Mehdi Momeni
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Mehdi Momeni has a total of 69 co-authors. Here are the top co-authors by number of shared publications:
- Payman Moallem
Department of Electrical Engineering Faculty of Engineering University of Isfahan Isfahan Iran
7 shared publications Amirhassan Monadjemi
Department of Computer Science, School of Computing, National University of Singapore, Singapore, Singapore
3 shared publicationsAkbar Rahimi
2 shared publications- Amir Rahimi
Department of Chemical Engineering, College of Engineering, University of Isfahan, P.O. Box 81746-73441, Isfahan, Iran
2 shared publications - Iman Khosravi
Department of Remote Sensing & GIS, Faculty of Geography, University of Tehran, Tehran, Iran
2 shared publications Kamal Shahtalebi
2 shared publications
Latest Publications
- Open Access
- 7 May 2026
- Published by Elsevier in SSRN Electronic Journal
A QA-Driven Framework for Gap Reconstruction in MAIAC Aerosol Optical Depth Using Geostatistical and Machine Learning Approaches
Long-term satellite records of Aerosol Optical Depth (AOD) from satellite products such as MAIAC algorithm, provide a unique observational basis for air-quality monitoring and climate research. However, the presence of complex, QA-dependent retrieval gaps limits the usability of these historical datasets and challenges conventional gap-filling approaches. This study introduces a Quality Assurance (QA)-based framework that both classifies and reconstructs MAIAC AOD gaps, explicitly linking gap mechanisms to reconstruction performance. We define three gap classes: (1) complete retrieval absence (Type-1), (2) QA-flagged, high-uncertainty retrievals at boundaries (Type-2), and (3) algorithm-induced discontinuities at aerosol-model transitions (Type-3). These gap types are simulated to preserve the statistical and algorithmic structure of real MAIAC artifacts, including heavy-tailed residual behavior in Type-2 gaps modeled via a Laplace error distribution and edge artifacts at dust/smoke model boundaries for Type-3. Based on this gap classification, four representative reconstruction methods Ordinary Kriging, Geographically Weighted Regression (GWR), Support Vector Regression (SVR), and XGBoost are applied to MAIAC AOD over the study region for 2021–2023. Models are trained exclusively on valid retrievals and evaluated at simulated gap locations across varying gap sizes, aerosol loading conditions, and spatial heterogeneity, which represent common drivers of AOD gap variability. Machine-learning models incorporate auxiliary predictors (land cover, elevation, and key meteorological variables such as air temperature, relative humidity, and wind speed, alongside spatial information. Reconstruction performance is quantified using RMSE, MAE, maximum error, R2, and error variance, and inter-model differences are assessed via one-way ANOVA (F ≈ 20, p < 0.001) and Tukey’s Honestly Significant Difference (HSD) tests. Results show that model performance is condition-dependent. Under high-AOD, spatially heterogeneous conditions, Kriging attains the lowest median RMSE and narrowest error dispersion, significantly outperforming SVR (Tukey HSD, p < 0.05), while SVR exhibits higher and more variable errors. Under low-AOD conditions, XGBoost and SVR yield statistically indistinguishable median RMSE and both significantly outperform GWR, whereas Kriging provides intermediate accuracy. These findings demonstrate that local spatial dependency is the primary driver of reconstruction accuracy under challenging regimes, whereas machine-learning models can match or exceed geostatistical methods under cleaner, less heterogeneous conditions. The proposed QA-driven gap classification and comparative assessment offer practical guidance for selecting AOD gap-filling strategies tailored to gap type and atmospheric regime, thereby enhancing the reliability and usability of MAIAC AOD for downstream applications.
- 7 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessA QA-Driven Framework for Gap Reconstruction in MAIAC Aerosol Optical Depth Using Geostatistical and Machine Learning Approaches
Long-term satellite records of Aerosol Optical Depth (AOD) from satellite products such as MAIAC algorithm, provide a unique observational basis for air-quality monitoring and climate research. However, the presence of complex, QA-dependent retrieval gaps limits the usability of these historical datasets and challenges conventional gap-filling approaches. This study introduces a Quality Assurance (QA)-based framework that both classifies and reconstructs MAIAC AOD gaps, explicitly linking gap mechanisms to reconstruction performance. We define three gap classes: (1) complete retrieval absence (Type-1), (2) QA-flagged, high-uncertainty retrievals at boundaries (Type-2), and (3) algorithm-induced discontinuities at aerosol-model transitions (Type-3). These gap types are simulated to preserve the statistical and algorithmic structure of real MAIAC artifacts, including heavy-tailed residual behavior in Type-2 gaps modeled via a Laplace error distribution and edge artifacts at dust/smoke model boundaries for Type-3. Based on this gap classification, four representative reconstruction methods Ordinary Kriging, Geographically Weighted Regression (GWR), Support Vector Regression (SVR), and XGBoost are applied to MAIAC AOD over the study region for 2021–2023. Models are trained exclusively on valid retrievals and evaluated at simulated gap locations across varying gap sizes, aerosol loading conditions, and spatial heterogeneity, which represent common drivers of AOD gap variability. Machine-learning models incorporate auxiliary predictors (land cover, elevation, and key meteorological variables such as air temperature, relative humidity, and wind speed, alongside spatial information. Reconstruction performance is quantified using RMSE, MAE, maximum error, R2, and error variance, and inter-model differences are assessed via one-way ANOVA (F ≈ 20, p < 0.001) and Tukey’s Honestly Significant Difference (HSD) tests. Results show that model performance is condition-dependent. Under high-AOD, spatially heterogeneous conditions, Kriging attains the lowest median RMSE and narrowest error dispersion, significantly outperforming SVR (Tukey HSD, p < 0.05), while SVR exhibits higher and more variable errors. Under low-AOD conditions, XGBoost and SVR yield statistically indistinguishable median RMSE and both significantly outperform GWR, whereas Kriging provides intermediate accuracy. These findings demonstrate that local spatial dependency is the primary driver of reconstruction accuracy under challenging regimes, whereas machine-learning models can match or exceed geostatistical methods under cleaner, less heterogeneous conditions. The proposed QA-driven gap classification and comparative assessment offer practical guidance for selecting AOD gap-filling strategies tailored to gap type and atmospheric regime, thereby enhancing the reliability and usability of MAIAC AOD for downstream applications.
- 12 March 2025
- Published by Springer Nature in International Journal of Environmental Science and Technology
- Vol. 22 (9) , 7797-7814
Enhancing PM2.5 modeling with reinforcement learning: dynamic ensembling of multi-graph attention networks and deep recurrent models
- 12 March 2025
- Published by Springer Nature in International Journal of Environmental Science and Technology
- Vol. 22 (9) , 7797-7814
Enhancing PM2.5 modeling with reinforcement learning: dynamic ensembling of multi-graph attention networks and deep recurrent models
- 23 August 2024
- Published by Springer Nature in Environmental Science and Pollution Research
- Vol. 31 (40) , 53140-53155
A new attention-based CNN_GRU model for spatial–temporal PM2.5 prediction
Accurately predicting the spatial-temporal distribution of PM2.5 is challenging due to missing data and selecting an appropriate modeling method. Effective imputation of missing data must consider the relationships between variables while preserving their inherent variability and uncertainty. In this study, we employed machine learning techniques to impute missing data by analyzing the relationships between meteorological variables and other pollutants. Subsequently, we introduced an innovative spatiotemporal hybrid model, AC_GRU, which integrates a one-dimensional convolutional neural network (CNN), GRU, and an attention-based network to predict PM2.5 concentrations in urban areas. The AC_GRU model utilizes meteorological variables, PM2.5 concentrations from nearby air quality monitoring stations, and concentrations of other pollutants as inputs. This approach allows the model to learn spatiotemporal correlations within the time-series data, enhancing the accuracy of PM2.5 predictions. Additionally, the attention mechanism improves prediction accuracy by automatically weighting the past input variables based on their importance for future PM2.5 predictions. The experimental results demonstrate that our AC_GRU model outperforms state-of-the-art methods, making it a valuable tool for urban air quality management and public health protection.
- 23 August 2024
- Published by Springer Nature in Environmental Science and Pollution Research
- Vol. 31 (40) , 53140-53155
A new attention-based CNN_GRU model for spatial–temporal PM2.5 prediction
Accurately predicting the spatial-temporal distribution of PM2.5 is challenging due to missing data and selecting an appropriate modeling method. Effective imputation of missing data must consider the relationships between variables while preserving their inherent variability and uncertainty. In this study, we employed machine learning techniques to impute missing data by analyzing the relationships between meteorological variables and other pollutants. Subsequently, we introduced an innovative spatiotemporal hybrid model, AC_GRU, which integrates a one-dimensional convolutional neural network (CNN), GRU, and an attention-based network to predict PM2.5 concentrations in urban areas. The AC_GRU model utilizes meteorological variables, PM2.5 concentrations from nearby air quality monitoring stations, and concentrations of other pollutants as inputs. This approach allows the model to learn spatiotemporal correlations within the time-series data, enhancing the accuracy of PM2.5 predictions. Additionally, the attention mechanism improves prediction accuracy by automatically weighting the past input variables based on their importance for future PM2.5 predictions. The experimental results demonstrate that our AC_GRU model outperforms state-of-the-art methods, making it a valuable tool for urban air quality management and public health protection.
- 6 August 2024
- Published by Springer Nature in Plant and Soil
- Vol. 509 (1-2) , 377-397
SWIR based estimation of TIR emissivity of bare soil surfaces using deep conditional generative adversarial network in Landsat data
- 6 August 2024
- Published by Springer Nature in Plant and Soil
- Vol. 509 (1-2) , 377-397
SWIR based estimation of TIR emissivity of bare soil surfaces using deep conditional generative adversarial network in Landsat data
- Open Access
- 19 June 2024
- Published by Springer Nature in Research Square
A New Attention-Based a convolutional neural network-gated recurrent unit Model for Spatial-Temporal PM2.5 Prediction
Accurately predicting the spatiotemporal distribution of PM2.5 using real data is a challenge, as missing data must be imputed and a suitable modeling method selected. When imputing missing data, the relationships between the variables must be taken into account while preserving their inherent variability and uncertainty. In this study, the imputation of missing data considering the relationships between meteorological variables and other pollutants was performed using different machine learning methods. After handling the missing data, the selection of an appropriate modeling method becomes imperative. In this paper, we attempt to present an innovative spatiotemporal hybrid model based on GRU that aims to obtain and extract spatiotemporal relationships. A dataset collected from 2017 to 2023 in Tehran metropolis was used for the study. The proposed model with (R2 = 0.84) can well determine the changes in PM2.5 concentration for each station and visualize the level of air pollution on a diffusion map.
- 19 June 2024
- Published by Springer Nature in Research Square
Open AccessA New Attention-Based a convolutional neural network-gated recurrent unit Model for Spatial-Temporal PM2.5 Prediction
Accurately predicting the spatiotemporal distribution of PM2.5 using real data is a challenge, as missing data must be imputed and a suitable modeling method selected. When imputing missing data, the relationships between the variables must be taken into account while preserving their inherent variability and uncertainty. In this study, the imputation of missing data considering the relationships between meteorological variables and other pollutants was performed using different machine learning methods. After handling the missing data, the selection of an appropriate modeling method becomes imperative. In this paper, we attempt to present an innovative spatiotemporal hybrid model based on GRU that aims to obtain and extract spatiotemporal relationships. A dataset collected from 2017 to 2023 in Tehran metropolis was used for the study. The proposed model with (R2 = 0.84) can well determine the changes in PM2.5 concentration for each station and visualize the level of air pollution on a diffusion map.
- Open Access
- 21 April 2024
- Published by Wiley
Advancing Drought Modeling: Harnessing Innovative GEE-derived Parameters for Improved Fusion-based Analysis at Regional Scale
This study proposes a novel fusion-based approach using Google Earth Engine (GEE) to accurately model distinct drought phenomena in various environmental regions, with Iran as a case study. The method integrates Environmental Parameters (EPs) - Vegetation (Veg), Temperature (Tem), Humidity (Hum), and Evaporation-Transpiration (ET) - with auxiliary parameters such as land-cover, topography, and wind-speed to improve modeling accuracy. A 39-year (1982-2020) time series of 14 Remote Sensing (RS) indices and 18 Ground-Based (GB) drought indices were used as input and output, respectively, for training and testing machine learning algorithms in three phases. The results obtained during the test period (2015-2020) from each phase were employed in the next phase based on a hereditary procedure to evaluate the potential of the proposed indices and their associated EPs (group of indices). The Consolidated Fusion-Based Drought Model (CFDM) was developed as the final product, which demonstrated superior accuracy and stability compared to other models, with an overall accuracy (OA) higher than 90% for all GB indices. As a result, the CFDM has utilized for generating drought maps in Iran. Furthermore, the effectiveness of the simultaneous use of auxiliary parameters and EPs was demonstrated through the Prevalent Fusion-Based Drought Model (PFDM). Our approach entails a systematic framework that incorporates innovative parameters (indices/indicators) to accurately model drought phenomena. This can contribute to the development of effective management strategies.
- 21 April 2024
- Published by Wiley
Open AccessAdvancing Drought Modeling: Harnessing Innovative GEE-derived Parameters for Improved Fusion-based Analysis at Regional Scale
This study proposes a novel fusion-based approach using Google Earth Engine (GEE) to accurately model distinct drought phenomena in various environmental regions, with Iran as a case study. The method integrates Environmental Parameters (EPs) - Vegetation (Veg), Temperature (Tem), Humidity (Hum), and Evaporation-Transpiration (ET) - with auxiliary parameters such as land-cover, topography, and wind-speed to improve modeling accuracy. A 39-year (1982-2020) time series of 14 Remote Sensing (RS) indices and 18 Ground-Based (GB) drought indices were used as input and output, respectively, for training and testing machine learning algorithms in three phases. The results obtained during the test period (2015-2020) from each phase were employed in the next phase based on a hereditary procedure to evaluate the potential of the proposed indices and their associated EPs (group of indices). The Consolidated Fusion-Based Drought Model (CFDM) was developed as the final product, which demonstrated superior accuracy and stability compared to other models, with an overall accuracy (OA) higher than 90% for all GB indices. As a result, the CFDM has utilized for generating drought maps in Iran. Furthermore, the effectiveness of the simultaneous use of auxiliary parameters and EPs was demonstrated through the Prevalent Fusion-Based Drought Model (PFDM). Our approach entails a systematic framework that incorporates innovative parameters (indices/indicators) to accurately model drought phenomena. This can contribute to the development of effective management strategies.
- Open Access
- 1 February 2024
- Published by Springer Nature in Research Square
SWIR based estimation of TIR emissivity of bare soil surfaces using deep conditional generative adversarial network in Landsat data
Aims Land surface emissivity (LSE) is an important variable in soil studies. Although there are various remote sensing methods to estimate LSE, accurately predicting LSE still remains a major challenge. Typically, the correlation between LSE and the Visible Near Infrared bands is employed for LSE estimation. However, some studies have raised some concerns about this correlation, especially in bare soil areas. Therefore, it is necessary to conduct further investigation to determine if there exists a nonlinear relationship between the LSE and other spectral bands, which was not detected by simple linear correlation/regression. Methods In this study, firstly, a deep Auto-encoder (DAE) network has been used to investigate the correlation between LSE and other spectral bands. Subsequently, we have applied a Conditional Generative Adversarial Network (CGAN) to estimate the LSE. The proposed CGAN was trained using the Landsat and ECOSTRESS satellite datasets. The performance of the developed network was then compared with NDVI-based method on satellite/simulated-based bare soil pixels. Results For satellite data, the RMSE (Root Mean Squared Error) and correlation coefficient (R) between the estimated LSE using proposed CGAN and ECOSTRESS LSE are 0.005 and 0.97, respectively. For the simulated data, the RMSE and R between the estimated LSE and the simulated one are 0.01 and 0.92, respectively. Conclusion The results of the DAE show considerable relationship between the LSE and Short-Wave Infrared bands which not be seen using simple linear correlation.
- 1 February 2024
- Published by Springer Nature in Research Square
Open AccessSWIR based estimation of TIR emissivity of bare soil surfaces using deep conditional generative adversarial network in Landsat data
Aims Land surface emissivity (LSE) is an important variable in soil studies. Although there are various remote sensing methods to estimate LSE, accurately predicting LSE still remains a major challenge. Typically, the correlation between LSE and the Visible Near Infrared bands is employed for LSE estimation. However, some studies have raised some concerns about this correlation, especially in bare soil areas. Therefore, it is necessary to conduct further investigation to determine if there exists a nonlinear relationship between the LSE and other spectral bands, which was not detected by simple linear correlation/regression. Methods In this study, firstly, a deep Auto-encoder (DAE) network has been used to investigate the correlation between LSE and other spectral bands. Subsequently, we have applied a Conditional Generative Adversarial Network (CGAN) to estimate the LSE. The proposed CGAN was trained using the Landsat and ECOSTRESS satellite datasets. The performance of the developed network was then compared with NDVI-based method on satellite/simulated-based bare soil pixels. Results For satellite data, the RMSE (Root Mean Squared Error) and correlation coefficient (R) between the estimated LSE using proposed CGAN and ECOSTRESS LSE are 0.005 and 0.97, respectively. For the simulated data, the RMSE and R between the estimated LSE and the simulated one are 0.01 and 0.92, respectively. Conclusion The results of the DAE show considerable relationship between the LSE and Short-Wave Infrared bands which not be seen using simple linear correlation.
- 4 October 2023
- Published by Elsevier in Atmospheric Pollution Research
- Vol. 15 (1)
Improving the accuracy of AOD by using multi-sensors data over the Red Sea and the Persian Gulf
- 4 October 2023
- Published by Elsevier in Atmospheric Pollution Research
- Vol. 15 (1)
Improving the accuracy of AOD by using multi-sensors data over the Red Sea and the Persian Gulf
- 14 August 2023
- Published by Elsevier in Applied Soft Computing
- Vol. 148
A fusion-based approach to improve hyperspectral images’ classification using metaheuristic band selection
- 14 August 2023
- Published by Elsevier in Applied Soft Computing
- Vol. 148
A fusion-based approach to improve hyperspectral images’ classification using metaheuristic band selection