PMC:3216509 / 10705-12003 JSONTXT

{"target":"https://pubannotation.org/docs/sourcedb/PMC/sourceid/3216509","sourcedb":"PMC","sourceid":"3216509","source_url":"https://www.ncbi.nlm.nih.gov/pmc/3216509","text":"To determine the likely location of open ocean feeding areas for our test salmon populations, we assessed the temporal covariance between scale δ13C values and SST in each one-degree grid square between 45–75°N latitude and 65°W–20°E longitude for the period 1985 to 20024950. Time series of tissue isotopes and SST are autocorrelated, violating the assumptions of serial independence demanded by most classical inference tests51. Autocorrelated time series are frequently de-trended to remove underlying low-frequency variations, however we are primarily interested in low frequency variations that are common to both time series, thus de-trending would be counter-productive51. Instead, the influence of high frequency fluctuations in the time series was reduced by applying LOESS smoothing with a span of 0.5 and polynomial order 2 (Fig. 3). Low frequency variations in the smoothed time series were correlated, and the influence of autocorrelation on significance tests was accounted for by adjusting the effective degrees of freedom for each time series pair using the modified Chelton method51. Areas with the highest correspondence between temporal variations in SST and measured δ13C values are suggested as the most likely marine feeding areas during deposition of scale collagen (Fig. 4).","tracks":[]}