Graph neural networks are a newly established category of machine learning algorithms dealing with relational data. They can be used for the analysis of both spatial and/or temporal data. They are capable of modeling how time series of nodes, which are located at different spatial positions, change by the exchange of information between nodes and their neighbors. As a result, time series can be predicted to future epochs.GNSS networks consist of stations at different locations, each producing time series of geodetic parameters, such as changes in their positions. In order to successfully apply graph neural networks to predict time series from GNSS networks, the physical properties of GNSS time series should be taken into account. Thus, we suggest a new graph neural network algorithm that has both a physical and a mathematical basis. The physical part is based on the fundamental concept of information exchange between nodes and their neighbors. Here, the temporal correlation between the changes of time series of the nodes and their neighbors is considered, which is computed by geophysical loading and/or climatic data. The mathematical part comes from the time series prediction by mathematical models, after the removal of trends and periodic effects using the singular spectrum analysis algorithm. In addition, it plays a role in the computation of the impact of neighboring nodes, based on the spatial correlation computed according to the pair-wise node-neighbor distance. The final prediction is the simple weighted summation of the predicted values of the time series of the node and those of its neighbors, in which weights are the multiplication of the spatial and temporal correlations.In order to show the efficiency of the proposed algorithm, we considered a global network of more than 18000 GNSS stations and defined the neighbors of each node as stations that are located within the range of 10 km. We performed several different analyses, including the comparison between different machine learning algorithms and statistical methods for the time series prediction part, the impact of the type of data used for the computation of temporal correlation (climatic and/or geophysical loading), and comparison with other state-of-the-art graph neural network algorithms. We demonstrate the superiority of our method to the current graph neural network algorithms when applied to time series of geodetic networks. In addition, we show that the best machine learning algorithm to use within our graph neural network architecture is the multilayer perceptron, which shows an average of 0.34 mm in prediction accuracy. Furthermore, we find that the statistical methods have lower accuracies than machine learning ones, as much as 44 percent.
The road traffic is highly sensitive to weather conditions. Accumulation of snow on the road can cause important safety problems. But road conditions monitoring is as hard as critical: in mid-latitude countries, on the one hand, the spatial variability of snowfall is high and on the other hand, accurate characterization of snow accumulation mainly relies on costly sensors. In recent decades, webcams have become ubiquitous along the road network. The quality of these webcams is variable but even low-resolution images capture information about the extent and the thickness of the snow layer. Their images are also currently used by forecasters to refine their analysis. The automatic extraction of relevant meteorological information is hence very useful. Recently, generic and efficient computer vision methods have emerged. Their application to image-based weather estimation has become an attractive field of research. However, the scope of existing work is generally limited to high-resolution images from one or a few cameras. In this study, we show that for a moderate effort of labelling, recent Machine Learning approaches allow us to predict quantitative indices of the snow depth for a large variety of webcam settings and illumination. Our approach is based on two datasets. The smallest one contains about 2.000 images coming from ten webcams that were set up near sensors devoted to snow depth measurements. The largest one contains 20,000 images coming from 200 cameras of the AMOS dataset. Meteorological standard rules of human observation and the specifics of the webcams have been taken into account to manually label each image. These labels are not only about the thickness and the extent of the snow layer but also describe the precipitation (rain or snow, presence of streaks), the optical range and the foreground noise. Both datasets contain night images (45%) and at least 15% of images corrupted by foreground noise (filth, droplets, and snowflakes on the lens). The labels of the AMOS subset allowed us to train ranking models for snow depth and visibility using a multi-task setting. The models are then calibrated on the smallest dataset. We tested several versions, built from pre-trained CNNs (ResNet152, DenseNet161, and VGG16). Results are promising with up to 85% accuracy for comparison tasks, but a 10% decrease can be observed when the test webcams have not been used during the training phase. A case study based on a widespread snow event over the French territory will be presented. We will show the potential of our method through a comparison with operational model forecasts.
Deep-seated gravitational slope deformations (DSGSD) gains new attention in Taiwan due to their catastrophic impacts on lives and infrastructures during Typhoon Morakot in 2009. As the main Taiwan island is located on a complex convergent plate boundary, conventional observations and analyses suggest that the island’s strong tectonic activity has, along with its subtropical climate and intense human activity in mountain areas, contributed to the formation of deep-seated landslides. It is especially so for high-altitude areas featuring Miocene to Eocene meta-sandstone and slate successions, where reactivations of landslide terrains are observed from field observation and some GPS sites after specific events. Among them, Tienchih, located in Lalong River of Kaohsiung, and Yakou, few km east in Taitung County were assessed as highly landslide-prone area after the heavy precipitation of Typhoon Morakot (over 2700 mm of rainfall within only 5 days). In this areas, several deep-seated landslides were identified according to geomorphological features seen in the 1-m resolution LiDAR DEM and InSAR preliminary results. In Tienchih area, a catastrophic 240-mm displacement sized 6.7 ha was recorded by a continuous GPS site, TENC, in 2016 after a heavy rainfall occurred on June 2. The correlation in the temporal variation of continuous GPS displacement time series and rainfall suggests that the movement is possibly related to gravitational load overlying water-saturated sediments. In addition, the average annual displacement rate of this downslope movement was measured at 20-40 mm/yr using the recently developed temporarily coherence points InSAR (TCPInSAR) technique based on ALOS/PALSAR imagery collected between 2007 and 2011. Apart therefrom, the high-angle thrust with highly fractured metamorphosed sandstone on the hanging wall; and the river incision and lateral river bank erosion are considered as the triggering factor of this catastrophic landslide. Similar triggering factors are responsible for Yakou landslide, where the 2018 landslide event exposed an outstanding cross section of the predisposing geological setting characterized by a tightly folded sequence of metamorphosed sandstone and slates. Spectacular gravitational deformation structures (i.e. kink folds and shear zones) are also found along this slope testifying a long-term displacement history and shedding light on possible kinematic mechanisms controlling its evolution. Through field data, remote sensing techniques and optical methods (i.e. digital image correlation, 3D LiDAR point cloud comparison), we compared the two landslide sites unravelling different deformation styles and identifying nested sectors possibly evolving to collapse. Our primary results demonstrate that valley erosion and deep-seated gravitational creep are significant to the deformation of slate, indicating a block movement with shear concentration at the basal sliding surface with a mainly rotational-translational movement in Tienchih and a translational failure mechanism in Yakou.
Seismic images provided by standard Reverse Time Migration are usually contaminated by artefacts associated with the migration of multiples. Multiples can corrupt seismic images by producing both false negatives, i.e. by destructively interfering with primaries, and false positives, i.e. by focusing energy at unphysical interfaces. Free-surface multiples particularly affect seismic images resulting from marine data, while internal multiples strongly contaminate both land and marine data. Multiple prediction / primary synthesis methods are usually designed to operate on point source gathers, and can therefore be computationally demanding when large problems, involving hundreds of gathers, are considered. In this contribution, a new scheme for fully data-driven retrieval of primary responses of plane-wave sources is presented. The proposed scheme, based on convolutions and cross-correlations of the reflection response with itself, extends a recently devised Marchenko point-sources primary retrieval method for to plane-wave source data. As a result, the presented algorithm allows fully data-driven synthesis of primary reflections associated with plane-wave source data. Once primary plane-wave responses are estimated, they are used for multiple-free imaging via standard reverse time migration. Numerical tests of increasing complexity demonstrate the potential of the proposed algorithm to produce multiple-free images only involving the migration of few datasets. The plane-wave source primary synthesis algorithm discussed in this contribution could then be used as an initial and unexpensive processing step, potentially guiding more expensive target imaging techniques. Moreover, the method could be applied to large 3D problems for which standard methods are prohibitively expensive from a computational point of view.
In most university geosciences curricula, structural geology and tectonics (SGT) form a core part of teaching. While only a small percentage of Earth science graduates will become structural geologists, many will someday use structural concepts and techniques to solve problems in fields such as nuclear waste storage, the geology of growing urban environments, geohazards, unconventional reservoirs, geothermal energy, CO2 sequestration, energy storage and more. A basic understanding of structural geology is thus part of a critical knowledge foundation in Earth sciences and many related disciplines. In addition, new tools and data are becoming available at a rapid pace, and enable more integrated, multi-dimensional assessments of the geosphere and our societal interfaces with it. All of this provides new opportunities and challenges for STG courses.In April 2019, a pre-EGU two-day workshop (TeachSGT21) was organized during which strengths and weaknesses of, and threats to current SGT curricula were analyzed. Participants of the workshop covered 11 European and 2 overseas countries, and came from academia as well as industry. On the basis of the workshop, we now outline educational demands from industry and research and discuss the role and significance of field training. Further, we review initiatives that use innovative tools and techniques in teaching. While not claiming to represent all aspects of modern SGT teaching, we expect that our observations can stimulate reflection on degrees and approach and may help making choices in curriculum renewal.
