Horning and Mellish: Predation on Eumetopias /ubatus by Somniosus pacificus in the Gulf of Alaska 
301 
of transmissions, as shown in Figure 2, D and E, indi- 
cates that the tag remained in the carcass as it cooled. 
This scenario suggests a nontraumatic death resulting 
from disease, starvation, or drowning due to entangle- 
ment. Events caused by disease or starvation addition- 
ally should be associated with antemortem tempera- 
tures outside of the normal range of 36.0-38.2°C. 
Type III: Gradual but quicker than predicted cool- 
ing with delayed sensing of light and air may indicate 
traumatic predator attacks and partial dismemberment 
even if tags remain embedded in tissue. 
Type IV: Very rapid drops in postmortem tempera- 
tures, as shown in Figure 2B, without immediate sens- 
ing of light or air indicate dismemberment and tag 
ingestion by a cold-blooded predator. Ingestion by a 
warm-blooded predator would result in no discernible 
changes in temperature and light until expulsion from 
the predator’s digestive tract. A type-IV scenario also 
could be caused by scavenging if it occurred and if the 
tags were ingested within 60 min post mortem (other- 
wise, the onset of gradual cooling would be detectable). 
In this case, scavenging would have been preceded im- 
mediately by death due to other causes. Any deaths 
caused by disease or starvation additionally should be 
associated with antemortem temperatures outside of 
the normal range of 36.0-38.2°C. 
Tag deployments and controls 
We implanted LHX tags into 36 juvenile Steller sea 
lions captured in Prince William Sound, Alaska, from 
2005 to 2011 (Horning and Mellish, 2012). Dual LHX 
tags were used in all but the first 2 animals to in- 
crease and determine probability of detection of a mor- 
tality event (the first 2 animals received only single 
implants). LHX tags were implanted intraperitoneally 
while the sea lions were under gas anesthesia and ac- 
cording to standard aseptic surgical techniques (Horn- 
ing et ah, 2008). All animals were released in Resurrec- 
tion Bay in the Kenai Fjords region, at ages of 14-26 
months and at a mean mass of 128.6 kg (standard 
deviation [SD] 28.6; range: 73-198 kg). In addition to 
live animal deployments, we conducted 10 simulation 
tests by implanting dual LHX tags in sea lion car- 
casses obtained from regional stranding networks and 
subsequently deposited ashore (n= 5) or at sea (n= 5) in 
California and Oregon. 
Previously published control studies with animals 
monitored during temporary captivity (21-71 d) and 
tracked for a mean postrelease period of 86 d (SD 55; 
range: 10-242 d; n= 35) for implanted animals and of 
76 d (SD 54; range: 11-194; n=30) for nonimplanted 
animals — with all animals having conventional satel- 
lite transmitters externally attached after their release 
(Wildlife Computers SDRT-16 and SPLASH-5 tags) — 
confirmed that 1) implant surgery results in mild to 
moderate wound healing responses and temporary el- 
evation of white cell counts and haptoglobin concen- 
trations, with full physiological recovery within 45 d 
following surgery; 2) tags and surgery result in zero 
mortality to 45 d; 3) postrelease foraging and ranging 
behavior does not differ between implanted and non- 
implanted animals or between temporarily captive and 
free-ranging animals; and 4) the cumulative survival of 
animals over the ages of 14-60 months was 0.415 (95% 
confidence intervals [Cl] =0.26—0.63 ), compared with 
0.413 (95% CI=0.27-0.55) for mark-resight studies 
based on hot iron branding conducted by the National 
Marine Fisheries Service, providing no evidence of any 
effects of LHX tags or implant surgery on survival to 
the age of 5 years (Mellish et ah, 2006, 2007; Thornton 
et ah, 2008; Petrauskas et ah, 2008; Walker et ah, 2009; 
Horning and Mellish, 2012). 
Received data and environmental data 
We estimated probability of event detection from the 
ratio of dual to single tag returns. We first estimated 
uplink failures (the combination of technical tag fail- 
ures and transmissions from a functional tag that did 
not reach any satellite because of tag exposure con- 
straints) as 
Pfail = ^single I ^single + 2Cjj ua i), 
where C s i ng i e = the count of single returns; and 
Cdual = the count of dual returns. 
Then, a correction factor F was derived as F = 1/(1- 
Pfail 2 ), and the corrected number of mortality events 
E corr was calculated as E corr = F (C s i ng i e + Cd ua i). From 
Pfaii, the probability of event detection Pdetect i n turn 
was derived as Pdetect = 1 -Pfaii 2 = 1/ F. Ranges contain- 
ing 95% of the likely variance for the estimate of Pf a n 
were derived from the cumulative distribution function 
of a Monte Carlo simulation (>2500 iterations) of ran- 
domly assigned individual tag failures for 0 < Pf a ii-simu- 
lated < 1 yielding Pf a p not exceeding the observed value 
without increasing E corr integer counts. The range of 
95% Cl for Pfaji in turn yielded confidence intervals for 
F and Ecorr* 
Sea-surface temperature (SST) composite data de- 
rived from the NOAA Geostationary Operational En- 
vironmental Satellites (GOES) and Polar-orbiting 
Environmental Satellites (POES) were obtained for 
specific dates from the Comprehensive Large Array- 
data Stewardship System (CLASS) (SST50 or SST100 
products available from http://www.nsof.class.noaa.gov/ 
saa/products/welcome, accessed June 2013). In situ SST 
data for specific dates were obtained from NOAA’s Na- 
tional Data Buoy Center (available from http://www. 
ndbc.noaa.gov/, accessed June 2013). Long-term aver- 
ages of water column temperature profiles from 1970 to 
2012 for the GAK1 location, at 59°50.7'N, 149°28.0'W, 
near the Kenai Fjords region of the northern Gulf of 
Alaska were obtained from a time-series data set of 
the University of Alaska’s Institute of Marine Science 
(available from http://www.ims.uaf.edu/gakl/, accessed 
June 2013). Temperature estimates for deeper waters 
