Abstract
Because of their intense patchiness, euphausiid spatial distributions and stock sizes are often assessed using echosounder surveys. However, statistical error bars appropriate for individual survey results are not well known. We quantified these by examining the statistical repeatability of acoustic estimates of total euphausiid biomass within two enclosed fjords adjoining the Strait of Georgia, British Columbia. Within each inlet, paired and replicated `mirror image' zig‐zag survey tracks provided sets of closely spaced backscatter profiles along the survey lines. Local stock density (biomass per unit area) was calculated by vertical integration across the euphausiid scattering layer. Total inlet population size was then estimated by horizontal interpolation and integration of the local measurements, both by block averaging and by geostatistical interpolation (kriging). By assuming no change in true population biomass over the short time interval separating replicate surveys, we could then estimate statistical precision by analysis‐of‐variance comparison among replicate survey grids. For the partial surveys (one or the other half of the mirror‐image paired grid) multiplicative error bars were about a factor of 1.5 for Jervis Inlet and 1.7 for Saanich Inlet. Use of the full surveys (both parts of the mirror‐image pair, roughly doubling the number of measurements in each estimate) reduced the error to about factor of 1.35 for Jervis but only to about 1.65 for Saanich. Statistical precision was similar for the block average vs. kriging interpolation and integration methods, however, kriging provided additional useful information about spatial pattern within the inlets.