Cirrus clouds, consisting entirely of ice crystals in the upper troposphere, cover approximately 25% of the globe at any given time. These clouds produce a net warming effect through substantial longwave trapping, despite reflecting some solar radiation. Ice crystal formation occurs through two pathways: homogeneous nucleation, which requires temperatures below -38 ℃ and high ice supersaturation, or heterogeneous nucleation on ice-nucleating particles (INPs) such as mineral dust and smoke at lower supersaturation levels. These distinct formation mechanisms result in different microphysical properties: heterogeneous nucleation typically produces fewer, larger crystals and optically thinner clouds, while homogeneous nucleation generates numerous small crystals that enhance infrared heating. Although the upper troposphere above oceanic regions is generally considered pristine, Asian dust transported over long distances can reach altitudes of 5?15 km and periodically seed cirrus formation in remote areas, including the mid-North Atlantic. However, the quantitative impact of such transported dust on ice formation—and consequently on cirrus radiative forcing—remains inadequately understood due to limited direct in-situ observations. This study examines representative oceanic cases to determine the relative prevalence of heterogeneous versus homogeneous nucleation in dust-influenced cirrus, providing essential constraints for climate models and reducing uncertainty in cirrus feedbacks.

In this study, we first interrogate CALIPSO/CALIOP profiles to locate dust?cirrus co-occurrences along the trans-Pacific/North American corridor into the North Atlantic. We use CALIOP Level-1 (v4.51) 532 nm attenuated backscatter and volume depolarization ratio to flag layered features, and Level-2 aerosol profiles (v4.2) to obtain extinction, particle depolarization, and the vertical feature mask for subtype identification (pure/polluted dust and cirrus). Because automated classification can confuse optically thin cirrus with dust, we apply additional screening based on depolarization thresholds and layer continuity, and collocate meteorology (temperature, pressure, relative humidity) from embedded MERRA-2 fields. To characterize cirrus microphysics, we use A-Train DARDAR products that combine CloudSat CPR with CALIOP. DARDAR-Cloud provides extinction, effective radius, and ice water density; DARDAR-Nice supplies ice crystal number concentration (ICNC) profiles for ice crystal diameters greater than 5, 25, and 100 μm at 60 m vertical and 1.7 km horizontal resolution. Recognizing lidar?radar size-sensitivity differences and known retrieval limitations, we assign a conservative factor-of-three uncertainty to ICNC. Dust-related ice-nucleating particles concentration (INPC) is derived with POLIPHON using dust extinction as input. Dust extinction is separated from total aerosol extinction via a one-step method and computed assuming a dust lidar ratio of 45 sr. We convert extinction to dust mass, coarse-mode number, and surface area concentrations with regionally constrained coefficients from nearby AERONET observations (Tudor Hill). Given the very low cirrus temperatures, we restrict nucleation to deposition mode and calculate INPC with the U17-D parameterization using dust surface area as predictor. We define dust?cirrus interaction when vertical or horizontal overlap is evident. Event provenance is corroborated with daily MERRA-2 dust column mass fields and HYSPLIT back-trajectories (GDAS-driven), linking the observed North Atlantic layers to Asian sources and establishing two rare, long-range transport cases for subsequent INPC?ICNC closure.

In the first event (May 18, 2007), CALIOP reveals a cirrostratus deck embedded within a pure dust layer between 9.4 and 10.9 km, exhibiting strong attenuated backscatter and high depolarization that indicate nonspherical ice and dust (Fig. 1). HYSPLIT traces the air mass to Asian deserts on May 9?10, with elevated layers tracked across the Pacific, over North America, and into the North Atlantic (Figs. 2?3). Within the overlap zone, DARDAR-Cloud retrievals demonstrate mean extinction near 0.10 km^-1, an effective radius of 39.2 μm, and an ice water density of 2.0 mg·m^-3 (Fig. 4). Using POLIPHON-derived dust extinction with the U17-D scheme, the dust INPCs measure 26.0 L^-1 at an ice saturation ratio of 1.15 and 483.5 L^-1 at 1.25, increasing toward the cloud top (Fig. 5). Collocated DARDAR-Nice indicates mean ICNCs of 32.3 L^-1 for diameters larger than 5 μm, 15.3 L^-1 for larger than 25 μm, and 2.4 L^-1 for larger than 100 μm, peaking approximately 0.3 km below the top (Fig. 5). Excluding the 1.25 case—where dust INPC exceeds ICNC, likely due to sedimentation and particle-size assumptions—the closure remains within one order of magnitude, supporting dust-driven heterogeneous nucleation. In the second event (April 25, 2008), a smaller cirrus system between 10.0 and 11.4 km exists adjacent to a polluted or aged dust layer. The aerosol exhibits weaker backscatter and lower volume depolarization ratio (0.1?0.2), while the cirrus maintains higher volume depolarization ratio above 0.3 (Fig. 6). Trajectories again suggest Asian sources and sustained lofted transport between roughly 7 and 11 km (Figs. 7?8). Microphysical retrievals indicate mean extinction near 0.25 km^-1, an effective radius of 56.9 μm, and an ice water density of 8.9 mg·m?3 (Fig. 9). Under temperatures of -54 to -57 ℃ and typical ice saturation ratios of 1.15?1.25, the U17-D scheme, reduced tenfold to reflect aging, yields mean dust INPC from 4.4 to 95.7 L^-1 (Fig. 10). The corresponding ICNC averages are 68.9 L^-1 (>5 μm), 34.1 L^-1 (>25 μm), and 6.6 L^-1 (>100 μm), with localized enhancements in nice,5μm approaching 500 L^-1 (Figs.9?10). The near-closure again indicates heterogeneous nucleation dominance; brief high-ICNC pockets likely reflect transient updrafts or partial INP depletion. These two trans-Pacific cases demonstrate that Asian dust maintains ice-nucleating activity after extensive transport and can significantly influence cirrus formation over the North Atlantic, with pure dust active at lower saturation and polluted dust requiring slightly higher saturation to produce comparable ice.

To reduce uncertainties in cloud radiative forcing and climate projections, models must explicitly represent how Asian dust is lofted by midlatitude westerlies, traverses the Pacific, and enters the North Atlantic, while maintaining ice-nucleation activity after long-range transport. Clay minerals such as kaolinite and illite provide efficient ice-nucleating sites at relatively warmer upper-tropospheric temperatures, increasing initial ice crystal numbers, shifting sizes toward larger particles, rapidly consuming supersaturation, and inhibiting subsequent homogeneous freezing. These modifications affect cirrus effective radius, ice water density, and the balance between longwave cooling and shortwave reflection, potentially influencing regional heat budgets and circulation. Therefore, global and regional models should incorporate prognostic dust mineralogy and aging-dependent INP efficiency, resolve competition between heterogeneous and homogeneous pathways under realistic updrafts, and evaluate against combined lidar?radar constraints (e.g., CALIPSO/CloudSat and EarthCARE) and reanalysis dust fields. Additional marine INP sources (sea salt, smoke, volcanic aerosols) should also be considered. Our North Atlantic cases demonstrate that dust-driven heterogeneous nucleation can dominate cirrus formation, emphasizing the importance of including these processes to reduce radiative-forcing uncertainty.