Biases in acoustic biomass density estimates used for calculating catch limits

We provide estimates of krill biomass density in three survey strata defined by the U.S. AMLR Program: Elephant Island (EI), the West Shelf (WA), and the Bransfield Strait (or South Area (SA)). We include density estimates from U.S. AMLR research vessel surveys conducted from 1996 to 2011 and from gliders deployed in 2018-19. We also include density estimates from fishery-independent surveys conducted by fishing vessels from 2013 to 2019. The combined set of estimates demonstrates that krill biomass density varies over time, including within a season, and among survey strata (Fig. 1). We do not currently know how much of this variation is attributable to measurement error. While SC-CAMLR and its working groups have devoted considerable time to ensuring that techniques for collecting and analyzing acoustic data are comparable across platforms, differences in the nets used during these surveys (e.g., mesh sizes) likely add uncertainty to the time series of biomass estimates. We show that using length-frequency distributions (LFDs) from biased samplers (e.g., krill in penguin diets) can subsequently bias estimates of krill biomass up when compared to estimates based on LFDs from the IKMT nets historically used by the U.S. AMLR Program (Fig. 5). The LFDs of krill collected from commercial trawls are, on average, more similar to LFDs from penguin diets than from IKMTs, likely because commercial trawls tend to have larger mesh sizes than scientific trawls. Therefore, the time series of biomass estimates from fishing vessels (starting in 2013) and gliders (2018/19) that rely on commercial trawls or predators should be considered the maximum potential biomass density in the region and used to document temporal changes in biomass time series, rather than as absolute estimates. Going forward, we propose that acoustic surveys conducted by fishing vessels use a standard gear, or at the least a standard cod-end liner of <5 mm, when conducting fishery-independent surveys. The temporal and spatial variability observed in biomass density estimates raises two concerns related to using individual biomass estimates for computing catch limits. First, if surveys are conducted infrequently, biomass estimates used to calculate catch limits might be either too high, increasing risks to krill-dependent predators, or too low, forgoing valuable catch. Second, if surveys are conducted frequently, catch limits will themselves be highly variable and survey costs will increase. At present, we do not know the optimal survey frequency. We propose WG-ASAM recommend that an average density, computed over time and multiple survey strata, be used to determine catch limits.