Comparison of recreational harvest estimates provided by onsite and offsite surveys: detecting bias and corroborating estimates

We compare estimates of recreational harvest provided concurrently by two fundamentally different survey designs: an offsite panel survey and an onsite aerial-access survey. Aerial-access estimates for the five most commonly caught species were 2% to 50% lower than the panel survey estimates, with the greatest differences apparent for the least commonly caught species. Boosted regression tree modelling of spatially and temporally disaggregated harvest estimates identified a consistent pattern of temporal bias that explained much of the difference between the two sets of estimates. An analysis of web camera-based traffic data collected concurrently at key boat ramps confirmed that the selection of days for the aerial-access survey was biased towards lower effort days in three out of four temporal strata. Some evidence of under-reporting of zero catch trips by panelists was also identified by boosted regression tree modelling of catch per trip data. Nonetheless, the estimates provided by the two surveys were still remarkably similar given the range of potential biases that the two contrasting survey approaches were potentially subject to. Comparative studies such as this are rare, but they provide greater insights than introspective evaluations of single surveys and therefore greater certainty in the future.