Cloud Detection Based on High Resolution Stereo Pairs of the Geostationary Meteosat Images

Abstract

Due to the considerable impact of clouds on the energy balance in the atmosphere and on the earth surface, they are of great importance for various applications in meteorology or remote sensing. An important aspect of the cloud research studies is the detection of cloudy pixels from the processing of satellite images. In this research, we investigated a stereographic method on a new set of Meteosat images, namely the combination of the high resolution visible (HRV) channel of the Meteosat-8 Indian Ocean Data Coverage (IODC) as a stereo pair with the HRV channel of the Meteosat Second Generation (MSG) Meteosat-10 image at 0° E. In addition, an approach based on the outputs from stereo analysis was proposed to detect cloudy pixels. This approach is introduced with a 2D-scatterplot based on the parallax value and the minimum intersection distance. The mentioned scatterplot was applied to determine/detect cloudy pixels in various image subsets with different amounts of cloud cover. Apart from the general advantage of the applied stereography method, which only depends on geometric relationships, the cloud detection results are also improved because: (1) The stereo pair is the HRV bands of the Spinning Enhanced Visible and InfraRed Imager (SEVIRI) sensor, with the highest spatial resolution available from the Meteosat geostationary platform; and (2) the time difference between the image pairs is nearly 5 s, which improves the matching results and also decreases the effect of cloud movements. In order to prove this improvement, the results of this stereo-based approach were compared with three different reflectance-based target detection techniques, including the adaptive coherent estimator (ACE), constrained energy minimization (CEM), and matched filter (MF). The comparison of the receiver operating characteristics (ROC) detection curves and the area under these curves (AUC) showed better detection results with the proposed method. The AUC value was 0.79, 0.90, 0.90, and 0.93 respectively for ACE, CEM, MF, and the proposed stereo-based detection approach. The results of this research shall enable a more realistic modelling of down-welling solar irradiance in the future.

Geometric estimation of volcanic eruption column height from GOES-R near-limb imagery – Part 2: Case studies

“…Two technical points are worth mentioning about the temperature methods used for comparison. First, the VolSatView information system (Bril et al, 2019;Girina et al, 2018;Gordeev et al 2016) operated by the Kamchatka Volcanic Eruption Response Team (KVERT) derives plume heights from Himawari-8 11 µm brightness temperatures (BT 11 ).…”

GEO–GEO Stereo-Tracking of Atmospheric Motion Vectors (AMVs) from the Geostationary Ring

“…This relaxed sampling requirement enables stereo winds to be produced from a heterogeneous constellation that includes pairs in the GEO ring, or LEO-GEO pairs filling gaps in the overlaps between GEO satellites. Other work [12,13] on stereo measurements of cloud-top heights (hence AMV height assignment) assumes synchronous observations, but Merucci et al [14] describes a method that adjusts for non-simultaneous observations when using the parallax to measure the height of volcanic ash plumes.…”

Cloud detection algorithm based on point by point refinement

“…Ji et al [5] proposed an F-CNN remote sensing image cloud detection method based on a fully convolutional neural network model, which can achieve cloud segmentation in large scale, high resolution, multi-channel remote sensing images, but the method is difficult to distinguish between clouds and snow. Dehnavi et al [6] proposed a cloud detection method based on stereo analysis output, which identifies and detects multicloud pixels with different cloud amounts through scatter plots, but the method covers complex regions with cloud It is difficult to determine a suitable threshold for accurate detection. Although the above work has greatly improved the accuracy of cloud detection, there is still much room for improvement for scenarios with variable cloud size and sparse distribution in remote sensing images.…”