FISHERY BULLETIN: VOL. 83. NO. 3 



kayak on a parallel course at distances of 50-150 m. 

 If only one observer was present in the kayak, these 

 periods were only included in the main data base 

 if very few animals were known to be in the study 

 area. 



STATISTICAL ANALYSIS 



For statistical analyses of these data we con- 

 sidered the potential application of log-linear and 

 multivariate models, principle component analysis 

 (PCA), factor analysis (FA), and a categorical data 

 (x^) procedure. Considerable differences in the 

 magnitude of x-variate variances would make results 

 from PCA or FA suspect (Maxwell 1977, ch. 4). Log- 

 linear and multivariate approaches were initially 

 attractive, but both have disadvantages. Tests for in- 

 dependence of x-variates are difficult in the latter, 

 and while these are facilitated by the former, results 

 obtained from log-linear models are often difficult to 

 interpret (SAS Institute Inc. 1979, p. 222). Further- 

 more, statistical advisers noted that use of the whole 

 data base was contraindicated in either method 

 because 1) there was of necessity inclusion of linear, 

 nonlinear, and enumeration data types, and 2) there 

 were significant numbers of empty data cells usually 

 resulting from poor weather when operation would 

 have been pointless. The consultants recommended 

 use of the categorical data procedure, not only for 

 the reasons outlined above, but also because the very 

 nature of the x-variates (e.g., lunar cycle and tidal 

 cycle) precluded the existence of complete indepen- 

 dence. From the ecological point of view it was con- 

 sidered more important to relate one dependent 

 variable (relative abundance of porpoises) to a group 

 of variables one at a time than to test for indepen- 

 dence in the latter when the result would likely be 

 spurious. 



We first determined (P = 0.13 -^ ) that no data set 

 from any x-variate was significantly associated with 

 one part of the "range" of any other by a series of 

 simple paired x^ tests. The main analyses were then 

 carried out on an IBM^ 360 mainframe computer 

 using the PROC FREQ program (SAS Institute Inc. 

 1979) which used a generalized least squares model 

 to generate x^ values for combinations of the 

 categorical levels between variables. In this case the 

 relative abundance was related to date, time of day, 

 tidal amplitude, tidal phase, lunar phase, extant wind 

 direction, and wind direction 24 h previously. 

 To examine changes in spatial distribution within 



^Reference to trade names does not imply endorsement by the Na- 

 tional Marine Fisheries Service, NOAA. 



the Study area, the location of 669 porpoise sightings 

 collected in 1973-75 were plotted with respect to 

 chronological time and time of tide. To avoid repeat 

 sightings only the initial sighting was used in this 

 context. To examine possible changes in distribution 

 by depth and location, the study area was divided in- 

 to three zones based on subsurface features: shallow 

 shelving areas, subsurface slopes and scarps, and 

 relatively deep basins and channels (Fig. 1). Such 

 data cannot be used to deduce actual patterns of 

 movement, but we were able to obtain some informa- 

 tion on these through selective observations of iden- 

 tifiable individuals. Five of these were known to be 

 females (consistently accompanied by calves), one a 

 male (seen briefly copulating with a known female), 

 and one of unknown sex. Each recorded sequential 

 position was plotted for movements within a specific 

 observation period, and for four of the known 

 females data on all observations were plotted and the 

 apparent ranges in consecutive seasons examined. In 

 each case the master sheets had a grid with ordinates 

 100 m apart. 



RESULTS 



Hours of Observation of Porpoises in 

 Fish Harbour Region 



About 550 h of observation were made between 

 June 1970 and September 1978, and 324 h of this 

 were during an "intensive" study phase encompass- 

 ing the summers of 1973-75. Estimates of changes in 

 relative abundance with respect to environmental 

 parameters were based on 181 h in optimal sighting 

 conditions in this period (see section on Methods). 

 Sporadic observations were maintained by our 

 research group subsequent to 1975 or by auxilliary 

 observers from Marine Research Associates of Lords 

 Cove in most months except for late-December to 

 mid-January. 



Changes in Relative Abundance 

 During the Year 



From low relative abundance in the spring and late 

 fall months, numbers of porpoises were highest dur- 

 ing August or September in each year of the study 

 period (Table 1). Highest values in ideal conditions 

 were 8.00/h in Fish Harbour and 7.72/h in Simpson's 

 Passage in September 1973 and August 1974 respec- 

 tively. Observations by Marine Research Associates 

 over the period 1971-77 in the same area, although 

 largely of a casual nature, confirmed the animals can 

 occur in small numbers in any month between Octo- 



430 



