Monthly Archives: January 2020

rstudio::conf(2020)

Maia Sosa Kapur attended this year’s rstudio::conference in San Francisco. Her attendance was supported by a Diversity Scholarship (from rstudio). Maia completed a 2-day workshop in machine learning in the tidyverse. While machine learning (ML) has not really been incorporated into stock assessment, the method may be useful for the analysis and development of simulation studies in our field. Indeed, a recent paper in PNAS applied a robotics-based algorithm which outperformed traditional MSY-based fishery management strategies. A relevant topic discussed was model ensembling, which in fisheries modeling typically refers to the process of averaging information from different model structures.

Following her exposure to the tidyverse, Maia produced a brief guide for automating simple CPUE standardizations hosted on her personal blog here. Aside from ML applications, the tidyverse is a very versatile framework for automating transparent modeling workflows.

 

Maia (center) with other attendees at the rstudio::conf Diversity Scholars luncheon.

Is ignoring predation mortality leading to an inability to achieve management goals in Alaska?

Is ignoring predation mortality leading to an inability to achieve management goals in Alaska?

Tuesday, January 28th at 2:00 PM (PST) in FSH 203

Grant Adams

Changes in the dynamics of exploited marine resources have long been understood to coincide with, or be the result of, trophic interactions. However, tactical management of exploited marine resources almost universally effectively ignores trophic interactions despite copious calls for Ecosystem Based Fisheries Management (EBFM). In particular, predation is thought to represent a large proportion of mortality of groundfish in Alaska, and research continues to be needed to identify the relevance of predation to management performance. Management strategy evaluation is used to examine the impact of ignoring predation mortality on the performance of the current management strategy for walleye pollock, pacific cod, and arrowtooth flounder in the Gulf of Alaska and Bering Sea. Multi-species age-structured operating models fitted to available data are used both systems. The current management approach of using single-species assessment models linked to single-species harvest control rules is evaluated under alternative predation and fishing scenarios. Performance for each scenario is evaluated by comparing the ability of the current management strategy to achieve single-species biological reference points, maximize catch, and reduce bias in biomass estimates.