Search Results (5)

The Vectorized Earth Observation Retrieval (VEOR) algorithm is a novel algorithm suited to the efficient supervised classification of large Earth Observation (EO) datasets. VEOR addresses shortcomings in well-established machine learning methods with an emphasis on numerical performance. Its characteristics include (1) derivation of classification probability; (2) objective selection of classification features that maximize Cohen’s kappa coefficient ($\kappa$) derived from iterative “leave-one-out” cross-validation; (3) reduced sensitivity of the classification results to imbalanced classes; (4) smoothing of the classification probability field to reduce noise/mislabeling; (5) numerically efficient retrieval based on a pre-computed look-up vector (LUV); and (6) separate parametrization of the algorithm for each discrete feature class (e.g., land cover). Within this study, the performance of the VEOR classifier was compared to other commonly used machine learning algorithms: K-nearest neighbors, support vector machines, Gaussian process, decision trees, random forest, artificial neural networks, AdaBoost, Naive Bayes and Quadratic Discriminant Analysis. Firstly, the comparison was performed using synthetic 2D (two-dimensional) datasets featuring different sample sizes, levels of noise (i.e., mislabeling) and class imbalance. Secondly, the same experiments were repeated for 7D datasets consisting of informative, redundant and insignificant features. Ultimately, the benchmarking of the classifiers involved cloud discrimination using MODIS satellite spectral measurements and a reference cloud mask derived from combined CALIOP lidar and CPR radar data. The results revealed that the proposed VEOR algorithm accurately discriminated cloud cover using MODIS data and accurately classified large synthetic datasets with low or moderate levels of noise and class imbalance. On the contrary, VEOR did not feature good classification skills for significantly distorted or for small datasets. Nevertheless, the comparisons performed proved that VEOR was within the 3–4 most accurate classifiers and that it can be applied to large Earth Observation datasets. Full article

Sea fog is a precarious weather disaster affecting transportation on the sea. The accuracy of the threshold method for sea fog detection is limited by time and region. In comparison, the deep learning method learns features of objects through different network layers and can therefore accurately extract fog data and is less affected by temporal and spatial factors. This study proposes a scSE-LinkNet model for daytime sea fog detection that leverages residual blocks to encoder feature maps and attention module to learn the features of sea fog data by considering spectral and spatial information of nodes. With the help of satellite radar data from Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP), a ground sample database was extracted from Moderate Resolution Imaging Spectroradiometer (MODIS) L1B data. The scSE-LinkNet was trained on the training set, and quantitative evaluation was performed on the test set. Results showed the probability of detection (POD), false alarm rate (FAR), critical success index (CSI), and Heidke skill scores (HSS) were 0.924, 0.143, 0.800, and 0.864, respectively. Compared with other neural networks (FCN, U-Net, and LinkNet), the CSI of scSE-LinkNet was improved, with a maximum increase of nearly 8%. Moreover, the sea fog detection results were consistent with the measured data and CALIOP products. Full article

Taking advantage of both the polar orbit active remote sensing data (from the Cloud-Aerosol Lidar with Orthogonal Polarization—CALIOP) and vertical information and the geostationary passive remote sensing measurements (from the Spinning Enhanced Visible and Infrared Imager) with large coverage, a methodology is developed for retrieving the volcanic ash cloud top height (VTH) from combined CALIOP and Spinning Enhanced Visible and Infrared Imager (SEVIRI) data. This methodology is a deep-learning-based algorithm through hybrid use of Stacked Denoising AutoEncoder (SDA), the Genetic Algorithm (GA), and the Least Squares Support Vector Regression (LSSVR). A series of eruptions over Iceland’s Eyjafjallajökull volcano from April to May 2010 and the Puyehue-Cordón Caulle volcanic complex eruptions in Chilean Andes in June 2011 were selected as typical cases for independent validation of the VTH retrievals under various meteorological backgrounds. It is demonstrated that using the hybrid deep learning algorithm, the nonlinear relationship between satellite-based infrared (IR) radiance measurements and the VTH can be well established. The hybrid deep learning algorithm not only performs well under a relatively simple meteorological background but also is robust under more complex meteorological conditions. Adding atmospheric temperature vertical profile as additional information further improves the accuracy of VTH retrievals. The methodology and approaches can be applied to the measurements from the advanced imagers onboard the new generation of international geostationary (GEO) weather satellites for retrieving the VTH science product. Full article

Cloud occurrences, vertical structures, and along-track horizontal scales over East Asia are studied using four years (2007–2010) of CloudSat 2B-CLDCLASS-LIDAR data. The CloudSat 2B-CLDCLASS-LIDAR data employs combined CloudSat radar and Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) measurements to provide by far the most accurate detections of cloud boundaries and their vertical structures. The mean cloud occurrence frequency over East Asia is 66.3%, which is 13.8% and 21.6% higher than that from the Cloud–Aerosol LIdar with Orthogonal Polarization (CALIOP) level 2 5-km cloud layer product and the CloudSat 2B-GEOPROF product, respectively. Cloud-top heights over East Asia show three local peaks at approximately 1.5 km, 10 km, and 15 km above ground level (AGL), indicating different mid-altitude cloud formation mechanisms from those over the tropics. Significant fractions of low-level cloud, mid-level cloud, and high-level cloud have thicknesses smaller than 0.5 km, indicating that models with vertical resolutions lower than 0.5 km have difficulties resolving those clouds. The average cloud along-track horizontal scale over East Asia is 82.0 km. Probability distribution functions (PDFs) of cloud along-track horizontal scales suggest that approximately 81.2% of the clouds over East Asia cannot be resolved by climate models with a grid resolution of 1°. The results from this study can be used to improve cloud parameterizations in climate models and validate model simulations of clouds over East Asia. Full article

To investigate the characteristics of Spinning Enhanced Visible and Infrared Imager (SEVIRI)-derived products from the UK Met Office algorithm, one year of cloud occurrence frequency (COF) and cloud-top height (CTH) data from May 2013 to April 2014 was analysed in comparison with Cloud Profiling Radar (CPR) and Cloud-Aerosol LiDAR with Orthogonal Polarization (CALIOP) cloud products observed from the A-Train constellation. Because CPR operated in daylight-only data collection mode, daytime products were validated in this study. It is important to note that the different sensor characteristics cause differences in CTH retrievals. The CTH of active instruments, CPR and CALIOP, is derived from the return time of the backscattered radar or LiDAR signal, while the infrared sensor, SEVIRI, measures a radiatively effective CTH. Therefore, some systematic differences in comparison results are expected. However, similarities in spatial distribution and seasonal variability of COFs were noted among SEVIRI, CALIOP, and CPR products, although COF derived by the SEVIRI algorithm showed biases of 14.35% and −3.90% compared with those from CPR and CALIOP measurements, respectively. We found that the SEVIRI algorithm estimated larger COF values than the CPR product, especially over oceans, whereas smaller COF was detected by SEVIRI measurements over land and in the tropics than by CALIOP, where multi-layer clouds and thin cirrus clouds are dominant. CTHs derived from SEVIRI showed better agreement with CPR than with CALIOP. Further comparison with CPR showed that SEVIRI CTH was highly sensitive to the CO₂ bias correction used in the Minimum Residual method. Compared with CPR CTHs, SEVIRI has produced stable CTHs since the bias correction update in November 2013, with a correlation coefficient of 0.93, bias of −0.27 km, and standard deviation of 1.61 km. Full article