Multi-campaign analysis of smoke properties and cloud interactions in the Southeast Atlantic using ORACLES, LASIC, and CLARIFY data with WRF-CAM5

The southeast Atlantic Ocean provides an excellent natural laboratory to study smoke-cloud interactions, a large driver of uncertainty in climate projections. The value of studying this in particular region is largely attributable to two factors—the expansive, bright, semi-permanent stratocumulus cloud deck and the fact that southern Africa is the largest source of biomassburning aerosols in the world. We study this region using the WRF-Chem model with CAM5 aerosols and in situ observations from the ORACLES, LASIC, and CLARIFY field campaigns, all of which overlapped in August 2017. Across these campaigns, we compare aerosol, cloud, and thermodynamic variables to quantify model performance and expand upon observational findings of aerosol-cloud effects. Specifically, our approach is to analyze aerosol and cloud properties along flight tracks, picking out uniform legs within tropospheric smoke plumes and in the boundary layer. This unique approach allows us to sample the high spatiotemporal variability that can get lost to large-scale averaging. It also allows process-level comparison of local cloud responses to aerosol conditions, and measure model performance in those same processes. Along with better quantifying model predictive power, we find and justify updates to model parameters and processes to better emulate observations, notably aerosol

size parameters. Preliminary results suggest that WRF-CAM5 is activating a smaller percentage of aerosols into cloud droplets than shown in observations, which could lead to biased modeling of aerosol indirect radiative effects on a larger scale. We explore this effect further with CCN activation tendency, updraft, particle sizing, and composition analysis, as well as broader dynamics like entrainment and removal rates. Comparing the model with similar instrument suites across multiple colocated campaigns also allows us to quantify instrument uncertainty in ways that a focus on a single campaign cannot and gives further context to the model performance.

PRESENTED AT:
MOTIVATION Aerosols and aerosol-cloud interactions are a huge source of uncertainty in global net radiative forcing. Aerosols can directly scatter and absorb radiation, nucleate cloud droplets, and modify the thermodynamic profile of the atmosphere.
From June-October each year, agricultural burning in southern Africa generates enormous emissions of biomass-burning aerosols (BBAs) that interact with the semi-permanent stratocumulus cloud deck in the southeastern Atlantic Ocean. These particles can act as cloud nuclei, increasing dropley number concentrations and thus cloud albedo and optical depth, illustrated in Fig. 1. This leads to a substantial perturbation to the regional energy balance (Zuidema et al., 2016).
Models implement a very wide range of aerosol properties and processes, especially involving BBAs. The varied explicit and implicit assumptions about BBAs and their interactions are partially a result of a lack of in situ observations used to directly update models.
It is especially important for the study of aerosol-cloud impacts that models accurately represent particle size, number, hygroscopicity, chemical composition, aging, and cloud activation tendencies.
The existing multi-model intercomparison studies often focus on broad trends in a few parameters of interest across a large spatial area, such as cloud thickness or total aerosol optical depth (AOD), without closely examining the range of processes that underlie those differences (Shinozuka et al., 2020).

METHODS AND MODEL DESCRIPTION
Three major observational campaigns coincided in August 2017, the period under study: ORACLES (NASA) 1-2 aircraft flying roughly 1,000km west of the coast, sampling smoke and clouds through the boundary layer and free troposphere; remote sensing and direct aerosol sampling CLARIFY (UK Met Office) Aircraft flying around Ascension Island, roughly 3,000km west of the coast, sampling through the boundary layer and free troposphere; remote sensing and direct aerosol sampling LASIC (US Dept. of Energy) Atmospheric Radiation Measurement (ARM) station on Ascension Island, 3km from coast; large remote sensing and direct sampling suite in the boundary layer, with radiosonde balloons.  [Haywood et al., 2021] Model configuration WRF-Chem with CAM5 aerosol and physics: Intermediate resolution (36km) between purely regional and global scale models 5hPa vertical layers (50 layers below 3km)

Cloud droplet nucleation
When aerosols can nudge clouds towards smaller, more numerous droplets, the clouds become brighter. We measure the cloud droplet number concentration against the aerosol concentration directly below that cloud to get a measure of this tendency (Fig. 4, below).
WRF-CAM5 is underestimating cloud activation by aerosols Aerosols in WRF are further from the 1:1 N :N line, and show a much wider range in droplet concentrations than observations do The main factors determining whether a given particle is able to activate are its diameter, its single-parameter hygroscopicity, and the ambient supersaturation (using turbulent kinetic energy as a measurable proxy).

Smoke Mean Diameter
WRF-CAM5 is roughly accurate in estimating FT smoke diameter, though biased somewhat high. AER CLD  https://agu2021fallmeeting-agu.ipostersessions.com/Default.aspx?s=FC... 6 of 20 1/3/2022, 12:58 Larger particles will activate into cloud droplets more easily than smaller ones. Here, the U. Hawaii UHSAS is the most reliable report of geometric mean diameter as it has been corrected the most extensively. WRF is thus slightly overestimating average diameter, but significantly overlaps with UHSAS mean and range. The variability between instruments is possibly due to calibration differences or anisokinetic inlet setup.
Observations in the BL show a persistent coarse mode and nucleation mode at low and medium smoke loading The lower free troposphere, which is often smoky, shows no major coarse mode CN counters show there is no major nucleation mode in the FT AGU -iPosterSessions.com (agu-vm-0) https://agu2021fallmeeting-agu.ipostersessions.com/Default.aspx?s=FC... 7 of 20 1/3/2022, 12:58 Hygroscopicity WRF-CAM5 underestimates hygroscopicity overall, which is likely leading to the cloud underactivation tendency above.
WRF also shows a much narrower range of hygroscopicity than CCN estimates.  The single hygroscopicity parameter is calculated by either a weighted volume average, as done with either the AMS and SP2 or in WRF, or from integrating the size distribution with the UHSAS and CCN counter and using a lookup table [Petters & Kreidenweiss, 2008].
WRF-CAM5 in this configuration also has a lower fixed hygroscopicity and lower density than suggested by experimental literature, and both of these are also likely to evolve as the particles age and oxidize over the course of several days [Dinar et al., 2006;Kuwata et al., 2011]. This is an under-studied topic for long-range biomass-burning smoke transport.
Between 0.1% and 0.2% supersaturation, K changes drastically, indicating different composition between 50-200nm. Within the accumulation mode, particles are not uniformly, internally mixed. This is an ongoing research topic.
Chemical composition WRF-CAM5 estimates an accurate composition in the smoky free troposphere, especially the volume fraction of OA and sulfate (Fig. 9, below). FT smoke is well-approximated by the accumulation mode alone.

The model underestimates SO in the boundary layer
Note: This build of WRF-Chem has not implemented aerosol-form NH and NO , although it is possible. Chemical composition measurements from LASIC at Ascension Island show again that WRF is consistently underestimating the sulfate fraction in the boundary layer (Fig. 10, below).
The relative enhancement of BL SO during clean periods indicates that the background aerosol mass below 1μm is sulfate-dominated in the MBL.

Turbulence
Supersaturation and cloud mixing are driven by turbulent updrafts in the marine boundary layer, so here we describe the turbulent kinetic energy, which can be measured more directly by high-resolution anemometers.
The model is substantially overestimating subgrid turbulence in the boundary layer. Turbulent updrafts allow air parcels to cool and supersaturate with water vapor, and are a key to cloud activation. Cloud activation efficiency is low, despite this overestimation, thus this is likely not the dominant factor. The 'characteristic updraft' w* is derived from the standard deviation of the vertical wind distribution over short periods, sampled by anemometers on the campaign aircraft [Morales & Nenes, 2010].

Aging trends
Smoke particles can evolve over time in diameter, composition, optical properties, hygroscopicity, and more. Many models have limited ability to capture these trends.
WRF has aerosols mostly growing with time, while observations show shrinkage (Fig. 12) OA is decreasing relative to SO significantly with age. WRF does not capture this relative loss of OA (Fig. 13) This divergence between WRF-CAM5 and observations could explain the population differences in diameter and the relatively static WRF hygroscopicity. 4 Fig. 12: Mean particle diameter aging trends in observations and WRF-CAM5. Age is calculated by tracers added to a WRF-AAM forecasting model build.