Two models for predicting near-surface flux and concentration footprints are compared concerning their performance in the presence of heterogeneous surface conditions. One of the models is a conventional Lagrangian backward model and the second an LES model with an embedded Lagrangian footprint model. The latter model reveals generation of thermally induced secondary circulation under such surface heterogeneity. The conventional Lagrangian backward model with simple parameterization of flow conditions mostly performs well in footprint predictions for concentrations and somewhat worse for fluxes, and fails only in areas where flow patterns are dominated by pronounced secondary circulations.