Effect of reduction in spatial survey effort on indices of bottom temperature for the eastern Bering Sea

Abstract

— Bottom temperature is routinely measured as part of the bottom-trawl survey conducted every summer on the continental shelf of the eastern Bering Sea by the NOAA Alaska Fisheries Science Center. These data are widely used in ecosystem, stock assessment, and ocean modeling. We assessed the effect of alternative sampling designs and effort reduction on the quality of bottom-temperature information from the survey. Simple-random and stratified-random sampling were simulated and compared with the systematic sampling of fixed stations in the regular grid used in the standard survey, with respect to the use of survey data in the estimation of bottom temperatures and related indices. The effort simulated ranged from 34% to 100% of the full effort. In the simulated surveys, the use of each of the 3 sampling designs resulted in values of bottom-temperature metrics that are close to those from the real survey, even with as little as half the effort. Lower effort resulted in larger and more variable prediction errors of the indices. The decrease in prediction performance is most noticeable at the 34% effort level. Systematic sampling performed slightly better than simple-random and stratified-random sampling. One reason for this difference in performance is that random sampling may have been less effective than the standard sampling in capturing a small cold pool that is characteristic of the current warm ocean state.