科研之友
微信
新浪微博
Facebook
分享链接
摘要
Due to the logistics of monitoring remote and diffuse fisheries, typically only a small portion of the annual fishing effort can be observed, and a random sample is often impractical. We used a large data set (over 250,000 observations) from time-lapse cameras placed at 53 lakes across the province of British Columbia, Canada, to better understand the relative influence of different temporal strata (time of day, day type, and month) on the relative number of fishing boats observed on lakes. The high temporal frequency data available from cameras were used to evaluate alternative effort monitoring and estimation strategies for aerial surveys that are commonly used in recreational fisheries. We tested the predictive performance of simple mean count expansion factors versus generalized linear mixed models (GLMMs) under random and stratified (targeting high-effort strata) sampling regimes. Out-of-sample cross validation showed that the current sampling protocol and estimation method (mean count expansion) used for aerial surveys in British Columbia have low predictive power and are biased. A stratified sampling design combined with the GLMM estimator was an efficient strategy that could easily be applied to effort monitoring programs using aircraft. Moreover, the GLMM approach is much more flexible to interpret data obtained from nonstandard sampling regimes and can estimate angler effort over any temporal scale. We outline several summary statistics on accuracy, precision, and bias of estimates at different sample sizes so that managers can evaluate expected precision versus sampling effort in their management jurisdictions. Furthermore, we suggest that our general approach of using camera data to parameterize predictive models applied to data from other low-frequency effort monitoring methods (such as aircraft) can easily be replicated elsewhere.
- 出版日期2018-2
