Byrd et at: Strandings as indicators of marine mammal biodiversity and human interactions 
5 
evaluated by using a maximum likelihood generalized 
linear model (GLM) (SAS PROC GENMOD) with a log 
link function and a Poisson error distribution, with 
month as the categorical predictor variable and num- 
ber of strandings as the response variable (McFee et 
al., 2006). For each model, the month with the low- 
est mean count was used as the reference group. The 
model was evaluated for over-dispersion and goodness 
of fit by examining the magnitude of the deviance and 
the Pearson chi-square statistic divided by its degrees 
of freedom and by computing the probability of obtain- 
ing the observed chi-square statistic for each test. If 
over-dispersion or lack of fit occurred with either cri- 
terion, the model was rerun with a negative binomial 
error distribution and the fit was similarly evaluated. 
The lowest Akaike information criterion (AIC) values 
were used to determine the best fit. Harbor porpoise 
and harbor seal strandings were highly seasonal, caus- 
ing nonconvergence in the GLM analysis. As a result, 
their seasonal patterns were described qualitatively. 
Sample sizes were too small to test for month effects 
for balaenopterids, dwarf sperm whales, and non -Phoca 
pinnipeds. 
For species represented by a preponderance of small 
individuals, each stranded animal was assigned a ma- 
turity state on the basis of length relative to published 
length estimates: humpback whales ( Megaptera nouae- 
angliae ) (Rice, 1963), common minke whales ( Balaenop - 
tera acutorostrata\ hereafter ‘minke whales’) (Boyd et 
al., 1999), harbor porpoises (Read and Gaskin, 1990; 
Lockyer, 1995), harbor seals (Boulva and McLaren, 
1979), hooded seals (Cystophora cristata ) (Reeves et al., 
1992), gray seals ( Halichoerus grypus) (Bonner, 1981), 
and harp seals (Pagophilus groenlandicus ) (Reeves et 
al., 1992; Hammill et ah, 1995). 
Spatial patterns 
Spatial patterns were evaluated after stratifying the 
data into strandings recovered ocean-side or inshore. 
Relatively few strandings were recovered inshore; 
therefore, only qualitative results were presented. For 
the ocean-side analysis, the coast was stratified into 4 
segments (A-D) with boundaries at the VA-NC line, 
each cape (Hatteras, Lookout, and Fear), and at the 
NC-SC line (Fig. 1). The coastline length of each seg- 
ment was calculated in ArcGIS 10 software (Esri, Red- 
lands, CA). Within each taxonomic group, a chi-square 
analysis was used to compare the number of observed 
strandings with an expected value proportional to the 
segment length. When a significant difference (P=0.05) 
was found, standardized residuals were examined to 
determine which coastal segments had significantly 
more or fewer strandings (i.e., standardized residuals 
>| 1.96 | ). Delphinid species included in the taxonomic 
group of non -Kogia pelagic odontocetes were further 
divided into groupings that could be plotted in ArcGIS 
to qualitatively examine patterns not testable due to 
small sample sizes: 1) pilot whales, 2) pelagic delphi- 
nids generally associated with cool northern waters, 
3) pelagic delphinids generally associated with warm 
southern waters, and 4) pelagic delphinids with a more 
cosmopolitan distribution in areas north and south 
of NC. Species whose distribution in the western North 
Atlantic are not well defined were assigned to one 
of the aforementioned categories on the basis of 
what is generally known of their distribution (e.g., 
melon-headed whales [Peponocephala electra\ are gen- 
erally a tropical-subtropical species; Perryman et ah, 
1994). 
Because coastal bottlenose dolphin strandings were 
numerous enough to be investigated in more detail, 
the coastline was divided into 7 segments by dividing 
segments A-C in half (Al, A2, Bl, B2, Cl, C2) (Fig. 
1). Segment D was not divided because it was already 
less than half the length of segments A-C. As with the 
4-segment analyses, a chi-square analysis was used to 
determine whether the number of observed strandings 
per segment was significantly different from an expect- 
ed value proportional to the segment length (P=0.05); 
standardized residuals were examined when a signifi- 
cant difference was found in order to identify which 
segments had significantly more or fewer strandings 
(i.e., standardized residuals > | 1.96 | ). 
To visualize the distribution of all strandings on a 
finer scale, the NC coast was divided into 10-km sec- 
tions by using ArcGIS; the last (most southern) section 
was 7.6 km. Strandings were assigned to one of these 
54 sections with ArcGIS. The mean annual strandings 
per section were graphed for coastal bottlenose dol- 
phins and all other species combined. 
Human interactions 
Stranding data were stratified into the HI categories 
mentioned previously. The evidence for HI was re- 
viewed for strandings categorized as HI-FI and HI- 
other. Unless FI lesions were noted as healed, they 
were assumed to be fresh. For HI-FI animals recovered 
with attached gear, totals were produced by gear type. 
Results 
During 1997 through 2008, 1847 individual marine 
mammal strandings were reported for 1777 events 
comprising 9 families and 34 species (Table 1). Species 
could not be determined for 67 strandings. The major- 
ity of individual strandings were coastal bottlenose 
dolphins (56%), harbor porpoises (13%), and harbor 
seals (4%) (Fig. 2). Nineteen of 185 bottlenose dolphins 
tested were confirmed genetically as being the offshore 
morphotype. Kogia spp. represented 5% of strandings; 
of those, pygmy sperm whales were found to be more 
common than dwarf sperm whales. There were 19 
group stranding events, primarily those of Kogia spp. 
and coastal bottlenose dolphins, and mostly mother- 
calf pairs (Table 2). For 3 of the 9 mother-calf pairs of 
