Satellite Measurements of Carbon Dioxide: Analysis and Reduction of Scattering Related Retrieval Errors

Abstract

The greenhouse gas carbon dioxide (CO2) is the most important human-made contributor to global warming. Despite its importance, our knowledge about its sources and sinks has large gaps. This limits a reliable climate prediction. Satellite observations of atmospheric CO2 combined with modelling can help to reduce these knowledge gaps. However, this requires to meet demanding accuracy and precision requirements for the satellite instrument, the retrieval algorithm and the model. One of the most important error sources for satellite retrievals of CO2 from reflected and backscattered solar radiation is unaccounted scattering by aerosols and clouds. In this context, the objectives of this thesis are to assess the quality of an existing satellite-based CO2 data set focussing on the investigation of these error source and to generate and validate an improved CO2 data set. The CO2 data bases on measurements of the passive remote sensing satellite instrument SCIAMACHY on-board ENVISAT, which has performed more than 10 years of radiance measurements in the short wave infrared spectral region. In the period 2003-2009, SCIAMACHY was the only satellite instrument measuring CO2 with high sensitivity down to the Earth's surface where the sources and sinks of CO2 are located. Therefore, the SCIAMACHY measurements are important in terms of generating an accurate global long-term atmospheric CO2 data set. Starting point for this thesis was an analysis of an existing 7-year (2003-2009) data set of column-averaged dry airmole-fraction of CO2, denoted XCO2, which was generated with version 2.1 of the Weighting Function Modified - Differential Optical Absorption Spectroscopy (WFM-DOAS) retrieval algorithm (WFMDv2.1). In order to study if differences between satellite-derived and modelled XCO2 are from scattering related retrieval errors, the SCIAMACHY XCO2 data set has been compared with the output of NOAA's modelling and assimilation system CarbonTracker. It has been investigated to what extent the differences between SCIAMACHY and CarbonTracker XCO2 are temporally and spatially correlated with global aerosol and cloud data sets. For this purpose, aerosol information from the European GEMS project and cloud information of NASA's CALIPSO satellite have been utilised. In this analysis, significant correlations with thin clouds especially over tropical and southern hemispheric regions have been found. The maximum temporal (r2=54%) and spatial (r2=46%) correlations have been found for Darwin, Australia. Large temporal correlations with thin clouds have also been observed over other regions of the Southern Hemisphere (e.g. r2=43% for South America and r2=31% for South Africa). Over the Northern Hemisphere the temporal correlations are typically much lower. An exception is India, where large temporal correlations with clouds and aerosols have also been found. These results indicate, that the SCIAMACHY WFMDv2.1 XCO2 data set suffers from scattering related retrieval errors caused especially by thin clouds. In order to reduce the scattering related retrieval errors, a new version of the WFM-DOAS retrieval algorithm has been developed (WFMDv2.2), which is based on a new cloud filtering and correction method. This method is based on radiances from the 1.4 µm strong water vapour absorption and the 0.76 µm O2-A band. The new version of the WFM-DOAS retrieval algorithm has been used to generate an improved SCIAMACHY XCO2 data set covering the period 2003-2009. The new data set has been validated using ground-based Fourier Transform Spectrometer (FTS) observations from the Total Carbon Column Observing Network (TCCON). The validation shows significant improvements of the new version WFMDv2.2 in comparison to the previous version WFMDv2.1. For instance, the standard deviation of the difference to TCCON over Darwin has been improved from 4 ppm to 2 ppm. Overall, the validation of the SCIAMACHY WFMDv2.2 XCO2 can be summarised by a single measurement precision of 3.8 ppm, a regional scale precision of monthly averages of 1.6 ppm and an estimated regional scale relative accuracy of 0.8 ppm. In order to investigate the differences between the new SCIAMACHY WFMDv2.2 XCO2 data set and CarbonTracker XCO2, a comparison between these data sets has been performed. It has been shown that the new data set agrees better with CarbonTracker than the previous version. The new SCIAMACHY XCO2 data set has already been used for interesting applications, for example, for an assessment of regional enhancements of atmospheric CO2 and trends over major anthropogenic CO2 source regions. The data set has also been used as part of the Ensemble Median Algorithm (EMMA). EMMA is a promising candidate for generating a CO2 data set which fulfils the demanding requirements needed to obtain information on regional CO2 surface fluxes via inversion modelling.