Similar results were found by Soots and Parnell (1976) at laughing gull colo- 

 nies in North Carolina. The quarter method was ineffective at the heronries 

 we tested. A possible bias in our results may have arisen from the manner in 

 which the sampling lines were selected. In most cases, diagonals were used 

 within the sample plot instead of randomly-selected lines. The need to ran- 

 domly select line orientations and to measure the colony area limit the util- 

 ity of this method under most conditions. 



The belt transect sampling method provided more accurate estimates in all 

 but two colonies (Long Point and Penikese I.) than did the point-centered 

 quarter method. Although only two colony samples were taken, the method 

 seemed well-suited for use in heronries. Portnoy (1977) used the method 

 effectively in large mangrove heronries along the Gulf of Mexico, An advan- 

 tage of this method is that only the total length of the transect needs to be 

 measured, not the entire colony area. Counting all nests in 2-m wide belts 

 that completely cross the colony at random points along the transect bearing 

 line enhances the probability that an unbiased sample of the total colony is 

 taken. Some problems inherent in belt transect methods are discussed by 

 Mclntyre (1953), Eberhard (1968), and Anderson and Pospahala (1970). One bias 

 that we became aware of during the sampling was a tendency to "steer" the belt 

 transect toward nests seen in front of the observer even though compass orien- 

 tation was supposedly being followed. This would result in an overestimation 

 of the population, the opposite effect to that found by Anderson and Pospahala 

 (1970) where hidden waterfowl nests are often missed on the edges of the 

 transect (however, their transects were much wider than our 2-m belts). 



In summary, sampling techniques such as quadrat, belt transect, or point- 

 centered quarter can be useful for estimating nest numbers if the level of 

 accuracy needed is within 10 to 20% of the true population size. Sample size 

 should be increased however, where highly patchy nest distributions are en- 

 countered or where high vegetation density precludes obtaining a good sample 

 of the colony. 



LARGE-SCALE, "AMONG COLONY" METHODS OF ASSESSING POPULATION SIZE 



Due to the large geographic area to be covered and the short nesting 

 season, separate teams of field observers were selected to inventory different 

 sections of the coast for this study (Table 21). A coastal survey was taken 

 in late May 1976 using a U.S. Fish & Wildlife Service float-equipped De 

 Havilland Beaver. Emphasis was placed on locating colonies along the coast. 

 Parts of Long Island and Connecticut were ground censused by volunteers. The 

 Connecticut census was incomplete. Field censuses were performed primarily 

 using ground (boat, foot) methods, except in New York, parts of Massachusetts 

 (helicopter), and Maryland-Virginia (fixed-wing in many areas). Based on the 



1976 experience, all areas were covered during early June using helicopters in 



1977 (supplemented with fixed-wing aircraft in Virginia). Subsequent ground 

 checks were made at selected colonies throughout June and in some cases in 

 July. As a result, the 1977 inventory information is more complete than in 

 1976, especially in Connecticut and New Jersey. As in 1976, a fixed-wing 

 aircraft survey was flown in June 1977 to obtain aerial photographs. 



38 



