skip to main content


This content will become publicly available on September 9, 2025

Title: Aerosol Indirect Effects on Cirrus Clouds Based on Global-Scale Airborne Observations and Machine Learning Models
Cirrus cloud formation and evolution are subject to the influences of thermodynamic and dynamic conditions and aerosol indirect effects (AIEs). This study developed near global-scale in-situ aircraft observational datasets based on 12 field campaigns that spanned from the polar regions to the tropics, from 2008 to 2016. Cirrus cloud microphysical properties were investigated at temperatures ≤ ‑40 °C, including ice water content (IWC), ice crystal number concentration (Ni), and number-weighted mean diameter (Di). Positive correlations between the fluctuations of ice microphysical properties and the fluctuations of aerosol number concentrations for larger (> 500 nm) and smaller (> 100 nm) aerosols (i.e., Na500 and Na100, respectively) were found, with stronger AIE from larger aerosols than smaller ones. Machine learning (ML) models showed that using relative humidity with respect to ice (RHi) as a predictor significantly increases the accuracy of predicting cirrus occurrences compared with temperature, vertical velocity (w), and aerosol number concentrations. The ML predictions of IWC fluctuations showed higher accuracies when larger aerosols were used as an predictor compared with smaller aerosols, indicating the stronger AIE from larger aerosols than smaller ones, even though their AIEs are more similar when predicting the occurrences of cirrus. It is also important to capture the spatial variabilities of large aerosols at smaller scales as well as those of smaller aerosols at coarser scales to accurately simulate IWC in cirrus. These results can be used to improve understanding of aerosol-cloud interactions and evaluate model parameterizations of cirrus cloud properties and processes.  more » « less
Award ID(s):
1744965 1642291
NSF-PAR ID:
10542148
Author(s) / Creator(s):
; ; ; ;
Publisher / Repository:
EGU publications
Date Published:
Journal Name:
Atmospheric chemistry and physics discussion
ISSN:
1680-7375
Subject(s) / Keyword(s):
Cirrus clouds Aircraft-based Observations Machine Learning Aerosol Indirect Effects
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Aerosols affect cirrus formation and evolution, yet quantification of these effects remain difficult based on in-situ observations due to the complexity of nucleation mechanisms and large variabilities in ice microphysical properties. This work employed a method to distinguish five evolution phases of cirrus clouds based on in-situ aircraft-based observations from seven U.S. National Science Foundation (NSF) and five NASA flight campaigns. Both homogeneous and heterogeneous nucleation were captured in the 1-Hz aircraft observations, inferred from the distributions of relative humidity in the nucleation phase. Using linear regressions to quantify the correlations between cirrus microphysical properties and aerosol number concentrations, we found that ice water content (IWC) and ice crystal number concentration (Ni) show strong positive correlations with larger aerosols (> 500 nm) in the nucleation phase, indicating strong contributions of heterogeneous nucleation when ice crystals first start to nucleate. For the later growth phase, IWC and Ni show similar positive correlations with larger and smaller (i.e., > 100 nm) aerosols, possibly due to fewer remaining ice nucleating particles in the later growth phase that allows more homogeneous nucleation to occur. Both 200-m and 100-km observations were compared with the nudged simulations from the National Center for Atmospheric Research (NCAR) Community Atmosphere Model version 6 (CAM6). Simulated aerosol indirect effects are weaker than the observations for both larger and smaller aerosols. Observations show stronger aerosol indirect effects (i.e., positive correlations between IWC, Ni and Na) in the Southern Hemisphere (SH) compared with the Northern Hemisphere (NH), while the simulations show negative correlations in the SH. The simulations underestimate IWC by a factor of 3 – 30 in the early/later growth phase, indicating that the low bias of simulated IWC was due to insufficient ice particle growth. Such hypothesis is consistent with the model biases of lower frequencies of ice supersaturation and lower vertical velocity standard deviation in the early/later growth phases. Overall, these findings show that aircraft observations can capture the competitions between heterogeneous and homogeneous nucleation, and their contributions vary as cirrus clouds evolve. Future model development is also recommended to evaluate and improve the representation of water vapor and vertical velocity on the sub-grid scale to resolve the insufficient ice particle growth. 
    more » « less
  2. Abstract. Aerosols affect cirrus formation and evolution, yet quantificationof these effects remain difficult based on in situ observations due to thecomplexity of nucleation mechanisms and large variabilities in icemicrophysical properties. This work employed a method to distinguish fiveevolution phases of cirrus clouds based on in situ aircraft-basedobservations from seven U.S. National Science Foundation (NSF) and five NASAflight campaigns. Both homogeneous and heterogeneous nucleation werecaptured in the 1 Hz aircraft observations, inferred from the distributionsof relative humidity in the nucleation phase. Using linear regressions toquantify the correlations between cirrus microphysical properties andaerosol number concentrations, we found that ice water content (IWC) and icecrystal number concentration (Ni) show strong positive correlations withlarger aerosols (>500 nm) in the nucleation phase, indicatingstrong contributions of heterogeneous nucleation when ice crystals firststart to nucleate. For the later growth phase, IWC and Ni show similarpositive correlations with larger and smaller (i.e., >100 nm)aerosols, possibly due to fewer remaining ice-nucleating particles in thelater growth phase that allows more homogeneous nucleation to occur. Both200 m and 100 km observations were compared with the nudged simulations fromthe National Center for Atmospheric Research (NCAR) Community AtmosphereModel version 6 (CAM6). Simulated aerosol indirect effects are weaker thanthe observations for both larger and smaller aerosols for in situ cirrus,while the simulated aerosol indirect effects are closer to observations inconvective cirrus. The results also indicate that simulations overestimatehomogeneous freezing, underestimate heterogeneous nucleation andunderestimate the continuous formation and growth of ice crystals as cirrusclouds evolve. Observations show positive correlations of IWC, Ni and icecrystal mean diameter (Di) with respect to Na in both the Northern and SouthernHemisphere (NH and SH), while the simulations show negative correlations inthe SH. The observations also show higher increases of IWC and Ni in the SHunder the same increase of Na than those shown in the NH, indicating highersensitivity of cirrus microphysical properties to increases of Na in the SHthan the NH. The simulations underestimate IWC by a factor of 3–30 in theearly/later growth phase, indicating that the low bias of simulated IWC wasdue to insufficient continuous ice particle formation and growth. Sucha hypothesis is consistent with the model biases of lower frequencies of icesupersaturation and lower vertical velocity standard deviation in theearly/later growth phases. Overall, these findings show that aircraftobservations can capture both heterogeneous and homogeneous nucleation, andtheir contributions vary as cirrus clouds evolve. Future model developmentis also recommended to evaluate and improve the representation of watervapor and vertical velocity on the sub-grid scale to resolve theinsufficient ice particle formation and growth after the initial nucleationevent. 
    more » « less
  3. Abstract

    Three climate models are evaluated using in situ airborne observations from the Southern Ocean Clouds, Radiation, Aerosol Transport Experimental Study (SOCRATES) campaign. The evaluation targets cloud phases, microphysical properties, thermodynamic conditions, and aerosol indirect effects from −40°C to 0°C. Compared with 580‐s averaged observations (i.e., 100 km horizontal scale), the Community Atmosphere Model version 6 (CAM6) shows the most similar result for cloud phase frequency distribution and allows more liquid‐containing clouds below −10°C compared with its predecessor—CAM5. The Energy Exascale Earth System Model (E3SM) underestimates (overestimates) ice phase frequencies below (above) −20°C. CAM6 and E3SM show liquid and ice water contents (i.e., LWC and IWC) similar to observations from −25°C to 0°C, but higher LWC and lower IWC than observations at lower temperatures. Simulated in‐cloud RH shows higher minimum values than observations, possibly restricting ice growth during sedimentation. As number concentrations of aerosols larger than 500 nm (Na500) increase, observations show increases of LWC, IWC, liquid, and ice number concentrations (Nliq, Nice). Number concentrations of aerosols larger than 100 nm (Na100) only show positive correlations with LWC and Nliq. From −20°C to 0°C, higher aerosol number concentrations are correlated with lower glaciation ratio and higher cloud fraction. From −40°C to −20°C, large aerosols show positive correlations with glaciation ratio. CAM6 shows small increases of LWC and Nliqwith Na500and Na100. E3SM shows small increases of Nicewith Na500. Overall, CAM6 and E3SM underestimate aerosol indirect effects on ice crystals and supercooled liquid droplets over the Southern Ocean.

     
    more » « less
  4. null (Ed.)
    Abstract. Cirrus cloud radiative effects are largely affected byice microphysical properties, including ice water content (IWC), ice crystalnumber concentration (Ni) and mean diameter (Di). These characteristics varysignificantly due to thermodynamic, dynamical and aerosol conditions. Inthis work, a global-scale observation dataset is used to examine regionalvariations of cirrus cloud microphysical properties, as well as several keycontrolling factors, i.e., temperature, relative humidity with respect toice (RHi), vertical velocity (w) and aerosol number concentrations (Na).Results are compared with simulations from the National Center forAtmospheric Research (NCAR) Community Atmosphere Model version 6 (CAM6).Observed and simulated ice mass and number concentrations are constrained to≥62.5 µm to reduce potential uncertainty from shattered ice indata collection. The differences between simulations and observations arefound to vary with latitude and temperature. Comparing with averagedobservations at ∼100 km horizontal scale, simulations arefound to underestimate (overestimate) IWC by a factor of 3–10 in theNorthern (Southern) Hemisphere. Simulated Ni is overestimated in mostregions except the Northern Hemisphere midlatitudes. Simulated Di isunderestimated by a factor of 2, especially for warmer conditions(−50 to −40 ∘C), possibly due tomisrepresentation of ice particle growth/sedimentation. For RHi effects, thefrequency and magnitude of ice supersaturation are underestimated insimulations for clear-sky conditions. The simulated IWC and Ni show bimodaldistributions with maximum values at 100 % and 80 % RHi, differing fromthe unimodal distributions that peak at 100 % in the observations. For weffects, both observations and simulations show variances of w (σw) decreasing from the tropics to polar regions, but simulations show muchhigher σw for the in-cloud condition than the clear-sky condition.Compared with observations, simulations show weaker aerosol indirect effectswith a smaller increase of IWC and Di at higher Na. These findings provide anobservation-based guideline for improving simulated ice microphysicalproperties and their relationships with key controlling factors at variousgeographical locations. 
    more » « less
  5. Cirrus cloud radiative effects are largely affected by ice microphysical properties, including ice water content (IWC), ice crystal number concentration (Ni) and mean diameter (Di). These characteristics vary significantly due to thermodynamic, dynamical and aerosol conditions. In this work, a global-scale observation dataset is used to examine regional variations of cirrus cloud microphysical properties, as well as several key controlling factors, i.e., temperature, relative humidity with respect to ice (RHi), vertical velocity (w), and aerosol number concentrations (Na). Results are compared with simulations from the National Center for Atmospheric Research (NCAR) Community Atmosphere Model version 6 (CAM6). The differences between simulations and observations are found to vary with latitude and temperature. Specifically, simulations are found to underestimate IWC by a factor of 5–30 in all regions. Simulated Ni is overestimated in most regions except Northern Hemisphere midlatitude and polar regions. Simulated Di is underestimated, especially for warmer conditions (−50 °C to −40 °C) and higher Na, possibly due to less effective ice particle growth/sedimentation and weaker aerosol indirect effects, respectively. For RHi effects, the frequency and magnitude of ice supersaturation is underestimated in simulations for clear-sky conditions, and the simulated IWC and Ni show maximum values at 80 % RHi instead of 110 % as observed. For w effects, both observations and simulations show variances of w (σw) decreasing from tropics to polar regions, but simulations show much higher σw for in-cloud condition than clear-sky condition. These findings provide an observation-based guideline for improving simulated ice microphysical properties and their relationships with key controlling factors at various geographical locations. 
    more » « less