REGRESSION OF ABUNDANCE ESTIMATES FROM MARK-RECAPTURE SURVEYS AGAINSTENVIRONMENTAL COVARIATES

Two alternative approaches are presented for regressing abundance esti mates from mark-recapture surveys against environmental covariates. Th e first is based on weighted least squares under the assumption of nor mality of square-root transformed estimates of abundance. A second app roach based on maximum likelihood estimation and analysis of deviance is presented that inherently weights the precision of the survey resul ts and is relatively distribution free. The analytic approaches are il lustrated using a study whose objective is to relate 8 years of abunda nce estimates of chinook salmon smelt production versus spring river f low conditions. Analysis of deviance and maximum likelihood estimation procedures are recommended for the analysis of replicated mark-recapt ure studies because they intrinsically provide proper weighting of the capture data and an opportunity to construct realistic response model s independent of distributional assumptions of the data.