This work presents a numerical study of biomass pyrolysis in turbulent riser flow. Eulerian Lagrangian simulations of unbounded sedimenting gas-solid flows are performed to isolate the effects of particle clustering on the production of syngas and tar. This configuration provides a framework to resolve the relevant length- and time-scales associated with thermal, chemical and multiphase processes taking place in the fully-developed region of a circulating fluidized bed riser. A four-step kinetic scheme is employed to model the devolatilization of biomass particles and secondary cracking of tar. Two-way coupling between the phases leads to clusters of sand particles that generate and sustain gas-phase turbulence and transport biomass particles. Neglecting the heterogeneity caused by clusters was found to lead to a maximum over-prediction of syngas yield of 33%. Further, it was found that two-dimensional simulations over-predict the level of clustering, resulting in an under-prediction of syngas and tar yields. (C) 2019 Elsevier Ltd. All rights reserved.