EDUCE: WP 6 & 7 D evelopment of an Algorithm to Detect Spikes and Distortions in UV Spectra Charikleia Meleti, Alkis Bais Aristotle University of Thessaloniki Laboratory of Atmospheric Physics
Goals Independent of other measurements Expose anomalies above statistical noise Run automatically –Producing flags –Correcting spectra Applicable to cloud-induced distortions
Description of Procedure Construction of a Reference spectrum α(λ, i): statistical coefficients to derive E(λ) from its neighboring wavelengths (SZA dependent) w(i-λ): weights inversely proportional to the difference from the central wavelength
Description of Procedure Construction of a Reference spectrum α(λ, i): statistical coefficients to derive E(λ) from its neighboring wavelengths (SZA dependent) w(i-λ): weights inversely proportional to the difference from the central wavelength
Description of Procedure Construction of a Reference spectrum α(λ, i): statistical coefficients to derive E(λ) from its neighboring wavelengths (SZA dependent) w(i-λ): weights inversely proportional to the difference from the central wavelength
Detection of Spikes Compute the ratio between E(λ) and E R (λ) Ratios higher than 1.5 indicate the presence of a spike
Flagging Spectral Distortions Compute the correlation coefficient (r 2 ) between E(λ) and E R (λ) –Accepted spectra r 2 > 0.99 –Suspicious spectra 0.89 < r 2 < 0.99 Small spikes Distortion by clouds –Highly distorted spectra r 2 < 0.89 Wavelength shifts Spikes Noise Zeros
Concluding Remarks All spectra flagged “accepted” were actually correct (by visual inspection) Small uncertainties still exist at low wavelengths The method does not depend on absolute calibration so it can be applied to raw data