Abstract. Several modeling studies reported elevated carbon emissions from historical land use change (LUC) by including bi-directional transitions at the sub-grid scale (termed gross land use change). This has implication on the estimation of so-called residual land CO2 sink over undisturbed lands. However, in most dynamic global vegetation models (DGVM), forests and/or other land use types are represented with a single sub-grid tile, without accounting for secondary lands that are often involved in shifting cultivation or wood harvest. As a result, land use change emissions (ELUC) are likely overestimated, because it is high-biomass mature forests instead of low-biomass secondary forests that are cleared. Here we investigated the effects of including sub-grid forest age dynamics in a DGVM on historical ELUC over 1501–2005. We run two simulations, one with no forest age (Sageless) and the other with sub-grid secondary forests of different age classes whose demography is driven by historical land use change (Sage). Estimated global ELUC for 1501–2005 are 179 Pg C in Sage compared to 199 Pg C in Sageless. The lower emissions in Sage arise mainly from shifting cultivation in the tropics, being of 27 Pg C in Sage against 46 Pg C in Sageless. Estimated cumulative ELUC from wood harvest in the Sage simulation (31 Pg C) are however slightly higher than Sageless (27 Pg C), because secondary forests simulated in Sage are insufficient to meet the prescribed harvest area, leading to the harvest of old forests. This result depends on pre-defined forest clearing priority rules given a simulated portfolio of differently aged forests in the model. Our results highlight that although gross land use change as a former missing emission component is included by a growing number of DGVMs, its contribution to overall ELUC tends to be overestimated, unless low-biomass secondary forests are properly represented.