Workshop on Optimization of Biological Sampling (WKBIOPTIM3)

The third Workshop on Optimization of Biological Sampling (WKBIOPTIM3) convened to discuss practical aspects of optimization of sampling and provide a compilation of methods so that this information is well documented and organized for end users. New optimization algorithms were presented alongside developments and improvements of the work carried out in the earlier 2017 and 2018 workshops. Six different approaches/tools were presented: code developed under WKBIOPTIM 1 and 2 for optimization at sample level (SampleLevelOptim), SampleOptim R-tool to optimize fish sampling for biological parameters, code for quantifying robustness of a length frequency distribution shape (SampleReferenceLevel), SDTool and BioSimTool as contri-butions from the STREAM Project to the optimization at sample-level and sampling design level, respectively, WKBIOPTIM code used for optimization at sampling design level (SimPop). Additionally, data exploration and biological simulation tools developed under the framework of fishPi2 (FishPi4WKBioptim). R-scripts for the different tools are available on the WKBIOPTIM3 GitHub (https://github.com/ices-eg/wk_WKBIOPTIM3).

Usability testing, including thorough instructions, were a focus of the workshop. As such, the group used a two-tiered system to evaluate the methods. The first subgroup tested the different scripts with their own data to see if they were of easy use. This first group also evaluated the documentation needed and whether it was clear how to interpret the results properly, how to decide on the best procedure according to their objectives, and how to compare results obtained using similar approaches. Based on this assessment, the second subgroup then cleaned and doc-umented the code, discussing the approaches used in the different scripts and potential options for integration. In parallel, some participants developed a first draft of standardized notation aimed at harmonizing the documentation of simulation procedures used in the different codes, tested a new quality indicator for length frequency and developed scripts that demonstrate the effects of the common options of resampling ‘with replacement’ (wr) and ‘without replacement’ (wor) in the precision and bias of estimates. The workshop concluded with participants high-lighting that specific documentation regarding quality indicators concepts and functions could also be very practical and useful for end users. Work of WKBIOPTIM will continue intersession-ally towards a final discussion of methodologies and results in 2020.