Next Article in Journal
A Multi-Task Learning Framework with Enhanced Cross-Level Semantic Consistency for Multi-Level Land Cover Classification
Previous Article in Journal
Spatial and Temporal Expansion of Photovoltaic Sites and Thermal Environmental Effects in Ningxia Based on Remote Sensing and Deep Learning
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Remote Sensing
Volume 17
Issue 14
10.3390/rs17142439
remotesensing-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Assessment of Aerosol Optical Depth, Cloud Fraction, and Liquid Water Path in CMIP6 Models Using Satellite Observations

by
Jiakun Liang
and
Jennifer D. Small Griswold
*
Atmospheric Sciences Department, University of Hawaii at Mānoa, 2525 Correa Rd., Honolulu, HI 96822, USA
*
Author to whom correspondence should be addressed.
Remote Sens. 2025, 17(14), 2439; https://doi.org/10.3390/rs17142439 (registering DOI)
Submission received: 1 June 2025 / Revised: 9 July 2025 / Accepted: 12 July 2025 / Published: 14 July 2025
(This article belongs to the Special Issue Advances in Atmospheric Aerosol Monitoring Based on Lidar and Satellites)
Browse Figures
Versions Notes

Abstract

Aerosols are critical to the Earth’s atmosphere, influencing climate through interactions with solar radiation and clouds. However, accurately replicating the interactions between aerosols and clouds remains challenging due to the complexity of the physical processes involved. This study evaluated the performance of Coupled Model Intercomparison Project phase 6 (CMIP6) models in simulating aerosol optical depth (AOD), cloud fraction (CF), and liquid water path (LWP) by comparing them with satellite observations from MODIS and AMSR-E. Using 30 years of CMIP6 model simulations and available satellite observations during the satellite era, the results show that most CMIP6 models underestimate CF and LWP by 24.3% for LWP in the Northern Hemisphere. An assessment of spatial patterns indicates that models generally align more closely with observations in the Northern Hemisphere than in the Southern Hemisphere. Latitudinal profiles reveal that while most models capture the overall distribution patterns, they struggle to accurately reproduce observed magnitudes. A quantitative scoring system is applied to evaluate each model’s ability to replicate the spatial characteristics of multi-year mean aerosol and cloud properties. Overall, the findings suggest that CMIP6 models perform better in simulating AOD and CF than LWP, particularly in the Southern Hemisphere.
Keywords:
CMIP6; satellite; aerosol; cloud

1. Introduction

Aerosols play a critical role in Earth’s climate system, influencing both radiative and cloud-related processes [1]. They can affect the atmosphere directly by modifying Earth’s energy budget through scattering and absorbing solar radiation [2]. Additionally, aerosols act as cloud condensation nuclei (CCN), indirectly affecting the atmosphere by altering cloud properties, coverage, and lifespan through modifications in microphysical processes [3]. These aerosol-induced changes influence cloud radiative properties, impacting their ability to reflect and absorb solar radiation [4]. Aerosols also influence cloud dynamics by affecting cloud formation, vertical motion, and mixing, particularly in convective clouds [5]. Increased aerosol concentrations can also suppress precipitation under certain conditions, influencing cloud longevity [3,6,7]. In particular, aerosols influence clouds through cloud responses to aerosol-induced radiative effects, where the absorption of shortwave radiation alters humidity and atmospheric stability, leading to heating at the top of the boundary layer and the dissipation of clouds [6]. This process increases atmospheric warming while reducing surface solar radiation, which suppresses convection and limits new cloud formation [8]. Simulating aerosol–cloud interactions remains a critical challenge in climate modeling due to the complexity of atmospheric processes [9,10,11].
The Coupled Model Intercomparison Project (CMIP) is an international initiative designed to improve and evaluate climate models by coordinating simulations across multiple research groups worldwide. Over successive phases, the CMIP has advanced understanding of climate system responses to both natural and anthropogenic forcing. The latest phase, CMIP6 [12], introduces substantial improvements over previous iterations, particularly in the representation of clouds and water vapor [13]. A central objective of CMIP6 is to simulate historical climate changes from 1850 to 2014, providing a basis for assessing model performance in reproducing observed trends driven by variations in atmospheric composition and other climate forces. While these comparisons offer only a partial evaluation of model accuracy, they are essential for constraining climate sensitivity and improving the reliability of future projections. Enhanced representation of aerosol–cloud interactions in CMIP6 contributes to reducing uncertainties in climate simulations, ultimately refining predictions of climate change.
The complex relationship between clouds, aerosols, and radiation necessitates simplification and parameterization within global-scale models [14,15]. Achieving realistic global mean climate projections and representing microphysical processes while determining priority areas remains a challenge in current global models [16,17,18]. A major source of uncertainty in these simulations is the inadequate representation of clouds, particularly marine boundary layer clouds, which are key to cloud feedback and global cloud forcing predictions [19]. These warm low-level clouds are especially important, as their radiative properties are highly sensitive to their vertically integrated liquid water content, making them crucial in determining projected global cloud forcing changes [20]. To assess the impact of aerosol on cloud properties over large regions, satellite-derived cloud, and aerosol properties, such as cloud fraction (CF), liquid water path (LWP), and aerosol optical depth (AOD), can be used as proxies for cloud amount, cloud water content, and CCN, respectively [21,22]. The inclusion of cloud–aerosol interactions in CMIP6 enables the evaluation of model performance in simulating the spatial and temporal variations in clouds, aerosols, and their coupled processes.
Improving our understanding of aerosol–cloud interactions (ACIs) remains a central challenge in climate science and a major source of uncertainty in climate projections. CMIP6 involves a diverse set of models that differ substantially in their treatment of aerosol and cloud microphysics. Models that incorporate interactive aerosol schemes and detailed cloud microphysical representations tend to simulate more realistic indirect aerosol effects, such as CESM2, MRI-ESM2-0, and NorESM2-MM, while models employing simplified aerosol formulations often systematically overestimate AOD and exhibit weaker skill of aerosol–cloud coverability [23], such as INM-CM4-8 and IPSL-CM5A2-INCA. These structural differences are critical to the interpretation of performance variations in aerosol and cloud simulations. Recent studies assessed ACI in CMIP6 simulations. Wang et al. [24] found that stronger cloud feedback in CMIP6 models is associated with enhanced aerosol-induced cooling, which contributes to hemispheric warming asymmetry. In addition, regional evaluations identified notable biases in AOD, particularly over East Asia, along with inconsistencies in CF and LWP, despite the multi-model mean generally outperforming individual models [25,26]. These findings underscore the ongoing need to improve the representation of aerosol and cloud processes in CMIP6, especially in microphysics and emissions inputs. Despite the advances offered by CMIP6, substantial inter-model variability in ACI representation remains. As Earth system modeling advances, it is critical to identify which models offer robust ACI performance and where improvements are needed. Satellite observations provide near-global coverage and serve as a valuable constraint, yet CMIP6 models remain essential for assessing long-term variability and projecting future climate states under various forcing scenarios. To make reliable use of CMIP6 in both research and policy contexts, it is necessary to evaluate model fidelity across cloud and aerosol variables and to diagnose structural deficiencies.
This study provides a quantitative evaluation of cloud and aerosol variables simulated by CMIP6, using satellite observations from the Moderate Resolution Imaging Spectroradiometer (MODIS) and the Advanced Microwave Scanning Radiometer for EOS (AMSR-E). A performance scoring system is introduced to assess the model results, which is provided for all applicable CMIP6 models. The analysis focuses on the climatological mean variations of CF, LWP, and AOD, with a particular emphasis on the annual mean and correlation between modeled and observed fields. Evaluations are conducted at the global scale and separately for the Northern Hemisphere (NH) and Southern Hemisphere (SH), with special attention given to the spatial distribution of CF, LWP, and AOD and the corresponding model performance in each region.

2. Data

2.1. CMIP6 Climate Models

This study utilized a total of 56 models from the CMIP6 archive, including 54 models with CF, 52 models with LWP, and 35 models with AOD (Table 1) at the time of analysis. The od550aer (AOD) is used to represent aerosol properties, while cloud-related variables include clt (total cloud area fraction), lwp (LWP), clivi (ice water path), and clwvi (total water path). In CMIP6 models, lwp, clivi, and clwvi are vertically integrated variables, where lwp can be derived by subtracting clivi from clwvi. This approach is applied to models where independent LWP data are unavailable, ensuring consistency in the analysis.
For each CMIP6 model, monthly outputs from 1985 to 2014 are analyzed, providing a 30-year time frame to align with the World Meteorological Organization (WMO) standard for climate analysis. This duration ensures a robust assessment of long-term climate trends while minimizing the influence of short-term variability. The satellite observations used for evaluation, specifically from MODIS Aqua, cover 2002 to 2014. This allowed for a direct comparison between model outputs and observational records. The variant label for the CMIP6 data is r1i1p1f1, which is the most commonly available for all models, and historical runs were used for all aerosol- and cloud-related variables. These simulations incorporated time-dependent atmospheric conditions, including anthropogenic and volcanic emissions, solar forcing, and land use changes [27], enabling a comprehensive assessment of climate model accuracy.
Table 1. CMIP6 models used in this study.
Table 1. CMIP6 models used in this study.
ModelResolutionVariables
CSIRO-ARCCSS, ACCESS-CM2 [28]1.25° × 1.875°, (144 × 192)AOD, CF
CSIRO-ARCCSS, ACCESS-ESM1-5 [29] 1.25° × 1.875°, (144 × 192)AOD, CF
AWI, CM-1-1-MR [30]0.9375° × 0.9375°, (192 × 384)CF, LWP 1
AWI, ESM-1-1-LR [31]1.875° × 1.875°, (96 × 192)AOD, CF, LWP 1
AWI, ESM-1-REcoM [32]1.875° × 1.875°, (96 × 192)AOD, CF, LWP 1
BCC, CSM2-MR [33]1.125° × 1.125°, (160 × 320)CF, LWP 1
BCC, ESM1 [34]2.8125° × 2.8125°, (64 × 128)AOD, CF, LWP 1
CAMS, CSM 1-0 [35]1.125° × 1.125°, (160 × 320)CF, LWP 1
CAS, ESM2-0 [36]1.40625° × 1.40625°, (128 × 256)CF, LWP 1
NCAR, CESM2 [37]0.9375° × 1.25°, (192 × 288)AOD, CF, LWP
NCAR, CESM2-WACCM [37]0.9375° × 1.25°, (192 × 288)AOD, CF, LWP
NCAR, CESM2-FV2 [37]1.875° × 2.5°, (96 × 144)AOD, CF, LWP
NCAR, CESM2-WACCM-FV2 [37]1.875° × 2.5°, (96 × 144)AOD, CF, LWP
THU Department of Earth
System Science China, CIESM [38]		0.9375° × 1.25°, (192 × 288)CF, LWP 1
CMCC, CM2-HR4 [39]0.9375° × 1.25°, (192 × 288)CF, LWP 1
CMCC, CM2-SR5 [39]0.9375° × 1.25°, (192 × 288)AOD, CF, LWP 1
CMCC, ESM2 [39]0.9375° × 1.25°, (192 × 288)AOD, CF, LWP 1
CCCMa, CanESM5 [40]2.8125° × 2.8125°, (64 × 128)AOD, CF, LWP
CCCMa, CanESM5-1 [40]2.8125° × 2.8125°, (64 × 128)AOD, CF, LWP
E3SM-Project, E3SM-1-0 [41]1° × 1°, (180 × 360)AOD, CF, LWP 1
E3SM-Project, E3SM-1-1 [41]1° × 1°, (180 × 360)AOD, CF, LWP 1
E3SM-Project, E3SM-1-1-ECA [41]1° × 1°, (180 × 360)AOD, CF, LWP 1
E3SM-Project, E3SM-2-0 [41]1° × 1°, (180 × 360)AOD, CF, LWP 1
EC-Earth3, EC-Earth3 [42]0.703125° × 0.703125°, (256 × 512)AOD, CF, LWP 1
EC-Earth3, EC-Earth3-AerChem [42]0.703125° × 0.703125°, (256 × 512)AOD, CF, LWP
EC-Earth3, EC-Earth3-CC [42]0.703125° × 0.703125°, (256 × 512)CF, LWP 1
EC-Earth3, EC-Earth3-Veg [42]0.703125° × 0.703125°, (256 × 512)AOD, CF, LWP 1
EC-Earth3, EC-Earth3-Veg-LR [42]1.125° × 1.125°, (160 × 320)CF, LWP
CAS, FGOALS-f3-L [43]1° × 1.25°, (180 × 288)CF, LWP 1
CAS, FGOALS-g3 [44]2.25° × 2°, (80 × 180)CF, LWP 1
FIO-QNLM, FIO-ESM-2-0 [45]0.9375° × 1.25°, (192 × 288)CF, LWP 1
GFDL, CM4 [46]1° × 1.25°, (180 × 288)AOD
GFDL, ESM4 [47]1° × 1.25°, (180 × 288)AOD, CF, LWP
GISS, E2-1-G [48]2° × 2.5°, (90 × 144)CF, LWP 1
GISS, E2-1-H [48]2° × 2.5°, (90 × 144)CF, LWP 1
GISS, E2-2-G [49]2° × 2.5°, (90 × 144)CF, LWP 1
GISS, E2-2-H [48]2° × 2.5°, (90 × 144)CF, LWP 1
GISS, E3-G [48]1° × 1.25°, (180 × 288)CF
CCCR, IITM-ESM [50]~1.915° × 1.875°, (94 × 192)CF, LWP 1
INM, CM4-8 [51]1.5° × 2°, (120 × 180)AOD, CF, LWP
INM, CM5-0 [52]1.5° × 2°, (120 × 180)AOD, CF, LWP
IPSL, CM5A2-INCA [53]~1.259° × 2.5°, (143 × 144)LWP 1
IPSL, CM6A-LR [53]~1.259° × 2.5°, (143 × 144)AOD, CF, LWP
IPSL, CM6A-LR-INCA [53]~1.259° × 2.5°, (143 × 144)AOD, CF, LWP
NIMS-KMA, KACE-1-0-G [54]1.25° × 1.875°, (144 × 192)AOD, CF, LWP 1
KIOST, ESM [55]1.875° × 1.875°, (96 × 192)CF, LWP 1
MIROC6 [56]1.40625° × 1.40625°, (128 × 256)AOD, CF, LWP 1
MPI, ESM1.2-HAM [57]1.875° × 1.875°, (96 × 192)AOD, CF, LWP
MPI, ESM1.2-HR [58]0.9375° × 0.9375°, (192 × 384)AOD, CF, LWP 1
MPI, ESM1.2-LR [57]1.875° × 1.875°, (96 × 192)AOD, CF, LWP 1
MRI, ESM2-0 [59]1.125° × 1.125°, (160 × 320)AOD, CF, LWP 1
NUIST, NESM3 [60]1.875° × 1.875°, (96 × 192)CF, LWP 1
NCC, NorESM2-LM [61]1.875° × 2.5°, (96 × 144)AOD, CF, LWP
NCC, NorESM2-MM [61]0.9375° × 1.25°, (192 × 288)AOD, CF, LWP
SNU, SAM0-UNICON [62]0.9375° × 1.25°, (192 × 288)CF, LWP 1
AS-RCEC, TaiESM [63]0.9375° × 1.25°, (192 × 288)AOD, CF, LWP
1 LWP is calculated from the total water path and ice water path (clwvi − clivi).

2.2. Satellite Data

2.2.1. MODIS Data

MODIS is a key instrument onboard the satellite Aqua, which was launched on 4 May 2002 [64]. MODIS provides critical observations of Earth’s atmosphere, land, and oceans, including measurements of aerosols and cloud properties, which are crucial for climate and environmental research.
This study utilized the MODIS Aqua Collection 6.1 Level 3 to evaluate the representation of cloud and aerosol properties in CMIP6 climate simulations. The Level 3 data, derived from Level 2 observations, consist of calibrated and geolocated satellite measurements that were aggregated and gridded for improved spatial and temporal consistency. The dataset used in this study has a spatial resolution of 1° × 1° and a monthly temporal resolution [65]. CF is assessed using the “Cloud_Fraction_Mean_Mean” variable from the MODIS cloud product, which is based on the MOD35 cloud mask and MOD06 cloud retrievals [65]. AOD is obtained from the “AOD_550_Dark_Target_Deep_Blue_Combined_Mean_Mean” variable, which merges retrievals from the dark target (DT) and deep blue (DB) algorithms. The DT algorithm retrieves AOD over dark ocean and vegetated surfaces, while DB is optimized for bright or arid land regions. The merged product leverages the strengths of both approaches to provide more robust global AOD estimates [61]. Previous studies indicate that CF retrieved from MODIS carries regional uncertainties up to 30%, particularly in polar regions due to limitations in cloud detection sensitivity [66]. AOD retrievals show an uncertainty of approximately 0.03 + 0.20 × AOD over ocean [67]. Specifically, MODIS AOD retrievals exhibit sensitivity to the solar zenith angle, leading to significant discrepancies at high latitudes when compared with other satellite-based measurements (e.g., AMSR-E [68]). To minimize these uncertainties, this study focused on the 70°S–70°N latitude range, encompassing tropical and mid-latitude regions where MODIS data provide more reliable aerosol and cloud observations.

2.2.2. AMSR-E Data

LWP is a crucial parameter for characterizing climate systems and their variability, influencing both hydrological processes and radiative transfer [69,70,71,72]. Remote sensing instruments such as MODIS, equipped with multiple spectral bands, commonly estimate LWP using the solar reflectance of two visible-shortwave infrared (SWIR) channels, which are used to derive cloud optical depth and effective radius [73,74]. However, previous studies noted that high biases exist in MODIS-derived LWP, particularly associated with variations in the solar zenith angle, leading to substantial overestimations at high latitudes [75,76,77]. The primary source of this bias is the cloud-top height variability, which results in three-dimensional radiative interactions that affect retrieval accuracy. Radiative transfer simulations confirm that cloud-top inhomogeneity is the dominant factor contributing to the solar zenith angle-dependent bias in MODIS LWP estimates [78]. These uncertainties present significant challenges in constraining climate model simulations using MODIS-based LWP observations [79,80].
Given the unresolved biases in MODIS LWP retrievals, this study instead utilized LWP data from AMSR-E. AMSR-E, also onboard the Aqua satellite, provides atmospheric and surface observations, including cloud-related variables. Unlike MODIS, which relies on shortwave reflectance, AMSR-E is a microwave radiometer and retrieves LWP using Wentz’s algorithm, which estimates column-integrated LWP based on multi-frequency brightness temperature measurements [81,82]. This study employs AMSR-E Level 3 data from version 4 (June 2002–July 2010) and version 5 (August 2010–September 2011) [81]. AMSR-E LWP retrievals exhibit typical uncertainties of 15–30 g/m2, particularly in low-LWP regions or over surfaces with complex emissivity, and offer a spatial resolution of 0.25° × 0.25° with demonstrated retrieval accuracy exceeding 95% [83]. The instrument operated from its launch on 4 May 2002 until 4 October 2011, when it ceased functioning due to a power failure. Compared to MODIS, AMSR-E has a shorter operational period of approximately nine and a half years.

3. Methodology

Following the methodology of Jiang et al. [79], we implemented a quantitative scoring system to evaluate the performance of CMIP6 models in reproducing satellite-observed climatological means. The assessment focused on the agreement between each model’s multi-year monthly mean and the corresponding satellite-derived multi-year mean, using spatial averages. The scoring approach, adapted from Douglass et al. [84], assigns a performance score between 0 to 1. A perfect score of 1 is awarded when a model’s monthly mean exactly matches the 30-year satellite climatology. Scores decrease as deviations from observations increase, with deviations normalized by the standard deviation of the satellite monthly means. If a model deviates by more than three standard deviations, it receives a score of zero. These performance scores are calculated for AOD, CF, and LWP across the global domain, NH, and SH:
S c o r e = 1 − 1 12 ∑ t = 1 12 V C M I P 6   M o d e l − V S a t e l l i t e   O b s e r v a t i o n 3 σ S a t e l l i t e   O b s e r v a t i o n ,
where V represents the multi-year monthly mean of AOD, CF, and LWP, σ denotes the monthly standard deviation, and t refers to each of the 12 months. A score of 1 indicates perfect agreement between the model and observations, while a score of 0 indicates no skill.

4. Comparisons of CMIP6 Models’ Simulations and Satellite Observations

In this section, we examine the multi-year, multi-model averages of CMIP6 simulations to assess their discrepancies relative to satellite observations, both collectively and on an individual model basis. We begin with the global average maps to explore the spatial distribution of both CMIP6 model outputs and satellite data, along with their differences. This is followed by a model-by-model analysis, comparing each CMIP6 simulation against satellite observations at the global scale and separately for the NH and SH, which provides a comprehensive evaluation of model performance and highlights regional variations in the representation of key variables.

4.1. Multi-Year Averages of AOD, CF, and LWP

Multi-year averages from the CMIP6 models and satellite data were computed over distinct periods: 30 years (January 1985 to December 2014) for the CMIP6 models; 12 years and 5 months (July 2002 to December 2014) for MODIS; and 9 years and 3 months (June 2002 to September 2011) for AMSR-E. These time frames were selected to support a robust climate-scale evaluation of model performance, offering long-term context while aligning with the available temporal coverage of the satellite datasets.
To ensure consistency in spatial resolution across all data sources, each dataset was re-gridded onto a uniform 1° × 1° (360 × 180) latitude–longitude grid. Standardizing grid resolution is essential for meaningful comparisons across datasets. Model outputs and satellite observations often differ in their native resolutions, and without re-gridding, spatial mismatches can introduce artificial errors in the evaluation process. To ensure that the differences reflected actual physical discrepancies rather than inconsistencies in spatial representation, both datasets were interpolated to the same grid using linear interpolation. To preserve the integrity of the original data and avoid introducing unnecessary interpolation artifacts, all subsequent analyses, including the computation of climatological means and performance scores, were conducted using the data in its original resolution. This approach prevents artificial smoothing or distortion of small-scale features, ensuring that the analyses reflect the true characteristics of the datasets.
Figure 1 presents the multi-year, multi-model mean values, and differences in AOD, CF, and LWP derived from CMIP6 simulations and satellite observations from MODIS and AMSR-E. The AMSR-E LWP represents the total water vapor content that is integrated from the surface to the top of the atmosphere and is available only over the ocean. Therefore, to ensure a consistent comparison, LWP analysis for CMIP6 models is restricted to oceanic areas, aligning with AMSR-E observations.
Both observational data and climate models indicate that the highest annual average AOD is in the Middle East and north Africa, particularly across the Arabian Peninsula and north Africa. Our results support this pattern, with CMIP6 simulations showing elevated AOD values over north Africa and northwest China. However, these models tend to simulate substantially higher AOD levels in north Africa compared to satellite observations, with an overestimation of 0.054 (Figure 1a). Conversely, CMIP6 models underestimate AOD over southern Africa (by 0.044), the Himalayan region (by 0.058), and northeast China (by 0.019).
Previous studies showed that MODIS observations identify the North China Plain as having the highest AOD levels, followed by central China, with the lowest values observed over the Tibetan Plateau, Qinghai, and the Gobi Desert. The elevated AOD levels in these regions are largely attributed to significant emissions of sand and dust from natural sources, particularly the Taklamakan and Gobi Deserts, with peak emissions occurring in spring [85]. In southern Africa, aerosols originate from both natural and anthropogenic sources, including desert dust, biomass burning, and industrial pollution. In contrast, aerosol emissions in north India and northeast China are primarily driven by industrial activity, transportation, and coal combustion.
Zhao et al. [26] found that significant uncertainties and inter-model variability exist in the simulation of global dust processes across different models, particularly in capturing spatial patterns and seasonal variations. Compared to previous model generations, CMIP6 simulations exhibit a broader range in global dust emission, but struggle to reproduce key regional dust distribution features, such as the accumulation of dust along the southern edge of the Himalayas. While multi-year, multi-model CMIP6 simulations indicate aerosol accumulation in the Himalayan region, the boundary is less distinct compared to satellite observations (Figure 1a,b).
For CF (Figure 1d,e), CMIP6 models generally simulate higher CF values than satellite observations, particularly at higher latitudes in both hemispheres. Consistent with the findings of Vignesh et al. [86], notable spatial discrepancies between model simulations and observations are evident in the tropics. In particular, the CMIP6 models fail to capture the stratocumulus decks along the west coasts of North and South America, which aligns with the LWP discrepancies shown in Figure 1g.
Furthermore, most CMIP6 models do not adequately represent cloud structures in the South Pacific Convergence Zone. In some cases, these features are absent from the simulations. Instead, the models tend to produce higher CF values over the western Pacific, indicative of an anthropogenic “double ITCZ” (Figure 1f). This feature is also apparent in the LWP simulations and becomes more pronounced in the Supplementary Figures for individual models (Figure S1), see Supplementary Material.
Compared to AOD and CF, CMIP6 models generally underestimate LWP, particularly in the Intertropical Convergence Zone (ITCZ) region. However, a slight overestimation of LWP is observed along the west coast of Africa (Figure 1i). Supplementary Figures (Figure S1) provide multi-year mean maps of AOD, CF, and LWP for each CMIP6 model, highlighting the considerable inter-model variability and systematic deviations from satellite observations.
Specifically, for AOD, 26 out of 35 CMIP6 models (74.3%) overestimate AOD over north Africa, suggesting an overrepresentation of dust transport from the Sahara Desert. Despite these individual biases, the ensemble mean substantially reduces errors and provides a more accurate representation of climatological aerosol and cloud distributions, as shown in Figure 1.
To evaluate the performance of individual CMIP6 models, multi-year mean differences between model simulations and satellite observations were analyzed, as shown in Figure 2. These comparisons were conducted separately for the global domain and for the two hemispheres.
Among the 35 CMIP6 models with available AOD data, 27 of them exhibit globally negative differences between simulated and observed values (model simulations minus observations), indicating a general tendency to underestimate AOD compared to satellite measurements. When examined by hemisphere, 28 models show negative mean differences in the NH, while 20 do so in the SH, suggesting that AOD underestimation is more prevalent in the NH. These findings are consistent with the multi-model mean differences shown in Figure 2 (last column), where the NH differences are systematically negative. This persistent underestimation in the NH suggests that regional compensations may influence the global mean. Substantial deviations are observed at the regional scale, particularly in the NH, where model biases are more pronounced than in the SH. Among the three evaluated variables, the SH exhibits the smallest multi-model mean AOD bias, at −0.0009, highlighting relatively better model performance for AOD in this region.
The multi-year mean differences between CMIP6 model simulations and satellite observations for CF and LWP are negative for both global and hemispheric scales, indicating a general tendency for the models to underestimate these variables. Among the 54 CMIP6 models analyzed for CF, only three (CESM2, CESM2-WACCM, and INM-CM5-0) exhibit overestimations globally and in both hemispheres, whereas 35 models consistently underestimate CF. In the SH, only four models show positive biases compared to 36 in the NH, suggesting a tendency for CMIP6 models to underestimate CF in the NH while overestimating it in the SH. Overall, model biases are more pronounced in the SH, implying that CMIP6 models simulate CF with greater accuracy in the NH.
Similarly, for LWP, a majority of models (41 out of 52) display negative biases, indicating a systematic underestimation. Eleven models show positive biases in the NH, which is comparable to the SH (nine models). The average multi-year mean differences for the NH (−0.024 kg/m2) and the SH (−0.027 kg/m2) are about the same, suggesting that the magnitude of LWP bias is similar across both hemispheres.
Among the three variables evaluated, AOD exhibits the smallest discrepancies between model simulations and satellite observations when compared to CF and LWP. In general, the models tend to perform better in the NH than in the SH for both CF and LWP, with a general tendency to underestimate values. For AOD, models tend to underestimate concentrations in the NH and overestimate them in the SH, although overall simulation performance is slightly better in the SH. Specifically, the E3SM-2-0 model shows the smallest AOD difference from the global mean and satellite observations, with a difference of −0.001. The CanESM5-1 model has the smallest difference of 0.0013 from the NH mean and satellite observations, while the CMCC-CM2-SR5 model shows the smallest AOD difference of 0.00017 from the SH mean and satellite observations. Collectively, the E3SM-2-0 model demonstrates the smallest differences from satellite observations both globally and in two hemispheres, outperforming all other AOD simulations across the 35 CMIP6 models.
For CF, the KACE-1-0-G model exhibits the closest agreement with satellite observations at the global scale, with a minimal difference of 0.00033. Regionally, the INM-CM4-8 model performs best in the NH and the SH at −0.00306 and 0.00233, respectively. Among the 54 CMIP6 models evaluated for CF, the INM-CM4-8 model demonstrates consistently low biases relative to satellite observations across global and hemispheric scales. For LWP, the MRI-ESM2-0 model yields the smallest deviations from satellite observations at the global and hemispheric levels, outperforming all other LWP simulations among the 52 CMIP6 models assessed.

4.2. Latitude Profiles of AOD, CF, and LWP

To examine the meridional distribution of AOD, CF, and LWP, the latitudinal profiles derived from CMIP6 model outputs were analyzed. Figure 3 shows the multi-year mean latitudinal distributions of these three variables (panels a–c), alongside the correlation coefficients between CMIP6 simulations and satellite observations (d). The observed AOD exhibits a pronounced hemispheric asymmetry, with substantially higher values and greater variability in the NH compared to the SH. This pattern aligns with the spatial distribution of anthropogenic activities, where higher population densities, industrial emissions, and biomass burning in regions such as north Africa, the Middle East, and South and East Asia contribute to elevated AOD in the NH. A prominent AOD maximum is observed near 20°N, coinciding with the locations of major desert regions, notably the Sahara, as well as areas of intense agricultural activity and urban emissions.
Most CMIP6 models capture the enhanced AOD in the NH and successfully reproduce the peak near 20°N. However, several models deviate from this observed feature. For example, CanESM5 and CanESM5-1 simulate a peak further north, around 40°N, while INM-CM4-8 and INM-CM5-0 place the peak at approximately 50°N. Although CESM2-WACCM-FV2 correctly positions the AOD maximum at 20°N, it substantially overestimates the magnitude, with values reaching 0.5. In the SH, many models simulate a secondary peak around 55°S, which is consistent with observational data. Notably, INM-CM4-8 and INM-CM5-0 again exhibit a latitudinal bias, placing this SH peak further south at around 60°S and overestimating the amplitude (~0.45).
The latitudinal distribution of CF (Figure 3b) reveals that most CMIP6 models are generally consistent with satellite observations in terms of the overall spatial structure. Satellite data exhibit prominent CF maxima near 60°N and 60°S, corresponding to the climatological positions of subpolar low-pressure systems. While many models capture the southern peak near 60°S, they frequently fail to reproduce the corresponding northern peak near 60°N with similar accuracy. Near the equator, observations show a distinct double-peak structure associated with the ITCZ. This feature is reasonably well represented in most models; however, both GISS-E2-2-G and GISS-E2-2-H fail to resolve the double-peak structure, instead simulating a single broad maximum that overlooks key spatial variability. The CMCC-CM2-HR4 model consistently underestimates CF across most latitudes. Despite this overall underestimation, the model captures the relative positioning and amplitude of CF maxima and minima, resulting in a relatively strong correlation with satellite observations, as indicated in Figure 3d.
Figure 3c illustrates the latitudinal distribution of LWP observed and simulated by CMIP6 models. Observational data reveal prominent LWP maxima in the mid-latitudes of both hemispheres and a secondary peak in the tropics, reflecting climatological cloud regimes associated with extratropical cyclones and convective activity near the ITCZ, respectively. While CMIP6 models broadly capture the spatial variability of LWP, they tend to underestimate its magnitude overall. A closer examination reveals that many models overestimate LWP in the mid-latitudes while simultaneously underestimating it in the tropics. Notably, the IITM-ESM model fails to simulate enhanced LWP in the mid-latitude regions, diverging significantly from observed patterns. Moreover, it overestimates LWP globally and lacks the tropical peak entirely. Models such as SAM0-UNICON reproduce mid-latitude maxima, but do not simulate the tropical enhancement. While several models, including FGOALS-f3-L, replicate the overall structure of the latitudinal LWP profile, they exhibit significant underestimation, particularly in the mid-latitude zones, resulting in weak agreement with observational data. Collectively, these discrepancies contribute to the persistent limitations in CMIP6 models’ ability to accurately simulate the observed meridional distribution of AOD, CF, and LWP, particularly with respect to amplitude and latitudinal positioning.

4.3. CF–AOD Relationship

This section presents a summary analysis evaluating the performance of CMIP6 models in reproducing the observed relationship between CF and AOD at global and hemispheric scales. Observational data show generally higher AOD values in the NH, although most values remain below 1.0. To reduce the likelihood of misclassifying clouds and aerosols, AOD values are binned linearly from 0 to 1.0, following the methodology of Koren et al. [21]. Multi-year mean CF and AOD values are paired at each grid point for global, NH, and SH analyses.
Figure 4 displays the mean CF–AOD relationships derived from each CMIP6 model alongside MODIS satellite observations. To represent variability, standard errors are used in place of standard deviations, as they reflect sampling uncertainty more appropriately across spatial and temporal scales. The observed pattern follows a characteristic trend: CF increases with AOD at lower AOD levels, which is commonly interpreted as the evidence of aerosol–cloud microphysical interactions associated with the first and second indirect effects, followed by a decline in CF at higher AOD, suggesting radiative suppression of cloud formation linked to the second indirect effect, consistent with the findings of Koren et al. [21].
At the global scale, most CMIP6 models can capture the characteristic CF–AOD relationship, as shown in Figure 4d. However, several models fail to reproduce the initial increase in CF that is typically attributed to aerosol microphysical effects. For instance, KACE-1-0-G, MIROC6, and TaiESM predominantly exhibit a decreasing CF trend with increasing AOD, indicative of radiative effects alone. Similarly, the INM-CM4-8 and INM-CM5-0 models do not show the expected microphysical signature—specifically, an increase in CF with AOD in the range of 0.15 to 0.4—displaying instead a consistent decline in CF beyond an AOD threshold of 0.45. The CMIP6 multi-model mean is broadly consistent with MODIS observations, which also suggest a weak or ambiguous microphysical response at the global scale.
In the NH (Figure 4e), several models exhibit atypical behavior in the CF–AOD relationship. While most models tend to reflect a stronger radiative effect, only a few, such as KACE-1-0-G, MPI family, NorESM2-LM, and NorESM2-MM, display indications of microphysical effects. Consistent with the global analysis, both INM-CM4-8 and INM-CM5-0 fail to reproduce a microphysical signature, instead producing an unrealistic peak in CF over the AOD range of 0.15 to 0.8. The multi-model mean fails to exhibit the microphysical effect. It only shows a monotonic decrease in CF with increasing AOD, suggesting a dominant radiative response across the ensemble.
In the SH, most CMIP6 models exhibit a general increase in CF with AOD, particularly up to an AOD of approximately 0.12. Beyond this threshold, the data density decreases substantially, resulting in a noisier and less consistent signal. Consistently with the global analysis, the TaiESM model fails to capture any notable microphysical effect. The CMCC models display only a modest increase in CF with rising AOD, indicating a weak representation of microphysical processes. Unlike the global and NH cases, several models in the SH demonstrate only a monotonic increase in CF with AOD, suggesting the absence of radiative effects in their simulations. The CMIP6 multi-model mean shows good agreement with MODIS satellite observations, capturing a clear and structured CF–AOD relationship.
Overall, most CMIP6 models exhibit a recognizable relationship between CF and AOD in the global analysis, although substantial inter-model variability remains. The TaiESM model, for example, tends to overemphasize radiative effects while failing to reproduce the microphysical response evident in satellite observations. In the NH, the CMIP6 multi-model mean displays a trend consistent with radiative effects, whereas satellite data suggest a dominant microphysical influence. However, neither effect is distinctly captured by the ensemble, likely due in part to the re-gridding of model outputs to coarser spatial resolutions, which may obscure the finer-scale features critical for resolving aerosol–cloud interactions. Notably, the INM-CM4-8 and INM-CM-5-0 models consistently produce an anomalous enhancement, or “bump,” in CF across the AOD range of 0.15 to 0.8 in both the global and NH analyses, which lacks physical justification and is inconsistent with observational data. In contrast, in the SH, some models, including members of the E3SM and EC-Earth3 model families, show evidence of capturing microphysical effects at lower AOD values (AOD < 0.1), demonstrating a closer alignment with observed behavior in this region.
These results provide valuable insights, yet underscore limitations of the current analysis. The complex relationship between AOD and CF cannot be fully explained by aerosol availability alone. AOD impacts often occur through precipitation suppression and cloud lifetime effects [87]. Spatial and temporal variations in aerosol properties [88], improvements in retrieval algorithms [89], and uncertainties related to cloud masking [90] further complicate aerosol–cloud interactions. These factors, along with regional meteorology and aerosol–radiation feedback, underscore the complexity of interpreting satellite-derived CF–AOD relationships. This study focused on spatial patterns and did not address the specific microphysics parameterizations of each model. Investigating how individual schemes implement aerosol–cloud interactions, including microphysical and radiative processes, represents a promising direction for future research.

5. Quantitative Assessments of CMIP6 Models’ Performances

In this section, we evaluate the performance of CMIP6 models by comparing their outputs with multi-year satellite observations. The analysis is restricted to the latitudinal band between 70°S and 70°N to minimize the influence of solar zenith angle variations on MODIS retrievals. Additionally, the assessment of LWP is confined to oceanic regions, corresponding to the spatial coverage of AMSR-E data.

5.1. Performance of the CMIP6 Models in Terms of Spatial Averaging

We assessed the ability of CMIP6 models to simulate the spatial mean, variance, and spatial correlation of CF, LWP, and AOD, using satellite observations as a reference. Figure 5 presents the relationships between multi-year means of AOD and CF, as well as AOD and oceanic LWP, for the global domain and each hemisphere. Black dots represent the multi-year mean values derived from MODIS and AMSR-E observations, while the accompanying horizontal and vertical lines, along with shaded gray regions, represent observational uncertainties. These uncertainty bounds are calculated using the mean absolute deviation (MAD), which quantifies the average absolute deviation of each data point from the dataset mean and serves as a robust measure of internal variability. Colored markers (including dots, circles, triangles, and squares) correspond to the multi-year means from individual CMIP6 models, and open black circles indicate the ensemble mean across all models.
Overall, the inter-model spread in CF and AOD is narrower in the NH than in the SH when compared to LWP. As shown in Figure 5a–c, CMIP6 models generally reproduce CF more accurately than AOD, with all models falling within the observational uncertainty bounds for CF in the global domain and both hemispheres. In contrast, AOD simulations show greater variability: MIROC6 is the only model outside the observational uncertainty range for AOD in the global and NH domains, indicating an underestimation of AOD. In the SH, three models (INM-CM4-8, INM-CM5-0, and MIROC6) fall outside the uncertainty bounds, with INM-CM5-0 exhibiting the largest deviation, primarily due to a significant overestimation of AOD. The multi-model mean tends to underestimate CF across most regions, especially in the SH, while for AOD it aligns more closely with the observed trends in that hemisphere. These results indicate that despite considerable inter-model variability, the CMIP6 multi-model mean captures key spatial correlations reasonably well when compared with satellite observations.
As illustrated in Figure 5d–f, most CMIP6 models exhibit weaker performance in simulating LWP compared to AOD. The multi-model mean generally underestimates LWP relative to satellite observations, and in the SH falls outside the bounds of observational uncertainty. A total of 14 models fall within the observational uncertainty range for the global domain, 17 models for the NH, and only 7 for the SH, highlighting a clear decline in model performance from the NH to the SH. Several models lie near the edge of the uncertainty bounds, including members of the E3SM model family for the global domain, GFDL-ESM4 for the NH, and members of the CanESM5 model family and CESM2-WACCM-FV2 for the SH. MIROC6 falls outside the extended observational bounds in both the global and NH domains due to its overestimation of LWP and underestimation of AOD. INM-CM4-8 and INM-CM-5-0 show even greater deviation in the SH, with pronounced overestimation of AOD and substantial underestimation of LWP. Overall, model performance declines in the SH, as indicated by the smaller number of models falling within the observational uncertainty range and proximity of the multi-model mean to the edge of the uncertainty bounds. Notably, IPSL-CM6A-LR consistently overestimates LWP across all regions, placing it outside the uncertainty range in each case.
In general, CMIP6 models tend to underestimate both CF and LWP across the global domain and in both hemispheres, with the largest biases occurring in the SH. The multi-year, multi-model means of CMIP6-simulated AOD and oceanic LWP display a broader spread compared to CF, indicating greater inter-model variability and disagreement in the representation of these variables. Model performance is the highest in the NH, where CF is underestimated by only 2.6% compared to 5.6% in the SH. LWP is particularly poorly represented, with the multi-model mean exhibiting a systematic underestimation globally and the largest bias occurring in the SH, where the underestimation reaches 32.8%.

5.2. Statistical Evaluation of Model Performance

Taylor diagrams for AOD, CF, and LWP are shown in Figure 6. Among all 35 models with available AOD, INM-CM4-8 and INM-CM5-0 show consistently negative correlation with observations across the globe and both hemispheres. CanESM5 and CanESM5-1 show negative correlations in the NH (Figure 6b), while IPSL-CM6A-LR is negatively correlated in the SH (Figure 6c). CESM2-WACCM-FV2 shows the largest CRMSE globally and in the NH, indicating substantial deviation from observed spatial patterns. Overall, model performance for AOD is highest in the SH, with generally smaller CRMSE range (less than 0.3).
For LWP, most CMIP6 models exhibit strong positive correlations with observations, particularly in the NH, with a large number of models negatively correlated in the SH (Figure 6f), indicating variability in amplitude representation among models. Some models, such as GFDL-ESM4, GISS-E2-1-H, GISS-E2-2-G, and GISS-E2-2-H, show consistently high CRMSE across all domains. Model agreement with observations is generally better for CF (Figure 6g–i), with the majority of models clustering near the reference point, suggesting strong spatial pattern agreement and consistent variability. Correlation coefficients are high, with a number of models exceeding 0.95 across all domains, with reduced scatter in standard deviation, particularly in the NH. This indicates the overall better performance and consistency among CMIP6 models in simulating CF compared to AOD and LWP.
Overall, the Taylor diagram analysis reveals distinct differences in model performance across AOD, CF, and LWP. While most CMIP6 models capture CF patterns well across all domains, AOD performance varies more widely among models, with notably better agreement in the SH. In contrast, LWP simulations exhibit the most inconsistency, particularly in the SH, where many models show negative correlations and high CRMSE, suggesting challenges in capturing LWP variability. These differences underscore that while CMIP6 models generally perform well for CF, substantial uncertainties remain in the representation of aerosol optical properties and cloud water content, particularly in the SH.

5.3. The Scoring System

Figure 7 provides a comprehensive overview of the performance scores for all 56 CMIP6 models, using color-coded values to represent model skill in simulating AOD, CF, and LWP over the global domain, NH, and SH. The results show that most models perform better in the NH than in the SH. Although the scores across AOD, CF, and LWP are not directly comparable due to differing observational uncertainties and variable characteristics, a consistent pattern emerges that CF is generally simulated with higher accuracy than AOD and LWP across all regions. Nevertheless, certain models, such as CMCC-CM2-HR4, show large errors in CF simulations, which are accompanied by relatively low performance in LWP as well. Notably, four models (MPI-ESM-1-2-HAM, MRI-ESM2-0, NorESM2-LM, and NorESM2-MM) achieve scores exceeding 0.85 for all three variables and in all spatial domains, highlighting their overall robustness.
For AOD, CMIP6 models generally perform comparably in both the NH and SH, with 16 models achieving scores above 0.9 in each hemisphere. Among the 35 models evaluated, INM-CM4-8 and INM-CM-5-0 exhibit particularly low scores in the SH, consistent with the latitude profile in Figure 3, which highlights these models’ inability to capture the tropical AOD distribution and their substantial overestimation near 60°S. In contrast, most models perform better in simulating CF, except CMCC-CM2-HR4, which consistently underestimates CF across most parts of the global domain, as evident in its latitudinal profile. The wide range of model performance for LWP, both globally and between hemispheres, reflects the high degree of uncertainty and variability associated with the representation of cloud microphysical processes. This issue is particularly pronounced in the SH, where more than half of the models yield LWP scores below 0.5. Furthermore, several models that score well for CF exhibit poor performance for LWP, indicating inconsistencies in how models represent the spatial mean properties of clouds.
The ensemble means (Figure 7, last column) generally outperform individual models for AOD and CF across the global domain and both hemispheres. However, the mean LWP score in the SH remains below 0.5, reflecting a substantial bias in the spatially averaged LWP relative to satellite observations. This low score highlights the difficulty many models face in accurately representing LWP in the SH. In contrast, the multi-model mean effectively compensates for individual model biases in other variables, resulting in overall performance that often surpasses that of most individual models. This result reinforces the established practice in climate modeling of relying on multi-model ensembles, as ensemble means typically offer greater reliability and closer alignment with observational benchmarks than any single model, consistent with earlier model evaluation studies (e.g., Gleckler et al. [91]).
To systematically assess CMIP6 model performance across spatial scales, a composite performance score was calculated for each model. These overall scores, along with their corresponding rankings, are provided in Supplementary Table S3.
Among all models evaluated, IPSL-CM6A-LR-INCA achieved the highest global AOD score (0.989), closely followed by GFDL-ESM4 (0.988). For global CF, the top-performing models were E3SM-1-0, INM-CM4-8, and KACE-1-0-G (all 0.996), followed by E3SM-1-1, E3SM-1-1-ECA, and E3SM-2-0. Overall, CMIP6 models show better performance in simulating CF compared to AOD and LWP. Notably, only two models scored below 0.8: FGOALS-g3 (0.71) and CMCC-CM2-HR4 (0.56). In terms of global LWP, MRI-ESM2-0 yielded the highest score (0.988), followed by CIESM (0.978).
Over the NH, E3SM-2-0 recorded the highest AOD score (0.985), with ACCESS-CM2 ranking second (0.984). For CF in the NH, INM-CM4-8 again led with a score of 0.993, closely followed by CESM2-FV2 and CIESM (both 0.991). MRI-ESM2-0 continued to lead for LWP in the NH (0.960), followed by NorESM2-MM (0.955).
In the SH, CMCC-CM2-SR5 achieved the highest AOD score (0.988), with GFDL-CM4 close behind (0.987). For CF in the SH, INM-CM4-8 again ranked highest (0.994), followed by ACCESS-CM2 and TaiESM (both 0.984). For LWP, MRI-ESM2-0 remained the top-performing model (0.959), with MPI-ESM1.2-HAM a close second (0.958).
Overall, CMIP6 models demonstrate their strongest performance in simulating CF, whereas LWP remains comparatively challenging. A hemispheric comparison indicates generally lower model performance in the SH than in the NH, with the notable exception of AOD, for which models tend to perform better in the SH.
Analysis of individual models reveals particularly strong performances in specific variables. INM-CM4-8 consistently achieves the highest CF scores across all spatial scales, with an average near unity (0.994). Likewise, MRI-ESM2-0 excels in simulating LWP, maintaining the highest scores across all regions with an average score of 0.969, which is an especially notable result given the known challenges in simulating LWP. Nevertheless, certain outliers warrant attention: both FGOALS-f3-L and IPSL-CM5A2-INCA yield negative LWP scores in the SH, reinforcing the persistent challenges CMIP6 models face in accurately representing LWP in this region.
To derive an overall performance metric for each CMIP6 model, the individual scores for AOD, CF, and LWP were averaged. As not all models provide outputs for all three variables, only those with available data for AOD, CF, and LWP were included in the comprehensive performance assessment, resulting in a total of 32 models. As shown in Table 2, MRI-ESM2-0 achieved the highest comprehensive performance score at the global scale (0.931), followed by NorESM2-MM (0.923) and NorESM2-LM (0.919). The lowest global score was recorded by AWI-ESM-1-REcoM (0.732). In the NH, NorESM2-MM ranked the highest (0.941), with NorESM2-LM (0.931) and MRI-ESM2-0 (0.928) ranking second and third, respectively. MIROC6 had the lowest performance score in the NH (0.716). In the SH, IPSL-CM6A-LR-INCA ranked the highest (0.909), followed by MPI-ESM1.2-HAM (0.902) and MRI-ESM2-0 (0.902), with CESM2 close behind (0.901). The weakest-performing models in the SH were INM-CM4-8 (0.609) and INM-CM5-0 (0.601), primarily due to their poor simulation of LWP, with scores below 0.3.
As discussed earlier, most CMIP6 models exhibit notably lower skill in simulating LWP compared to AOD and CF. Consequently, LWP accuracy has a significant influence on a model’s overall performance score. Models that perform well for LWP tend to rank higher on comprehensive evaluation. This relationship explains why MRI-ESM2-0 consistently ranks among the top-performing models globally and across both hemispheres—it is the top-performing model for LWP.

6. Conclusions

This study assessed the multi-year mean distributions of aerosol and cloud properties simulated by CMIP6 models using satellite observations from MODIS and AMSR-E. Substantial variability is found across variables, with the largest discrepancies in LWP, which is consistently underestimated globally and across both hemispheres. AOD is generally overestimated, while CF is moderately well simulated. More specifically, we found that high-performing models, such as CESM2 and MRI-ESM2-0, contribute to more accurate LWP and AOD simulations, as the result of employing advanced aerosol and cloud schemes. Conversely, models with simpler aerosol representations often exhibit substantial biases, particularly in the SH (such as the INM model family). These inter-model differences reflect both external forcing and internal structural choices in aerosol activation, cloud formation, and radiative coupling.
Regional and zonal patterns further illustrate model limitations. More than half of the models overestimate AOD over north Africa and underestimate it over southern Africa. Correlation analysis shows that CMIP6 models generally achieve stronger agreement in the NH than in the SH, particularly for AOD and LWP. Specifically, the INM-CM4-8 and INM-CM5-0 models show limited skill in simulating AOD, while the FGOALS-f3-L and IPSL-CM5A2-INCA models exhibit poor agreement for LWP.
Analysis of the CF–AOD relationship reveals that models incorporating more detailed microphysical processes tend to reproduce the expected increase in CF with higher aerosol loading, an indication of the first indirect aerosol effect. However, some models, such as TaiESM, primarily reflect radiative suppression of clouds, raising concerns about how radiative forcing mechanisms are implemented in these models. In addition, results in the NH are more variable and less consistent with observations, and may be partially attributed to the re-gridding process, which can obscure localized features and reduce the fidelity of spatial signals.
A quantitative assessment of model performance based on spatial means reveals that CF is generally simulated with higher fidelity than AOD or LWP. Only one model, CMCC-CM2-HR4, falls significantly below the others in terms of performance score. In particular, LWP simulations in the SH are notably deficient, likely due to large underestimations in the tropics and the presence of the “double ITCZ” in the simulations. The multi-model mean supports these findings, with the SH LWP achieving a performance score below 0.5.
Overall, CMIP6 models show broad agreement with satellite observations, though substantial improvements are still needed, especially in the representation of LWP in the SH. Part of this discrepancy may stem from observational limitations, such as known biases in satellite AOD retrievals over the Southern Ocean, as well as the relative scarcity of ground-based validation data in the SH. While the evaluation methodology employed in this study is relatively straightforward, it provides a clear and quantitative framework for assessing model skill in simulating aerosol and cloud fields. Continued advancements in spatial and temporal resolution, as well as refinement of physical parameterizations, are essential for improving the credibility of aerosol–cloud interaction processes in global climate simulations.

7. Future Work

This study primarily investigated the spatial distribution of AOD, CF, and LWP in CMIP6 models by analyzing multi-year mean fields. However, a more comprehensive evaluation of model performance requires considerations of temporal variability. Future research should therefore expand to include the temporal evolution of these variables, with a focus on seasonal cycles, interannual variability, and short-term perturbations. Accurately capturing the timing and dynamics of CF, AOD, and LWP is essential for improving the predictive capabilities of climate models.
Additionally, our results reveal that many models struggle to reproduce the observed spatial patterns of these key climate variables. This underscores the need for further investigation into the physical processes that govern aerosol–cloud interactions. Improving the representation of such processes, particularly cloud microphysics, aerosol activation, and convective dynamics, could substantially improve the realism of simulated cloud and aerosol properties.
In summary, future work should prioritize both the refinement of physical parameterization and evaluation of model skill across a range of temporal scales. Advancing these areas is critical for reducing uncertainties in climate projections and enhancing our ability to simulate and anticipate future changes in the Earth’s climate system.

Supplementary Materials

The following supporting information can be downloaded at https://www.mdpi.com/article/10.3390/rs17142439/s1. Figure S1: Multi-Year Mean and Model–Satellite Differences of AOD, CF, and LWP from Individual CMIP6 Models and Satellite Observations; Table S1: Multi-Year Mean ± Standard Deviation of AOD, CF, and LWP from Individual CMIP6 Models over the Globe, Northern Hemisphere (NH), and Southern Hemisphere (SH); Table S2: Performance Scores and Rankings of Individual CMIP6 Models for AOD, CF, LWP, and the Combined Evaluation across the Globe, Northern Hemisphere (NH), and Southern Hemisphere (SH).

Author Contributions

Data curation, J.L.; software, J.L.; supervision, J.D.S.G.; writing—original draft, J.L.; writing—review and editing, J.D.S.G. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Data Availability Statement

All datasets are available online. The CMIP6 data used for this study can be downloaded from https://esgf-ui.ceda.ac.uk/cog/search/cmip6-ceda/ (accessed on 8 February 2025). The satellite datasets can be downloaded from https://ladsweb.modaps.eosdis.nasa.gov/search/ (accessed on 11 January 2025) for MODIS data and https://nsidc.org/data/amsre (accessed on 25 February 2025) for AMSR-E data. For additional questions regarding data availability, please contact the corresponding author at smalljen@hawaii.edu.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Rosenfeld, D. Aerosols, Clouds, and Climate. Science 2006, 312, 1323–1324. [Google Scholar] [CrossRef]
  2. Intergovernmental Panel on Climate Change (Ed.) Anthropogenic and Natural Radiative Forcing. In Climate Change 2013—The Physical Science Basis; Cambridge University Press: Cambridge, UK, 2014; pp. 659–740. [Google Scholar]
  3. Albrecht, B.A. Aerosols, Cloud Microphysics, and Fractional Cloudiness. Science 1989, 245, 1227–1230. [Google Scholar] [CrossRef]
  4. Li, J.; Carlson, B.E.; Yung, Y.L.; Lv, D.; Hansen, J.; Penner, J.E.; Liao, H.; Ramaswamy, V.; Kahn, R.A.; Zhang, P.; et al. Scattering and Absorbing Aerosols in the Climate System. Nat. Rev. Earth Environ. 2022, 3, 363–379. [Google Scholar] [CrossRef]
  5. Altaratz, O.; Koren, I.; Remer, L.; Hirsch, E. Review: Cloud Invigoration by Aerosols—Coupling between Microphysics and Dynamics. Atmos. Res. 2014, 140–141, 38–60. [Google Scholar] [CrossRef]
  6. Hansen, J.; Sato, M.; Ruedy, R. Radiative Forcing and Climate Response. J. Geophys. Res. Atmos. 1997, 102, 6831–6864. [Google Scholar] [CrossRef]
  7. Ackerman, A.S.; Toon, O.B.; Stevens, D.E.; Heymsfield, A.J.; Ramanathan, V.; Welton, E.J. Reduction of Tropical Cloudiness by Soot. Science 2000, 288, 1042–1047. [Google Scholar] [CrossRef]
  8. Hansen, J.; Sato, M.; Kharecha, P.; Russell, G.; Lea, D.W.; Siddall, M. Climate Change and Trace Gases. Philos. Trans. R. Soc. A Math. Phys. Eng. Sci. 2007, 365, 1925–1954. [Google Scholar] [CrossRef]
  9. Fan, J.; Wang, Y.; Rosenfeld, D.; Liu, X. Review of Aerosol–Cloud Interactions: Mechanisms, Significance, and Challenges. J. Atmos. Sci. 2016, 73, 4221–4252. [Google Scholar] [CrossRef]
  10. Seinfeld, J.H.; Bretherton, C.; Carslaw, K.S.; Coe, H.; DeMott, P.J.; Dunlea, E.J.; Feingold, G.; Ghan, S.; Guenther, A.B.; Kahn, R.; et al. Improving Our Fundamental Understanding of the Role of Aerosol−Cloud Interactions in the Climate System. Proc. Natl. Acad. Sci. USA 2016, 113, 5781–5790. [Google Scholar] [CrossRef]
  11. Proske, U.; Ferrachat, S.; Lohmann, U. Developing a Climatological Simplification of Aerosols to Enter the Cloud Microphysics of a Global Climate Model. Atmos. Chem. Phys. 2024, 24, 5907–5933. [Google Scholar] [CrossRef]
  12. Eyring, V.; Bony, S.; Meehl, G.A.; Senior, C.A.; Stevens, B.; Stouffer, R.J.; Taylor, K.E. Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) Experimental Design and Organization. Geosci. Model Dev. 2016, 9, 1937–1958. [Google Scholar] [CrossRef]
  13. Jiang, J.H.; Su, H.; Wu, L.; Zhai, C.; Schiro, K.A. Improvements in Cloud and Water Vapor Simulations Over the Tropical Oceans in CMIP6 Compared to CMIP5. Earth Space Sci. 2021, 8, e2020EA001520. [Google Scholar] [CrossRef]
  14. Randall, D.A.; Wood, R.A.; Bony, S.; Colman, R.; Fichefet, T.; Fyfe, J.; Kattsov, V.; Pitman, A.; Shukla, J.; Srinivasan, J.; et al. Climate Models and Their Evaluation. In Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change; Cambridge University Press: Cambridge, UK, 2007; pp. 589–662. [Google Scholar]
  15. Intergovernmental Panel on Climate Change. Climate Change 2014: Synthesis Report. Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change; IPCC: Geneva, Switzerland, 2014. [Google Scholar]
  16. Stephens, G.L.; L’ECuyer, T.; Forbes, R.; Gettelmen, A.; Golaz, J.; Bodas-Salcedo, A.; Suzuki, K.; Gabriel, P.; Haynes, J. Dreary State of Precipitation in Global Models. J. Geophys. Res. Atmos. 2010, 115, 2010JD014532. [Google Scholar] [CrossRef]
  17. Golaz, J.; Horowitz, L.W.; Levy, H. Cloud Tuning in a Coupled Climate Model: Impact on 20th Century Warming. Geophys. Res. Lett. 2013, 40, 2246–2251. [Google Scholar] [CrossRef]
  18. Suzuki, K.; Stephens, G.L.; Lebsock, M.D. Aerosol Effect on the Warm Rain Formation Process: Satellite Observations and Modeling. J. Geophys. Res. Atmos. 2013, 118, 170–184. [Google Scholar] [CrossRef]
  19. Bony, S.; Dufresne, J. Marine Boundary Layer Clouds at the Heart of Tropical Cloud Feedback Uncertainties in Climate Models. Geophys. Res. Lett. 2005, 32, 2005GL023851. [Google Scholar] [CrossRef]
  20. Turner, D.D.; Clough, S.A.; Liljegren, J.C.; Clothiaux, E.E.; Cady-Pereira, K.E.; Gaustad, K.L. Retrieving Liquid Wat0er Path and Precipitable Water Vapor from the Atmospheric Radiation Measurement (ARM) Microwave Radiometers. IEEE Trans. Geosci. Remote Sens. 2007, 45, 3680–3690. [Google Scholar] [CrossRef]
  21. Koren, I.; Martins, J.V.; Remer, L.A.; Afargan, H. Smoke Invigoration Versus Inhibition of Clouds over the Amazon. Science 2008, 321, 946–949. [Google Scholar] [CrossRef]
  22. Huang, J.; Lin, B.; Minnis, P.; Wang, T.; Wang, X.; Hu, Y.; Yi, Y.; Ayers, J.K. Satellite-based Assessment of Possible Dust Aerosols Semi-direct Effect on Cloud Water Path Over East Asia. Geophys. Res. Lett. 2006, 33, 2006GL026561. [Google Scholar] [CrossRef]
  23. Intergovernmental Panel on Climate Change (IPCC). Annex II: Glossary (Table AII.5). In Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change; Masson-Delmotte, V., Zhai, P., Pirani, A., Connors, S.L., Péan, C., Berger, S., Caud, N., Chen, Y., Goldfarb, L., Gomis, M.I., et al., Eds.; Cambridge University Press: Cambridge, UK, 2021; pp. 2247–2258. [Google Scholar] [CrossRef]
  24. Wang, C.; Soden, B.J.; Yang, W.; Vecchi, G.A. Compensation Between Cloud Feedback and Aerosol-Cloud Interaction in CMIP6 Models. Geophys. Res. Lett. 2021, 48, e2020GL091024. [Google Scholar] [CrossRef]
  25. Chen, G.; Wang, W.; Bao, Q.; Li, J. Evaluation of Simulated Cloud Diurnal Variation in CMIP6 Climate Models. J. Geophys. Res. Atmos. 2022, 127, e2021JD036422. [Google Scholar] [CrossRef]
  26. Zhao, A.; Ryder, C.L.; Wilcox, L.J. How Well Do the CMIP6 Models Simulate dust Aerosols? Atmos. Chem. Phys. 2022, 22, 2095–2119. [Google Scholar] [CrossRef]
  27. Taylor, K.E.; Stouffer, R.J.; Meehl, G.A. An Overview of CMIP5 and the Experiment Design. Bull. Am. Meteorol. Soc. 2012, 93, 485–498. [Google Scholar] [CrossRef]
  28. Bi, D.; Dix, M.; Marsland, S.; O’fArrell, S.; Sullivan, A.; Bodman, R.; Law, R.; Harman, I.; Srbinovsky, J.; Rashid, H.A.; et al. Configuration and Spin-up of ACCESS-CM2, the New Generation Australian Community Climate and Earth System Simulator Coupled Model. J. South. Hemisph. Earth Syst. Sci. 2020, 70, 225–251. [Google Scholar] [CrossRef]
  29. Ziehn, T.; Chamberlain, M.A.; Law, R.M.; Lenton, A.; Bodman, R.W.; Dix, M.; Stevens, L.; Wang, Y.-P.; Srbinovsky, J. The Australian Earth System Model: ACCESS-ESM1.5. J. South. Hemisph. Earth Syst. Sci. 2020, 70, 193–214. [Google Scholar] [CrossRef]
  30. Sidorenko, D.; Rackow, T.; Jung, T.; Semmler, T.; Barbi, D.; Danilov, S.; Dethloff, K.; Dorn, W.; Fieg, K.; Goessling, H.F.; et al. Towards Multi-Resolution Global Climate Modeling with ECHAM6–FESOM. Part I: Model Formulation and Mean Climate. Clim. Dyn. 2014, 44, 757–780. [Google Scholar] [CrossRef]
  31. Semmler, T.; Danilov, S.; Gierz, P.; Goessling, H.F.; Hegewald, J.; Hinrichs, C.; Koldunov, N.; Khosravi, N.; Mu, L.; Rackow, T.; et al. Simulations for CMIP6 With the AWI Climate Model AWI-CM-1-1. J. Adv. Model. Earth Syst. 2020, 12, e2019MS002009. [Google Scholar] [CrossRef]
  32. Danek, C.; Gürses, Ö.; Gierz, P.; Andrés-Martínez, M.; Hauck, J. AWI AWI-ESM1REcoM Model Output Prepared for CMIP6 ScenarioMIP ssp245, Version 20200212; Earth System Grid Federation, 2023. [Google Scholar] [CrossRef]
  33. Wu, T.; Lu, Y.; Fang, Y.; Xin, X.; Li, L.; Li, W.; Jie, W.; Zhang, J.; Liu, Y.; Zhang, L.; et al. The Beijing Climate Center Climate System Model (BCC-CSM): The Main Progress from CMIP5 to CMIP6. Geosci. Model Dev. 2019, 12, 1573–1600. [Google Scholar] [CrossRef]
  34. Wu, T.; Zhang, F.; Zhang, J.; Jie, W.; Zhang, Y.; Wu, F.; Li, L.; Yan, J.; Liu, X.; Lu, X.; et al. Beijing Climate Center Earth System Model Version 1 (BCC-ESM1): Model Description and Evaluation of Aerosol Simulations. Geosci. Model Dev. 2020, 13, 977–1005. [Google Scholar] [CrossRef]
  35. Rong, X.; Li, J.; Chen, H.; Xin, Y.; Su, J.; Hua, L.; Zhou, T.; Qi, Y.; Zhang, Z.; Zhang, G.; et al. The CAMS Climate System Model and a Basic Evaluation of Its Climatology and Climate Variability Simulation. J. Meteorol. Res. 2018, 32, 839–861. [Google Scholar] [CrossRef]
  36. Chinese Academy of Sciences. WCRP CMIP6: Chinese Academy of Sciences (CAS) FGOALS-g3 Model Output for the "Hist-GHG" Experiment. 2021. Available online: https://catalogue.ceda.ac.uk/uuid/0f392b23d8934abc86c005feb15240f1 (accessed on 12 July 2025).
  37. Danabasoglu, G.; Lamarque, J.; Bacmeister, J.; Bailey, D.A.; DuVivier, A.K.; Edwards, J.; Emmons, L.K.; Fasullo, J.; Garcia, R.; Gettelman, A.; et al. The Community Earth System Model Version 2 (CESM2). J. Adv. Model. Earth Syst. 2020, 12, e2019MS001916. [Google Scholar] [CrossRef]
  38. Lin, Y.; Huang, X.; Liang, Y.; Qin, Y.; Xu, S.; Huang, W.; Xu, F.; Liu, L.; Wang, Y.; Peng, Y.; et al. Community Integrated Earth System Model (CIESM): Description and Evaluation. J. Adv. Model. Earth Syst. 2020, 12, e2019MS002036. [Google Scholar] [CrossRef]
  39. Cherchi, A.; Fogli, P.G.; Lovato, T.; Peano, D.; Iovino, D.; Gualdi, S.; Masina, S.; Scoccimarro, E.; Materia, S.; Bellucci, A.; et al. Global Mean Climate and Main Patterns of Variability in the CMCC-CM2 Coupled Model. J. Adv. Model. Earth Syst. 2019, 11, 185–209. [Google Scholar] [CrossRef]
  40. Swart, N.C.; Cole, J.N.S.; Kharin, V.V.; Lazare, M.; Scinocca, J.F.; Gillett, N.P.; Anstey, J.; Arora, V.; Christian, J.R.; Hanna, S.; et al. The Canadian Earth System Model Version 5 (CanESM5.0.3). Geosci. Model Dev. 2019, 12, 4823–4873. [Google Scholar] [CrossRef]
  41. Golaz, J.; Caldwell, P.M.; Van Roekel, L.P.; Petersen, M.R.; Tang, Q.; Wolfe, J.D.; Abeshu, G.; Anantharaj, V.; Asay-Davis, X.S.; Bader, D.C.; et al. The DOE E3SM Coupled Model Version 1: Overview and Evaluation at Standard Resolution. J. Adv. Model. Earth Syst. 2019, 11, 2089–2129. [Google Scholar] [CrossRef]
  42. Döscher, R.; Acosta, M.; Alessandri, A.; Anthoni, P.; Arsouze, T.; Bergman, T.; Bernardello, R.; Boussetta, S.; Caron, L.-P.; Carver, G.; et al. The EC-Earth3 Earth System Model for the Coupled Model Intercomparison Project 6. Geosci. Model Dev. 2022, 15, 2973–3020. [Google Scholar] [CrossRef]
  43. He, B.; Yu, Y.; Bao, Q.; Lin, P.; Liu, H.; Li, J.; Wang, L.; Liu, Y.; Wu, G.; Chen, K.; et al. CAS FGOALS-f3-L Model Dataset Descriptions for CMIP6 DECK Experiments. Atmos. Ocean. Sci. Lett. 2020, 13, 582–588. [Google Scholar] [CrossRef]
  44. Li, L.; Yu, Y.; Tang, Y.; Lin, P.; Xie, J.; Song, M.; Dong, L.; Zhou, T.; Liu, L.; Wang, L.; et al. The Flexible Global Ocean-Atmosphere-Land System Model Grid-Point Version 3 (FGOALS-g3): Description and Evaluation. J. Adv. Model. Earth Syst. 2020, 12, e2019MS002012. [Google Scholar] [CrossRef]
  45. Bao, Y.; Song, Z.; Qiao, F. FIO-ESM Version 2.0: Model Description and Evaluation. J. Geophys. Res. Oceans 2020, 125, e2019JC016036. [Google Scholar] [CrossRef]
  46. Held, I.M.; Guo, H.; Adcroft, A.; Dunne, J.P.; Horowitz, L.W.; Krasting, J.; Shevliakova, E.; Winton, M.; Zhao, M.; Bushuk, M.; et al. Structure and Performance of GFDL’s CM4.0 Climate Model. J. Adv. Model. Earth Syst. 2019, 11, 3691–3727. [Google Scholar] [CrossRef]
  47. Dunne, J.P.; Horowitz, L.W.; Adcroft, A.J.; Ginoux, P.; Held, I.M.; John, J.G.; Krasting, J.P.; Malyshev, S.; Naik, V.; Paulot, F.; et al. The GFDL Earth System Model Version 4.1 (GFDL-ESM 4.1): Overall Coupled Model Description and Simulation Characteristics. J. Adv. Model. Earth Syst. 2020, 12, e2019MS002015. [Google Scholar] [CrossRef]
  48. Kelley, M.; Schmidt, G.A.; Nazarenko, L.S.; Bauer, S.E.; Ruedy, R.; Russell, G.L.; Ackerman, A.S.; Aleinov, I.; Bauer, M.; Bleck, R.; et al. GISS-E2.1: Configurations and Climatology. J. Adv. Model. Earth Syst. 2020, 12, e2019MS002025. [Google Scholar] [CrossRef]
  49. Rind, D.; Orbe, C.; Jonas, J.; Nazarenko, L.; Zhou, T.; Kelley, M.; Lacis, A.; Shindell, D.; Faluvegi, G.; Romanou, A.; et al. GISS Model E2.2: A Climate Model Optimized for the Middle Atmosphere—Model Structure, Climatology, Variability, and Climate Sensitivity. J. Geophys. Res. Atmos. 2020, 125, e2019JD032204. [Google Scholar] [CrossRef]
  50. Swapna, P.; Krishnan, R.; Sandeep, N.; Prajeesh, A.G.; Ayantika, D.C.; Manmeet, S.; Vellore, R. Long-Term Climate Simulations Using the IITM Earth System Model (IITM-ESMv2) with Focus on the South Asian Monsoon. J. Adv. Model. Earth Syst. 2018, 10, 1127–1149. [Google Scholar] [CrossRef]
  51. Volodin, E.M.; Mortikov, E.V.; Kostrykin, S.V.; Galin, V.Y.; Lykossov, V.N.; Gritsun, A.S.; Diansky, N.A.; Gusev, A.V.; Iakovlev, N.G.; Shestakova, A.A.; et al. Simulation of the Modern Climate Using the INM-CM48 Climate Model. Russ. J. Numer. Anal. Math. Model. 2018, 33, 367–374. [Google Scholar] [CrossRef]
  52. Volodin, E.M.; Mortikov, E.V.; Kostrykin, S.V.; Galin, V.Y.; Lykossov, V.N.; Gritsun, A.S.; Diansky, N.A.; Gusev, A.V.; Iakovlev, N.G. Simulation of the Present-Day Climate with the Climate Model INMCM5. Clim. Dyn. 2017, 49, 3715–3734. [Google Scholar] [CrossRef]
  53. Boucher, O.; Servonnat, J.; Albright, A.L.; Aumont, O.; Balkanski, Y.; Bastrikov, V.; Bekki, S.; Bonnet, R.; Bony, S.; Bopp, L.; et al. Presentation and Evaluation of the IPSL-CM6A-LR Climate Model. J. Adv. Model. Earth Syst. 2020, 12, e2019MS002010. [Google Scholar] [CrossRef]
  54. Lee, J.; Kim, J.; Sun, M.-A.; Kim, B.-H.; Moon, H.; Sung, H.M.; Kim, J.; Byun, Y.-H. Evaluation of the Korea Meteorological Administration Advanced Community Earth-System model (K-ACE). Asia-Pacific J. Atmos. Sci. 2019, 56, 381–395. [Google Scholar] [CrossRef]
  55. Pak, G.; Noh, Y.; Lee, M.-I.; Yeh, S.-W.; Kim, D.; Kim, S.-Y.; Lee, J.-L.; Lee, H.J.; Hyun, S.-H.; Lee, K.-Y.; et al. Korea Institute of Ocean Science and Technology Earth System Model and Its Simulation Characteristics. Ocean Sci. J. 2021, 56, 18–45. [Google Scholar] [CrossRef]
  56. Tatebe, H.; Ogura, T.; Nitta, T.; Komuro, Y.; Ogochi, K.; Takemura, T.; Sudo, K.; Sekiguchi, M.; Abe, M.; Saito, F.; et al. Description and Basic Evaluation of Simulated Mean State, Internal Variability, and Climate Sensitivity in MIROC6. Geosci. Model Dev. 2019, 12, 2727–2765. [Google Scholar] [CrossRef]
  57. Mauritsen, T.; Bader, J.; Becker, T.; Behrens, J.; Bittner, M.; Brokopf, R.; Brovkin, V.; Claussen, M.; Crueger, T.; Esch, M.; et al. Developments in the MPI-M Earth System Model version 1.2 (MPI-ESM1.2) and Its Response to Increasing CO2. J. Adv. Model. Earth Syst. 2019, 11, 998–1038. [Google Scholar] [CrossRef]
  58. Müller, W.A.; Jungclaus, J.H.; Mauritsen, T.; Baehr, J.; Bittner, M.; Budich, R.; Bunzel, F.; Esch, M.; Ghosh, R.; Haak, H.; et al. A Higher-resolution Version of the Max Planck Institute Earth System Model (MPI-ESM1.2-HR). J. Adv. Model. Earth Syst. 2018, 10, 1383–1413. [Google Scholar] [CrossRef]
  59. Mizuta, R.; Yoshimura, H.; Murakami, H.; Matsueda, M.; Endo, H.; Ose, T.; Kamiguchi, K.; Hosaka, M.; Sugi, M.; Yukimoto, S.; et al. Climate Simulations Using MRI-AGCM3.2 with 20-km Grid. J. Meteorol. Soc. Jpn. Ser. II 2012, 90A, 233–258. [Google Scholar] [CrossRef]
  60. Cao, J.; Wang, B.; Yang, Y.-M.; Ma, L.; Li, J.; Sun, B.; Bao, Y.; He, J.; Zhou, X.; Wu, L. The NUIST Earth System Model (NESM) Version 3: Description and Preliminary Evaluation. Geosci. Model Dev. 2018, 11, 2975–2993. [Google Scholar] [CrossRef]
  61. Seland, Ø.; Bentsen, M.; Olivié, D.; Toniazzo, T.; Gjermundsen, A.; Graff, L.S.; Debernard, J.B.; Gupta, A.K.; He, Y.-C.; Kirkevåg, A.; et al. Overview of the Norwegian Earth System Model (NorESM2) and Key Climate Response of CMIP6 DECK, Historical, and Scenario Simulations. Geosci. Model Dev. 2020, 13, 6165–6200. [Google Scholar] [CrossRef]
  62. Park, S.; Shin, J. SNU SAM0-UNICON Model Output Prepared for CMIP6 CMIP, Version 20230720; World Data Center for Climate (WDCC): Hamburg, Germany, 2019. [Google Scholar] [CrossRef]
  63. Tsai, I.-C.; Lee, W.-L.; Hsu, H.-H. AS-RCEC TaiESM1.0 Model Output Prepared for CMIP6 AerChemMIP, Version 20220610; World Data Center for Climate (WDCC): Hamburg, Germany, 2020. [Google Scholar] [CrossRef]
  64. Barnes, W.; Pagano, T.; Salomonson, V. Prelaunch Characteristics of the Moderate Resolution Imaging Spectroradiometer (MODIS) on EOS-AM1. IEEE Trans. Geosci. Remote. Sens. 1998, 36, 1088–1100. [Google Scholar] [CrossRef]
  65. Platnick, S.; Meyer, K.G.; King, M.D.; Wind, G.; Amarasinghe, N.; Marchant, B.; Arnold, G.T.; Zhang, Z.; Hubanks, P.A.; Holz, R.E.; et al. The MODIS Cloud Optical and Microphysical Products: Collection 6 Updates and Examples from Terra and Aqua. IEEE Trans. Geosci. Remote. Sens. 2016, 55, 502–525. [Google Scholar] [CrossRef]
  66. Kotarba, A.Z. Calibration of Global MODIS Cloud Amount Using CALIOP Cloud Profiles. Atmos. Meas. Tech. 2020, 13, 4995–5012. [Google Scholar] [CrossRef]
  67. Sayer, A.M.; Munchak, L.A.; Hsu, N.C.; Levy, R.C.; Bettenhausen, C.; Jeong, M.-J. MODIS Collection 6 Aerosol Products: Comparison between Aqua’s e-Deep Blue, Dark Target, and “Merged” Data Sets, and Usage Recommendations. J. Geophys. Res. Atmos. 2014, 119, 13965–13989. [Google Scholar] [CrossRef]
  68. Greenwald, T.J. A 2 Year Comparison of AMSR-E and MODIS Cloud Liquid Water Path Observations. Geophys. Res. Lett. 2009, 36, 2009GL040394. [Google Scholar] [CrossRef]
  69. Bojinski, S.; Verstraete, M.; Peterson, T.C.; Richter, C.; Simmons, A.; Zemp, M. The Concept of Essential Climate Variables in Support of Climate Research, Applications, and Policy. Bull. Am. Meteorol. Soc. 2014, 95, 1431–1443. [Google Scholar] [CrossRef]
  70. Heymsfield, A.J.; Matrosov, S.; Baum, B. Ice Water Path–Optical Depth Relationships for Cirrus and Deep Stratiform Ice Cloud Layers. J. Appl. Meteorol. Clim. 2003, 42, 1369–1390. [Google Scholar] [CrossRef]
  71. Sengupta, M.; Clothiaux, E.E.; Ackerman, T.P.; Kato, S.; Min, Q. Importance of Accurate Liquid Water Path for Estimation of Solar Radiation in Warm Boundary Layer Clouds: An Observational Study. J. Clim. 2003, 16, 2997–3009. [Google Scholar] [CrossRef]
  72. Shupe, M.D.; Intrieri, J.M. Cloud Radiative Forcing of the Arctic Surface: The Influence of Cloud Properties, Surface Albedo, and Solar Zenith Angle. J. Clim. 2004, 17, 616–628. [Google Scholar] [CrossRef]
  73. Nakajima, T.; King, M.D. Determination of the Optical Thickness and Effective Particle Radius of Clouds from Reflected Solar Radiation Measurements. Part I: Theory. J. Atmos. Sci. 1990, 47, 1878–1893. [Google Scholar] [CrossRef]
  74. Platnick, S.; King, M.D.; Ackerman, S.A.; Menzel, W.P.; Baum, B.A.; Riédi, J.C.; Frey, R.A. The MODIS Cloud Products: Algorithms and Examples from Terra. IEEE Trans. Geosci. Remote Sens. 2003, 41, 459–473. [Google Scholar] [CrossRef]
  75. Kato, S.; Marshak, A. Solar Zenith and Viewing Geometry-Dependent Errors in Satellite Retrieved Cloud Optical Thickness: Marine Stratocumulus Case. J. Geophys. Res. Atmos. 2009, 114, 2008JD010579. [Google Scholar] [CrossRef]
  76. Seethala, C.; Horváth, Á. Global Assessment of AMSR-E and MODIS Cloud Liquid Water Path Retrievals in Warm Oceanic Clouds. J. Geophys. Res. Atmos. 2010, 115, 2009JD012662. [Google Scholar] [CrossRef]
  77. Grosvenor, D.P.; Wood, R. The Effect of Solar Zenith Angle on MODIS Cloud Optical and Microphysical Retrievals Within Marine Liquid Water Clouds. Atmos. Chem. Phys. 2014, 14, 7291–7321. [Google Scholar] [CrossRef]
  78. Khanal, S.; Wang, Z. Uncertainties in MODIS-Based Cloud Liquid Water Path Retrievals at High Latitudes Due to Mixed-Phase Clouds and Cloud Top Height Inhomogeneity. J. Geophys. Res. Atmos. 2018, 123, 11154–11172. [Google Scholar] [CrossRef]
  79. Jiang, J.H.; Su, H.; Zhai, C.; Perun, V.S.; Del Genio, A.; Nazarenko, L.S.; Donner, L.J.; Horowitz, L.; Seman, C.; Cole, J.; et al. Evaluation of Cloud and Water Vapor Simulations in CMIP5 Climate Models Using NASA “A-Train” Satellite Observations. J. Geophys. Res. Atmos. 2012, 117, 2011JD017237. [Google Scholar] [CrossRef]
  80. Lauer, A.; Hamilton, K. Simulating Clouds with Global Climate Models: A Comparison of CMIP5 Results with CMIP3 and Satellite Data. J. Clim. 2013, 26, 3823–3845. [Google Scholar] [CrossRef]
  81. Wentz, F.J. A Well-calibrated Ocean Algorithm for Special Sensor Microwave/Imager. J. Geophys. Res. Ocean. 1997, 102, 8703–8718. [Google Scholar] [CrossRef]
  82. Wentz, F.; Meissner, T. AMSR-E/Aqua L2B Global Swath Ocean Products Derived from Wentz Algorithm, Version 2; NASA National Snow and Ice Data Center: Boulder, CO, USA, 2004. [Google Scholar] [CrossRef]
  83. Deeter, M.N.; Vivekanandan, J. New Dual-frequency Microwave Technique for Retrieving Liquid Water Path Over Land. J. Geophys. Res. Atmos. 2006, 111. [Google Scholar] [CrossRef]
  84. Douglass, A.R.; Prather, M.J.; Hall, T.M.; Strahan, S.E.; Rasch, P.J.; Sparling, L.C.; Coy, L.; Rodriguez, J.M. Choosing Meteorological Input for the Global Modeling Initiative Assessment of High-speed Aircraft. J. Geophys. Res. Atmos. 1999, 104, 27545–27564. [Google Scholar] [CrossRef]
  85. Ali, A.; Bilal, M.; Wang, Y.; Qiu, Z.; Nichol, J.E.; de Leeuw, G.; Ke, S.; Mhawish, A.; Almazroui, M.; Mazhar, U.; et al. Evaluation and Comparison of CMIP6 Models and MERRA-2 Reanalysis AOD Against Satellite Observations from 2000 to 2014 over China. Geosci. Front. 2022, 13, 101325. [Google Scholar] [CrossRef]
  86. Vignesh, P.P.; Jiang, J.H.; Kishore, P.; Su, H.; Smay, T.; Brighton, N.; Velicogna, I. Assessment of CMIP6 Cloud Fraction and Comparison with Satellite Observations. Earth Space Sci. 2020, 7, e2019EA000975. [Google Scholar] [CrossRef]
  87. Rosenfeld, D.; Lohmann, U.; Raga, G.B.; O’Dowd, C.D.; Kulmala, M.; Fuzzi, S.; Reissell, A.; Andreae, M.O. Flood or Drought: How Do Aerosols Affect Precipitation? Science 2008, 321, 1309–1313. [Google Scholar] [CrossRef]
  88. Jin, S.; Ma, Y.; Li, H.; Liu, B.; Fan, R.; Zhang, M.; Lopatin, A.; Dubovik, O.; Hu, X.; Gong, W.; et al. Characterizing Aerosol Optical Properties and Direct Radiative Effects from the Perspective of Components: A Synergy Retrieval Study Based on Sun Photometer and Lidar in Central China. Geophys. Res. Lett. 2025, 52, e2024GL113448. [Google Scholar] [CrossRef]
  89. Wang, L.; Su, X.; Wang, Y.; Cao, M.; Lang, Q.; Li, H.; Sun, J.; Zhang, M.; Qin, W.; Li, L.; et al. Towards Long-Term, High-Accuracy, and Continuous Satellite Total and Fine-Mode Aerosol Records: Enhanced Land General Aerosol (e-LaGA) Retrieval Algorithm for VIIRS. ISPRS J. Photogramm. Remote Sens. 2024, 214, 261–281. [Google Scholar] [CrossRef]
  90. Wang, L.; Lang, Q.; Wang, Z.; Feng, L.; Zhang, M.; Qin, W. Quantifying and Mitigating Errors in Estimating Downward Surface Shortwave Radiation Caused by Cloud Mask Data. IEEE Trans. Geosci. Remote Sens. 2024, 62, 1–15. [Google Scholar] [CrossRef]
  91. Gleckler, P.J.; Taylor, K.E.; Doutriaux, C. Performance Metrics for Climate Models. J. Geophys. Res. Atmos. 2008, 113, 2007JD008972. [Google Scholar] [CrossRef]
Remotesensing 17 02439 g001
Figure 1. Multi-year mean AOD (a,b), CF (d,e), and LWP (g,h) from CMIP6 (a,d,g), MODIS (b,e), and AMSR-E (h). The color bars on the right indicate the magnitude of CF, AOD, and LWP, with yellow representing higher values and blue representing lower values. Subplots (c,f,i) show the differences between the multi-year, multi-model mean from CMIP6 and satellite observations for AOD, CF, and LWP, respectively. The different color bars on the right highlight discrepancies, where red denotes positive biases and blue denotes negative biases in the model simulations compared to observations.
Figure 1. Multi-year mean AOD (a,b), CF (d,e), and LWP (g,h) from CMIP6 (a,d,g), MODIS (b,e), and AMSR-E (h). The color bars on the right indicate the magnitude of CF, AOD, and LWP, with yellow representing higher values and blue representing lower values. Subplots (c,f,i) show the differences between the multi-year, multi-model mean from CMIP6 and satellite observations for AOD, CF, and LWP, respectively. The different color bars on the right highlight discrepancies, where red denotes positive biases and blue denotes negative biases in the model simulations compared to observations.
Remotesensing 17 02439 g001
Remotesensing 17 02439 g002
Figure 2. Differences between the multi-year mean values from CMIP6 models and satellite observations for AOD, CF, and LWP across global, NH, and SH regions. The last column represents the average difference between the CMIP6 multi-year multi-model mean and satellite observations for each variable. The y-axis represents AOD, CF, and LWP, categorized by region (G: global; N: Northern Hemisphere; S: Southern Hemisphere), while the x-axis represents different CMIP6 models. Gray-shaded areas indicate models that do not include the particular variable. The LWP values were converted from g/m2 to kg/m2 to align with the magnitude of AOD and CF values in the color bar.
Figure 2. Differences between the multi-year mean values from CMIP6 models and satellite observations for AOD, CF, and LWP across global, NH, and SH regions. The last column represents the average difference between the CMIP6 multi-year multi-model mean and satellite observations for each variable. The y-axis represents AOD, CF, and LWP, categorized by region (G: global; N: Northern Hemisphere; S: Southern Hemisphere), while the x-axis represents different CMIP6 models. Gray-shaded areas indicate models that do not include the particular variable. The LWP values were converted from g/m2 to kg/m2 to align with the magnitude of AOD and CF values in the color bar.
Remotesensing 17 02439 g002
Remotesensing 17 02439 g003
Figure 3. The latitude profiles of multi-year mean AOD (a), CF (b), and LWP (c) for CMIP6 models (colored lines) and satellite observations (black lines, (a,b) from MODIS, (c) from AMSR-E). The gray-shaded areas indicate the multi-year mean satellite observation ± STD. (d) Correlations between the latitude profiles of the CMIP6 models and the satellite observations for AOD, CF, and LWP for global, NH, and SH, respectively. The y-axis indicates the three variables AOD, CF, and LWP, and different zones (G for global; N for Northern Hemisphere; S for Southern Hemisphere). The x-axis indicates different CMIP6 models. Converted from g/m2 to kg/m2 to match the magnitude of AOD and CF values in the color bar.
Figure 3. The latitude profiles of multi-year mean AOD (a), CF (b), and LWP (c) for CMIP6 models (colored lines) and satellite observations (black lines, (a,b) from MODIS, (c) from AMSR-E). The gray-shaded areas indicate the multi-year mean satellite observation ± STD. (d) Correlations between the latitude profiles of the CMIP6 models and the satellite observations for AOD, CF, and LWP for global, NH, and SH, respectively. The y-axis indicates the three variables AOD, CF, and LWP, and different zones (G for global; N for Northern Hemisphere; S for Southern Hemisphere). The x-axis indicates different CMIP6 models. Converted from g/m2 to kg/m2 to match the magnitude of AOD and CF values in the color bar.
Remotesensing 17 02439 g003
Remotesensing 17 02439 g004
Figure 4. CF versus AOD plots generated for different zones: (a,d) global; (b,e) NH; (c,f) SH. Subplots (a–c) represent the amount of data in the CMIP6 models and satellite observations datasets. For subplots (d–f), the x-axis represents AOD and is on a logarithmic scale to provide a closer view of the details with small AOD values. The y-axis represents the percentage of mean CF. The legend identifies the models represented by each colored line, with MODIS indicated by a black line. Error bars are included to display the standard error, calculated by dividing the standard deviation of each bin by the square root of the number of observations in each bin.
Figure 4. CF versus AOD plots generated for different zones: (a,d) global; (b,e) NH; (c,f) SH. Subplots (a–c) represent the amount of data in the CMIP6 models and satellite observations datasets. For subplots (d–f), the x-axis represents AOD and is on a logarithmic scale to provide a closer view of the details with small AOD values. The y-axis represents the percentage of mean CF. The legend identifies the models represented by each colored line, with MODIS indicated by a black line. Error bars are included to display the standard error, calculated by dividing the standard deviation of each bin by the square root of the number of observations in each bin.
Remotesensing 17 02439 g004
Remotesensing 17 02439 g005
Figure 5. Multi-year mean CF versus AOD (a–c) and LWP versus AOD (d–f) averaged across the globe (a,d), NH (b,e), and SH (c,f). Observational multi-year means are represented by the black dots, with the gray area indicating the observational uncertainties. Multi-year means from the CMIP6 models are shown as colored dots/triangles and open circles/triangles, with black circles denoting the CMIP6 multi-model means.
Figure 5. Multi-year mean CF versus AOD (a–c) and LWP versus AOD (d–f) averaged across the globe (a,d), NH (b,e), and SH (c,f). Observational multi-year means are represented by the black dots, with the gray area indicating the observational uncertainties. Multi-year means from the CMIP6 models are shown as colored dots/triangles and open circles/triangles, with black circles denoting the CMIP6 multi-model means.
Remotesensing 17 02439 g005
Remotesensing 17 02439 g006aRemotesensing 17 02439 g006b
Figure 6. Taylor diagrams showing the multi-year mean AOD (a–c), LWP (d–f), and CF (g–i) from CMIP6 models across the global and two hemispheres. The horizontal axis represents the standard deviation of the models. The correlation coefficients are expressed by the numbers on the black arc. The centered root mean square error (CRMSE) is shown as green arcs, indicating the difference between the modeled and observed values.
Figure 6. Taylor diagrams showing the multi-year mean AOD (a–c), LWP (d–f), and CF (g–i) from CMIP6 models across the global and two hemispheres. The horizontal axis represents the standard deviation of the models. The correlation coefficients are expressed by the numbers on the black arc. The centered root mean square error (CRMSE) is shown as green arcs, indicating the difference between the modeled and observed values.
Remotesensing 17 02439 g006aRemotesensing 17 02439 g006b
Remotesensing 17 02439 g007
Figure 7. Color-coded overview of the performance scores of CMIP6 models for simulating AOD, CF, and LWP. These scores are determined using the evaluation method introduced by Douglass et al. [84]. Each model’s performance is assessed based on its spatial mean. The y-axis denotes three variables—AOD, CF, and LWP—and different zones (G for global; N for Northern Hemisphere; S for Southern Hemisphere). The x-axis indicates different CMIP6 models. The color-bar scale ranges from 1 (dark green, indicating perfect skill) to 0 (dark red, indicating no skill).
Figure 7. Color-coded overview of the performance scores of CMIP6 models for simulating AOD, CF, and LWP. These scores are determined using the evaluation method introduced by Douglass et al. [84]. Each model’s performance is assessed based on its spatial mean. The y-axis denotes three variables—AOD, CF, and LWP—and different zones (G for global; N for Northern Hemisphere; S for Southern Hemisphere). The x-axis indicates different CMIP6 models. The color-bar scale ranges from 1 (dark green, indicating perfect skill) to 0 (dark red, indicating no skill).
Remotesensing 17 02439 g007
Table 2. Overall scores and ranks for the CMIP6 models for all three variables (AOD, CF, and LWP).
Table 2. Overall scores and ranks for the CMIP6 models for all three variables (AOD, CF, and LWP).
ModelGlobalNorthern HemisphereSouthern Hemisphere
ScoreRankScoreRankScoreRank
AWI, ESM-1-1-LR0.741270.777230.68627
AWI, ESM-1-REcoM0.732290.770250.67728
BCC, ESM10.853160.886130.80516
CESM20.91750.92140.9013
CESM2-WACCM0.91840.92140.8974
CESM2-FV20.89870.90680.8755
CESM2-WACCM-FV20.89190.898100.86110
CMCC, CM2-SR50.748240.763260.72621
CMCC, ESM20.747250.762270.72422
CanESM50.89380.90190.82712
CanESM5-10.889100.90770.83111
E3SM-1-00.876130.883140.8687
E3SM-1-10.885110.898100.8668
E3SM-1-1-ECA0.875140.887120.8639
E3SM-2-00.889100.90860.8706
EC-Earth30.773210.811190.70825
EC-Earth3-AerChem0.791180.835160.72023
EC-Earth3-Veg0.773210.810200.71024
GFDL, ESM40.867150.890110.81115
INM, CM4-80.751220.826180.60930
INM, CM5-00.749230.828170.60131
IPSL, CM6A-LR0.778200.756280.79717
IPSL, CM6A-LR-INCA0.882120.855150.9091
KACE-1-0-G0.834170.835160.78418
MIROC60.741270.716290.75120
MPI, ESM1.2-HAM0.91560.91750.9022
MPI, ESM1.2-HR0.733280.776240.66829
MPI, ESM1.2-LR0.743260.780220.68926
MRI, ESM2-00.93110.92830.9022
NorESM2-LM0.91930.93120.82014
NorESM2-MM0.92320.94110.82513
TaiESM0.790190.789210.76719
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Liang, J.; Griswold, J.D.S. Assessment of Aerosol Optical Depth, Cloud Fraction, and Liquid Water Path in CMIP6 Models Using Satellite Observations. Remote Sens. 2025, 17, 2439. https://doi.org/10.3390/rs17142439

AMA Style

Liang J, Griswold JDS. Assessment of Aerosol Optical Depth, Cloud Fraction, and Liquid Water Path in CMIP6 Models Using Satellite Observations. Remote Sensing. 2025; 17(14):2439. https://doi.org/10.3390/rs17142439

Chicago/Turabian Style

Liang, Jiakun, and Jennifer D. Small Griswold. 2025. "Assessment of Aerosol Optical Depth, Cloud Fraction, and Liquid Water Path in CMIP6 Models Using Satellite Observations" Remote Sensing 17, no. 14: 2439. https://doi.org/10.3390/rs17142439

APA Style

Liang, J., & Griswold, J. D. S. (2025). Assessment of Aerosol Optical Depth, Cloud Fraction, and Liquid Water Path in CMIP6 Models Using Satellite Observations. Remote Sensing, 17(14), 2439. https://doi.org/10.3390/rs17142439

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop