The EarthExplorer 10 mission Harmony by the European Space Agency ESA, scheduled for launch around 2029–2030, consists of two passive C-band synthetic-aperture-radar companion satellites flying in a flexible constellation with one Sentinel-1 radar satellite as an illuminator. Sentinel-1 will serve as transmitter and receiver of radar waves, and the two Harmonys will serve as bistatic receivers without the ability to transmit. During the first and last year of the 5-year mission, the two Harmony satellites will fly in a cross-track interferometric constellation, such as that known from TanDEM-X, about 350 km ahead or behind the assigned Sentinel-1. This constellation will provide 12-day repeat DEMs, among other regions, over most land-ice and permafrost areas. These repeat DEMs will be complemented by synchronous lateral terrain displacements from the well-established offset tracking method. In between the cross-track interferometry phases, one of the Harmony satellites will be moved to the opposite side of the Sentinel-1 to form a symmetric bistatic “stereo” constellation with ±~350 km along-track baseline. In this phase, the mission will provide opportunity for radar interferometry along three lines of sight, or up to six when combining ascending and descending acquisitions, enabling the measurement of three-dimensional surface motion, for instance sub- and emergence components of ice flow, or three-dimensional deformation of permafrost surfaces or slow landslides. Such measurements would, for the first time, be available for large areas and are anticipated to provide a number of novel insights into the dynamics and mass balance of a range of mass movement processes.

We present a workflow for investigating large, slow-moving landslides which combines the synthetic aperture radar (SAR) technique, GIS post-processing, and airborne laser scanning (ALS), and apply it to Fels landslide in Alaska, US. First, we exploit a speckle tracking (ST) approach to derive the easting, northing, and vertical components of the displacement vectors across the rock slope for two five-year windows, 2010–2015 and 2015–2020. Then, we perform post-processing in a GIS environment to derive displacement magnitude, trend, and plunge maps of the landslide area. Finally, we compare the ST-derived displacement data with structural lineament maps and profiles extracted from the ALS dataset. Relying on remotely sensed data, we estimate that the thickness of the slide mass is more than 100 m and displacements occur through a combination of slumping at the toe and planar sliding in the central and upper slope. Our approach provides information and interpretations that can assist in optimizing and planning fieldwork activities and site investigations at landslides in remote locations.

The timing of snowmelt is an important turning point in the seasonal cycle of small Arctic catchments. The TerraSAR-X (TSX) satellite mission is a synthetic aperture radar system (SAR) with high potential to measure the high spatiotemporal variability of snow cover extent (SCE) and fractional snow cover (FSC) on the small catchment scale. We investigate the performance of multi-polarized and multi-pass TSX X-Band SAR data in monitoring SCE and FSC in small Arctic tundra catchments of Qikiqtaruk (Herschel Island) off the Yukon Coast in the Western Canadian Arctic. We applied a threshold based segmentation on ratio images between TSX images with wet snow and a dry snow reference, and tested the performance of two different thresholds. We quantitatively compared TSX- and Landsat 8-derived SCE maps using confusion matrices and analyzed the spatiotemporal dynamics of snowmelt from 2015 to 2017 using TSX, Landsat 8 and in situ time lapse data. Our data showed that the quality of SCE maps from TSX X-Band data is strongly influenced by polarization and to a lesser degree by incidence angle. VH polarized TSX data performed best in deriving SCE when compared to Landsat 8. TSX derived SCE maps from VH polarization detected late lying snow patches that were not detected by Landsat 8. Results of a local assessment of TSX FSC against the in situ data showed that TSX FSC accurately captured the temporal dynamics of different snow melt regimes that were related to topographic characteristics of the studied catchments. Both in situ and TSX FSC showed a longer snowmelt period in a catchment with higher contributions of steep valleys and a shorter snowmelt period in a catchment with higher contributions of upland terrain. Landsat 8 had fundamental data gaps during the snowmelt period in all 3 years due to cloud cover. The results also revealed that by choosing a positive threshold of 1 dB, detection of ice layers due to diurnal temperature variations resulted in a more accurate estimation of snow cover than a negative threshold that detects wet snow alone. We find that TSX X-Band data in VH polarization performs at a comparable quality to Landsat 8 in deriving SCE maps when a positive threshold is used. We conclude that TSX data polarization can be used to accurately monitor snowmelt events at high temporal and spatial resolution, overcoming limitations of Landsat 8, which due to cloud related data gaps generally only indicated the onset and end of snowmelt.

The characterization of landslides located in remote areas poses significant challenges due to the costs of reaching the sites and the lack of reliable subsurface data to constrain geological interpretations. In this paper, the advantages of combining field and remote sensing techniques to investigate the deformation and stability of rock slopes are demonstrated. The characterization of the Fels landslide, a large, slowly deforming rock slope in central Alaska, is described. Historical aerial imagery is used to highlight the relationship between glacier retreat and developing instability. Airborne laser scanning (ALS) and Structure-from-Motion (SfM) datasets are used to investigate the structural geological setting of the landslide, revealing a good agreement between structural discontinuities at the outcrop and slope scales. The magnitude, plunge, and direction of slope surface displacements and their changes over time are studied using a multi-temporal synthetic aperture radar speckle-tracking (SAR ST) dataset. The analyses show an increase in displacement rates (i.e., an acceleration of the movement) between 2010 and 2020. Significant spatial variations of displacement direction and plunge are noted and correlated with the morphology of the failure surface reconstructed using the vector inclination method (VIM). In particular, steeper displacement vectors were reconstructed in the upper slope, compared to the central part, thus suggesting a change in basal surface morphology, which is largely controlled by rock mass foliation. Through this analytical approach, the Fels landslide is shown to be a slow-moving, compound rockslide, the displacement of which is controlled by structural geological features and promoted by glacier retreat.

We propose a simple modification to standard ResNet architectures--L2 normalization over feature space--that substantially improves out-of-distribution (OoD) performance on the previously proposed Deep Deterministic Uncertainty (DDU) benchmark. We show that this change also induces early Neural Collapse (NC), an effect linked to better OoD performance. Our method achieves comparable or superior OoD detection scores and classification accuracy in a small fraction of the training time of the benchmark. Additionally, it substantially improves worst case OoD performance over multiple, randomly initialized models. Though we do not suggest that NC is the sole mechanism or a comprehensive explanation for OoD behaviour in deep neural networks (DNN), we believe NC's simple mathematical and geometric structure can provide a framework for analysis of this complex phenomenon in future work.

We demonstrate that L2 normalization over feature space can produce capable performance for Out-of-Distribution (OoD) detection for some models and datasets. Although it does not demonstrate outright state-of-the-art performance, this method is notable for its extreme simplicity: it requires only two addition lines of code, and does not need specialized loss functions, image augmentations, outlier exposure or extra parameter tuning. We also observe that training may be more efficient for some datasets and architectures. Notably, only 60 epochs with ResNet18 on CIFAR10 (or 100 epochs with ResNet50) can produce performance within two percentage points (AUROC) of several state-of-the-art methods for some near and far OoD datasets. We provide theoretical and empirical support for this method, and demonstrate viability across five architectures and three In-Distribution (ID) datasets.