NSF-OOI Spectral Absorption and Attenuation - Baseline Drift Corrected - Relative, 2014-2026

This dataset consists of the absorption and attenuation data made available in the NSF-OOI Spectral Absorption and Attenuation, 2014-2026 repository (21402583), but adjusted for drift using a Moving Window Minimum Value morphological baseline at each interpolated wavelength. Additional information about this methodology can be found in the pybaselines documentation and in Yaroshchyk and Eberhardt, 2014. An example of this drift correction application can be found here. The pre-deployment and post-deployment user-collected pure water calibrations provided by the NSF-OOI are difficult to access/process or may not be consistently collected for all ac-s deployments. The MWMV drift removal method is experimental with ac-s data and does not consider these user calibrations. The MWMV method is one of many that may be used to account for instrument related drift. Method Application This method is only applied to the a_pg, a_pg_baseline, a_pg_proportional, and c_pg variables in each dataset (see 21402583 for a table describing file structure). A median filter with a window size of 7(~7 minutes) is applied along the time dimension for non-profiling datasets. A median filter with a window size of 5 (~2.5 meters) is applied along the depth dimension for profiling datasets. A window size equivalent to 8 days is then used for the baseline drift window. 8 days was selected because it appears to remove gradual changes (over many weeks) in signal while also preserving short-term changes which could represent particle pulse events. This assumes that gradual changes are due to instrument drift or biofouling and not changes in the mean particle and dissolved state of the waters, which may be a poor assumption depending on the intended application. Co-located camera imagery could be used to verify at some OOI sites. The baseline drift is then removed from the dataset. This effectively shifts each time-series at each wavelength relative to the minimum value of the time-series. The variable *_minimum provides the minimum value at each wavelength should users wish to add it and shift the time-series to a non-zero value. This shift should be sufficient for relative analyses. Analyses wishing to use absolute values may be less robust and users should take care when re-implementing the minimum value.