Dynamic global vegetation models (DGVMs) are widely used to assess the impacts of climate change on forests, but their projections of future forest dynamics remain uncertain. A key source of this uncertainty is the representation of tree mortality, particularly under drought conditions. Improving drought-related mortality schemes in DGVMs is therefore essential for more reliable projections of future forest dynamics. To address this, we extended the DGVM LPJmL-5 by implementing a new drought-induced mortality formulation based on water stress, vapour pressure deficit, and phenology. This scheme introduces two Plant Functional Type (PFT)-specific parameters, <em>α</em> and <em>β</em>, which control the onset and severity of drought-induced tree mortality. These parameters were calibrated using a multi-objective optimisation approach that considered both observed drought–mortality and drought-Gross Primary Production (GPP) relationships. We find that the new formulation improves the simulated drought–mortality relationship compared to the baseline model, particularly for tropical and temperate PFTs. While large-scale mean GPP and latent heat fluxes remain largely unchanged, the new scheme leads to stronger drought responses and impacts in several regions and reduces the strong post-drought recovery in GPP found in the baseline formulation of LPJmL-5. Overall, the results show that this relatively simple drought mortality formulation improves the representation of drought-induced mortality responses in LPJmL-5 while maintaining realistic large-scale carbon and water fluxes. Including drought-induced mortality therefore represents an important step towards more realistic simulations of vegetation responses to increasing drought stress under future climate change.