Scanning imaging absorption spectrometer for atmospheric chartography carbon monoxide total columns: statistical evaluation and comparison with chemistry transport model results

Abstract

This paper presents a detailed statistical analysis of one year (September 2003 to August 2004) of global Scanning Imaging Absorption Spectrometer for Atmospheric Chartography (SCIAMACHY) carbon monoxide (CO) total column retrievals from the Iterative Maximum Likelihood Method (IMLM) algorithm, version 6.3. SCIAMACHY provides the first solar reflectance measurements of CO and is uniquely sensitive down to the boundary layer. SCIAMACHY measurements and chemistry transport model (CTM) results are compared and jointly evaluated. Significant improvements in agreement occur, especially close to biomass burning emission regions, when the new Global Fire Emissions Database version 2 (GFEDv2) is used with the CTM. Globally, the seasonal variation of the model is very similar to that of the SCIAMACHY measurements. For certain locations, significant differences were found, which are likely related to modeling errors due to CO emission uncertainties. Statistical analysis shows that differences between single SCIAMACHY CO total column measurements and corresponding model results are primarily explained by random instrument noise errors. This strongly suggests that the random instrument noise errors are a good diagnostic for the precision of the measurements. The analysis also indicates that noise in single SCIAMACHY CO measurements is generally greater than actual variations in total columns. It is thus required to average SCIAMACHY data over larger temporal and spatial scales to obtain valuable information. Analyses of monthly averaged SCIAMACHY measurements over 3° × 2° geographical regions indicates that they are of sufficient accuracy to reveal valuable information about spatial and temporal variations in CO columns and provide an important tool for model validation. A large spatial and temporal variation in instrument noise errors exists which shows a close correspondence with the spatial distribution of surface albedo and cloud cover. This large spatial variability is important for the use of monthly and annual mean SCIAMACHY CO total column measurements. The smallest instrument noise errors of monthly mean 3° × 2° SCIAMACHY CO total columns measurements are 0.01 × 1018 molecules/cm2 for high surface albedo areas over the Sahara. Errors in SCIAMACHY CO total column retrievals due to errors other than instrument noise, like cloud cover, calibration, retrieval uncertainties and averaging kernels are estimated to be about 0.05–0.1 × 1018 molecules/cm2 in total. The bias found between model and observations is around 0.05–0.1 1018 molecules/cm2 (or about 5%) which also includes model errors. This thus provides a best estimate of the currently achievable measurement accuracy for SCIAMACHY CO monthly mean averages.