A new aerosol profile retrieval algorithm for high-altitude MAX-DOAS measurements and the application to the long-term observation on Zugspitze, Germany

Kurzfassung

In this work, we analyze the long-term spectra measured by the multi-axis differential optical absorption spectroscopy (MAX-DOAS) instrument at the Environmental Research Station Schneefernerhaus (UFS), located near Zugspitze, at an altitude of 2,650 m. It is a background site with usually unpolluted air. We present a new cloud screening method in Chapter 4. The method is based on the color index (CI) of zenith spectra of the ultraviolet (UV) band. We determined CI thresholds for cloud screening for different solar zenith an- gles (SZAs) according to the long-term frequency distribution of calibrated CIs. For each SZA, calibrated CIs show bimodal frequency distribution, and the two peaks correspond to the measurements under overcast and clear skies. Therefore, the threshold for each SZA was defined as the calibrated CI with the minimum probability between the two distribution peaks. The cloud screening method was applied to the entire measurement period from Feb 2012 to Feb 2016. The percentage of cloudy measurements is highest in summer and lowest in winter. In total, ∼60% of the zenith measurements were determined as cloudy scenes. In Chapter 5, we present the retrieval of aerosol profiles. We found the algorithms based on the optimal estimation method are not suitable for high-altitude measurements. Therefore, we developed a new retrieval algorithm based on the parametrization approach. The sensitivities of O4 absorption to several parameters were first investigated. Aerosol profiles were parametrized as the aerosol extinction coefficients of three layers. We defined a profiles set which is assumed to include all possible aerosol profiles under cloud-free conditions. O4 differential slant column densities (DSCDs) at 360 and 477 nm were simulated with all the profiles in the profile set and all possible viewing geometries. The simulated data were stored in a look-up table (LUT), which was used as the forward model. In the retrieval of each measurement cycle, simulated O4 DSCDs corresponding to all the possible profiles were obtained from the LUT. The cost function of each possible profile was then calculated according to the simulated and measured O4 DSCDs as well as the measurement uncertainties. We performed a comprehensive error analysis to estimate the total uncertainty, in which seven error sources were considered. Valid profiles were selected according to the cost functions. The optimal solution was defined as the weighted mean of the valid profiles. Based on the assumption that the LUT covers all possible profiles, we determined O4 DSCD scaling factors for different elevation angles and wavelengths. The retrieval algorithm was applied to synthetic measurement data, and the retrieved profiles could reproduce the true profiles. In addition, the retrieval is insensitive to measurement noise. The aerosol optical depths (AODs) retrieved from the long-term measurements were compared to the sun photometer observations at the UFS. High correlation coefficients of 0.733 and 0.798 were found for measurements at 360 and 477 nm, respectively. However, especially in summer the sun photometer AODs are systematically higher than the MAX-DOAS retrievals by a factor of ∼2. The MAX-DOAS measurements indicate the aerosol extinction decreases with increasing altitude during all seasons, which agrees with the co-located ceilometer measurements. Our results also show maximum AOD and maximum ̊Angström exponent in summer, which is consistent with the observations at an Aerosol Robotic Network (AERONET) station located ∼43 km from the UFS. In Chapter 6, we present the retrieval of the total VCDs of O3 and NO2 from the zenith spectra measured during twilight periods. The air mass factors (AMFs) of O3 and NO2 were obtained from LUTs developed at the Belgian Institute for Space Aeronomy (BIRA-IASB). Langley plots of O3 and NO2 were applied to the long-term measurements at the UFS. The Langley plots of O3 show that for most measurements, the AMFs obtained from the LUT are highly correlated with the DSCDs. While for NO2, the correlation between DSCDs and AMFs is on average weaker than O3. For both O3 and NO2, the SCD of the reference spectrum (Sref) derived from Langley plots varies in a large range. We determined the values of Sref from the long-term distributions using Gaussian fit. Then we calculated the VCDs of O3 and NO2 by directly dividing the SCDs (derived by adding Sref to the twilight DSCDs) by the AMFs. The long-term VCDs of O3 and NO2 derived by both Langley plots and direct dividing were compared to satellite measurements. The comparisons show that the VCDs of O3 measured by MAX-DOAS in both mornings and evenings agree well with the satellite results, and the results calculated using direct dividing agree with the satellite data better than the results derived by Langley plots. For the comparison of NO2, the agreement is weaker than O3, regardless of the calculation method. The long-term measurements show that the VCD of O3 is highest in spring and lowest in autumn, and the VCD of NO2 is highest in summer and lowest in winter. The VCD of NO2 has a much larger yearly amplitude than the VCD of O3.