Long-range Saharan dust transport affects radiation, clouds, and air quality, yet its vertical structure and entrainment
into the boundary layer remain challenging to simulate. We investigate four contrasting dust events over southwestern
Germany using ground-based lidar, sun-photometer and in situ aerosol measurements, satellite observations, backward trajectories,
and ICON-ART simulations. The events represent diverse atmospheric conditions, including a six-day cloud-free
5 episode, a cloudy episode involving dust–cloud interactions, and an intense outbreak with aerosol optical depth approaching
1.4. A depolarization-based lidar method separated dust backscatter from contributions by other aerosols, enabling a like-forlike
evaluation of the dust represented by ICON-ART and providing a useful approach for evaluating dust models in mixed
aerosol layers. ICON-ART reproduced the large-scale transport and arrival of the dust layers, with mean differences of approximately
20 min in arrival time and 50 m in layer height. Daily re-initialization of the global simulations improved correlations
10 with observed near-surface particulate matter concentrations from 0.65–0.69 to 0.86–0.88 and better reproduced the timing
and persistence of boundary-layer dust. The benefits of higher resolution (2 and 1 km) become more apparent when combined
with model reinitialization. However, ICON-ART consistently overestimated particles of 0.5–2 μm and underestimated particles
larger than 2 μm. These results demonstrate the complementary benefits of accurate meteorological initialization and
convection-permitting resolution while identifying size-dependent emission, sedimentation, and deposition as priorities for
15 model development. By distinguishing errors in dust transport from those arising from aerosol size representation, this study
provides an observation-constrained framework for evaluating and improving convection-permitting dust forecasts.