Aerosol Optical Thickness (AOT) data have many important applications including atmospheric correction of satellite imagery and monitoring of particulate matter air pollution. Current data products are generally available at a kilometre-scale resolution, but many applications require far higher resolutions. For example, particulate matter concentrations vary on the scale of tens of metres, and thus data products at a similar scale are required to provide accurate assessments of particle densities and allow effective monitoring of air quality and analysis of local air quality effects on health. This thesis describes the development of a novel method which retrieves per-pixel AOT values from high-resolution (30m) satellite data, and this method is the main novel contribution to scientific knowledge of this PhD. This method is designed to work over a wide range of land covers including both bright and dark surfaces - and requires only standard visible bands, making it applicable to a range of data from sensors such as Landsat, DMC, SPOT and Pleiades. The method is based upon an extension of the Haze Optimized Transform (HOT), which was originally designed for estimating the haziness of each pixel in a satellite image, based upon the distance from a `Clear Line' in feature space. In this research, the HOT method is adapted and used to estimate AOT instead. Significant extensions include Monte Carlo estimation of the `Clear Line', object-based correction for land cover, and modelling of the HOT-AOT relationship using radiative transfer models. Validation against ground and satellite measurements, as well as simulated data, shows that 40-50% of the pixels have an error within ±0.1, not much lower than many presently available low-resolution products, with further work likely to improve the accuracy. Two example applications show the potential of this method for per-pixel atmospheric correction and monitoring the spatial pattern of particulate matter pollution. This novel method will enable many new applications of AOT data that were impossible with low-resolution data.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:640777 |
| Date | January 2015 |
| Creators | Wilson, Robin |
| Contributors | Milton, Edward ; Noble, Jason ; Nield, Joanna |
| Publisher | University of Southampton |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | https://eprints.soton.ac.uk/374707/ |
Page generated in 0.0156 seconds