Accounting for the chemical processes in large-scale Computational Fluid Dynamic (CFD) simulations is essential to understand industrial reacting flows. The implementation of a detailed kinetic mechanism, involving a large number of chemical species and elementary reactions, can be challenging and may require unrealistic computational resources. In addition, the level of detail provided by these comprehensive mechanisms might be excessive, especially when accounting for the modeling uncertainties involved in the simulations. Proposed below is a systematic strategy to reduce a detailed kinetic model into a global model that will: (i) contain many fewer reaction steps, (ii) use lumped variables that combine species of similar chemical nature, and (iii) maintain the predictive capabilities of the detailed mechanism for the quantities of interest. The approach is demonstrated here in the context of biomass gasification. Partially stirred reactor (PaSR) models are used in conjunction with a detailed chemical kinetic mechanism to generate a database of gas phase detailed compositions likely to occur in an actual reactor. From the analysis of this database, representative lumped species are identified, for which effective elemental formula, molecular weight, and thermodynamical properties are determined. An appropriate set of global reactions describing the evolution of these lumped species is then proposed, whose rate coefficients in Arrhenius format are fitted to match production rates formed from the sampled detailed compositions and detailed kinetic scheme. Validation is performed by comparing the PaSR dynamics predicted using the detailed and global models.