Articles
Atmospheric Measurement Techniques (18671381)18(6)pp. 1415-1439
Multi-layer aerosol optical depth (AOD) estimation with sufficient spatial and temporal resolution is crucial for effective aerosol monitoring, given the significant variations over time and space. While ground-based observations provide detailed vertical profiles, satellite data are essential for addressing the spatial and temporal gaps. This study utilizes profiles from the Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) and data from the Spinning Enhanced Visible and Infrared Imager (SEVIRI) to estimate vertical AOD values at 1.5, 3, 5, and 10 km layers. These estimations are achieved with spatial and temporal resolutions of 3 km_3 km and 15 min, respectively, over the European troposphere. We employed machine learning models - XGBoost (XGB) and random forest (RF) - trained on SEVIRI data from 2017 to 2018 for the estimations. Validation using CALIOP AOD retrievals in 2019 confirmed the reliability of our findings, emphasizing the importance of wind speed (Ws) and wind direction (Wd) in improving AOD estimation accuracy. A comparison between seasonal and annual models revealed slight variations in accuracy, leading to the selection of annual models as the preferred approach for estimating SEVIRI multi-layer AOD values. Among the annual models, the XGB model demonstrated superior performance over the RF model at all four layers, yielding more reliable AOD estimations with R2 values of 0.99, 0.97, 0.98, and 0.98 for the four layers from low- to high-altitude layers. Further validation using data from European Aerosol Research Lidar Network (EARLINET) stations across Europe in 2020 indicated that the XGB model still achieved better agreement with EARLINET AOD profiles, with R2 values of 0.86, 0.80, 0.75, and 0.59 and RMSE values of 0.022, 0.012, 0.015, and 0.005. We performed a qualitative validation of multi-layer AOD estimations by comparing spatial trends with CALIOP AOD retrievals for SEVIRI pixels on four dates in 2019, showing strong agreement across varying AOD levels. Additionally, the model successfully estimated AOD at 15 min intervals for two real events - a Saharan dust plume and the Mount Etna eruption - revealing consistent physical characteristics, including long-range transport in the upper layers and a gradual increase in AOD from lower to higher tropospheric layers during volcanic events. The results demonstrate that the proposed method facilitates comprehensive monitoring of AOD behavior throughout the four vertical layers of the troposphere, offering important insights into the dynamics of aerosol occurrence. © Author(s) 2025.
PFG - Journal of Photogrammetry, Remote Sensing and Geoinformation Science (25122819)93(4)pp. 335-350
Interferometric Synthetic Aperture Radar (InSAR) has become a reliable method for estimating ground surface displacement. Change point detection in time series is one of the key post-processing tools for displacement analysis. Due to the complex characteristics of displacement time series, researchers have used statistical and artificial intelligence-based methods to identify change points in the trend of time series and face many limitations. Regarding the lack of reliable ground-truth, a data simulation scheme for the learning stage in deep learning approaches is a challenging issue that is the first focus of this work. Inspired by the generative adversarial network (GAN), we proposed a workflow to simulate datasets, where the real-world time series component with known change points is used in the generative stage and the multilayer perceptron in the discriminator stage. Additionally, a novel method called Moving Average Large Kernel Convolutional Neural Network (MALkCNN) is proposed for trend change point detection, which has a large kernel size to maintain global trend changes while reducing the effects of fluctuations. The results show that the MALkCNN model achieved an F1 score of 0.8321 and an accuracy of 0.9095, outperforming the Time Gated Long Short-Term Memory (TG-LSTM) model as a benchmark while performing tasks seven times faster. The validity of the proposed method has been checked according to the sorting data from the European Ground Motion Service (EGMS) portal located in Germany and Italy. According to our findings, our approach provides a promising solution for identifying trend change points in complex InSAR time series. © Deutsche Gesellschaft für Photogrammetrie, Fernerkundung und Geoinformation (DGPF) e.V. 2025.
Atmospheric Pollution Research (13091042)15(1)
Aerosol Optical Depth (AOD) retrieved using Cloud-Aerosol LIDAR with Orthogonal Polarization (CALIOP) and Moderate Resolution Imaging Spectroradiometer (MODIS) plays a crucial role in aerosol sensing. However, the performance of these products varies across different aerosol concentrations. This research assessed the performance of CALIOP and MODIS AOD products over a three-year period (January 2017 to June 2019) under various aerosol concentrations in the Red Sea and the Persian Gulf. The products were compared with ship-based AERONET (Aerosol Robotic Network) measurements conducted in the study area. The findings reveal that MODIS AOD products exhibit greater reliability during clear days, whereas CALIOP AOD products are more accurate in regions characterized by high aerosol concentrations. However, CALIOP's limited product resolution prevents it from providing adequate spatial-temporal coverage under heavy aerosol concentrations. To enhance the accuracy of MODIS AOD, this paper proposes a methodology based on various machine learning (ML) algorithms, including Random Forest (RF), Multi-layer Perceptron (MLP), Extreme Gradient Boosting (XGBoost), Support Vector Machine (SVR), and Bayesian network. The study explores the performance of MODIS and CALIOP AOD products during moderate and pollution days in the Persian Gulf and Red Sea regions. The ML models are ranked based on their accuracy, with the order being XGBoost > MLP > SVRrbf > RF > Bayesian > SVRlinear for the annual and two semi-annual networks. In other words, the Bayesian method demonstrates higher efficiency (R2 = 0.97, 0.91, and 0.90) in handling seasonal subsets. The proposed model's outcomes are validated using ship-based AERONET AOD data. By comparing the estimated AOD values with the ship-based AERONET AOD values and analyzing the overall correlation results, it is evident that machine learning techniques effectively provide accurate AOD estimates for moderate and pollution days. © 2023 Turkish National Committee for Air Pollution Research and Control
Atmospheric Pollution Research (13091042)15(7)
Aerosol Optical Depth (AOD) across various altitudes is crucial for gaining a comprehensive understanding of aerosol dynamics. However, current methodologies utilizing passive remote sensing and active sensors have limitations in providing precise vertical coverage. In our methodology, we introduce a Seasonal-Independent model employing Machine Learning (ML) algorithms to retrieve AOD values at both 1.5 km and 3 km layers. The propose approach is assessed the performance of various ML algorithms, including XGBoost, Random Forest (RF), Multilayer Perceptron (MLP), and Support Vector Regression (SVR). Remarkably, our study successfully overcame seasonal constraints, yielding impressive R2 values of 0.94, 0.93, 0.93, and 0.87 for the 1.5 km layer, and 0.83, 0.79, 0.82, and 0.78 for the 3 km layer across the mentioned models for 2017, 2018 and 2019 data. Evaluating the proposed Seasonal-Independent XGBoost model against CALIOP AOD values for the 2020 data, we observed substantial agreement with R2 values of 0.93 and 0.81, and minimal RMSE values of 0.002 and 0.004 for the AODs at 1.5 km and 3 km, respectively. Furthermore, a comparative analysis of trends between estimated and CALIOP AODs revealed a strong resemblance in both altitude layers. © 2024 Turkish National Committee for Air Pollution Research and Control
Earth Science Informatics (18650473)16(3)pp. 2529-2543
Classifying urban land use/cover types poses significant challenges due to the complex and heterogeneous nature of urban landscapes. Recent years have witnessed notable advancements in land use/cover classification, driven by improvements in classification methods and the utilization of data from multiple sources. Deep learning networks, especially, have gained prominence in various image analysis tasks, including land use/cover classification. However, when it comes to urban areas, the classification of urban land use/cover encounters additional obstacles, including the complexity of classes, limited training data, and the presence of numerous urban categories. To overcome the limitations arising from similar classes and insufficient training data, we propose a novel approach that integrates hyperspectral and LiDAR data through a Conditional Generative Adversarial Network (CGAN) for semantic segmentation. Our methodology leverages the UNet + + generator and PatchGAN discriminator to achieve accurate segmentation. The CGAN-generated segmented images are then processed by a fully connected neural network (FCN) to classify 20 land use/cover classes. By validating our approach on the 2018 GRSS Data Fusion dataset, our study demonstrates its exceptional operational performance. The CGAN architecture outperforms previous algorithms in terms of class diversity and training data volume. By generating synthetic data that closely resembles the ground truth, the CGAN enhances the classification performance. Clear visual distinctions are observed among various urban features, such as vegetation, trees, buildings, roads, and cars. Classes associated with healthy grass, stressed grass, bare earth, and stadium seats achieve high accuracy. However, road and railway classes exhibit poorer performance due to their similarity with sidewalk, highway, major thoroughfare, and crosswalk classes. Overall, our study showcases a significant improvement in classification accuracy, achieving an approximate accuracy of 96.98% compared to the winning articles presented in the 2018 competition, which achieved accuracies of 64.95% and 76.54%, respectively. This improvement in accuracy can be attributed to the effective extraction and combination of high and low-level urban land cover/land use features within our proposed architecture. © 2023, The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.
