The Probabilistic Collocation Method (PCM) has been proven to be an efficient general method of uncertainty analysis in atmospheric models (Tatang et al 1997, Cohen&Prinn 2011). However, its application has been mainly limited to urban- and regional-scale models and chemical source-sink models, because of the drastic increase in computational cost when the dimension of uncertain parameters increases. Moreover, the high-dimensional output of global models has to be reduced to allow a computationally reasonable number of polynomials to be generated. This dimensional reduction has been mainly achieved by grouping the model grids into a few regions based on prior knowledge and expectations; urban versus rural for instance. As the model output is used to estimate the coefficients of the polynomial chaos expansion (PCE), the arbitrariness in the regional aggregation can generate problems in estimating uncertainties. To address these issues in a complex model, we apply the probabilistic collocation method of uncertainty analysis to the aerosol representation in MOZART-4, which is a 3D global chemical transport model (Emmons et al., 2010). Thereafter, we deterministically delineate the model output surface into regions of homogeneous response using the method of Principal Component Analysis. This allows the quantification of the uncertainty associated with the dimensional reduction. Because only a bulk mass is calculated online in Mozart-4, a lognormal number distribution is assumed with a priori fixed scale and location parameters, to calculate the surface area for heterogeneous reactions involving tropospheric oxidants. We have applied the PCM to the six parameters of the lognormal number distributions of Black Carbon, Organic Carbon and Sulfate. We have carried out a Monte-Carlo sampling from the probability density functions of the six uncertain parameters, using the reduced PCE model. The global mean concentration of major tropospheric oxidants did not show a significant variation in response to the variation in input parameters. However, a substantial variation at regional and temporal scale has been found. Tatang M. A., Pan W., Prinn R G., McRae G. J., An efficient method for parametric uncertainty analysis of numerical geophysical models, J. Gephys. Res., 102, 21925-21932, 1997. Cohen, J.B., and R.G. Prinn, Development of a fast, urban chemistry metamodel for inclusion in global models,Atmos. Chem. Phys., 11, 7629-7656, doi:10.5194/acp-11-7629-2011, 2011. Emmons L. K., Walters S., Hess P. G., Lamarque J. -F., P_ster G. G., Fillmore D., Granier C., Guenther A., Kinnison D., Laepple T., Orlando J., Tie X., Tyndall G., Wiedinmyer C., Baughcum S. L., Kloster J. S., Description and evaluation of the Model for Ozone and Related chemical Tracers, version 4 (MOZART-4). Geosci. Model Dev., 3, 4367, 2010.