Matched CalCOFI samples did not resolve a residual oxygen variance change at 200 m between 1984-2002 and 2003-2020

We tested whether year-to-year oxygen spread changed at 200 m off southern California after separate means and linear trends were removed. We used observed CalCOFI bottle profiles from a fixed archive and 224 matched station-quarter groups at 56 stations. There were 7,360 annual values in 1984-2020. The late-to-early ratio of equal-group mean residual variances was 1.111 for 2003-2020 versus 1984-2002. Its conditional 95% three-year block interval was 0.860 to 1.434. Thus this test did not resolve a spread change. All planned 200 m sampling and block checks also left the direction uncertain. A planned secondary check at 300 m gave a ratio of 0.691, with limits of 0.521 to 0.871. Different depth samples and multiple comparisons limit that result. The study does not show constant variance, stable extremes or a physical cause. Quarterly sampling, method changes, missing years, detrending and short dependent periods limit inference. The package gives attributed profiles, exact source hashes, code, results, main resample statistics, a figure and numerical checks. The raw archive stays external.

CalCOFI · oxygen · interannual variance · matched sampling · block bootstrap · null result