Novel informatic software for automated aerosol component property predictions and ensemble predictions for direct model - measurement comparison

点击翻译按钮获取中文摘要

Atmospheric aerosol particles, or particulate matter suspended in the atmosphere, are highly important yet highly uncertain components of the earths climate system and key determinants of air quality. Properties which determine these highly uncertain impacts are linked at the most fundamental level to the chemical components which may reside in the particle. Both inorganic and organic material can transfer between the gas and particle phase. Inorganic material is restricted to a few well-understood compounds. However, organic material can comprise many thousands, as yet largely unidentified, compounds with a vast range of properties. Owing to the complexity and diversity of atmospheric aerosol components, quantification of the properties that determine their highly uncertain climatic and human health impacts requires the development and application of novel technological applications such as the informatic software proposed here. Firstly, we must be able to predict how ever many thousands of components can exist in particulate matter. Specifically, predicting the evolution of aerosol requires calculation of the distribution of all components between the gas and particle phases which in turn requires knowledge of all component vapour pressures and other thermodynamic properties. Furthermore, the physical properties of the aerosol determining their climatic impacts require detailed knowledge of fundamental properties of all components. The many thousands of individual aerosol components ensure that explicit manual calculation of these properties is laborious, time-consuming and often impossible. Thus, automation is necessary. Secondly, to identify key components and resolve their environmental impacts we must be able to replicate chemical characteristics measured in real/simulated atmospheres. A comprehensive experimental determination of individual organic components of atmospheric aerosols is not available, leading to indirect measurements on 'chemical signatures' of mixtures. Through automation of component property estimation, combined with a gas/aerosol transfer model, these 'chemical sigmatures' as determined by state-of-the-science atmospheric sampling instrumentation will be predicted. This will be achieved by calculating instrument response functions with the predicted abundance of all components. Again, the prediction (and combination) of instrument response functions for each individual component lends itself to automation due to the vast numbers involved. The informatics suite will be built using a flexible high-level portable programming language and an open source chemical informatics package that is designed to allow extraction of appropriate sub-molecular information relevant for each property estimation method.