Comparison and Analysis of CALIPSO Aerosol Optical Depth and AERONET Aerosol Optical Depth Products in Asia from 2006 to 2023

Abstract

Aerosol optical depth (AOD) serves as a significant parameter in aerosol research. With the increasing utilization of satellite data in AOD research, it is crucial to evaluate the satellite AOD data. Using Aerosol Robotic Network (AERONET) in situ measurements, this study investigates the accuracy and applicability of Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) AOD data in Asia from June 2006 to June 2023. By matching the CALIPSO AOD data in a 1° × 1° area around the selected AERONET sites, various statistical metrics were used to create a comprehensive evaluation system. The results show that: (1) There is a high correlation between the AODs of CALIPSO and AERONET (R = 0.636), and the AOD values of CALIPSO are only 1.7% higher than those of AERONET on average. The MAE (0.215) and RMSE (0.358) suggest that the error level of CALIPSO AOD is relatively low; (2) In most of the 25 sites throughout Asia CALIPSO AOD have high matching accuracies with the AERONET AOD, and only in three sites has a validation accuracy of ‘Poor’; (3) The accuracy varies across the four seasons, ranked as follows: winter demonstrates the highest accuracy, followed by autumn, spring, and summer; (4) The accuracy varies with surface elevation, with better matching in lowest altitude (<50 m) and high altitude (>500 m) areas, but slightly worse matching in medium altitude (200–500 m) areas and low altitude (50–200 m). The uncertainty in the CALIPSO AOD retrievals varies in seasons, altitudes, and aerosol characteristics.

1. Introduction

Aerosols constitute a significant element of the Earth’s atmospheric composition. They consist of both liquid and solid particles with different phases, sizes, and chemical compositions [1]. The sources of aerosols are diverse and mainly divided into human-induced sources, such as emissions generated by industrial activities, straw combustion, automobile emissions, and construction projects, as well as natural sources that include sea salt particulates, forest wildfires, volcanic explosions, and windblown dust [2,3,4,5]. Aerosols span four orders of magnitude in size, ranging from molecules to microns [5]. Although most aerosols are invisible to the naked eye, their substantial concentrations can significantly influence air quality. Furthermore, aerosols play a critical role in the assessment of environmental pollution [6], human health [7,8], and ecosystems [9].

Aerosol optical depth (AOD) is a critical parameter in the study of aerosol, with accuracy, continuity, and specificity essential for atmospheric gas monitoring [6,10,11]. AOD is primarily monitored through ground-based observations and satellite remote sensing [12]. Satellite remote sensing provides global aerosol monitoring and extensive field coverage, resulting in the development of monitoring products such as Multiangle Imaging Spectroradiometer (MISR), Moderate Resolution Imaging Spectroradiometer (MODIS), Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) and Polarization and Directionality of the Earth Reflectances. Ground-based monitoring is considered highly accurate due to the ground-based Sun photometry technique. Data from this method exhibit a significant level of accuracy [13,14], which is considered the truth for AOD measurements [15,16]. Aerosol Robotic Network (AERONET) is a ground-based network and is widely used for its global distribution and long operational period [13,17]. Additionally, its measurements are also characterized by high accuracy and low uncertainty, and numerous studies have utilized AERONET as foundational data for accuracy validation [18]. Filonchy et al. [18] used AERONET to validate MODIS aerosol products for different land use classifications over Eastern Europe and China, getting the most favorable results in Eastern European stations and the least favorable results in coastal regions. Mielonen et al. [19] screened the global AERONET aerosol subtype classification and compared it with the aerosol types measured by CALIOP, finding a 70% concordance between the two datasets. Bibi et al. [20] validated MISR, CALIPSO, OMI, and MODIS AOD measurements at four sites in the Indo-Gangetic region using AERONET, finding higher satellite AOD values during the summer months (June to August, JJA). Aldabash et al. [21] compared AERONET data with MERRA-2 and MODIS C6.1 AOD data from Turkey, concluding that MODIS outperforms MERRA-2 in overall, seasonal, and daily statistics.

The CALIOP is the primary instrument on the CALIPSO satellite, and it is calibrated using well-characterized algorithms [22]. CALIOP is the first spaceborne lidar to continuously observe the vertical structure of atmospheric aerosols and clouds with high vertical resolution. It possesses unique advantages in high vertical resolution, extensive observational coverage, and long-term temporal observation [23,24,25]. Numerous studies have compared CALIPSO AOD with AERONET AOD data and have calculated quantitative metrics related to accuracy. On a global scale, Omar et al. [26] conducted a regression analysis utilizing AERONET data from 2006 to 2010 in conjunction with CALIPSO AOD, finding a correlation coefficient of 0.59 between the two datasets at land stations. Mielonen et al. [19] conducted a comparative analysis of CALIOP secondary aerosol subtypes about AERONET data inversions, finding a 70% concordance in daily mean aerosol types between the two datasets. On the local scale, Ogunjob et al. [27] validated AOD measurements from MODIS, OMI, CALIPSO, and MISR sensors against AERONET data in West Africa, concluding that CALIPSO exhibited the best correlation observed for Ouagadougou. Similarly, Nelli et al. [28] reported that AOD measurements obtained from CALIPSO and AERONET in the United Arab Emirates exhibited a level of agreement with correlations between 0.4 and 0.8.

As the most populous continent, Asia is the predominant source of aerosols and their associated gases. Aerosols in Asia potentially exert a significant influence on global climate dynamics [29]. Investigating the spatial and temporal variations of AOD across Asia is essential. However, little research has been found focused on validating CALIPSO AOD across Asia, especially over an extended time series. The limitation may be attributed to CALIOP’s characteristics as a near-Earth observing instrument with a relatively small receiver footprint. This results in fewer coincident observations with ground validation points than other satellite sensors [22]. In this study, 25 AERONET sites were selected in Asia, characterized by a substantial number of matching points from June 2006 to June 2023. AOD from these sites was utilized to assess the applicability and accuracy of CALIPSO AOD data.

2. Materials and Methods

2.1. Study Area

This study selected 25 AERONET sites in Asia to evaluate the accuracy of CALIPSO data. These sites are located in 14 Asian countries, such as Turkey, India, Pakistan, and China, among others. The distribution of climate types for these AERONET sites is illustrated in Figure 1.

Table 1. Classification and criteria for delineating the Köppen climate zones *.

Climate Zone	Subcategory	Subdivision	Characteristic
A (Tropical)			Tcold ≥ 18
	f (Rainforest)		Tropical rainforest climate: Pdry ≥ 60 mm
	m (Monsoon)		Tropical monsoon climate: Pdry ≤ 60 mm and Pdry ≤ (100-MAP/25) mm
	W (Savanna, dry winter, and dry summer)		Tropical open forest climate: Pdry ≤ 60 mm and Pdry ≥ (100-MAP/25) mm
B (Dry)			MAP < 10 Pth
	W (Arid Desert)	h (Hot)	Desert climate: MAP < 5 Pth
	S (Semi-Arid Steppe)	k (Cold)	Steppe climate: MAP ≥ 5 Pth
C (Temperate)			Thot > 10 °C and 0 °C < Tcold < 18 °C
	w (Dry winter)	a (Hot summer)	The dry and warm climate in winter: Pwdry < Pswet/10
	f (No dry season)	B (Warm summer)	Dry summer and warm climate: Neither Cw nor Cf
	s (Dry summer)	c (Cold summer)	Dry summer and warm climate: Pwdry < Pswet/10
D (Continental)			Thot > 10 °C and Tcold ≤ 0 °C
	w (Dry winter)	a (Hot summer)	Sub-frigid monsoon climate: Pwdry < Pswet/10
	f (No dry season)	b (Warm summer)	Normally humid and cold temperature climate: Neither Ds nor Dw
	s (Dry summer)	c (Cold summer)	Subarctic continental climate: Psdry < 40 mm and Psdary < Pwwet/3
		d (Very cold winter)
E (Polar)			Thot < 10 °C
		T (Tundra)	Ice climate: Thot > 0 °C
		F (Ice Cap)	Tundra climate: Thot ≤ 0 °C

* Note: T_cold = average temperature of the month with the lowest temperatures, T_hot = average temperature of the month with the highest temperatures, P_dry = driest monthly precipitation, P_sdry = driest monthly precipitation in summer, P_wdry = driest monthly precipitation in winter, P_swet = wettest monthly precipitation in summer, P_wwet = wettest monthly precipitation in winter, and MAP = mean annual precipitation.

presents the classification and criteria for delineating the Köppen climate zones.

2.2. Data

2.2.1. CALIPSO AOD

CALIPSO was launched in April 2006. It was developed in collaboration with the French National Centre for Space Studies as part of the National Aeronautics and Space Administration (NASA)s’ Earth System Science Pathfinder program [31]. The mission aimed to fill gaps in observations of the global distribution and properties of aerosols and clouds [32,33,34]. CALIPSO ended on 1 August 2023 after a 17-year mission. A primary scientific objective of the CALIPSO mission was to generate a global dataset that would significantly enhance the accuracy of estimates regarding both direct and indirect climate forcing attributable to aerosols [33]. CALIOP, integrated into the CALIPSO platform, provided unique measurements of the vertical atmospheric profile of the Earth on a global scale [29]. CALIOP algorithms are used to retrieve the AOD with the selection of an appropriate lidar ratio based on an assumed aerosol type. CALIPSO was developed to provide capabilities for detecting atmospheric vertical profiles [35]. The CALIOP sensor aboard the CALIPSO satellite is a dual-wavelength backscatter lidar operating at 1064 nm and 532 nm, with polarization ratio sensitivity [36]. This enables it to capture detailed atmospheric profiles across multiple layers, providing critical information about aerosol and cloud structures [36]. For many years, CALIOP was the only space-based sensor equipped to observe and document the vertical distribution of global aerosols and their spatial and optical characteristics [37]. Its products offer considerable advantages for aerosol research [37]. For this study, the latest version of the CALIPSO Lidar Level 2 aerosol profiles, Version 4-51, was used (https://earthdata.nasa.gov/, accessed on 14 June 2024).

2.2.2. AERONET AOD

AERONET is a ground-based optical aerosol monitoring network and data repository [14]. It was established by NASA’s Earth Observing System and further developed through collaboration with various non-NASA organizations [4]. The network comprises over 1700 permanent and temporary stations distributed globally. Its primary objective is to facilitate automated data acquisition, transmission, and processing for Earth satellite systems to characterize aerosols at regional and global levels. The network supports national and global initiatives related to data transmission, radiation, and international collaboration [13,38,39,40]. AERONET data are available at three levels: Level 1.0 (unscreened), Level 1.5 (cloud screened), and Level 2.0 (cloud screened and quality assured) [20]. Level 2.0 incorporates the additional application of prefield and postfield calibration coefficients, along with expert verification [40]. The Level 2.0 dataset exhibits a high accuracy, with error margins ranging from ±0.01 to 0.02 [41]. The Level 2.0 dataset from AERONET version 3 was selected for this study (https://aeronet.gsfc.nasa.gov/, accessed on 14 June 2024).

Table 2. Information of AOD ground-based observation sites. (Note: Nes_Ziona and Weiz mann_Institute are seen as different names for the same site at different times.).

Sites	Country	Lon. (°E)	Lat. (°N)	Period of Time	Elevation (m)
Bac_Lieu	Vietnam	105.73	9.28	2007–2008, 2013–2019	<50
Dhabi	United Arab Emirates	54.382	24.48	2006–2008
IMS-METU-ERDEMLI	Turkey	34.255	36.565	2006–2019
Kuwait_University	Kuwait	47.971	29.325	2009–2010, 2019–2021
Masdar_Institute	United Arab Emirates	54.617	24.441	2018, 2020–2021
Nes_Ziona	Israel	34.789	31.922	2010–2015
Eilat	Israel	34.918	29.502	2007–2022
Karachi	Pakistan	67.136	24.946	2008–2023
Xianghe	China	116.962	39.754	2007–2021
Beijing	China	116.381	39.977	2006–2019	50–200
Manila_Observatory	Philippines	121.077	14.635	2018–2021
Weizmann_Institute	Israel	34.81	31.907	2015–2021
Silpakorn_Univ	Thailand	100.041	13.819	2007–2020
Gandhi_College	India	84.128	25.817	2006–2009, 2011–2021
Yonsei_University	South Korea	126.934	37.564	2011–2022
Seoul_SNU	South Korea	126.951	37.458	2012–2022
Yakutsk	Russia	129.367	61.661	2008–2021
EPA-NCU	China	121.185	24.968	2006, 2008–2009, 2011–2012,
				2014–2022
Kanpur	India	80.231	26.512	2006–2021
Mukdahan	Thailand	104.676	16.607	2006–2009
Mezaira	United Arab Emirates	53.754	23.105	2007–2019	200–500
Chiang_Mai_Met_Sta	Thailand	98.972	18.771	2008–2018
Jaipur	India	75.806	26.906	2009–2021
Yekaterinburg	Russia	59.545	57.038	2006–2007, 2011, 2013, 2018, 2021
Pokhara	Nepal	83.975	28.187	2018–2019, 2021–2022	>500
SACOL	China	104.137	35.946	2007–2013

shows the information of AOD ground-based observation sites.

2.3. Methods

In this study, CALIPSO AOD data from 25 different sites in Asia between June 2006 and June 2023 were compared with AERONET AOD data. Given the high confidence in the calibration of CALIOP at a wavelength of 532 nm, the CALIPSO AOD measurements at this wavelength were utilized [32]. This analysis involved comparing the CALIPSO AOD measurements at 532 nm with ground-based AERONET AOD data, which were interpolated at 532 nm using the Ångström exponent (α) [42] across the wavelength range of 440 to 870 nm. The interpolation of AOD within the AERONET framework is articulated through Equations (1) and (2).

$$\alpha =-{\displaystyle \frac{\ln \left({\tau }_{440}/{\tau }_{870}\right)}{\ln \left(440/870\right)}}$$

(1)

$${\tau }_{532}={\tau }_{870}{\left(532/870\right)}^{-\alpha }$$

(2)

AERONET provides ground-based measurements at fixed points, while CALIPSO, as a satellite, operates to assess specific areas at regular intervals. To match the temporal and spatial scales of AERONET AOD and CALIPSO AOD, we calculated the spatial average of CALIPSO AOD data within a 1° latitude by 1° longitude grid centered on the AERONET site. Additionally, for this study, we selected the temporal average of AERONET data collected in a two-hour window preceding and following the satellite’s passage [19,37]. The indicators used for validation are listed below.

(1)

Pearson correlation coefficient (R)

R serves as a significant metric in statistical analysis for assessing the linear correlation between two variables. Its values can vary from −1 to 1.

$$R={\displaystyle \frac{{\displaystyle {\sum }_{i=1}^n\left(AO{D}_{(CALIPSO)i}-\overline{AO{D}_{(CALIPSO)}}\right)}\left(AO{D}_{(AERONET)i}-\overline{AO{D}_{(AERONET)}}\right)}{\sqrt{{\displaystyle {\sum }_{i=1}^n{\left(AO{D}_{(CALIPSO)i}-\overline{AO{D}_{(CALIPSO)}}\right)}^2{\displaystyle {\sum }_{i=1}^n{\left(AO{D}_{(AERONET)i}-\overline{AO{D}_{(AERONET)}}\right)}^2}}}}}$$

(3)

(2)

Relative Mean Deviation (RMB)

The RMB is utilized to assess the extent of divergence between the inverse value and the true value. A value of RMB approaching 1 indicates minimal deviation, which signifies higher accuracy in the inversion results.

$$RMB={\displaystyle \frac{{\overline{AOD}}_{CALIPSO}}{{\overline{AOD}}_{AERONET}}}$$

(4)

(3)

Mean Absolute Error (MAE)

MAE is a significant metric employed in statistical analysis to evaluate the efficacy of predictive models. It provides a visual representation of the average magnitude of prediction errors.

$$MAE={\displaystyle \frac{1}{n}}{\displaystyle \sum _{i=1}^n\left|AO{D}_{\left(CALIPSO\right)i}-AO{D}_{\left(AERONET\right)i}\right|}$$

(5)

(4)

Root Mean Square Error (RMSE)

RMSE is an essential metric within the discipline of statistics, utilized to assess the discrepancy between predicted values and observed outcomes.

$$RMSE=\sqrt{{\displaystyle \frac{1}{n}}{{\displaystyle \sum _{i=1}^n\left(AO{D}_{\left(CALIPSO\right)i}-AO{D}_{\left(AERONET\right)i}\right)}}^2}$$

(6)

(5)

Total Accuracy Rating

To develop an evaluation framework that reduces bias from the use of individual indicators, a multifaceted evaluation system incorporating R, RMB, MAE, and RMSE has been established [41]. The system categorizes indicators according to numerical ranges, facilitating an assessment of data accuracy [43].

3. Results

3.1. Overall Analysis of the Comparison Between CALIPSO AOD and AERONET AOD

Figure 2 presents the statistical analysis of AOD data from CALIPSO and AERONET for the period spanning June 2006 to June 2023, based on 1152 matching points. The mean AOD values recorded from CALIPSO and AERONET are closely aligned at 0.469 and 0.461, respectively. This reflects a significant degree of consistency between the datasets. Furthermore, the median CALIPSO AOD value is 0.313, slightly lower than the median value of AERONET AOD. This indicates that the majority of the CALISPO AOD data is lower than that of the AERONET AOD data. Figure 3 illustrates the correlation between the AOD measurements obtained from AERONET and CALIPSO. The relationship between these two measurements exhibits a strong correlation, evidenced by an R value of 0.636. RMB is 1.017, this indicates that the AOD values derived from the CALIPSO product are, on average, 1.7 percent higher than those obtained from AERONET. Additionally, the MAE is recorded at 0.215, while the RMSE is 0.358, both of which suggest a low average level of error in the measurements. In conclusion, CALIPSO AOD products show a considerable level of accuracy and applicability within the Asia region.

3.2. Determination of Accuracy Class for Different Sites

Following a comprehensive assessment of four indicators at each site, Bac_Lieu, Masdar_Institute, Nes_Ziona (Weizmann_Institute), Eilat, Karachi, Yakutsk, Kanpur, Mukdahan, Chiang_Mai_Met_Sta, Pokhara, and SACOL were rated as “Excellent” (Figure 4,

Table 3. Criteria table for classification.

R	RMB	MAE	RMSE	Reevaluation
0.50–0.55	>1.55/<0.9	>0.29	>0.4	1
0.55–0.65	0.9–0.95/1.05–1.55	0.24–0.29	0.35–0.40	2
0.65–0.75	1–1.05/0.95–0.99	0.18–0.24	0.30–0.35	3
>0.75	0.99–1	<0.18	<0.3	4
Total reclassification		Accuracy
≥12		Excellent
10–11		Very good
7–9		Good
≤6		Poor

and

Table 4. R, RMB, MAE, RMSE, and accuracy level of CALIPSO AOD at individual sites.

Sites	R	RMB	MAE	RMSE	Accuracy
Bac_Lieu	0.867	1.149	0.087	0.117	Excellent
Dhabi	0.690	1.338	0.271	0.429	Good
IMS-METU-ERDEMLI	0.607	0.713	0.135	0.172	Very good
Kuwait_University	0.618	1.138	0.252	0.434	Good
Masdar_Institute	0.755	1.265	0.172	0.354	Excellent
Nes_Ziona	0.839	0.917	0.113	0.195	Excellent
Eilat	0.852	0.858	0.072	0.100	Excellent
Karachi	0.770	1.216	0.197	0.285	Excellent
Xianghe	0.724	0.777	0.262	0.482	Good
Beijing	0.617	0.663	0.353	0.575	Poor
Manila_Observatory	0.641	2.282	0.222	0.312	Good
Weizmann_Institute	0.528	0.981	0.101	0.179	Excellent
Silpakorn_Univ	0.527	1.195	0.266	0.363	Good
Gandhi_College	0.509	1.256	0.345	0.481	Poor
Yonsei_University	0.646	0.775	0.196	0.324	Good
Seoul_SNU	0.653	0.833	0.189	0.301	Very good
Yakutsk	0.733	1.056	0.147	0.229	Excellent
EPA-NCU	0.586	0.590	0.255	0.402	Poor
Kanpur	0.770	1.216	0.197	0.285	Excellent
Mukdahan	0.772	1.365	0.196	0.320	Excellent
Mezaira	0.854	1.538	0.210	0.364	Very good
Chiang_Mai_Met_Sta	0.703	0.989	0.187	0.296	Excellent
Jaipur	0.568	1.463	0.274	0.433	Good
Yekaterinburg	0.509	0.725	0.101	0.117	Very good
Pokhara	0.888	0.651	0.144	0.182	Excellent
SACOL	0.600	0.936	0.159	0.256	Excellent

). The analysis of the sites was conducted by integrating the matching point distribution frequency of

Table 5. Seasonal match point frequency at individual sites.

Name of Station	N	Frequency
		MAM	JJA	SON	DJF
Bac_Lieu	27	0.630	0.148	0.074	0.148
Dhabi	13	0.077	0.462	0.385	0.077
IMS-METU-ERDEMLI	101	0.208	0.337	0.218	0.238
Kuwait_University	18	0.444	0.167	0.222	0.167
Masdar_Institute	10	0.200	0.300	0.100	0.400
Nes_Ziona	49	0.224	0.265	0.265	0.245
Eilat	89	0.270	0.281	0.258	0.191
Karachi	103	0.340	0.146	0.214	0.301
Xianghe	108	0.278	0.176	0.259	0.287
Beijing	102	0.275	0.176	0.225	0.324
Manila_Observatory	14	0.643	0.071	0.071	0.214
Weizmann_Institute	51	0.196	0.275	0.353	0.176
Silpakorn_Univ	93	0.323	0.054	0.183	0.441
Gandhi_College	68	0.397	0.162	0.191	0.250
Yonsei_University	77	0.325	0.208	0.234	0.234
Seoul_SNU	53	0.321	0.226	0.208	0.245
Yakutsk	31	0.290	0.419	0.258	0.032
EPA-NCU	44	0.341	0.250	0.227	0.182
Kanpur	103	0.340	0.146	0.214	0.301
Mukdahan	19	0.474	0.105	0.211	0.211
Mezaira	101	0.168	0.168	0.337	0.327
Chiang_Mai_Met_Sta	105	0.314	0.057	0.257	0.371
Jaipur	82	0.341	0.183	0.195	0.280
Yekaterinburg	9	0.222	0.667	0.111	0.000
Pokhara	14	0.429	0.000	0.214	0.357
SACOL	68	0.294	0.250	0.250	0.206

Note: Nes_Ziona and Weizmann_Institute are seen as different names for the same site at different times. The frequency is determined by the number of matching points for each season divided by the total number of matching points.

. Among them, Bac_Lieu, Chiang_Mai_Met_Sta, and Mukdahan are situated in the tropical monsoon climate, characterized by heavy rainfall during JJA and autumn (September to November, SON). The frequency of matching points at these three sites during JJA and SON was relatively low. Masdar_Institute, Eilat, and Karachi are categorized as desert climates, which are characterized by arid conditions and minimal annual precipitation. Nes_Ziona (Weizmann_Institute) is located in a mediterranean climate, which is characterized by high precipitation during winter (December to February, DJF) and spring (March to May, MAM). During the study period, matching points occurred less frequently in DJF and MAM. Kanpur experiences a dry and warm climate in DJF, with hot and arid in JJA, and precipitation is primarily observed during the DJF months. Approximately 70% of the matching points for this site are observed during relatively dry periods outside of DJF. Similarly, Yakutsk is located in the Subarctic continental climate. The rainy season occurs in JJA, and approximately 60% of the matching points for this site are observed during the other three seasons. SACOL is situated in a steppe climate, characterized by low precipitation and relatively dry conditions during MAM, SON, and DJF, with the majority of rainfall concentrated in JJA. Matching points occur less frequently in JJA, with values of only 0.250. In summary, it can be found that the matching points for the sites that performed “Excellent” were all mostly found in arid climatic conditions.

3.3. Comparative Analysis of CALIPSO and AERONET AOD Across Different Seasons

Figure 5 shows the statistical differences in AOD data from CALIPSO and AERONET across different seasons. From a seasonal perspective, it can be observed that both CALIPSO AOD and AERONET AOD have higher mean values in MAM and JJA compared to SON and DJF. In comparison between the two datasets, the mean value of CALIPSO AOD is slightly higher in MAM and JJA, while AERONET AOD shows a slightly higher mean value in SON and DJF. In terms of median values, the median of AERONET AOD is higher than that of CALIPSO AOD in all seasons. This indicates that it is common for CALIPSO AOD data to be lower than AERONET AOD data and is not limited to a particular season. Figure 6 presents the validation results of CALIPSO AOD and AERONET AOD across 25 sites in different seasons. The matching accuracy shows a high degree of consistency in all seasons. Valid matching points are most frequent during MAM and DJF, and less frequent during JJA and SON. The highest correlation is observed in DJF, with an R value of 0.781, indicating a strong relationship between the two datasets during this period. Conversely, the lowest R value of 0.497 is recorded in JJA, with the weakest correlation observed in August. The RMB and RMSE values for August also exhibit poor performance, as shown in Figure 7. This can be attributed to the prevalence of elevated temperatures and high humidity levels during August. Such climatic conditions enhance the hygroscopic properties of aerosols, thereby adversely reducing the effectiveness of satellite monitoring [37,44]. This reduction is particularly pronounced given that the majority of the stations involved in this study are situated within the Asian monsoon climate.

Combining Figure 5 and Figure 6, it can be observed that the satellite-derived AOD values most closely align with ground stations’ measurements in DJF and SON. In contrast, the discrepancies between the satellite and ground station AOD values are more pronounced during the MAM and JJA, with the most significant divergence occurring in JJA. The MAE is lowest in DJF, followed by SON, while MAM and JJA show higher MAE values of 0.245 and 0.302, respectively. In terms of the RMSE, SON and DJF demonstrate superior performance with lower values of 0.316 and 0.238, respectively. MAM shows a marginally higher RMSE of 0.378, and JJA presents the highest RMSE at 0.487. The reduced MAE and RMSE values during SON and DJF indicate that the CALIPSO AOD exhibits less error in the two seasons. In contrast, increased error is observed in MAM and JJA. In conclusion, the accuracy of the total matching points for CALIPSO AOD and AERONET AOD across different seasons can be ranked as follows: DJF > SON > MAM > JJA.

3.4. Comparison of CALIPSO and AERONET AOD at Different Elevations

The 25 sites were categorized into four distinct classes based on varying altitudes, and the accuracy of these sites in each class was evaluated independently (Figure 8). The R values are all around 0.6, and they vary very slightly among different altitude classes. Notably, the R value reaches its peak within the altitude range of below 50 m, suggesting a stronger correlation between AERONET AOD and CALIPSO AOD data in this class. Conversely, the R value is lowest for sites between 50 and 200 m, indicating a greater inconsistency between the two measurements at this altitude range.

The RMB shows a tendency to rise and then fall as the altitude increases. In the lower altitude categories of less than 50 m and between 50 to 200 m, RMB are 0.936 and 0.987, respectively, suggesting minimal deviation. However, in the sites within the altitude range of 200 to 500 m, RMB is 1.283, indicating a large deviation between the two AOD products. RMB is 0.877 in the sites of the altitudes >500 m. The MAE and RMSE also display an increasing–decreasing trend as the altitude increases. Both MAE and RMSE are higher in medium altitudes (200–500 m) and low altitudes (50–200 m), indicating a higher degree of error. MAE and RMSE values decreased at altitudes greater than 500 m.

The terrain data for each site, including mean (DEM_MEAN), maximum (DEM_MAX), and minimum (DEM_MIN) elevations within the surrounding 1° × 1° grid cells, is provided in

Table 6. Site elevation and mean surrounding terrain altitude within 1° × 1° Grid.

Sites	Elevation (m)	DEMMEAN (m)	DEMMAX (m)	DEMMIN (m)
Bac_Lieu	10.0	1.246	23	−13
Dhabi	15.0	47.235	151	−24
IMS-METU-ERDEMLI	3.0	559.675	2518	0
Kuwait_University	42.0	50.499	219	−12
Masdar_Institute	4.0	32.533	166	−24
Nes_Ziona	40.0	193.518	1011	0
Eilat	15.0	657.144	1733	0
Karachi	49.0	81.924	687	−12
Xianghe	36.0	39.140	937	−1
Beijing	92.0	164.172	1351	6
Manila_Observatory	63.0	163.430	1517	−26
Weizmann_Institute	73.0	207.845	1011	0
Silpakorn_Univ	72.0	18.413	498	−4
Gandhi_College	60.0	66.041	40	94
Yonsei_University	97.0	104.270	1126	−8
Seoul_SNU	116.0	95.846	1126	−1
Yakutsk	118.5	188.268	325	59
EPA-NCU	144.0	310.689	3271	−15
Kanpur	123.0	127.288	152	87
Mukdahan	166.0	196.021	620	120
Mezaira	201.0	127.542	253	61
Chiang_Mai_Met_Sta	312.0	673.659	2566	227
Jaipur	450.0	386.087	788	258
Yekaterinburg	300.0	373.562	717	209
Pokhara	800.0	1496.801	7726	111
SACOL	1965.8	2056.306	3635	1421

. Among the research sites, most exhibit an elevation close to their DEM_MEAN values, indicating relatively smooth terrain within the 1 × 1° grid surrounding these locations. However, there are still some sites where the DEM_MEAN value has a significant difference from the site’s elevation, such as IMS-METU-ERDEMLI, Weizmann Institute, and Mezaira. This variation is due to significant elevation differences within the grid, as shown in Table 6.

4. Discussion

The comprehensive results of CALIPSO AOD site comparison are presented in Figure 2 and Figure 3, indicating that the CALIPSO AOD product demonstrates a high level of accuracy in Asia. Overall, the median AOD value from AERONET is higher than that derived from CALIPSO, with the majority of AERONET AOD are higher than CALIPSO AOD. This discrepancy is also noted by Omar et al. [26]. However, the mean AOD value from CALIPSO is slightly higher than the mean from AERONET. This is likely due to the influence of specific matching points. Two primary factors may explain these variations. First, the daytime CALIPSO AOD data used in this study, affected by solar background noise, has higher uncertainty compared to nighttime data [26]. Second, CALIPSO and AERONET use different methods to derive AOD. The CALIOP algorithm selects an appropriate lidar ratio based on predefined aerosol classifications, while AERONET directly measures AOD [26]. Although the mean of CALIPSO AOD values is slightly higher than the mean of AERONET AOD values in terms of overall comparison, CALIPSO products remain sufficiently accurate for AOD monitoring across large terrain areas [45,46].

Through comprehensive evaluation of indicators at 25 sites, CALIPSO AOD at Bac_Lieu, Masdar_Institute, Nes_Ziona (Weizmann_Institute), Eilat, Karachi, Yakutsk, Kanpur, Mukdahan, Chiang_Mai_Met_Sta, Pokhara, and SACOL is rated as “Excellent”, as shown in Figure 3. This suggests an excellent match in accuracy between satellite AOD and ground station point AOD for these locations. The majority of matching points for these sites occur under arid conditions. Under this condition, the negative impacts on satellite monitoring, such as high humidity and aerosol hygroscopicity are reduced to a certain extent [37,40]. High humidity and aerosol hygroscopicity negatively affect satellite measurements. In humid conditions, aerosol particles absorb moisture, leading to particle swelling. This alters their radiative properties and impairs the accuracy of satellite-based AOD measurements. Additionally, the single-scattering albedo of aerosol particles is strongly affected by hygroscopic growth, further reducing the precision of satellite AOD retrievals [47,48]. Therefore, sites located in arid conditions generally exhibit higher accuracy in AOD measurements.

As shown in Figure 5, the mean AOD values of CALIPSO and AERONET in MAM and JJA are both higher than those in DJF and SON. This is likely attributable to higher aerosol concentrations during MAM and JJA [35]. Firstly, the generally higher temperature and humidity in MAM and JJA enhance the hygroscopicity of aerosols, which increases AOD values [48]. Secondly, the higher frequency of forest fires during MAM and JJA contributes substantial smoke particles to the atmosphere, significantly raising aerosol concentrations [49]. However, this increase in smoke aerosols may impair the optical detection capabilities of satellites, complicating aerosol classification [49,50]. These results indicate that variations in aerosol concentration influence the degree of air pollution, which in turn impacts the retrieval of CALIPSO AOD measurements [35]. Such fluctuations can reduce the monitoring effectiveness of remote sensing data, leading to lower accuracy in MAM and JJA, as shown in Figure 6. Among the four seasons, JJA shows the highest RMSE, due not only to the factors mentioned above but also to the high density of surface vegetation and surface albedo during this period. These seasonal changes complicate satellite data inversion processes and increase uncertainty in satellite retrieval [51,52].

The accuracy of CALIPSO AOD varies at various elevations. The highest accuracy is at altitudes below 50 m, with the next highest accuracy being at altitudes > 500 m. The altitudes at 200–500 m and 50–200 m are slightly less accurate. Although the DEM_MEAN values of some sites are much higher than the elevation of the sites, no significant difference in accuracy has been found between these sites and others. The difference in accuracy across the altitude gradients in this study is not significant, which may be attributed to the limited number of matching points and the insignificant altitude differences among the studied sites. However, this still reflects the fact that CALIPSO monitoring exhibits varying advantages at different altitude gradients in Asia. The R and RMSE values were optimal in the <50 m altitude range; the CALIPSO AOD and AERONET AOD values were closest in the 50–200 m altitude range (RMB = 0.987), while the MAE values were minimal in the >500 m altitude range.

5. Conclusions

This study provides an intercomparison of CALIPSO AOD against AERONET AOD from 25 Asian sites during 2006 and 2023. The findings of the study are presented below.

(1)

Overall, CALIPSO retrievals show a high degree of consistency with AERONET observation. They have a strong correlation evidenced by an R value of 0.636. At eleven sites, CALIPSO AOD accuracy was performed as “Excellent” and at four sites as “Very good”. Additionally, at seven sites CALIPSO AOD accuracy was performed as “Good”, while at three sites as “Poor”. Different climate conditions influence the accuracy of various sites.

(2)

The accuracy of CALIPSO AOD is influenced by seasonal variations. CALIPSO AOD indicated a stronger correlation with AERONET AOD in DJF and SON, with higher accuracy. In JJA, forest fires and varying surface vegetation lead to uncertainty of CALIPSO AOD inversion. The CALIPSO AOD accuracy is lowest in JJA.

(3)

The accuracy of sites varies at different elevations. CALIPSO and AERONET AOD show the strongest consistency at sites below 50 m. However, in the sites within the altitude range of 200 to 500 m, RMB is 1.283, indicating a large deviation between the two AOD products.