Table 2 Quality control tests and associated flag names in table GLORIA_qc_flags.csv.
Flag name Number of cases | Description and method |
|---|---|
Noisy_red 40 | High-frequency variability, potentially instrument noise, near the red end: spectra were standardized to zero mean and unit standard deviation. A 4th order polynomial was fitted over the interval 750–900 nm. Spectra with a root-mean square error (RMSE) >0.2 were flagged. This threshold was determined using visual inspection of the distribution of RMSEs with respect to spectral shapes. |
Noisy_blue 15 | High-frequency variability, potentially instrument noise, near the blue end: spectra were standardized to zero mean and unit standard deviation. A 4th order polynomial was fitted over the interval 350–400 nm. Cases where root-mean square error >0.15 were flagged (threshold determined using visual inspection of the distribution of RMSEs with respect to spectral shapes). |
Baseline_shift 164 | Spectra shifted up are those where the minimum Rrs is 60% of its median. This percentage corresponds approximately to 1.5 times the interquartile range above the upper quartile of the baseline-percent distribution of the entire GLORIA dataset. Spectra shifted down are those with at least 20 negative values and either: • a negative linear slope in the interval 765–900 nm <−8.75 × 10−7 sr−1 nm−1 (the slope threshold was determined as the bound of the lower quartile) and >50% negative Rrs values in this spectral region; or • >70% negative Rrs in the interval 765 nm-900 nm; or • at least 20 negative Rrs at in the interval 350–450 nm. |
Oxygen_signal 1311 | Spectra where Oxygen_peak_height >0.1 (Table 3). This threshold was determined using visual inspection of the distribution of peak heights with respect to spectral shapes. |
Negative_uv_slope 139 | Negative slopes in the ultraviolet to blue end: The spectra were standardized to zero mean and unit standard deviation. A straight line was fitted over the interval 350–420 nm and spectra with slopes <−0.005 were flagged. This threshold was determined using visual inspection of the distribution of slope values. |
QWIP_fail 278 | Spectra failing a statistical quality control metric based on Apparent_visible_wavelength (Table 3). The QWIP score exceeded a value of |0.2|. |
Suspect 226 | Spectra identified during expert elicitation as potentially fraught with measurement problems. |
Flagged 1779 | A one in this column indicates the presence of at least one flag from the tests described in this table. |
