Smaller global and regional carbon emissions from gross land use change when considering sub-grid secondary land cohorts in a global dynamic vegetation model

Abstract. Several modelling studies reported elevated carbon emissions from historical land use change (E_LUC) by including bidirectional transitions on the sub-grid scale (termed gross land use change), dominated by shifting cultivation and other land turnover processes. However, most dynamic global vegetation models (DGVMs) that have implemented gross land use change either do not account for sub-grid secondary lands, or often have only one single secondary land tile over a model grid cell and thus cannot account for various rotation lengths in shifting cultivation and associated secondary forest age dynamics. Therefore, it remains uncertain how realistic the past E_LUC estimations are and how estimated E_LUC will differ between the two modelling approaches with and without multiple sub-grid secondary land cohorts – in particular secondary forest cohorts. Here we investigated historical E_LUC over 1501–2005 by including sub-grid forest age dynamics in a DGVM. We run two simulations, one with no secondary forests (S_ageless) and the other with sub-grid secondary forests of six age classes whose demography is driven by historical land use change (S_age). Estimated global E_LUC for 1501–2005 is 176 Pg C in S_age compared to 197 Pg C in S_ageless. The lower E_LUC values in S_age arise mainly from shifting cultivation in the tropics under an assumed constant rotation length of 15 years, being 27 Pg C in S_age in contrast to 46 Pg C in S_ageless. Estimated cumulative E_LUC values from wood harvest in the S_age simulation (31 Pg C) are however slightly higher than S_ageless (27 Pg C) when the model is forced by reconstructed harvested areas because secondary forests targeted in S_age for harvest priority are insufficient to meet the prescribed harvest area, leading to wood harvest being dominated by old primary forests. An alternative approach to quantify wood harvest E_LUC, i.e. always harvesting the close-to-mature forests in both S_ageless and S_age, yields similar values of 33 Pg C by both simulations. The lower E_LUC from shifting cultivation in S_age simulations depends on the predefined forest clearing priority rules in the model and the assumed rotation length. A set of sensitivity model runs over Africa reveal that a longer rotation length over the historical period likely results in higher emissions. 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 E_LUC remains uncertain and tends to be overestimated when models ignore sub-grid secondary forests.