Grigg et al.: Spatial and seasonal relationships between Phoca vitulina richardii and their prey 
361 
are customarily removed from the data 
set, and filtering methods are used to 
remove improbable locations and improve 
the mean accuracy of the remaining loca- 
tions. Vincent et al. (2002) and Hays et 
al. (2001) found that locations in classes 
B and 0 were inaccurate, but class A lo- 
cations were more accurate and compa- 
rable to class 1 locations (Table 1). We 
therefore removed location estimates in 
classes B and 0 from the record of each 
harbor seal’s movements, but did not au- 
tomatically remove locations in class A. 
All points that fell on land were removed, 
and then the remaining locations were 
filtered according to the speed neces- 
sary for a harbor seal to move between 
two successive locations, calculated in 
a geographic information system (GIS). 
Any location that would have required a 
travel speed greater than 10 km/hr, or 
2.78 m/s (Lowry et al., 2001), was flagged 
for inspection. These questionable har- 
bor seal locations were assessed by their 
spatial and temporal relationship to the 
prior or subsequent location with a great- 
er accuracy rating, and unlikely locations 
were removed. Points that would have re- 
quired an isolated movement away from 
and immediately returning to the same 
area, necessitating a narrow V-shaped 
movement track, were also eliminated 
(see Keating, 1994). Locations within 
1 km of a haul-out site were removed 
in order to eliminate locations associ- 
ated with haul-out site use or under- 
water movements unrelated to foraging 
(Thompson and Miller, 1990). Finally, 
for these analyses, we analyzed spatial overlap 
between harbor seal locations and abundance of 
potential prey species around SFB, rather than 
analyzing sequential tracks of movement by in- 
dividual harbor seals. To improve independence 
of point location estimates for the correlation 
analyses, locations recorded within one hour of 
another location for the same harbor seal were 
removed from the data set. 
Correlation analyses 
Records of prey distribution and abundance during the 
study period were obtained from the Interagency Eco- 
logical Program for the San Francisco Estuary and from 
the San Francisco Bay Study, California Dep. Fish and 
Game. Monthly samples of fish, crab, and shrimp species 
were collected by CDFG at 39 sampling stations located 
around SFB (Fig. 1), using two sampling methods: an 
otter trawl (OT) and a midwater trawl (MWT). The OT 
was used to sample bottom-dwelling fish, shrimp, and 
crab, the MWT was used to sample mid-water fish, and 
122 <, 30'0"W 
122°20'0"W 
122°10'0"W 
122WW 
The San Francisco Bay, CA, study area, showing Castro Rocks and 
other primary harbor seal ( Phoca vitulina richardii) haul-out sites, and 
39 fish sampling stations around the bay sampled by the California 
Department of Fish and Game(CDFG). Catch-per-unit-of-effort data 
from the CDFG sampling stations were used to build seasonal harbor 
seal prey distribution maps for San Francisco Bay. Coastline data 
layer created by NOAA/NOS/ORR/CPRD (available online). 
both trawls yielded quantitative data on fish abundance. 
For this study, we focused on eight species of fish known 
to be prey of harbor seals in SFB, combined crab species 
(primarily Dungeness crab, Cancer magister), and com- 
bined shrimp species (primarily Crangon spp.). Given 
concerns about harbor seals foraging on salmonids, 
CDFG data on the abundance and distribution of Chi- 
nook salmon (Oncorhynchus tshawytscha) in SFB were 
included in the analysis. Catch per unit of effort (CPUE) 
was calculated as follows for each station, month, and 
species, by using gear-specific formulas from CDFG: 
OT CPUE = 
(number caught/tow area) x 10,000, 
(1) 
