Reconstructing Saharan Dust–Cloud Scenes with WRF-L: Initial Evaluation of Aerosol-Aware Ice Nucleation Schemes ^†

Abstract

This study explores the role of mineral dust in ice nucleation using WRF-L model simulations during the ASKOS-ESA and CPEX-CV campaigns (Cabo Verde, 2022). Numerical experiments are carried out to examine dust impacts and secondary ice production via the Hallett–Mossop process. The results show variability in ice and liquid water paths, with the modeled aerosol optical depth aligning well with AERONET data. A case study of 15 September 2022 reveals notable cloud structure differences in aerosol-aware simulations. These findings can inform future LES simulations with assimilated aerosol fields and radar comparisons, emphasizing the importance of accurately representing aerosol–cloud interactions in atmospheric models.

1. Introduction

Mineral dust aerosols play a critical role in cloud microphysics, particularly by acting as ice-nucleating particles (INPs) in mixed-phase and cirrus clouds [1]. Their presence can influence cloud lifetime, radiative properties, and precipitation processes, making them an important component of weather and climate systems. Despite their significance, the representation of dust-related ice nucleation in atmospheric models remains uncertain, partly due to limited observational constraints and simplified parameterizations [2].

In this study, we use the WRF-L model, a dust-sensitive version of WRF-Chem [3], to simulate cloud scenes observed during the ASKOS-ESA and CPEX-CV campaigns in Cabo Verde (2022) [4,5]. A new aerosol-aware ice nucleation parameterization [6] is implemented and evaluated through mesoscale simulations, aiming to better capture the impact of Saharan dust on cloud formation and evolution.

2. Methods

We replaced the temperature-dependent ice nucleation process in the Morrison microphysics scheme [7,8] with a new aerosol-aware parameterization based on the relationship proposed in [6]. The Morrison scheme also includes secondary ice production through the Hallet and Mossop process [9]. Employing the modified Morison scheme, we conducted initial mesoscale simulations (15 km × 15 km resolution) with the WRF-L model [3]. The model domain was appropriately defined in order to simulate the Saharan dust emission and transport towards the Atlantic Ocean, covering the mineral dust sources of Sahara and the central Atlantic Ocean.

Meteorological forcing is provided by ECMWF-IFS model reanalysis (ERA5) datasets, updated every 6 h. The complete configuration options for the run are listed in

Table 1. Model configuration.

Parameterization	Scheme	Parameterization	Scheme
Surface Model	Noah [10]	sf_surface_physics	2
Surface Layer	MM5 [11]	sf_sfclay_physics	2
Radiation (SW and LW)	RRTMG [12]	ra_sw(lw)_physics	4
Microphysics	Morrison 2-moment [8]	mp_physics	10
Cumulus	Grell-3 [13]	cu_physics	5
Boundary Layer	MYNN 2.5 [14]	bl_pbl_physics	5
Chemistry	GOCART simple [15,16]	chem_opt	300
Dust Scheme	AFWA [16]	dust_opt	3

.

These simulations were designed to assess (a) the impact of the new ice nucleation parameterization and (b) the effect of secondary ice production (SIP), incorporating the new aerosol-aware scheme for primary ice nucleation. A summary of these tests is presented in

Table 2. Model experiments performed in this study using the WRF-L model.

MODEL EXPERIMENTS	DUST_IN	SIP
NoAERO-NoSIP	No	No
AERO-NoSIP	yes	No
AERO-SIP	yes	Yes

.

The primary goal of these simulations was to identify cases where notable differences arise from the inclusion or exclusion of specific physical processes. For these purposes, the liquid water path and the ice water path from the model were calculated as the sum of the integral of the concentration of all the different liquid hydrometeors (cloud droplets and rain) or all the ice hydrometeors (snow, ice, and graupel), respectively. The simulations covered the period of the CPEX-CV campaign, held during September 2022, a sub-period of the ASKOS-ESA campaign, during which multiple NASA DC-8 aircraft flights took place.

3. Results and Discussion

3.1. Simulated Aerosol Field

To examine the impact of the indirect aerosol effect in the simulation of a cloud scene by WRF-L, it is important to first accurately simulate the aerosol fields within the domain. In our simulations, the AOD is adequately represented, as shown in Figure 1, except for some days in late August, when the model overpredicts dust.

3.2. Simulated LWP and IWP

To identify cases of interest where significant differences between model simulations occur, we analyzed the time series of the simulated LWP and IWP, as shown in Figure 2 and Figure 3. Examining the differences in the simulation where we applied either the aerosol-aware scheme or the temperature-aware scheme for primary ice formation in WRF-L (Figure 2), we found three days where the simulated IWP differed significantly (14, 15, and 30 September 2022) and six days where the LWP differed significantly (9, 10, 14, 15, 16, and 30 September 2022). For the experiments shown in Figure 3 (i.e., where the dust-aware parameterization is on, and the secondary ice production through the Hallet and Mossop process is on/off), we found 3 days where the IWP differed significantly (22, 23, and 29 September 2022) and seven days where the LWP differed (9, 10, 14, 20, 26, 29, and 30 September 2022).

3.3. Simulated Cloud Scenes Against Aircraft Observations

An example of these days is presented in Figure 4. In the left column of the figure, the simulated cold cloud (represented by blue contours) and dust field (illustrated with colored contour lines) from different model experiments (NoAERO-NoSIP, AERO-NoSIP, and AERO-SIP) collocated to the flight of 15-09-2022 are depicted. The right column depicts the High-Altitude Lidar Observatory (HALO) observations from the NASA DC-8 of cloud screened, aerosol volume backscatter coefficient at 532 nm, relative humidity with respect to ice (using MERRA-2 reanalysis temperature fields), and dust mixing ratio for the same flight [17]. The dust layer depth is well captured within the observed altitude range (0–4.5 km). Between 5.5 and 8 km, regions with relative humidity (RH) with respect to ice greater than 100% are observed, indicating the presence of cold clouds. In all three model experiments, the cold cloud scene is reproduced. However, differences arise in the simulated IWC across the simulations. In the AERO-NoSIP and AERO-SIP cases, the cold cloud appears less dense and more fragmented than in the NoAERO-NoSIP case. This can be attributed to the fact that temperature-aware parameterization tends to overestimate ice crystal concentrations [18]. The cold clouds simulated in the AERO-NoSIP and AERO-SIP experiments consist of broken cloud structures, featuring three and two distinct cores, respectively.

4. Future Steps

Building on the initial mesoscale simulations and the implementation of the aerosol-aware ice nucleation scheme, the next phase of this work will focus on further analysis and refinement. Specifically, we plan to carry out the following:

• Perform detailed comparisons of model outputs with available ground-based and airborne radar measurements to evaluate the accuracy of cloud and ice structure simulations. This will help validate the sensitivity of the modeled IWP and LWP to aerosol–cloud interactions.
• Conduct large-eddy simulations (LESs) for selected days that exhibited significant differences between simulation setups in order to more accurately resolve cloud dynamics and microphysical processes influenced by mineral dust.
• Incorporate assimilated aerosol fields into the LES framework to better constrain the representation of dust layers and improve model initializations.
• Extend the simulations using available data from more recent campaigns.

These steps will strengthen the linkage between observed and modeled cloud structures and contribute to an improved understanding and representation of dust-driven ice nucleation in regional and global climate models.