Using satellite observations to evaluate model representation of Arctic mixed-phase clouds

Clouds play an important role in determining Arctic warming, but remain difficult to constrain with available observations. We use two satellite-derived cloud phase metrics to investigate the vertical structure of Arctic clouds in global climate models that use the Community Atmosphere Model version 6 (CAM6) atmospheric component. We produce a set of constrained model runs by adjusting model microphysical variables to match the cloud phase metrics. Models in this small ensemble have variable representation of cloud amount and phase in the winter, while uniformly underestimating total cloud cover in the spring and overestimating it in the summer. We find a consistent correlation between winter and spring cloud cover simulated for the present-day and the longwave cloud feedback parameter.

small ensemble uniformly overestimate total cloud fraction in the summer, but have vari-23 able representation of cloud fraction and phase in the winter and spring. By relating mod-24 elled cloud phase metrics and changes in low-level liquid cloud amount under warming 25 to longwave cloud feedback, we show that mixed-phase processes mediate the Arctic cli-26 mate by modifying how wintertime and springtime clouds respond to warming. liquid-rich cloud tops with their icy interiors. We describe a significant model error that 37 limits the formation of new ice crystals. We also find that global climate models repro-38 duce observations, and that a range of model parameters produce results consistent with 39 observations. Changes in cloud fraction resulting from these adjustments mostly occur 40 in the winter and spring, and cause the models to trap longwave radiation differently.

41
The results of this study highlight the need to capture seasonal changes in cloud phase 42 and amount in order to successfully predict future changes to the Arctic climate. ] 43 At temperatures between approximately −37°C and 0°C, cloud ice forms via het-63 erogeneous nucleation processes that are dependent on temperature, in-cloud vapor pres-64 sure, and the presence of ice nucleating particles (INPs) (Korolev, 2007). Cloud ice and 65 water can coexist as mixed-phase clouds in this regime. The fraction of supercooled liq-66 uid water in a mixed-phase cloud layer can be referred to as the supercooled liquid frac-67 tion (SLF) (Komurcu et al., 2014). Observations show that Arctic mixed-phase clouds 68 are both common and long-lived (Matus & L'Ecuyer, 2017;H. Morrison et al., 2012), 69 due in part to their vertical structure in which INP-limited liquid cloud tops are sepa-70 rated from glaciated interiors, preventing ice from quickly depleting cloud water and al-71 lowing clouds to persist for several days (Hobbs & Rangno, 1998). Through this effect 72 on cloud lifetime and opacity, cloudtop phase mediates the resulting long-and shortwave 73 cloud feedbacks.

74
Model representation of mixed-phase clouds relies on uncertain parameters. Sev-75 eral studies of mixed-phase clouds in Version 5 of the Community Atmosphere Model 76 (CAM5) have found that the Wegener-Bergeron-Findeisen (WBF) process time scale 77 has the largest role in determining liquid cloud fraction (Tan & Storelvmo, 2016;McIl-78 hattan et al., 2017;Huang et al., 2021), with Huang et al. (2021)  This study, however, did not examine whether the global model adjustments created a 117 reasonable representation of cloud phase in the Arctic itself or distinguish between re-118 mote and local drivers of Arctic feedbacks (Feldl et al., 2020). We address this concern 119 by focusing our model adjustments and analysis on the Arctic and assessing model per-120 formance with additional observational constraints. Atmosphere-only simulations iso-121 late how the microphysical representation of mixed-phase clouds impacts Arctic warm-122 ing. Whereas the fully-coupled simulations of Tan and Storelvmo (2019)  To investigate the vertical structure of mixed-phase clouds, we filter by overlying 128 cloud optical thickness (COT) to produce two SLF metrics. We obtain one metric (here-

173
To assess the importance of this model mechanism on cloud properties and ice num-174 ber concentration and size, we disable the ice number limit at mixed-phase temperatures 175 (−37°C < T < 0°C) in CAM6-Oslo, restoring heterogeneous ice production as well as 176 secondary ice production through the Hallet-Mossop process, and producing an additional 177 model variation we label as CAM6-OsloIce. To isolate the impact of heterogeneous nu-178 cleation, we also cap the ice number tendency variable from secondary ice production 179 in CAM6-OsloIce to avoid strong secondary production in the absence of the ice num-  Table 1 summarizes the three base models and four "fitted" models 192 presented in this work, as well as six ancillary simulations in which single parameters were 193 tuned to the "fitted" values. When the ice limit is in place, large INP multipliers increase 194 ice crystal size and decrease the ice number concentration (CAM6 Fit 4 vs CAM6,  Oslo vs CAM6-Oslo(1,10)), demonstrating non-physical behavior caused by the model 196 error. Conversely, runs without an ice number limit have smaller ice crystals and higher 197 concentrations than those with the limit in place. Ice crystal size and concentration vari-198 ables in the constrained runs (Table 1)   We use surface radiative kernels from Soden et al. (2008)     from May through October ( Figure S4).

282
To investigate why different parametrizations give rise to these feedbacks, we pro-283 pose that the slope of the SLF curves in Fig. 1  heterogeneous and secondary nucleation processes from creating new ice crystals and find 305 that cloud water is significantly reduced when these nucleation processes are able to op-306 erate freely.

307
All models produce insufficient cloud fraction in the spring and excess cloud frac-  to these winter feedbacks.

321
Our results demonstrate the need to capture local cloud phase processes in order 322 to understand how mixed-phase cloud processes impact Arctic warming. Future stud-323 ies should use multiple atmospheric components, and use fully-coupled models to deter-324 mine whether the proposed constraint is valid in a dynamic climate system.  3. Figures S1 -S5 Text S1. Validation of cloud-bulk SLF metric. We use new cloud products (Guzman et al., 2017) to study the CALIOP's ability to sample Arctic clouds (Fig. S1). On the annual mean, opaque clouds make up 56% of cloudy scenes and are sampled through an average depth of 1.17km. While more opaque clouds are present in the summer and fall, the sampling depth never falls below 1km, indicating that the cloud-bulk metric samples a distinct thermodynamic regime below the supercooled liquid layer for all seasons. Text S3. Description of limit on secondary ice nucleation. The ice number tendency variable from secondary nucleation processes ("nsacwi") is limited to 10 6 kg −1 per microphysics timestep (5 minutes) if it exceeds this value after the Hallet-Mossop secondary ice scheme runs. We only set a cap on the number production, which otherwise would have no limit, unlike the mass term that is subject to cloud mass conservation. We choose a very high cap in order to prevent errors from re-implementing the Hallet-Mossop parameterization that was effectively removed from the model due to the ice number error. Sensitivity tests without the secondary ice limit showed negligible changes in SLF, confirming the dominant contribution of ice from heterogeneous processes when the model error is removed.
Text S4. Tuning Methods. The rate of ice and snow growth via the WBF process is highly-dependent on in-cloud conditions (updraft speed, concentration of cloud droplets and ice crystals) (Korolev, 2007). Previous studies reduced the efficiency of the WBF process in CAM5 by factors up to 10 to increase cloud liquid (Tan et al., 2016;Huang et al., 2021). We perform an identical modification in CAM6 to modify the WBF rate. Tan et al. (2016) also modified the fraction of dust aerosols active as ice nuclei, presenting results with multipliers of 0.79 and 0.19. We perform a similar modification by scaling the aerosol concentration variables that are fed into the Hoose heterogeneous ice nucleation scheme ("total aer num", "coated aer num", "uncoated aer num", "total interstitial aer num" ,"total cloudborne aer num") (Hoose et al., 2008). We initially tested WBF rate multipliers between 0.1 and 10, and INP multipliers between 0.01 and 100. WBF multipliers significantly greater than 1 have not been previously used, and we found that values greater than 2 significantly reduced SLF in both metrics.   Table S1.
Annual model cloud biases for the region 66 • N-82 • N. Cloud cover biases are calculated relative to CALIOP GOCCP observations. Surface cloud radiative effect (CRE) biases are calculated relative to CERES-EBAF observations using a positive downward sign convention.