of
diminishing returns. However, it was apparent that older datasets
(early 1980s and before) did have diminishing returns in that the
effort to find current contacts, appropriate permissions, and
acquire data in readily available formats increased with the age of
the dataset.
Quantitative results of the quintile subtraction analysis showed
that the largest gaps in SEAMAP data holdings compared to what
was missing in the literature were in the Atlantic SW for all taxa
combined as well as for marine mammals in particular (Table 6).
In addition, the amount of marine mammal data was low for the
Pacific NW and seabird data holdings showed the largest
difference between references and records within the Atlantic
Antarctic and Pacific Antarctic. In contrast, SEAMAP had
relatively good coverage (small difference between the relative
number of references and published records) within the Atlantic
NE, where there were high numbers of records for all taxa
combined, and the Indian Antarctic, where there was a medium
amount of records for all taxa combined (Figure 4; Table 6). The
amount of sea turtle data on SEAMAP was also greater in relation
to the number of literature references in the Atlantic NW, where
the number of records was relatively high. It is important to note
that these results only give an estimate of regional distributions of
data within the literature for datasets not already published on
SEAMAP and the literature search was not comprehensive.
The majority of datasets published on SEAMAP were
contributed by government groups (n=150), followed by non-
government groups (n=50) and universities (n=34). When
breaking the number of records down by taxa, government
Table 2. Regional summary of tag, shipboard survey, and aerial survey trackline data published on SEAMAP.
Region (FAO Code)
Shipboard
survey (km)
Aerial
survey (km)Tag (km) Total (km)
Trackline length (km)
per million km2
Trackline
coverage (% area)
Arctic (18)0.00 0.00 2,166.322,166.32 296.76 1.50
Atlantic CE (34)143.550.00187,415.37 187,558.93 13,397.0733.02
Atlantic CW (31)173,706.37 125,772.93447,505.59746,984.8950,815.3028.14
Atlantic NE (27)508,823.860.00 105,519.58 614,343.4436,351.68 30.55
Atlantic NW (21) 265,028.59516,315.97246,527.561,027,872.11 197,667.7135.17
Atlantic SE (47)0.00 0.0043,382.4343,382.432,332.39 6.80
Atlantic SW (41)0.00 0.00 6,972.686,972.68 396.172.03
Atlantic-Antarctic (48)0.000.00395.23 395.2332.13 0.14
Indian E (57) 1,762.120.00 60,051.2761,813.39 2,074.276.35
Indian W (51) 7,678.140.0081,864.49 89,542.632,964.99 8.45
Indian-Antarctic (58)4,679.41 0.0071,189.58 75,868.99 6,021.3515.72
Mediterranean (37)59,091.06 0.00 1,168.2960,259.3620,086.45 5.91
Pacific CE (77) 372,602.97 0.00215,460.17 588,063.1412,025.8336.28
Pacific CW (71) 0.00 0.00 72,908.1672,908.16 2,196.0311.63
Pacific NE (67) 36,447.53130,635.27 6,547.17173,629.9723,150.66 14.32
Pacific NW (61)521.65 0.0048,504.9149,026.57 2,391.5412.67
Pacific SE (87) 150,428.730.003,622.74154,051.47 5,135.0519.04
Pacific SW (81)0.000.001,532.751,532.7553.97 0.35
Pacific-Antarctic (88)0.000.000.00 0.000.000.00
doi:10.1371/journal.pone.0012990.t002
Table 3. Environmental variables used to assess marine mammal, seabird, and sea turtle observations published on SEAMAP.
Variable SourceAvailability Cell sizeAveragingAvailable data
Bathymetry (Depth)S2004 (compilation of GEBCO [52]
and Smith/Sandwell [53])
All Years 1 arc-minuteNonen=92%
Sea Surface
Temperature (SST)
AVHRR v5 Sea Surface
Temperature [54]
1985–20060.04 degree Yearly, Monthly, WeeklyYear: 57% Month:
57% Week: 57%
Sea Surface Height (SSH) Aviso Maps of Absolute Dynamic
Topography (MADT; [55])
October 14,
1992–May 23, 2007
0.25 degreeWeeklyMonth: 26% Week:
26%
Chlorophyll (CHL) SeaWiFS Sea Surface
Chlorophyll [56]
September 4,
1997–June 1, 2006
0.08 degreeYearly, Monthly, 8 days Year: 13% Month:
8% 8 day: 6%
doi:10.1371/journal.pone.0012990.t003
OBIS-SEAMAP Gap Analysis
PLoS ONE | www.plosone.org9 September 2010 | Volume 5 | Issue 9 | e12990

Page 10

groups provide the largest number for seabirds and marine
mammals while non-government groups provide the most data for
sea turtles. For the four platforms, government groups provide the
most data by shipboard and aerial surveys, non-government
groups give the most satellite tag data, and universities provide the
most shore-based data. The number of datasets were mostly from
US-based organizations (n=170). When broken down by
platform, however, more shipboard and satellite tag data (records
and number of animals sighted) were from non-US contributors.
US-based organizations provided more records and animal
observations for marine mammals and sea turtles but not for
seabirds.
Results from the preliminary literature search showed that there
were numerous potential contacts within the government, non-
government, and university groups that have collected data
suitable for SEAMAP, including data to fill many temporal and
spatial gaps (Text S1). When targeting specific groups and
searching within these sectors, internet searches and conferences
proceedings were more effective than primary literature databases
as websites and conference presentations were more current and
often disseminated information through gray literature. Limits for
online searches include the existence of a website, the presence of
keywords to be retrieved by search engines, and enough available
information on research projects and key contacts to be
considered. The most common limitation for obtaining the data
presented at recent conferences was that the data were still being
worked on for peer review publication or other projects and
researchers were reluctant to make data available to the public
beforehand.
Comparing portal coverage with example species
As of January 25, 2010, the US Fish and Wildlife Service listed
28 marine mammals, 19 seabirds, and 6 sea turtles as endangered
or threatened based on the level of protection needed to reduce the
danger of the population’s extinction [10]. Of the 53 animals
listed, the SEAMAP project currently has no records for 9 marine
mammals (Table 7) and no records for 12 seabirds (Table 8), but
does have at least one record of each of the sea turtle species listed
(Table 9). With the exception of the bowhead whale (Balaena
mysticetus), the listed marine mammals and seabirds not included in
SEAMAP are known to be endemic to relatively small geographic
ranges. Other listed animals with low representation (,10 records)
in SEAMAP, and a more wide habitat range, include the
endangered North Pacific right whale (Eubalaena japonica) [11]
Table 4. Average value of environmental variables sampled by taxa and by platform.
Depth (m) Monthly SST (uC) Monthly SSH (cm)Monthly CHL (mg/m3)
Taxa
Marine mammals764 14.1129.8 2.04
Seabirds 311 10.0122.61.30
Sea turtles2,55021.4 179.9 0.29
Platform
Shipboard 329 11.7 116.0 1.24
Aerial 3217.3 154.7 2.53
Tag2,448 15.3 179.70.73
Shore 17 6.9 115.8 5.99
doi:10.1371/journal.pone.0012990.t004
Table 5. The mode and range (in parenthesis) of environmental variables sampled by observations of marine mammals, seabirds,
and sea turtles, separated by platform.
TaxaPlatformDepth (m) Monthly SST (degrees C)Monthly SSH (cm) Monthly CHL (mg/m3)
Marine mammals
Shipboard 1 (1–8,059)25.4 (22.9–35.18) 162.2 (216.5–246.7) 1.8 (0.03–37.15)
Aerial12 (1–4,575) 13.4 (22.9–29.5) 149.4 (96.5–199.9) 2.3 (0.09–21.38)
Tag1 (1–5,139) 6.4 (22.9–20.8)127.0 (80.8–139.4)2.3 (0.36–15.14)
Shore1 (1–149) 5.8 (22.9–18.2)118.5 (104.2–122.5)4.5 (1.78–44.16)
Seabirds
Shipboard53 (1–9,294)
22.9 (22.9–31.2) 99.6 (119.3–261.8)1.5 (0.03–25.41)
Aerial 1 (1–4,473)
22.9 (22.9–28.7)157.0 (139.8–182.7) 2.6 (0.35–30.20)
Tag 15 (1–7,761) 23.3 (22.9–27.3) 224.6 (119.3–261.8) 0.3 (0.03–14.62)
Sea turtles
Shipboard 1 (1–5,731)30.7 (3.4–31.7)184.8 (91.0–212.2) 0.2 (0.06–5.37)
Aerial1 (1–4,596) 24.2 (22.6–29.5)127.0 (98.8–166.8) 0.4 (0.14–7.59)
Tag5,614 (2–6,518)17.8 (21.7–31.0) 193.6 (107.9–314.8)0.1 (0.02–8.41)
doi:10.1371/journal.pone.0012990.t005
OBIS-SEAMAP Gap Analysis
PLoS ONE | www.plosone.org 10September 2010 | Volume 5 | Issue 9 | e12990

Page 11

and endangered beluga (Delphinapterus leucas) [12]. Preliminary
searches using literature databases showed that data from recent
research (within the last 25 years) were published for the 15
endangered and threatened marine mammals with little to no
representation on SEAMAP.
For each of the eight key marine species chosen to compare
among open-access databases, SEAMAP contributed varied
amounts of data compared to what was available through GBIF
and OBIS (Table 10). Although SEAMAP is a data node which
contributes to OBIS, which in turn contributes to GBIF, this
process is only periodically maintained and therefore not all
SEAMAP data can currently be found on OBIS and GBIF. The
most recent time when OBIS obtained data from SEAMAP was
on November 16, 2009, one day prior to the data download for
the gap analysis. As a result of the current data flow hierarchy,
SEAMAP and OBIS were most similar in data holdings while
GBIF and SEAMAP had a greater difference in numbers of
records for each marine species examined. SEAMAP contributed
the most data for species such as the bottlenose dolphin, great
shearwater (Puffinus gravis), and loggerhead sea turtle, but was not a
significant contributor for more rare species such as the sperm
whale (Physeter macrocephalus/P. catodon).
Most SEAMAP data for the eight key species were centralized
in the mid-latitudes, having heavy representation around North
America (Figures 8a–g), with the exception of the wandering
albatross (Diomedea exulans) whose habitat range is in the southern
oceans (Figure 8h). OBIS data not included on SEAMAP filled
many geographic gaps for most species, especially contributing
information in the Indian-Antarctic for Arctic tern (Sterna
paradisaea), killer whale (Orcinus orca), sperm whale, and wandering
albatross (Figures 8a, c, g, h, respectively). Around Australia,
bottlenose dolphin, sperm whale, and wandering albatross data
were lacking on SEAMAP, but were available through OBIS
(Figures 8b, g, h, respectively). Data found exclusively on GBIF
contributed some data in geographic gaps, most notably for Arctic
tern records in the Pacific NE and SE, Atlantic NE and CE, and
Arctic (Figure 8a). In addition, GBIF data filled data gaps in the
Mediterranean for bottlenose dolphin, leatherback sea turtle
Dermochelys coriacea, loggerhead sea turtle, and sperm whale
(Figures 8b, d, e, g, respectively). However, SEAMAP generally
had high data coverage for great shearwater, mostly lacking
records found on GBIF near Europe and OBIS in the southwest
Atlantic (Figure 8f).
Discussion
When assessing current knowledge on species distributions with
the use of multiple sources, it is important to take into account the
effects of temporal and geographic sampling biases. Overall,
SEAMAP dataset, record, and observation holdings per year have
been increasing which likely is a result of increased effort for
SEAMAP project data acquisition for more current data and not
increased abundances of species in the field. Community awareness
of the project and emphasis on data management for better
Figure 7. Kernel density estimates of environmental variables. Depth (A–C), sea surface temperature (D–F), sea surface height (G–I), and
chlorophyll a (J–L) density estimates for marine mammals (A, D, G, J), seabirds (B, E, H, K), and sea turtles (C, F, I, L). Observations collected by aerial
surveys (solid red line), shipboard surveys (dark green long dashed line), tag (dark blue short dashed line), and shore (light blue dashed and dotted
line, only for marine mammals) were used to sample environmental variables. Note that depth and chlorophyll a graphs are on a log scale.
doi:10.1371/journal.pone.0012990.g007
OBIS-SEAMAP Gap Analysis
PLoS ONE | www.plosone.org 11September 2010 | Volume 5 | Issue 9 | e12990

Page 12

archiving and/or dissemination has grown within recent years. On
the other hand, geographic biases in research can be due to site
accessibility, proximity to collecting agencies or highly human
populated areas, or interest (either targeting areas with the high
probability of species presence, or areas and species of high
conservation concern) [13]. Although the SEAMAP project has not
had a random approach to obtaining observations on marine
mammals, seabirds, and sea turtles, this gap analysis is a first
assessment to determine not only SEAMAP project gaps compared
to the greater research community, but also to look into possible
biases in research as a result of dataset holdings.
The majority of data contributed to the SEAMAP project has
been seabird observations (with peak numbers of records in the
1980s) followed by high numbers of marine mammals (mainly
occurring in the 1990s) and sea turtles (having increased
observations after 2000). It is unknown as to why high numbers
of records occurred when they did for each taxa or why seabird
data made up the majority, especially since data acquisition for the
SEAMAP project was sometimes on an ad hoc basis. A general
increased amount of publications with time was also found for
marine mammal diving and tracking studies, which may be
attributed to increased computer access and improved technology
for processing data [14]. One explanation for the differences in
numbers of records among taxa may partly be influenced by
inherent biases in detection based on taxa behavior and habitat
(time spent in the air or on land versus under water). Seabirds
usually occur in flocks and colonies, with higher numbers of
individuals relative to pods or groups of marine mammals. Sea
turtle observations may be least represented because they are
generally solitary at sea and are also usually smaller with a lower
profile than marine mammals, which makes detection more
difficult. In addition, detectability for both marine mammals and
sea turtles is low since they spend most of their time underwater.
Therefore, other complementary methods may be more effective
for informing habitat usage such as acoustic surveys (for marine
mammals [15,16]) and satellite telemetry [17]. Along with
detectability differences, the number of observations collected
over time may have been influenced by the number of active
researchers within each taxa community, the amount of resources
available for surveys (i.e., funding), and the general abundance of
species within the areas surveyed.
Although data from shipboard and aerial surveys, satellite tags,
and shore-based observations can be used to determine species
ranges, abundances, and density estimates, there are distinct
platform differences that need to be taken account. Shipboard
surveys have the advantage of allowing for closer inspection of
animals and additional biotic (i.e., prey abundance, acoustic
monitoring, etc.) or abiotic (sea surface temperature, salinity, etc.)
sampling. However, aerial surveys are less expensive and usually
cover a larger extent in a shorter period of time compared to
shipboard surveys. Avoidance or attraction to aerial or shipboard
platforms may also influence density estimates, depending on
certain species [18,19]. Higher densities of common animals are
generally found in aerial surveys due to greater visibility, but better
detection of rare or small species occur with shipboard surveys
[20,21]. Greater detection may also be found with shore-based
surveys for small, coastal marine mammals [22]. Furthermore,
satellite telemetry tags are critical for capturing areas where animals
traverse, filling in many of the spatial and temporal gaps left from
surveys [23,24]. For the data published on the SEAMAP website,
most observations were from shipboard surveys and far fewer were
from aerial surveys, tag, and shore. Shore-based surveys gave the
most animals per record since locations of species were usually
congregations (i.e., seal haul out sites and seabird colonies).
Depending on the priorities for filling gaps within SEAMAP, the
representation of certain species may be maximized by targeting
aerial or shore surveys while more observations of less commonly
sighted animals may be obtained through shipboard surveys.
While reviewing the full list of data already obtained for
SEAMAP, it was recognized that there were significant temporal
biases within a year, which may be a result of the combination of
seasonal affects and regional source of the data. Datasets have
primarily come from North America and northern Europe with
large geographic gaps existing in the Southern Hemisphere and
Indian Ocean. Since most data were located in the Northern
Hemisphere, the summer (June–August) and fall (September–
November) seasons had the highest number of observations and
animals probably due to a combination of greater survey effort and
animal activity. Summer weather conditions are usually more
favorable for surveys. Summer is also the breeding season for most
seabirds, where colonies flock onshore and are easily detected,
with fall being a time for feeding and fledging [25]. Many marine
mammals migrate towards the poles to feed during the summer
[26] and therefore, an increased likelihood of detection occurs if
surveys are planned in known migration or foraging areas. Sea
turtle nesting season often occurs in the summer to fall when adult
female sea turtles aggregate and come close to and onto the shore.
The higher number of seabird observations, published on
SEAMAP, found in winter (December–February) may be
attributed to the greater number of observations in the Southern
Hemisphere during breeding and birthing seasons (austral
summer; n=589,116) compared to fall (n=188).
The high number of records and animals published on
SEAMAP in the Northern Hemisphere, along with the relatively
high density of effort in the eastern US coast, can be attributed to
Table 6. Regional results from the quintile subtraction
analysis for all taxa; higher numbers (in bold) represent larger
data gaps in SEAMAP relative to the literature search.
Region (FAO Code)
All
taxa
Marine
mammal Seabird
Sea
turtle
Arctic (18)0110
Atlantic CE (34)
2110
24
Atlantic CW (31)0000
Atlantic NE (27)
21
21
221
Atlantic NW (21)00
220
Atlantic SE (47)0100
Atlantic SW (41)2211
Atlantic-Antarctic (48)
21020
Indian E (57)
210
21
21
Indian W (51)01
211
Indian-Antarctic (58)
22
22
210
Mediterranean (37)11
221
Pacific CE (77)
21
210
21
Pacific CW (71)0001
Pacific NE (67)0001
Pacific NW (61)13
21
22
Pacific SE (87)
21
2200
Pacific SW (81)0
2211
Pacific-Antarctic (88)
21
222
21
doi:10.1371/journal.pone.0012990.t006
OBIS-SEAMAP Gap Analysis
PLoS ONE | www.plosone.org 12September 2010 | Volume 5 | Issue 9 | e12990

Page 13

collaborators and cooperation from US government groups, which
were the greatest contributor (i.e., National Oceanic and
Atmospheric Administration). This bias was expected due to high
US-based research efforts and because the SEAMAP project is led
by Duke University, US, with more established partners in the US
and Europe than other regions. In addition, the SEAMAP website
is currently only available in English, which may serve as a barrier
in some regions. Even if requests for data were written in the
native language (non-English) of the recipient, some providers may
not receive the full benefits of SEAMAP if the majority of targeted
users cannot utilize an English-only website. A cost-benefit analysis
of the website translation for SEAMAP is needed to determine if
language is the greatest limitation for obtaining data in certain
regions. Based on current SEAMAP data holdings, non-US groups
are an important target community for filling gaps for specific taxa
(i.e., seabirds) and regions.
The values and ranges of environmental variables associated
with observations published on SEAMAP varied by taxa and can
be influenced by platform. By combining available data across
platforms, environmental ranges are more complete in relation to
available habitat for species when compared to using one platform
alone [27]. For example, satellite tag locations can cover many
spatial/habitat gaps when combined with traditional survey data
to acquire a better picture of the presence of animals in a region.
Currently, habitat and density modeling techniques for mixing
data from multiple platforms while incorporating abundance,
residence time, and detectability have not yet been fully developed
and present a worthwhile opportunity for future research.
The distribution of each environmental variable was only
analyzed for large taxa groups (i.e., marine mammals, seabirds,
and sea turtles), but the same analysis can be used for individual
species within a taxa to determine if SEAMAP data holdings
adequately cover the known range for that particular species. For
example, ecological niche models developed for blue whales
(Balaenoptera musculus) showed that occurrence from whaling data
was highest around waters 1–4 km deep [28], but only 25% of
Table 7. The number of marine mammal records and animals published on SEAMAP that are protected under the US Endangered
Species Act, based on the US Fish and Wildlife listings (E=endangered, T=threatened, E/T=populations are endangered and
threatened; [10]).
Status Species Common NameRecords (n) Animals (n)
T Arctocephalus townsendiGuadalupe fur seal11
E Balaena mysticetusBowhead whale00
E Balaenoptera borealisSei whale159446
E B. musculusBlue whale8,539 18,116
*E B. m. brevicaudaPygmy blue whale11
E B. physalusFin whale6,31914,294
E Delphinapterus leucasBeluga7 37
E Dugong dugonDugong34
T Enhydra lutrisSea otter2,042 13,667
*T E. l. nereis Southern sea otter1417
E Eschrichtius robustus Gray whale1,6083,688
E Eubalaena australisSouthern right whale13 42
E E. glacialis Northern right whale757 1,633
E E. japonica North Pacific right whale12
E/TEumetopias jubatus Steller’s sea lion 257543
E Lipotes vexillifer Chinese river dolphin00
E Lontra felinaMarine otter00
E L. provocaxSouthern river otter 34 54
EMegaptera novaeangliaeHumpback whale 12,040 24,045
E Monachus monachusMediterranean monk seal00
E M. schauinslandiHawaiian monk seal00
E Orcinus orca Killer whale 614 2,826
E Phoca hispida saimensisSaimaa seal00
E Phocoena sinus Vaquita26 43
E Physeter macrocephalusSperm whale 2,0757,188
E Platanista minorIndus river dolphin00
E Trichechus inunguisAmazon manatee00
E T. manatus Caribbean manatee9 11
T T. senegalensisWest African manatee00
T Ursus maritimus Polar bear 2533
*SEAMAP holds records for these subspecies.
doi:10.1371/journal.pone.0012990.t007
OBIS-SEAMAP Gap Analysis
PLoS ONE | www.plosone.org13September 2010 | Volume 5 | Issue 9 | e12990

Page 14

marine mammal sightings on SEAMAP sightings were in waters
.2 km deep. Other factors known to influence species occurrence
that were not examined in this study (i.e., prey distributions,
preferred benthic habitats) can also be used to predict areas of
relatively high abundance. Although many records and sightings
for blue whales are available on SEAMAP (Table 7), representa-
tion would be improved by focusing future dataset acquisition
and/or research efforts towards filling the gap in knowledge within
the species’ ecological niche.
Tracklines representing survey effort from shipboard and aerial
surveys, as well as satellite telemetry tag paths showed that
SEAMAP coverage is specifically biased towards coastal areas,
with particularly good coverage of the US Exclusive Economic
Zone (EEZ). Coastal areas are more accessible for research and
can be the primary habitat of many marine species. With the
exception of shipboard surveys in the Caribbean, highest densities
of tracklines published on SEAMAP were within US waters,
probably as a result of more contributions from US partners and
collaborators, many of which were gathering data under US
government mandates to survey the US EEZ for marine animals.
SEAMAP currently has few partners conducting research in
regions with the largest gaps, such as the Arctic, southern Atlantic,
Atlantic-Antarctic, Pacific CW and SW, and Pacific-Antarctic.
Building relationships with organizations who gather data in these
regions should be a priority, especially since data exist in these
regions, based on the preliminary literature search. Additionally,
obtaining data gathered further offshore (e.g., satellite tag data)
should be a priority in filling spatial gaps.
The challenges for exhaustive literature searches and reviews
typically include time constraints, language barriers, availability
(access or subscription) to literature, and limitations of available
electronic databases, catalogues, and search engines to detect all
published and unpublished resources related to the study. The
results of the preliminary literature search showed great potential
for the research community to contribute data to fill many of the
spatial and temporal gaps known within the SEAMAP data
holdings, especially for the Atlantic SW and Pacific NW regions.
Information gathered from direct contact with experts in the field
definitely supplements literature searches [9,29]. It was apparent
that the dissemination of new information for marine mammal,
seabird, and sea turtle distribution relied heavily on the taxa
networks, underlining the importance of avenues outside of
published literature (i.e., symposiums and conferences, e-mail
listservs, project and database information posted on the web, etc.).
Other sources that were extremely useful were published regional
Table 8. The number of seabird records and animals published on SEAMAP that are protected under the US Endangered Species
Act, based on the US Fish and Wildlife listings (E=endangered, T=threatened, E/T=populations are endangered and threatened;
[10]).
Status SpeciesCommon Name Records (n) Animals (n)
T Brachyramphus marmoratusMarbled murrelet1,677 3,236
E Fregata andrewsi Andrew’s frigatebird00
E Larus audouinii Audouin’s gull117513
E L. relictusRelict gull00
E Phoebastria albatrusShort-tailed albatross1,447 2,393
E Pterodroma aterrima Mascarene black petrel00
E P. axillaris Chatham Island petrel00
E P. cahow Bermuda petrel00
EP. macgillivrayi Fiji petrel00
E P. magentae Magenta petrel00
T P. phaeopygia Galapagos petrel24
E P. p. sandwichensisHawaiian dark-rumped petrel00
T Puffinus auricularisTownsend’s shearwater1 13
T P. auricularis newelliNewell’s shearwater00
T P. heinrothi Heinroth’s shearwater00
E Spheniscus mendiculus Galapagos penguin00
E Sterna antillarumLeast tern 48 453
E S. a. browniCalifornia least tern00
E/TS. dougallii dougallii Roseate tern 44 165
doi:10.1371/journal.pone.0012990.t008
Table 9. The number of sea turtle records and animals
published on SEAMAP that are protected under the US
Endangered Species Act, based on the US Fish and Wildlife
listings (E=endangered, T=threatened, E/T=populations are
endangered and threatened; [10]).
StatusSpecies
Common
Name
Records
(n)
Animals
(n)
T Caretta carettaLoggerhead59,443 61,197
E/T Chelonia mydas Green6,911 6,911
EDermochelys coriacea Leatherback4,127 4,173
E Eretmochelys imbricataHawksbill 2,713 2,713
E Lepidochelys kempiiKemp’s ridley 332349
E/TLepidochelys olivacea Olive ridley2,344 2,425
doi:10.1371/journal.pone.0012990.t009
OBIS-SEAMAP Gap Analysis
PLoS ONE | www.plosone.org 14September 2010 | Volume 5 | Issue 9 | e12990

Page 15

reviews of research on marine mammals, seabirds, or sea turtles
(e.g., [30–32]). For most data requests, referencing published
sources streamlined the process involved with identifying potential
data for publication on the SEAMAP website. Alongside primary
literature, internet searches for contributors (i.e., government
agencies, non-governmental organizations, individuals or groups
at universities) have been beneficial in identifying large datasets
not yet included on SEAMAP. Since the rate of finding potential
high quality datasets within the literature and online websites is
still high at this point, continuing these data acquisition methods
would be prudent in filling in the gaps.
When combining all available marine mammal, seabird, and
sea turtle data from open-access online databases (SEAMAP,
OBIS, and GBIF), it was revealed that SEAMAP ranged in its
percent contribution and spatial coverage. OBIS is an aggregating
portal and batch data flow from SEAMAP into OBIS, but
currently there is no systematic method for integrating data from
OBIS back into SEAMAP. This dichotomy seems to exist between
OBIS and GBIF as well; although OBIS is a data contributor for
GBIF, not all data on GBIF are available on OBIS. In addition,
data published on SEAMAP are not simultaneously posted to
OBIS, nor are OBIS data automatically synced with GBIF. While
all three databases have the same objective for making data freely
available, the SEAMAP project’s unique features (e.g., visualiza-
tion of survey effort, environmental data availability, and metadata
details) make it particularly valuable for data providers to have
their data published within SEAMAP and not only at OBIS or
GBIF. However, consistent data accessibility through all three
portals would be more beneficial for all users. In addition, a
systematic mechanism using unique identifiers should be enforced
at each node to ensure that data contributed to all three databases
are not duplicated.
SEAMAP provides a convenient means for data dissemination
and outreach to the public, researchers, and resource managers.
Along with data users, data providers benefit from publishing
datasets on SEAMAP because the results of their research,
including published reports or articles, are highlighted to reach a
broader audience and their data can be added to customized maps
where it can be viewed in the context of oceanographic
observations, other datasets, and in the context of global
distribution maps. In addition, SEAMAP provides an organized,
secure off-site archive, which can expedite future data requests and
queries. Making these observations and analysis tools available to
the global research community will facilitate improved ecological
understanding of marine mammals, seabirds, and sea turtles as
well as inform the management and conservation of these species.
The digital archive increases interoperability while building a
knowledge base that will enhance research possibilities and
support evidence-based conservation [33,34]. In the future,
mechanisms for easy bidirectional data transport among SEA-
MAP, OBIS, and GBIF would improve data archiving and
dissemination.
Baseline knowledge on marine species distributions will be
fundamental in documenting any responses to consequences of
environmental and human induced threats to marine mammal,
seabird, and sea turtle populations. Many cases of marine species
populations close to or already extinct go undocumented due to
the lack of research and available information on the geographic
ranges (breeding, adult, and vagrant), among other reasons [35].
Furthermore, data management becomes increasingly difficult as
information accumulates and a cataloging system for the
information needs to be established especially as scientific
researchers and personnel turnover. Based on responses from
major conservation agencies, the greatest limitation was lack of
time and resources to collate data to publish on SEAMAP.
Therefore, conservation projects should have objectives that
include making data publicly available to reduce information
loss. It is important to have the involvement of scientists in the
process of publishing their data within a comprehensive and
accessible inventory of marine species and their distributions in
order to fully evaluate future threats to marine mammal, seabird,
and sea turtle populations. This is particularly important for the
42 species already identified to be endangered or threatened
[10]. The aggregation of effort from multiple research projects at
a data center, such as SEAMAP, can broaden the data coverage
in space and time for a more global perspective. Although
thousands of records were available from SEAMAP for most of
the endangered or threatened species identified, in addition to
millions of records of more common species, there were many
species for which SEAMAP lacks data. These and other rarely
recorded species should be a priority for future data acquisition
efforts.
Table 10. The number of records downloaded per database with latitude/longitude data and percent of the total comprised of
SEAMAP records for eight key species; SEAMAP records on OBIS were identified by the information in the ‘‘Datapub’’ and
‘‘DataRight’’ columns, SEAMAP records on GBIF were identified by information in the ‘‘Res_name’’ and ‘‘Institutio’’ columns.
SpeciesCommon NameSEAMAPOBIS SEAMAP on OBISGBIF SEAMAP on GBIF
Marine mammal
Orcinus orca Killer whale6431,53152% 1,297 33%
Physeter macrocephalus;
P. catodon
Sperm whale 2,46020,097 12%19,444 8%
Tursiops truncatus Bottlenose dolphin 17,90318,73798% 14,81383%
Seabird
Diomedea exulans Wandering albatross2,36018,102 12%5,789 15%
Puffinus gravisGreat shearwater 16,613 16,46099% 24,55765%
Sterna paradisaeaArctic tern2,683 5,21555% 68,4341%
Sea turtle
Caretta carettaLoggerhead73,19165,49796% 20,54191%
Dermochelys coriacea Leatherback4,141 8,86747% 1,385 34%
doi:10.1371/journal.pone.0012990.t010
OBIS-SEAMAP Gap Analysis
PLoS ONE | www.plosone.org 15September 2010 | Volume 5 | Issue 9 | e12990