Authors: Stefano Alessandrini; François Vandenberghe; Joshua P. Hacker
Addresses: National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307-3000, USA ' National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307-3000, USA ' National Center for Atmospheric Research, P.O. Box 3000, Boulder, CO 80307-3000, USA
Abstract: Self-organising maps (SOMs) are used to extract typical days from a 30-year long record of 24-hour meteorology and concentration fields. The proposed methodology provides information regarding the probability of a typical time evolution of the concentration patterns (typical days), which could be important when estimating a priori the impact of a potential release of toxic substances. We have run the weather and research forecasting (WRF) model for a defined given month over a 30-year period to generate the required input for the second-order closure integrated puff diffusion model (SCIPUFF). An array for each day including the wind components, boundary layer height and integrated concentration over 24 hours at all the grid points is input to the SOM to perform an iterative learning process. The result is a number of typical days associated with different probabilities of occurrence. An assessment of the performance and reliability of this approach is presented.
Keywords: self-organising maps; SOMs; typical days; dispersion of hazardous materials; weather forecasting and research model.
International Journal of Environment and Pollution, 2017 Vol.62 No.2/3/4, pp.305 - 318
Received: 22 Aug 2016
Accepted: 11 Apr 2017
Published online: 16 Jan 2018 *