Characterisation of dust sources in Central Asia using remote sensing

Central Asian deserts are a significant source of dust in the middle latitudes, where economic activity and the health of millions of people are affected by dust storms. Detailed knowledge of sources of dust, controls on their activity, seasonality and atmospheric pathways are of crucial importance but to date, these data are limited. This thesis presents a detailed database ofsources ofdust emissions in Central Asia, from western China to the Caspian Sea, obtained by a multi-scale analysis of the Moderate Resolution Imaging Spectroradiometer (MODIS) satellite data. The multi-scale approach consists of the following steps: 1) MODIS Deep Blue Aerosol Optical Depth (DB AOD) at 10 km resolution, acquired between 2003 and 2014, is used to investigate the spatiotemporal distribution ofdust hotspots. 2) A dust enhancement algorithm was employed to obtain two composite images (Dust Enhancement Product, DEP) per day at 1 km resolution from MODIS Terra/Aqua acquisitions between 2003 and 2012, from which dust point sources (DPS) were detected by visual analysis of dust plumes and recorded in a database together with meteorological variables at each DPS location derived from the ERA-Interim reanalysis dataset. In all, more than 13500 DPS were identified. Using this multi-scale approach we provided a high resolution inventory of dust sources at sub-basin scale for Central Asia. Our analysis revealed several active source regions, the most active of which are the eastern part ofthe Taklmakan desert. An important finding was an increase in dust activity in the newly-formed desert ofthe Aralkum. Several ofthe identified dust source regions were not previously identified (e.g. sources in northern Afghanistan) or were not widely discussed in literature before (e.g. the Pre-Aral region in western Kazakhstan). Investigation of land surface characteristics and meteorological conditions at each source region revealed mechanisms for the formation of dust sources, including rapid desiccation of water bodies (e.g. Aral Sea), deflation of dust from fluvial sources (e.g. the Upper Amudarya region) and post-fire wind erosion (e.g. Pre-Aral and Lake Balkhash basins). Different seasonal patterns of dust emissions were observed as well as inter-annual trends. Comparison of DB AOD and DPS revealed a noticeable spatial bias in the AOD-based methods for detection of dust sources which is attributed to the fact that the highest atmospheric dust loadings are not always observed over the dust point sources.