ARTICLES https://doi.org/10.1038/s41561-019-0379-6 1 Department of Atmospheric and Oceanic Sciences, Laboratory for Atmospheric and Space Physics, University of Colorado, Boulder, CO, USA. 2 Department of Atmospheric and Planetary Science, Hampton University, Hampton, VA, USA. 3 Space Science Institute, Boulder, CO, USA. *e-mail: victoria.hartwick@gmail.com M iddle atmosphere ice clouds are observed year-round 1,2 and in 30–50% of the solar and stellar occultation measure- ments of the Martian middle atmosphere 3–8 (here defined as ~30–60 km above the surface or at pressures between ~100 and 1 Pa). Above 65 km, cloud layers are probably CO 2 ice 7–10 . However, between 30 and 60 km, water ice clouds are also observed 5,8,11–13 . General circulation models struggle to replicate the observed verti- cal distribution of water ice clouds and typically confine cloud lay- ers to pressures higher than approximately 100 Pa (heights below ~25 km) 14–16 . Inaccurate representations of clouds directly and indi- rectly impact the simulated climate. Optically thin clouds, particu- larly at high altitudes where the density scaled opacity is greatest 17 , can be a significant source of temperature variability that impacts the diurnal and semidiurnal thermal waves 18–21 and the large-scale meridional overturning circulation 14,17 . To accurately simulate the processes that control the nucleation, growth and evolution of mid- dle atmosphere clouds is, therefore, fundamental to capturing the true complexity of the Martian climate cycle 16 . The prevailing view of ice cloud formation on Mars is that sur- face mineral dust mixes aloft and serves as heterogeneous ice nuclei when the air is supersaturated with respect to water ice 22 . Although supersaturation is observed 23 and modelled 17 at high altitudes, the models predict limited nuclei above the planetary boundary layer. Few cloud particles nucleate in these models, and those that do nucleate grow quickly to large sizes and gravitationally sediment to lower altitudes. Some recent models include parameteriza- tions to enhance the vertical mixing of mineral dust; these include topographic upslope flows 24,25 and enhanced convective updrafts in ‘rocket dust storms’ 26 . These parameterizations enhance ice nuclei concentrations in bursts that should decay over several sols (Martian days) and therefore cannot easily explain the persistence of high cloud features. On Earth, optically thin noctilucent clouds nucleate on mete- oric smoke particles 27,28 . If a sufficient mass of interplanetary dust particles (IDPs) ablates in the Martian atmosphere, cloud formation by the same mechanism is likely. The MAVEN IUVS (Imaging Ultraviolet Spectrograph) instrument identified a persistent meteoroid ablation layer near ~80–90 km (ref. 29 ). Chemical ablation models predict that ~5–10% of the total intercepted meteoric mass will ablate and recondense 30 . However, as recondensed particles are small, their numbers are large even when the ablation rate is low. Meteoric smoke particles, therefore, represent an abundant and likely source of ice nuclei for cloud nucleation at high altitudes and in lower regions where the atmospheric load of surface mineral dust is depleted. In this study, we investigated the impact on water ice cloud for- mation of including a new ablated IDP fluence of 0.4 ton per Mars sol distributed evenly across our model top 30 . Observations with the MAVEN IUVS instrument 29 found an IDP fluence of 2–3–ton sol −1 . As discussed in the Supplementary Information, our model top is near 60 km, which is about 25 km below the observed ablation level. To account for coagulation and chemical processes that occur above our model top, we distributed the total ablated mass into the predicted 50 km size distribution for meteoric smoke or ‘dirty ice’ particles as illustrated by Plane et al. 30 . We compare the results with simulations that include no meteoric smoke. We simulated clouds using a size-resolving aerosol model cou- pled with a three-dimensional general circulation model for Mars (MarsCAM-CARMA). We demonstrated that IDPs are necessary to reproduce the vertical and horizontal distribution of the observed water ice clouds in the middle atmosphere. To capture the impact of IDPs under different dynamical regimes, we show results at both equinox (heliocentric longitude (L s = 180°)) and solstice (L s = 270°) as these represent extremes of the climate. We note that clouds form in cold troughs of the diurnal and semidiurnal tide, and, as a result, clouds in the middle atmosphere may act as signposts for local dynamics 19,20 . We show that water ice clouds are radiatively active and influence local and large-scale dynamical systems. We close with an analysis of the radiative-dynamical impact of IDP-induced clouds on local thermal tides as well as on the large-scale Hadley circulation. High-altitude water ice cloud formation on Mars controlled by interplanetary dust particles V. L Hartwick  1 *, O. B. Toon 1 and N. G. Heavens  2,3 Submicrometre-size meteoric smoke aggregates form when interplanetary dust particles ablate and re-coagulate in the Martian atmosphere. The MAVEN (Mars Atmosphere and Volatile Evolution) satellite has detected pervasive ionized metallic layers due to meteor ablation at an 80–90 km altitude, which suggests a continuous supply of meteoric smoke particles that settle to lower altitudes. Until now, meteoric smoke has been neglected in general circulation model studies of the formation of Martian water ice clouds. Here we show that when meteoric smoke is included in simulations of the atmospheric circulation on Mars, mesospheric water ice clouds form at low pressures and in discrete layers, polar hood clouds extend to higher altitudes and the seasonal Hadley cell is weakened. Furthermore, we find that the middle atmosphere water ice clouds respond to and influence the diurnal and semidiurnal migrating thermal tides. We conclude that Mars atmospheric simulations that neglect meteoric smoke do not reproduce the observed spatial distribution of water ice clouds and miss crucial radiative impacts on the overall atmospheric dynamics. NATURE GEOSCIENCE | www.nature.com/naturegeoscience