An open platform for Aerosol InfraRed Spectroscopy analysis-AIRSpec.

AIRSpec is a platform consisting of several chemometric packages developed for analysis of Fourier Transform Infrared (FTIR) spectra of atmospheric aerosols. The packages are accessible through a browser-based interface, which also generates the necessary input files based on user interactions for provenance management and subsequent use with a command-line interface. The current implementation includes the task of baseline correction, organic functional group (FG) analysis, and multivariate calibration for any analyte with absorption in the mid-infrared. The baseline correction module uses smoothing splines to correct the drift of the baseline of ambient aerosol spectra given the variability in both environmental mixture composition and substrates. The FG analysis is performed by fitting individual Gaussian line shapes for alcohol COH, carboxylic COOH, alkane CH, carbonyl CO and amine NH for each spectrum. The multivariate calibration model uses the spectra to estimate the concentration of relevant target variables (e.g., organic or elemental carbon) measured with different reference instruments. In each of these analyses, AIRSpec receives spectra and user choices on parameters for model computation; input files with parameters that can later be used with a command-line interface for batch computation are returned together with diagnostic figures and tables in text format. AIRSpec is built using the open source software consisting of R and Shiny and is released under the GNU Public License v3. Users can download, modify, and extend the package, or access its functionality through the web application (http://airspec.epfl.ch) hosted at the École Polytechnique Fédérale de Lausanne (EPFL). AIRSpec provides a unified framework by which different chemometric techniques can be shared and accessed, and its underlying suite of packages provides the basic functionality for extending the platform with new types of analyses. For example, basic functionality includes operations for populating and accessing spectra residing in in-memory arrays or relational databases, input and output of spectra and results of computation, and user interface development. This paper describes the modular architecture and provides examples of the implemented packages using the spectra of aerosol samples collected on PM2.5 Polytetrafluoroethylene (Teflon) filters. [ABSTRACT FROM AUTHOR]