Analysing atmospheric observational data, consisting of high numbers of variables monitored simultaneously over long periods, producing very large datasets and extended time series, often involving novel instrumentation for which no satisfactory off the shelf software packages are available. Consolidation and synchronisation of different time-series. Extensive use of multivariate data analysis and statistical methods for exploratory data analysis, information extraction and data visualisation. Coding of efficient algorithms in Matlab, R, Excel or SAS to enable processing of data using machine-learning techniques and semi-automated processing of observational data stream. Using clustering and classification techniques, e.g. for identifying profiles of chemical species arising from different pollution sources, or recognising individual particles/samples by chemical profiles. Using predictive analytics techniques, e.g. for identifying sources of atmospheric pollution and producing source contribution estimates, or relating chemical content of environmental samples to toxicity. Calibration and Spectral analysis/multivariate analysis of Fourier-transform infrared spectroscopy data. Keywords include principal component analysis (pca), positive matrix factorisation (pmf), gradient descent optimisation, uncertainty weighting, factor rotation, independent component analysis, time-series analysis, linear and partial least squares (pls) regression, logistic regression, neural networks, clustering, simca, k-means, discriminant analysis, classification trees, etc.