Retrieval of CH_4, CO, and CO_2 total column amounts from SCIAMACHY near-infrared nadir spectra: Retrieval algorithm and first results

摘要

SCIAMACHY is a UV/visible/near-infrared grating spectrometer on board the European environmental satellite ENVISAT that observes the atmosphere in nadir, limb, and solar and lunar occultation viewing geometries with moderate spectral resolution (0.2-1.5 nm). At the University of Bremen a modified DOAS algorithm (WFM-DOAS) is being developed primarily for the retrieval of CH_4, CO, CO_2, H_2O, N_2O, and O_2 total columns from SCIAMACHY near-infrared and visible nadir spectra. A first version of this algorithm has been implemented based on a fast look-up table approach. The algorithm and the look-up table is described along with an initial error analysis. Weighting functions and averaging kernels indicate that the SCIAMACHY near-infrared nadir measurements are highly sensitive to trace gas concentration changes even in the lowest kilometer of the atmosphere. The results presented have been obtained by applying WFM-DOAS to small spectral fitting windows focusing on CH_4, CO_2, CO, and O_2 column retrieval and CH_4 and CO_2 to O_2 column ratios (denoted XCH_4 and XCO_2, respectively). These type of data products are planned to be used within the EU research project EVERGREEN to constrain surface sources and sinks of CH_4 and CO_2 using inverse modeling techniques. This study discusses the first set of WTM-DOAS products generated for and to be further improved within EVERGREEN. Although no detailed validation has been performed yet we found that the retrieved columns have the right order of magnitude and show (at least qualitatively) the expected correlation of the well mixed gases CO_2 and CH_4 with O_2 and surface topography. The standard deviation of the dry air column averaged mixing ratio XCO_2 within 10° latitude bands is +-10 ppmv or +-2.7% (XCH_4: +-50 ppbv or +-2.8%) for measurements over land (over ocean the scatter is a factor of 2-4 larger). These values have been determined from ～25% of the ground pixels of one orbit which fulfill the following requirements: (nearly) cloud free, solar zenith angle < 75°, XCO_2 error < 4% (XCH_4 error < 6%). It has not yet been assessed how much of this variability can be attributed to real column changes. The observed variability is about three times larger than expected from (single spectra) signal-to-noise considerations but might be affected by limitations of the current implementation of the retrieval algorithm (e.g., sensitivity to surface reflectivity) and calibration issues (e.g., not yet considered ADC non-linearity correction). Especially the CO retrieval needs further study and improvement. The CO fit errors are 20-40% over land but typically significantly larger over the ocean. A clear identification of the weak CO lines is difficult as the CO fit residuals are dominated by relatively stable systematic artifacts (also observed in the CO_2 and CH_4 fitting windows) on the order of the weak CO absorption lines. This might be explained by the still preliminary calibration of the SCIAMACHY spectra and/or errors of the spectroscopic data.