Improving the precision of recreational fishing harvest estimates using two-part conditional general linear models

As recreational fishing continues to expand, the need to obtain precise harvest estimates is becoming increasingly important for the sustainable management of fisheries. Recreational fishing data are frequently zero-inflated which can present problems for commonly used analyses that assume a normal distribution. In this study, we analysed zero-inflated recreational fishing data collected from a bus-route access point survey in southeastern Queensland, Australia. Using the Time Interval Count method, we compared estimates of the proportion of boats fishing, fishing effort, harvest per unit effort (HPUE) and harvest using sample mean values and mean values derived from a two-part conditional general linear model (CGLM). The CGLM gave more precise estimates of the proportion of boats fishing, fishing effort and HPUE, which formed the basis of the harvest calculations. Differences in harvest estimates using the two methods ranged from 3 to 28% for the five recreational species examined. Relative standard errors for harvest estimated by the CGLM were 65–84% smaller. The results suggest that CGLMs may deliver more precise outputs in other types of recreational fishing surveys that derive effort and catch from zero-inflated data.

► We analysed recreational catch data from a bus-route boat ramp survey. ► We compared sample means and two-part conditional general linear model (CGLM) values. ► Relative standard errors for harvest from the CGLM were 65–84% smaller. ► Conditional models may improve precision in other recreational fishing surveys. ► Gains in precision improve the utility and acceptance of survey data.