The validity of the completed equation was then measured by an F test of 

 the significance of each regression coefficient. The use of the regression 

 equation as a predictive tool was determined from the multiple R value or 

 the coefficient of determination value, R . When the multiple R or R 

 value was large enough to indicate a substantial portion of the density 

 variation from quadrat to quadrat was accounted for by the regression 

 relationship, the equation was then used to interpolate from our sample 

 to the entire study area. This step consisted of sequentially substitut- 

 ing the value found at the quadrat midpoint for each statistically 

 significant variable into the regression equation in order to calculate 

 an expected density in this specific location. This interpolation procedure 

 was carried out for each of the 1,000 quadrats following each ship or 

 aerial transect survey and resulted in a picture of how and in what 

 numbers animals were distributed within the study area at the time of the 

 survey. Interpolations, however, did not extend into waters that were 

 insufficiently sampled (Fig. III-41). 



Comments on population enumerators 



It is generally agreed that aerial surveys of terrestrial wildlife 

 yield underestimates of total populations. The reliability of aerial 

 surveys, when utilized in the marine environment, becomes even more 

 difficult to assess. 



Cetacea generate their own set of handicaps to the investigator 

 engaged in population surveys. First, to be counted, the animal must be 

 at or near the surface; the smaller species of cetacea surface every 3-4 

 minutes or so, while the large ones may not surface for 10-12 minutes or 

 considerably longer. Secondly, when cetacea are disturbed or "flushed" 

 they, unlike birds, move downward in the water and become unavailable for 

 enumeration. Finally, it is well known that schooling cetacea are 

 stacked or layered in the water column so that only some portion of the 

 school are at the surface at any one time. Each of these phenomena 

 indeed add to the probability that population numbers are underestimated, 

 and/or that entire schools pass uncounted. 



Although aware of these conditions which lead to probable underesti- 

 mation, no attempt was made to establish a "fudge" factor for those 

 animals flushing away from the line of sight or those below the surface 

 at the time of the count. The rationale for this is twofold: we are 

 more comfortable with "hard" numbers (those representing actually 

 observed animals), and any arrived at "fudge" factor would only fit one 

 of many sets of conditions, leading to a series of such factors, each 

 more suspect than the last. Secondly, since any future surveys or mon- 

 itoring attempts will be faced with the same vexing conditions of 

 flushing, layering, and surface time, it seems reasonable for the sake of 

 comparability to utilize only observed numbers and relative indices of 

 abundance, rather than lean too heavily upon absolute population estim- 

 ates arrived at in a questionable or non-reproducible manner. 



A usual assumption made during a census of this type is that the 

 transect lines or search areas are randomly selected. Such was not 

 the case in this survey; our transect lines were fixed and we considered 



