Implications of maximum bin size selection in statistical length-structured stock assessment models

Length-structured stock assessment models divide a population into multiple size categories, or bins.  The maximum size bin designated for each data source may either be static across the time series or allowed to vary annually.  In the case of static bin selection, the maximum bin contains the largest animal typically observed across the entire time series for each data source; in the case of dynamic bin selection, the maximum bin will contain the largest animal typically observed annually for each data source.  However, the relative performance of static versus dynamic bin selection approaches has not been evaluated.  We simulated a series of lobster populations for which length structure had either remained stable or collapsed over time due to different fishery exploitation histories.  A statistical catch-at-length model was used to assess each population time series twice, using both static and dynamic maximum bin selection methods.  Model performance relative to simulated population dynamics was compared between bin selection methods.  When length structure of the population remained relatively stable across the time series, both methods performed similarly.  However, when population length structure expanded or collapsed, static bin selection outperformed dynamic bin selection if the appropriate maximum bin was selected.