WP 3: Traceability of Total Column Ozone

The aim of this work-package is to determine a comprehensive uncertainty budget for the atmospheric total column ozone measurement by ground-based instruments, taking into account all parameters affecting the retrieval of this atmospheric species.

The problem to be solved: Atmospheric total column ozone is currently retrieved by various methodologies using different measurement procedures, wavelength regions and ozone absorption cross-sections. In particular, the effect from cross-correlation when using different wavelength regions on the retrieval uncertainty will have to be assessed in order to develop an uncertainty model valid for discrete wavelength ozone retrievals as applied to Dobson and Brewer spectrophotometers and array spectroradiometers using high resolution solar spectra.

Another source of uncertainty in the ozone retrieval is the use of ozone absorption cross-sections. Both Brewer and Dobson spectrometers operate in the near UV (305-340 nm) where the Huggins ozone absorption band has distinct absorption minima and maxima (whose differences are exploited in a differential retrieval to derive ozone amounts). This band is also the spectral region where ozone cross-sections have strong temperature dependence so that uncertainties in the assumed atmospheric temperatures (or neglect of temperature variation in the retrieval as in the standard WMO Brewer and Dobson retrieval) add to the retrieval error (Vanicek et al., 2003).

In the WMO standard retrieval of Brewers and Dobsons to date Bass Paur ozone cross-sections from the 1980s are used. In the meantime newer measurements like those from Brion as well as the very recent data from the University of Bremen group (are available. These newer ozone cross-sections have an estimated error of about 2 % or worse (depending on the spectral region). However, the error budgets reported generally do not include all potential uncertainties and correlations and are rather too optimistic. In this work package a detailed error budget will be provided for the relevant cross-section data that are currently used in remote sensing, which can be then translated into error contribution to the ozone retrieval. Secondly, the new ozone cross-section data of the Bremen cover a wide spectral range and a large dynamic range (on the order of 107). As part of this work improved ozone cross-sections in the Huggins band will be provided to further reduce the uncertainties in the near UV spectral region.



Description of Task 3.1

Validation of high resolution solar reference spectra (SFI Davos, PTB, REG(ULL))

The aim of this task is to validate the high resolution solar spectrum used as reference for the retrieval of total column ozone from ground-based direct solar UV spectra. The prime candidate for this task will be the extra-terrestrial spectrum constructed in JRP ENV03 SolarUV D2.4.1. Since the retrieval of total column ozone will be performed in the wavelength range 300 to 350 nm, the validation will be restricted to this range. The goal is to validate this spectrum with a relative uncertainty of 1 % over this wavelength range, e.g. the absolute irradiance level is not relevant for the ozone retrieval. The raw measurements will be based on the direct solar irradiance spectra measured during the Task 2.3 field measurement campaign organised at the Izaña Observatory for atmospheric research, Tenerife, Canary Islands.


Description of Task 3.2

Assessed and validated ozone absorption cross-sections (REG(UniHB), PTB)

The aim of this task is to provide a detailed uncertainty budget of the most relevant ozone absorption cross‑sections that are currently available. In addition new measurements of the Huggins ozone band in the near UV with an improved experimental setup will be conducted at REG(UniHB) in order to further reduce the uncertainties in that spectral region.


Description of Task 3.3

Task 3.3 Uncertainty models for correlated quantities (Aalto, VSL, SFI Davos, REG(ULL))

The aim of this task is to develop two methods to calculate uncertainties from correlated input variables, without a-priori knowing the correlation matrices: The first method will be applied to the double-ratio technique used by Dobson and Brewer spectrophotometers to retrieve TOC from 4 discrete wavelengths; the second method will retrieve the uncertainties of TOC from measurements of the full solar spectrum, using Differential Optical Absorption Spectroscopy (DOAS) techniques.

Total column is retrieved from direct solar irradiance measurements using retrieval methodologies based on the Beer-Lambert Law, with varying complexity, essentially due to the number of free parameters in the description of the atmosphere and the number of spectral measurements. There are various practices on how to calculate the uncertainty for the derived total column ozone that propagate from the uncertainties of the input variables. However, to get realistic uncertainty estimates, one has to account for correlations between the uncertainties at different wavelengths. This is problematic if the correlations are not known, which is typically the case. Monte Carlo simulations by randomly varying the input variables and neglecting correlations is a method often used. However, the uncorrelated uncertainties behave like noise and average out in the calculation, which strongly underestimates the uncertainty. On the contrary, if we assume full correlation of the uncertainties at different wavelengths, the uncertainty is underestimated too, because the spectral shape is not affected. In practice, the situation is typically between the two extremes presented. A method is needed that gives limits or an estimate for the uncertainty taking into account that correlations may exist, although unknown.


Description of Task 3.4

Task 3.4: Comprehensive Uncertainty budget for Total Column Ozone (SFI Davos, REG(ULL), REG(UniHB), Aalto, VSL)

The aim of this task is to produce a comprehensive uncertainty budget for ground-based total column ozone retrieval from Dobson, Brewer and array spectroradiometer measurements, taking into account cross-correlations between measurements at different wavelengths.