Stavros-Andreas Logothetis
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Stavros-Andreas Logothetis has a total of 40 co-authors. Here are the top co-authors by number of shared publications:
- Andreas Kazantzidis
University of Patras, Greece
9 shared publications - Vassilis Amiridis
IAASARS, National Observatory of Athens, Greece
6 shared publications - Antonis Gkikas
Research Centre for Atmospheric Physics and Climatology, Academy of Athens, Athens, 11521, Greece
5 shared publications - Stelios Kazadzis
Physikalisch-Meteorologisches Observatorium Davos, World Radiation Center (PMOD/WRC), Davos 7260, Switzerland
5 shared publications Vasileios Salamalikis
4 shared publicationsAyman Mohammed Albar
3 shared publications
Latest Publications
- Open Access
- 18 September 2026
- Published by American Association for the Advancement of Science (AAAS) in Science Advances
- Vol. 12 (38)
A global decline in atmospheric dust during the 21st century
Mineral dust affects Earth’s energy budget, clouds, and biochemistry, yet the evolution of the global dust cycle in recent decades remains shrouded in uncertainty. Here we reconstruct the global dust cycle during the early 21st century by integrating a suite of global dust aerosol optical depth (DAOD) products from satellite-borne lidar, infrared sounders, and multispectral imagers; three aerosol reanalyses; and a compilation of dust-dominated AERONET sites into a large-ensemble inverse modeling framework. We show that North Africa has remained the dominant global dust source, yet it has experienced a decline in DAOD of −10 ± 8% between 2003 and 2023. A comparable reduction is observed across Asia (−10 ± 12%), together driving a global decrease of −10 ± 8%. Similar declines in dust loading, emissions, and deposition indicate a coherent reduction of the global dust cycle. Individual satellite products converge on overall similar trends, providing the strongest evidence to date that the planet has become less dusty in the early 21st century. This decline in atmospheric dust has exerted a positive effective radiative forcing of +0.02 ± 0.05 watts per square meter over 2003–2023, implying that declining atmospheric dust might have modestly contributed to the rapid warming of the past two decades.
- 18 September 2026
- Published by American Association for the Advancement of Science (AAAS) in Science Advances
- Vol. 12 (38)
Open AccessA global decline in atmospheric dust during the 21st century
Mineral dust affects Earth’s energy budget, clouds, and biochemistry, yet the evolution of the global dust cycle in recent decades remains shrouded in uncertainty. Here we reconstruct the global dust cycle during the early 21st century by integrating a suite of global dust aerosol optical depth (DAOD) products from satellite-borne lidar, infrared sounders, and multispectral imagers; three aerosol reanalyses; and a compilation of dust-dominated AERONET sites into a large-ensemble inverse modeling framework. We show that North Africa has remained the dominant global dust source, yet it has experienced a decline in DAOD of −10 ± 8% between 2003 and 2023. A comparable reduction is observed across Asia (−10 ± 12%), together driving a global decrease of −10 ± 8%. Similar declines in dust loading, emissions, and deposition indicate a coherent reduction of the global dust cycle. Individual satellite products converge on overall similar trends, providing the strongest evidence to date that the planet has become less dusty in the early 21st century. This decline in atmospheric dust has exerted a positive effective radiative forcing of +0.02 ± 0.05 watts per square meter over 2003–2023, implying that declining atmospheric dust might have modestly contributed to the rapid warming of the past two decades.
- Open Access
- 20 July 2026
- Published by Elsevier in SSRN Electronic Journal
A satellite-based framework for identifying glaciogenic cloud-seeding opportunities over Saudi Arabia
Cloud seeding is increasingly used as a water-resource management strategy in arid and semi-arid regions, yet comprehensive information on the occurrence of suitable target clouds remains limited. In this study, a satellite-based methodology was developed and evaluated for identifying potential glaciogenic cloud-seeding opportunities over the Kingdom of Saudi Arabia (KSA) using cloud macro- and microphysical properties retrieved from the Satellite Application Facility on Climate Monitoring (CM SAF) CLoud Property Dataset Using SEVIRI, Edition 3 (CLAAS-3). CLAAS-3 is derived from observations of the Spinning Enhanced Visible and Infrared Imager (SEVIRI) onboard the Meteosat Second Generation satellites. The methodology combines information on cloud phase, cloud optical thickness, cloud water path, cloud-top temperature, cloud-top height, cloud effective radius, and precipitation rate. Thresholds were constrained using in situ observations collected during operational cloud-seeding activities and dedicated research aircraft campaigns. Evaluation against an extensive more than 10,000 cloud-seeding events conducted during the 2022–2025 operational period of the Saudi Cloud Seeding Program (CSP) demonstrated that the methodology successfully identified operational seeding targets with an overall accuracy of 82.9%. This result indicates that the methodology can objectively reproduce, using satellite-derived cloud properties, many of the target areas independently selected by experienced cloud-seeding pilots and operational meteorologists. Independent comparisons with aircraft cloud microphysical measurements showed good agreement between satellite-derived and in situ cloud effective radius retrievals, supporting the applicability of the approach.The methodology was subsequently applied to five years (2021–2025) of CLAAS-3 observations to produce the first high-resolution satellite-based dataset of potential glaciogenic cloud-seeding opportunities across KSA. The analysis identifies enhanced frequencies of seedable cloud conditions over two key regions of KSA. The highest occurrence is found over the southwestern mountainous areas surrounding Abha, Al-Baha, and Taif, where annual frequencies reach up to 14.5%, while lower but still substantial frequencies occur across the central regions encompassing Riyadh, Qassim, and Hail, reaching up to 5.3%. The occurrence of satellite-detected opportunities exhibits pronounced seasonal variability, with the southwestern region experiencing its highest frequencies during summer and the central region during spring. The spatial distribution of seedable cloud conditions closely aligns with regional precipitation climatology, highlighting the importance of topographic and meteorological controls on cloud development. The resulting dataset provides a quantitative framework for identifying cloud-seeding opportunities from remote sensing instruments and can support strategic planning and operational optimization of rainfall-enhancement activities in KSA.
- 20 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessA satellite-based framework for identifying glaciogenic cloud-seeding opportunities over Saudi Arabia
Cloud seeding is increasingly used as a water-resource management strategy in arid and semi-arid regions, yet comprehensive information on the occurrence of suitable target clouds remains limited. In this study, a satellite-based methodology was developed and evaluated for identifying potential glaciogenic cloud-seeding opportunities over the Kingdom of Saudi Arabia (KSA) using cloud macro- and microphysical properties retrieved from the Satellite Application Facility on Climate Monitoring (CM SAF) CLoud Property Dataset Using SEVIRI, Edition 3 (CLAAS-3). CLAAS-3 is derived from observations of the Spinning Enhanced Visible and Infrared Imager (SEVIRI) onboard the Meteosat Second Generation satellites. The methodology combines information on cloud phase, cloud optical thickness, cloud water path, cloud-top temperature, cloud-top height, cloud effective radius, and precipitation rate. Thresholds were constrained using in situ observations collected during operational cloud-seeding activities and dedicated research aircraft campaigns. Evaluation against an extensive more than 10,000 cloud-seeding events conducted during the 2022–2025 operational period of the Saudi Cloud Seeding Program (CSP) demonstrated that the methodology successfully identified operational seeding targets with an overall accuracy of 82.9%. This result indicates that the methodology can objectively reproduce, using satellite-derived cloud properties, many of the target areas independently selected by experienced cloud-seeding pilots and operational meteorologists. Independent comparisons with aircraft cloud microphysical measurements showed good agreement between satellite-derived and in situ cloud effective radius retrievals, supporting the applicability of the approach.The methodology was subsequently applied to five years (2021–2025) of CLAAS-3 observations to produce the first high-resolution satellite-based dataset of potential glaciogenic cloud-seeding opportunities across KSA. The analysis identifies enhanced frequencies of seedable cloud conditions over two key regions of KSA. The highest occurrence is found over the southwestern mountainous areas surrounding Abha, Al-Baha, and Taif, where annual frequencies reach up to 14.5%, while lower but still substantial frequencies occur across the central regions encompassing Riyadh, Qassim, and Hail, reaching up to 5.3%. The occurrence of satellite-detected opportunities exhibits pronounced seasonal variability, with the southwestern region experiencing its highest frequencies during summer and the central region during spring. The spatial distribution of seedable cloud conditions closely aligns with regional precipitation climatology, highlighting the importance of topographic and meteorological controls on cloud development. The resulting dataset provides a quantitative framework for identifying cloud-seeding opportunities from remote sensing instruments and can support strategic planning and operational optimization of rainfall-enhancement activities in KSA.
- Open Access
- 13 March 2026
- Published by Copernicus GmbH
Analyzing cloud seeding operations using Cloud Water Inertial Probe (CWIP) measurements: A case study from the Regional Cloud Seeding Program in Saudi Arabia
The Kingdom of Saudi Arabia (KSA) initiated a Regional Cloud Seeding Program (RCSP) in 2022 to enhance precipitation over the southwestern and central parts of the country. This study explores the application of the Cloud Water Inertial Probe (CWIP) in RCSP cloud seeding operations. The CWIP is a high-resolution atmospheric sensing instrument deployed on cloud seeding aircraft, delivering 1-second measurements of key microphysical parameters, including liquid water content and three-dimensional wind velocity. These real-time data are essential for adaptive cloud seeding operations, enabling flight crews and ground-based teams to assess evolving cloud environments and make timely, data-informed decisions.This study focuses on the cloud seeding event conducted by the RCSP on April 29th, 2023, over the southwestern region of KSA. The case analysis focused on characterizing atmospheric conditions during short-duration penetrations of convective clouds. CWIP measurements are used to examine the variability of microphysical and dynamic parameters over short timescales, contributing to a better understanding of cloud structure and evolution in the region. The results highlight the utility of high-frequency in situ observations for capturing the transient nature of cloud processes in arid and semi-arid environments and for assessing the cloud seeding events.
- 13 March 2026
- Published by Copernicus GmbH
Open AccessAnalyzing cloud seeding operations using Cloud Water Inertial Probe (CWIP) measurements: A case study from the Regional Cloud Seeding Program in Saudi Arabia
The Kingdom of Saudi Arabia (KSA) initiated a Regional Cloud Seeding Program (RCSP) in 2022 to enhance precipitation over the southwestern and central parts of the country. This study explores the application of the Cloud Water Inertial Probe (CWIP) in RCSP cloud seeding operations. The CWIP is a high-resolution atmospheric sensing instrument deployed on cloud seeding aircraft, delivering 1-second measurements of key microphysical parameters, including liquid water content and three-dimensional wind velocity. These real-time data are essential for adaptive cloud seeding operations, enabling flight crews and ground-based teams to assess evolving cloud environments and make timely, data-informed decisions.This study focuses on the cloud seeding event conducted by the RCSP on April 29th, 2023, over the southwestern region of KSA. The case analysis focused on characterizing atmospheric conditions during short-duration penetrations of convective clouds. CWIP measurements are used to examine the variability of microphysical and dynamic parameters over short timescales, contributing to a better understanding of cloud structure and evolution in the region. The results highlight the utility of high-frequency in situ observations for capturing the transient nature of cloud processes in arid and semi-arid environments and for assessing the cloud seeding events.
- Open Access
- 13 March 2026
- Published by Copernicus GmbH
Satellite-Based Analysis of Cloud Properties Over Saudi Arabia in Support of Regional Cloud Seeding Program
Clouds play a vital role in modulating the Earth’s radiation budget and regulating regional hydrological cycles. In arid and semi-arid regions such as the Kingdom of Saudi Arabia (KSA), where annual rainfall is scarce, understanding cloud occurrence and microphysical properties is essential for developing effective cloud seeding strategies. In 2022, KSA launched the Cloud Seeding Program (CSP) under the National Center for Meteorology (NCM), aiming to enhance precipitation across the country. Recent advancements in satellite remote sensing have significantly improved the ability to monitor cloud properties in regions with sparse in situ observations. Despite these technological advancements, research dedicated to the spatiotemporal variability of cloud properties over KSA remains limited. This represents a critical gap, as such analyses are fundamental for understanding cloud–precipitation interactions and the associated atmospheric feedback mechanisms in dry environments.This study investigates the seasonal variability of cloud properties over the KSA using satellite-derived data from the CM SAF Cloud property dAtAset using SEVIRI (CLAAS-3) dataset. Cloud variability across KSA shows pronounced spatial and seasonal differences, particularly in terms of daytime cloud occurrence. In the central region, cloud cover is generally sparse and dominated by shallow cumulus and cirrus clouds, with peak activity in spring and winter (up to 30–35%), largely influenced by mid-latitude frontal systems and upper-level troughs. In contrast, the southwestern highlands (escarpment) exhibit significantly higher cloud occurrence during the summer, reaching up to 50%, driven by moist inflow from the Red Sea, the Indian Ocean monsoon extension, and orographic lifting. These factors promote the formation of deep convective clouds, capable of producing intense, short-lived rainfall events. Understanding these distinct patterns is essential for improving cloud seeding operations, weather forecasting, and water resource management in KSA.
- 13 March 2026
- Published by Copernicus GmbH
Open AccessSatellite-Based Analysis of Cloud Properties Over Saudi Arabia in Support of Regional Cloud Seeding Program
Clouds play a vital role in modulating the Earth’s radiation budget and regulating regional hydrological cycles. In arid and semi-arid regions such as the Kingdom of Saudi Arabia (KSA), where annual rainfall is scarce, understanding cloud occurrence and microphysical properties is essential for developing effective cloud seeding strategies. In 2022, KSA launched the Cloud Seeding Program (CSP) under the National Center for Meteorology (NCM), aiming to enhance precipitation across the country. Recent advancements in satellite remote sensing have significantly improved the ability to monitor cloud properties in regions with sparse in situ observations. Despite these technological advancements, research dedicated to the spatiotemporal variability of cloud properties over KSA remains limited. This represents a critical gap, as such analyses are fundamental for understanding cloud–precipitation interactions and the associated atmospheric feedback mechanisms in dry environments.This study investigates the seasonal variability of cloud properties over the KSA using satellite-derived data from the CM SAF Cloud property dAtAset using SEVIRI (CLAAS-3) dataset. Cloud variability across KSA shows pronounced spatial and seasonal differences, particularly in terms of daytime cloud occurrence. In the central region, cloud cover is generally sparse and dominated by shallow cumulus and cirrus clouds, with peak activity in spring and winter (up to 30–35%), largely influenced by mid-latitude frontal systems and upper-level troughs. In contrast, the southwestern highlands (escarpment) exhibit significantly higher cloud occurrence during the summer, reaching up to 50%, driven by moist inflow from the Red Sea, the Indian Ocean monsoon extension, and orographic lifting. These factors promote the formation of deep convective clouds, capable of producing intense, short-lived rainfall events. Understanding these distinct patterns is essential for improving cloud seeding operations, weather forecasting, and water resource management in KSA.
- Open Access
- 6 March 2026
- Published by Copernicus GmbH
Supplementary material to "Two Decades of Aerosol Optical Depth evolution from CAMS Reanalysis"
- 6 March 2026
- Published by Copernicus GmbH
Open AccessSupplementary material to "Two Decades of Aerosol Optical Depth evolution from CAMS Reanalysis"
- Open Access
- 6 March 2026
- Published by Copernicus GmbH in EGUsphere
- Vol. 2026, 1-61
Two Decades of Aerosol Optical Depth evolution from CAMS Reanalysis
Aerosol optical depth (AOD) is a key indicator for evaluating global climate change, and its long-term trends provide critical insights into the evolving climatic impact of atmospheric aerosols. In this study, we analyse the long-term (2003–2024) AOD record from the fourth-generation ECMWF Atmospheric Composition Reanalysis (EAC4) to investigate global and regional variability and long-term trends over the past two decades. EAC4 is evaluated against independent AERONET observations from 178 stations selected using strict data-availability criteria. The impact of satellite assimilation is explicitly quantified through comparison with a free-running control simulation (CTRL). EAC4 reproduces observed global AOD variability with high skill (R = 0.84, RMSE = 0.12, IOA = 0.90; >1.6 million collocations) and shows strong agreement with AERONET-derived trends at stations exhibiting statistically significant changes (R = 0.89), correctly capturing the trend sign at 96.7 % of sites. Negative trends are generally well represented, while positive trends are more frequently underestimated. Trend analyses across 18 regions of interest reveal significant AOD declines over eastern China, North Africa, Central Europe, and the eastern United States, alongside persistent increases over South Asia and the Middle East. Decadal analyses identify pronounced transitions, including a reversal from increasing to decreasing trends over eastern China and shifts in several dust- and biomass-burning-dominated regions. Sliding-window trend diagnostics further highlight temporal variability and regime changes. By integrating long-term reanalysis data, independent validation, and component-resolved diagnostics, this study provides a consistent assessment of recent global AOD evolution.
- 6 March 2026
- Published by Copernicus GmbH in EGUsphere
- Vol. 2026, 1-61
Open AccessTwo Decades of Aerosol Optical Depth evolution from CAMS Reanalysis
Aerosol optical depth (AOD) is a key indicator for evaluating global climate change, and its long-term trends provide critical insights into the evolving climatic impact of atmospheric aerosols. In this study, we analyse the long-term (2003–2024) AOD record from the fourth-generation ECMWF Atmospheric Composition Reanalysis (EAC4) to investigate global and regional variability and long-term trends over the past two decades. EAC4 is evaluated against independent AERONET observations from 178 stations selected using strict data-availability criteria. The impact of satellite assimilation is explicitly quantified through comparison with a free-running control simulation (CTRL). EAC4 reproduces observed global AOD variability with high skill (R = 0.84, RMSE = 0.12, IOA = 0.90; >1.6 million collocations) and shows strong agreement with AERONET-derived trends at stations exhibiting statistically significant changes (R = 0.89), correctly capturing the trend sign at 96.7 % of sites. Negative trends are generally well represented, while positive trends are more frequently underestimated. Trend analyses across 18 regions of interest reveal significant AOD declines over eastern China, North Africa, Central Europe, and the eastern United States, alongside persistent increases over South Asia and the Middle East. Decadal analyses identify pronounced transitions, including a reversal from increasing to decreasing trends over eastern China and shifts in several dust- and biomass-burning-dominated regions. Sliding-window trend diagnostics further highlight temporal variability and regime changes. By integrating long-term reanalysis data, independent validation, and component-resolved diagnostics, this study provides a consistent assessment of recent global AOD evolution.
- 31 October 2025
- Published by Springer Nature
- p. 127-136
New Challenges in Solar Energy Resource and Forecasting at Different Temporal and Spatial Scales
- 31 October 2025
- Published by Springer Nature
- p. 127-136
New Challenges in Solar Energy Resource and Forecasting at Different Temporal and Spatial Scales
- Open Access
- 1 October 2025
- Published by MDPI AG in Environmental and Earth Sciences Proceedings
- Vol. 35 (1) , 59
Deep 3D Scattering of Solar Radiation in the Atmosphere Due to Clouds-D3D
The three-dimensional (3D) structure of clouds is a key factor in atmospheric processes, profoundly influencing solar radiation transfer, weather patterns, and climate dynamics. However, accurately representing this complex structure in radiative transfer models remains a significant challenge. As part of the Deep 3D Scattering of Solar Radiation in the Atmosphere due to Clouds (D3D) project, we conducted a comprehensive study on the role of all-sky imagers (ASIs) in reconstructing observational 3D cloud fields and integrating them into advanced 3D cloud modeling. Since November 2022, a network of four ASIs has been operating across the broader Patras region in Greece, continuously capturing atmospheric measurements over an area of approximately 50 km2. Using simultaneously captured images from the ASIs within the network, a 3D cloud reconstruction was performed utilizing advanced image processing techniques, with a primary focus on cumulus cloud scenarios. The Structure from Motion (SfM) technique was employed to reconstruct the 3D structural characteristics of clouds from two-dimensional images. The resulting 3D cloud fields were then integrated into the MYSTIC three-dimensional radiative transfer model to simulate and reconstruct solar irradiance fields.
- 1 October 2025
- Published by MDPI AG in Environmental and Earth Sciences Proceedings
- Vol. 35 (1) , 59
Open AccessDeep 3D Scattering of Solar Radiation in the Atmosphere Due to Clouds-D3D
The three-dimensional (3D) structure of clouds is a key factor in atmospheric processes, profoundly influencing solar radiation transfer, weather patterns, and climate dynamics. However, accurately representing this complex structure in radiative transfer models remains a significant challenge. As part of the Deep 3D Scattering of Solar Radiation in the Atmosphere due to Clouds (D3D) project, we conducted a comprehensive study on the role of all-sky imagers (ASIs) in reconstructing observational 3D cloud fields and integrating them into advanced 3D cloud modeling. Since November 2022, a network of four ASIs has been operating across the broader Patras region in Greece, continuously capturing atmospheric measurements over an area of approximately 50 km2. Using simultaneously captured images from the ASIs within the network, a 3D cloud reconstruction was performed utilizing advanced image processing techniques, with a primary focus on cumulus cloud scenarios. The Structure from Motion (SfM) technique was employed to reconstruct the 3D structural characteristics of clouds from two-dimensional images. The resulting 3D cloud fields were then integrated into the MYSTIC three-dimensional radiative transfer model to simulate and reconstruct solar irradiance fields.
- Open Access
- 13 May 2024
- Published by MDPI AG in Atmosphere
- Vol. 15 (5) , 594
Forecasting the Exceedances of PM2.5 in an Urban Area
Particular matter (PM) constitutes one of the major air pollutants. Human exposure to fine PM (PM with a median diameter less than or equal to 2.5 μm, PM2.5) has many negative and diverse outcomes for human health, such as respiratory mortality, lung cancer, etc. Accurate air-quality forecasting on a regional scale enables local agencies to design and apply appropriate policies (e.g., meet specific emissions limitations) to tackle the problem of air pollution. Under this framework, low-cost sensors have recently emerged as a valuable tool, facilitating the spatiotemporal monitoring of air pollution on a local scale. In this study, we present a deep learning approach (long short-term memory, LSTM) to forecast the intra-day air pollution exceedances across urban and suburban areas. The PM2.5 data used in this study were collected from 12 well-calibrated low-cost sensors (Purple Air) located in the greater area of the Municipality of Thermi in Thessaloniki, Greece. The LSTM-based methodology implements PM2.5 data as well as auxiliary data, meteorological variables from the Copernicus Atmosphere Monitoring Service (CAMS), which is operated by ECMWF, and time variables related to local emissions to enhance the air pollution forecasting performance. The accuracy of the model forecasts reported adequate results, revealing a correlation coefficient between the measured PM2.5 and the LSTM forecast data ranging between 0.67 and 0.94 for all time horizons, with a decreasing trend as the time horizon increases. Regarding air pollution exceedances, the LSTM forecasting system can correctly capture more than 70.0% of the air pollution exceedance events in the study region. The latter findings highlight the model’s capabilities to correctly detect possible WHO threshold exceedances and provide valuable information regarding local air quality.
- 13 May 2024
- Published by MDPI AG in Atmosphere
- Vol. 15 (5) , 594
Open AccessForecasting the Exceedances of PM2.5 in an Urban Area
Particular matter (PM) constitutes one of the major air pollutants. Human exposure to fine PM (PM with a median diameter less than or equal to 2.5 μm, PM2.5) has many negative and diverse outcomes for human health, such as respiratory mortality, lung cancer, etc. Accurate air-quality forecasting on a regional scale enables local agencies to design and apply appropriate policies (e.g., meet specific emissions limitations) to tackle the problem of air pollution. Under this framework, low-cost sensors have recently emerged as a valuable tool, facilitating the spatiotemporal monitoring of air pollution on a local scale. In this study, we present a deep learning approach (long short-term memory, LSTM) to forecast the intra-day air pollution exceedances across urban and suburban areas. The PM2.5 data used in this study were collected from 12 well-calibrated low-cost sensors (Purple Air) located in the greater area of the Municipality of Thermi in Thessaloniki, Greece. The LSTM-based methodology implements PM2.5 data as well as auxiliary data, meteorological variables from the Copernicus Atmosphere Monitoring Service (CAMS), which is operated by ECMWF, and time variables related to local emissions to enhance the air pollution forecasting performance. The accuracy of the model forecasts reported adequate results, revealing a correlation coefficient between the measured PM2.5 and the LSTM forecast data ranging between 0.67 and 0.94 for all time horizons, with a decreasing trend as the time horizon increases. Regarding air pollution exceedances, the LSTM forecasting system can correctly capture more than 70.0% of the air pollution exceedance events in the study region. The latter findings highlight the model’s capabilities to correctly detect possible WHO threshold exceedances and provide valuable information regarding local air quality.
- Open Access
- 23 March 2024
- Published by MDPI AG in Remote Sensing
- Vol. 16 (7) , 1132
A Machine Learning Approach to Retrieving Aerosol Optical Depth Using Solar Radiation Measurements
Aerosol optical depth (AOD) constitutes a key parameter of aerosols, providing vital information for quantifying the aerosol burden and air quality at global and regional levels. This study demonstrates a machine learning strategy for retrieving AOD under cloud-free conditions based on the synergy of machine learning algorithms (MLAs) and ground-based solar irradiance data. The performance of the proposed methodology was investigated by applying different components of solar irradiance. In particular, the use of direct instead of global irradiance as a model feature led to better performance. The MLA-based AODs were compared to reference AERONET retrievals, which encompassed RMSE values between 0.01 and 0.15, regardless of the underlying climate and aerosol environments. Among the MLAs, artificial neural networks outperformed the other algorithms in terms of RMSE at 54% of the measurement sites. The overall performance of MLA-based AODs against AERONET revealed a high coefficient of determination (R2 = 0.97), MAE of 0.01, and RMSE of 0.02. Compared to satellite (MODIS) and reanalysis (MERRA-2 and CAMSRA) data, the MLA-AOD retrievals revealed the highest accuracy at all stations. The ML-AOD retrievals have the potential to expand and complement the AOD information in non-existing timeframes when solar irradiances are available.
- 23 March 2024
- Published by MDPI AG in Remote Sensing
- Vol. 16 (7) , 1132
Open AccessA Machine Learning Approach to Retrieving Aerosol Optical Depth Using Solar Radiation Measurements
Aerosol optical depth (AOD) constitutes a key parameter of aerosols, providing vital information for quantifying the aerosol burden and air quality at global and regional levels. This study demonstrates a machine learning strategy for retrieving AOD under cloud-free conditions based on the synergy of machine learning algorithms (MLAs) and ground-based solar irradiance data. The performance of the proposed methodology was investigated by applying different components of solar irradiance. In particular, the use of direct instead of global irradiance as a model feature led to better performance. The MLA-based AODs were compared to reference AERONET retrievals, which encompassed RMSE values between 0.01 and 0.15, regardless of the underlying climate and aerosol environments. Among the MLAs, artificial neural networks outperformed the other algorithms in terms of RMSE at 54% of the measurement sites. The overall performance of MLA-based AODs against AERONET revealed a high coefficient of determination (R2 = 0.97), MAE of 0.01, and RMSE of 0.02. Compared to satellite (MODIS) and reanalysis (MERRA-2 and CAMSRA) data, the MLA-AOD retrievals revealed the highest accuracy at all stations. The ML-AOD retrievals have the potential to expand and complement the AOD information in non-existing timeframes when solar irradiances are available.