1Hokkaido Fish Hatchery, Hokkaido 061-1433, Japan, 2Fisheries and Oceans Canada, Pacific Biological Station, British Columbia V9R 5K6, Canada
Multiple marking and recapture were conducted to estimate numbers of masu salmon Oncorhynchus masou smolts in a river in Hokkaido, northern Japan. A partial fence and a rotary-screw trap, located approximately 5km apart, were used to capture downstream migrating smolts from May to early July 1998 and 1999. Marking and recovery data were initially stratified at 7-day intervals although data for some weeks were subsequently pooled to increase sample sizes. Diagnostic chi-square tests indicated that pooled Petersen estimates, which were formed by pooling all of the marks applied and all of the recoveries, were not consistent, and the stratified-Petersen, i.e., maximum likelihood Darroch, estimates were appropriate. The stratified-Petersen estimate was 30% higher than the pooled Petersen in 1998, and was 16% higher in 1999. The migratory behavior of masu salmon smolts changed during each run and the capture efficiency of the trap varied as water flow fluctuated. Under these conditions, heterogeneity in the capture probability of smolts apparently occurred, which violated a critical assumption required for the pooled Petersen estimator.