Riming Leaves a Hidden Isotopic Fingerprint in Rain and Snow Worldwide

Every snowflake and raindrop carries a chemical diary of the cloud it fell from. For decades, scientists have read that diary using the isotopic composition of water—specifically the ratios of hydrogen-2 and oxygen-18—and a derived quantity called deuterium excess, or d-excess. This value has been treated as a messenger from the distant past, telling us about the temperature, humidity and windspeed over the ocean where the moisture that became rain or snow first evaporated. Now a new study published in Atmospheric Chemistry and Physics challenges one of the foundational assumptions behind those readings, and the implications ripple from modern weather radar to the interpretation of ice cores that record Earth’s ancient climates.

A team led by Pradeep K. Aggarwal, formerly of the Isotope Hydrology Section at the International Atomic Energy Agency in Vienna, together with Courtney Schumacher and Aaron Funk of Texas A&M University, Frederick J. Longstaffe of the University of Western Ontario, and Matthew D. Shupe of the University of Colorado and NOAA, examined what happens to d-excess when ice grows by a process called riming. In mixed-phase clouds, where supercooled liquid droplets, ice crystals and vapor coexist at temperatures above roughly minus 38 degrees Celsius, ice can grow in two distinct ways. In the Wegener–Bergeron–Findeisen process, liquid droplets never touch the ice; instead they evaporate, and the resulting vapor diffuses onto the ice crystals, which preferentially incorporate heavier isotopologues and end up with a d-excess as much as 30 per mil higher than the surrounding liquid. Riming is different: here, supercooled droplets collide directly with ice particles and freeze on their surfaces.

The long-standing assumption, embedded in isotope models since the early 1980s, was that riming occurs without isotopic fractionation—that rimed ice simply inherits the isotopic composition of the liquid it accretes. The new analysis shows this assumption is wrong. When droplets freeze onto a falling ice particle, the latent heat of fusion warms the particle surface, and under many conditions a liquid film forms that partially evaporates before freezing is complete. Because the semi-heavy isotopologue H2H16O diffuses more slowly than H218O in air, this evaporation preferentially removes oxygen-18 relative to deuterium from the accreted liquid, driving down the d-excess of the ice that ultimately forms. The result is rimed ice with a d-excess substantially lower than the parent liquid—by as much as 20 to 30 per mil in wet growth conditions.

To test this idea against real weather, the team needed an independent measure of riming intensity, and they found one in radar. Riming makes ice particles denser and rounder, which increases their terminal fall speed: a lightly rimed snowflake might fall at 0.5 to 1 meter per second, while dense graupel exceeds 1.5 meters per second. Vertically pointing Doppler radars measure the mean Doppler velocity of falling hydrometeors, and in stratiform clouds, where small upward and downward air motions cancel out when averaged over 20 to 30 minutes, this velocity approximates the terminal fall speed. The researchers correlated mean Doppler velocity with the d-excess of precipitation collected at six sites spanning the globe: Summit, Greenland; Ny-Ålesund and Andenes in Norway; Dumont d’Urville in Antarctica; Cazadero, California; and Rio Claro, Brazil.

The result was strikingly consistent. Across an observed d-excess range from minus 23 to plus 45 per mil, d-excess decreased as fall velocity increased at every location except winter precipitation at Summit. Pooling 211 samples from all sites except Summit, a standardized regression yielded a strong correlation coefficient of minus 0.68, with the slope negative at each individual site and tightly constrained at the best-sampled locations. The team converted fall velocities into rimed mass fractions using a published parameterization and found that each 10 percent increase in rimed mass corresponds to a d-excess decrease of roughly 8 per mil. Because the physics of droplet freezing is the same everywhere, this relationship holds from the tropics to the poles—a universality that argues the correlation reflects a genuine microphysical mechanism rather than a site-specific coincidence.

Skeptics might wonder whether sub-cloud evaporation of raindrops, which also lowers d-excess, could explain the pattern. The researchers addressed this head-on. For rainfall sites, Doppler velocities were taken from the snow region above the melting layer, well inside the cloud and untouched by sub-cloud processes. Calculations with an established raindrop evaporation model showed that evaporation could have lowered d-excess by only about 2 per mil during the key Cazadero event, and less than 1 per mil at Rio Claro where surface humidity exceeded 90 percent. Moreover, comparing fall velocities above and below the melting layer showed that the lowest d-excess rain samples plotted in the graupel and heavily rimed fields—exactly where riming, not evaporation, would place them.

Radiosonde profiles of temperature and humidity added independent atmospheric context. At Summit, winter temperatures throughout the precipitating column fell below minus 38 degrees Celsius, ruling out mixed-phase conditions and riming entirely; the low winter d-excess there instead arises from vapor deposition at very cold temperatures from vapor strongly depleted in oxygen-18, under high ice supersaturation. In summer, humidity near the cloud base approached liquid saturation at temperatures between minus 10 and minus 4 degrees Celsius, precisely the conditions for wet-growth riming. At Ny-Ålesund and Dumont d’Urville, the profiles revealed a two-layer structure: an upper, ice-supersaturated layer where vapor deposition dominates, and a lower, near-water-saturated layer where riming takes over. Case studies of individual storms traced the full spectrum, from a deposition-dominated event at Ny-Ålesund with d-excess of plus 34.5 per mil to a heavily rimed rainfall at minus 4.6 per mil.

The team then built a process-weighted framework treating precipitation d-excess as a mass-weighted mixture of four end-members: vapor deposition, diamond dust, wet-growth riming and dry-growth riming. Mixtures containing 20 to 80 percent vapor deposition, with the remainder riming, reproduced the observed d-excess ranges at all the maritime polar, mid-latitude and tropical sites, while pure vapor deposition reached the highest observed values near plus 44 per mil. At Summit in winter, mixtures of vapor deposition and diamond dust accounted for the observed range without any riming at all. The calculations demonstrate that in-cloud microphysics alone can generate the full spread of d-excess values seen in precipitation worldwide.

The consequences reach deep into paleoclimate science. Lower d-excess in Greenland and Antarctic ice cores during the Last Glacial Maximum, and during abrupt climate swings, has been interpreted as evidence that the geographical source of moisture shifted dramatically. The new findings suggest an alternative: glacial conditions meant colder precipitating columns, more isotopically depleted vapor and higher ice supersaturation—precisely the combination that suppresses d-excess through in-cloud fractionation, no source shift required. Similarly, the north-to-south d-excess gradient across Greenland and the contrast between coastal and interior Antarctic snow, traditionally attributed to moisture origin, could partly reflect systematic differences in riming intensity and deposition conditions along those transects. Even the low d-excess of tropical rainfall, usually blamed on sub-cloud evaporation or the amount effect, may carry a riming signal born in the ice phase high above the surface.

Practically, the work hands climate scientists a new tool. Where radar and isotope observations coexist, the d-excess–riming relationship offers a way to test whether the microphysics schemes in climate models—which currently treat riming as isotopically neutral—produce realistic riming rates. Where only isotope data exist, d-excess can be inverted to estimate the rimed mass fraction, a quantity otherwise nearly impossible to obtain from the ground. After more than forty years of assuming that freezing droplets leave isotopes untouched, the field now has both a reason and a method to look again at one of paleoclimatology’s most trusted proxies.

Subject of Research: Isotopic fractionation during riming ice growth in mixed-phase clouds and its effect on the deuterium excess of precipitation

Article Title: Isotopic fractionation during ice growth by riming and its effect on the d-excess of precipitation

Article References: Aggarwal, P. K., Schumacher, C., Longstaffe, F. J., Funk, A., & Shupe, M. D. (2026). Isotopic fractionation during ice growth by riming and its effect on the d -excess of precipitation. Atmospheric Chemistry and Physics, 26(19), 13661-13692. https://doi.org/10.5194/acp-26-13661-2026

Image Credits: AI Generated

DOI: 10.5194/acp-26-13661-2026

Keywords: deuterium excess, riming, mixed-phase clouds, isotopic fractionation, ice cores, paleoclimate, Doppler radar, cloud microphysics, precipitation, vapor deposition, graupel, climate models