Satellite based remote sensing of halogens in the Arctic troposphere, under the impact of Arctic Amplification

Abstract

Bromine plays a crucial role in polar atmospheric composition. During springtime, photochemistry converts bromine compounds originating from ice and snow into gaseous reactive bromine radicals (Br), which deplete ozone in the boundary layer, forming bromine oxides (BrO). Due to the autocatalytic nature of the reaction mechanism, it has been called bromine explosion. The strong relationship of bromine photochemistry and release from its sources to ozone depletion events (ODEs) was discovered in the late 1980s. Since then, and because of the importance of tropospheric ozone (the primary source of hydroxyl radical (OH), the major oxidizing agent of the atmosphere), many studies focused on the mechanisms which release bromine into the troposphere and the driving parameters which enhance BrO production and therefore ozone depletion. Arctic Amplification (AA) is the phenomenon that surface temperature in high latitudes increases more rapidly than at lower latitudes. One of the most profound consequences of AA is the significant changes in sea ice conditions. Sea ice extent, age and thickness are drastically changing. Inevitably, all aspects of the Arctic ecosystem are expected to be affected by Arctic Amplification. Remote sensing from satellites can be extremely useful for studying the Arctic region. By the end of the 1970s, sea ice concentration was successfully monitored by satellite sensors. Since 1995, we also have the ability to study atmospheric composition worldwide with data retrieved from nadir radiance spectra from a series of European satellite sensors: GOME on ERS-2 (1995 - 2003), SCIAMACHY on ENVISAT (2002 - 2012), GOME-2A on MetOpA (2007 - today) and GOME-2B on MetOpB (2012 - today). The focus of this thesis is twofold: firstly, to create the first consolidated and consistent long-term (1996 to 2017) tropospheric BrO dataset for the Arctic region and the Hudson Bay (a well-known hotspot for bromine explosion events), retrieved from the four ultraviolet-visible (UV-VIS) sensors mentioned above, in order to assess the changes and the impact that AA has on tropospheric BrO. Since the different satellite instruments have different instrumental attributes and the fact that BrO is a weak absorber in the UV spectral region, many sensitivity retrieval tests have been performed in order to identify the proper fitting settings for each instrument and to derive a high-quality BrO dataset with high consistency between the instruments. Vertical column densities (VCD) of tropospheric BrO are extracted from total (geometric) VCDs, using a climatology of stratospheric BrO VCDs from a chemical transport model. The BrO time-series (geometric and tropospheric VCDs) show remarkable agreement during the overlapping periods between the sensors (the best agreement being between SCIAMACHY and GOME-2A with a correlation coefficient squared of 0.97). Additionally, the agreement is verified by studying daily and monthly maps of geometric and tropospheric BrO VCD. This agreement allows us to create a merged tropospheric BrO VCD dataset, the basis for deriving geophysical conclusions on the impact of AA on the Arctic BrO atmospheric composition. By studying the trends of tropospheric BrO VCDs, we infer an increase of around 1% per year. The increase is significant during polar spring, reaching 1.5% per year. A similar increase can be observed for the Hudson Bay, where tropospheric BrO VCD have increased around 0.9% per year for the spring period and 2.3% per year for the winter period. However, the increasing trend is not monotonic, and variability on the tropospheric BrO VCD appear (e.g. 2016 and 2017 are lower than 2015 for the Arctic region), as many parameters influence bromine release and subsequent BrO formation. Secondly, the link of observed tropospheric BrO VCDs to possible bromine sources and favoring weather conditions was investigated. The primary source of inorganic bromine release in the Arctic atmosphere is sea ice, especially first year ice, which is rich in sea salts. Therefore, a long-term sea ice age dataset from NSIDC was compared to the tropospheric BrO VCD dataset. The increase of first year ice extent due to the changes in the Arctic climate is in general agreement with the observed increase in tropospheric BrO, with a moderate daily correlation coefficient of +0.32, but a strong yearly one of +0.62. Also, the increase of the occurrences of first year ice over some regions in the Arctic (i.e. northeast of Greenland) is correlated with the increase of tropospheric BrO VCD in these regions. Apart from the bromine sources, driving mechanisms, like air temperature, wind speed, boundary layer height, and cyclone activity, can contribute to the Amplification, transport, vertical uplifting and recycling of tropospheric BrO plumes. Consequently, similar comparisons have been performed between tropospheric BrO and meteorological data. We infer that the parameter with the most substantial influence on the formation of bromine explosion events is air temperature, with a correlation coefficient of -0.54 (anti-correlation) for the Arctic during spring and -0.78 for Hudson Bay spring. Furthermore, the spatial agreement and correlation between trend maps of tropospheric BrO VCD and air temperature verify the anti-correlation between the two quantities. However, the bromine release and the formation, transport, and re-cycling of BrO plumes are complex and dynamic phenomena. They depend on many geophysical parameters, and there are also complex relationships between these parameters. Therefore, single individual linear comparisons between tropospheric BrO VCDs and key parameters of BrO formation cannot fully represent the actual relationship between them. This leads us to the effort of employing, for the first time, an artificial intelligence neural network in order to model and predict Arctic tropospheric BrO VCDs using the parameters contributing to tropospheric BrO formation. By training the neural network only with one year of data, it can reproduce accurately (both spatially and in magnitude) many BrO plumes which occurred in other years. This ability of the neural network to efficiently model some bromine explosion events allows us to distinguish them between those occurring at the surface and those at higher altitudes. From studying the effect of each of the individual key parameters on the magnitude of modeled tropospheric BrO VCD, we conclude that air temperature and mean sea level pressure (which can describe the boundary conditions under which bromine is released) have the highest impact.