Lagrangian particle dispersion models (LPDMs) in backward mode are widely-used to quantify the impact of transboundary pollution on downwind sites. Most LPDM applications assume mixing of surface emissions in a boundary layer that is constant in height. The height of this mixing layer (ML), however, is subject to strong spatio-temporal variability. Neglecting this variability may introduce substantial errors in the quantification of source contributions. Here, we perform backward trajectory simulations with the FLEXPART model starting at Cyprus to quantify these errors. The simulations calculate the sensitivity to emissions of upwind pollution sources within the ML height. The emission sensitivity is used to quantify source contributions at the receptor and support the interpretation of ground measurements carried out during the CYPHEX campaign in July 2014. It is determined by two interacting factors: the dilution of pollutants within the ML and the number of trajectories impacted by the emissions. In this study, we calculate the emission sensitivity for a constant ML height of 300 m and a dynamical ML height to compare the resulting differences. The results show that the impact of emission sources is predominantly overestimated by the neglected dilution in expanded daytime ML heights. There is, however, substantial variability in the simulated differences. For shallow marine or nocturnal ML heights, for example, a ML assumed to high may lead to an underestimation of the intensive concentrations. This variability is predominantly caused by the spatio-temporal changes in ML heights and the meteorological conditions that drive the dispersion of the trajectories. In an application example, the impact of CO emissions from hypothetical forest fires is simulated and source contributions are compared for different ML heights. The resulting difference shows that the 300 m overestimates the total CO contributions from upwind sources by 16 %. Thus, it is recommended to generally implement a dynamic mixing layer height parametrization in LPDMs to prevent these errors.