Multivariate receptor models—current practice and future trends

Multivariate receptor models have been applied to the analysis of air quality data for sometime. However, solving the general mixture problem is important in several other fields. This paper looks at the panoply of these models with a view of identifying common challenges and projecting future developments. The challenges common to all fields are: (1) the need for better and less subjective methods to estimate the number of factors to include in the model; (2) better documentation of the algorithms in the open literature; and (3) user-friendly software to enlarge the user community. Future trends will likely center on the development of models based on self-modeling curve resolution (SMCR) and constrained weighted least squares methods such as positive matrix factorization.