The High Mountain Asia (HMA) continues to witness an increased frequency of glacial lake outburst floods (GLOFs), which is likely in response to continued global warming. In situ measurements to understand the triggers all across the region will remain inadequate given the vastness and lack of accessibility of the region. This work explores a data-driven logistic regression-based framework to evaluate potential GLOF triggers, such as the lake dam type, its surface area, aspect, distance, freeboard, slope, precipitation, and temperature. A comprehensive inventory of past GLOF events in the region has been compiled, with 25 events verified using pre-& post-event multispectral images acquired between 2016 and 2022. The logistic regression model is developed using samples of the positive class (lakes with confirmed GLOFs) and of the negative class (potentially dangerous lakes that have not experienced a GLOF event). The samples of the negative lake class were collected with resembling characteristics from the nearby areas of the positive class. We randomly keep 80% of the samples for training. The models performance is assessed using adjusted R2and the Akaike Information Criterion (AIC) on the test samples, which are 79% and 21.5, respectively. The classification accuracy is 90%, which is promising. In short, the proposed method is a useful tool to investigate risk of outburst flooding of glacial lakes.

A glacial lake outburst flood (GLOF) is typically a natural phenomenon caused by rapid discharge of water from a glacier, leading to a flood. The frequency of GLOFs has increased significantly in the northern areas of Pakistan, which demands identification and continuous monitoring of potentially dangerous glacial lakes. In this paper, an up-to-date inventory of glacial lakes in this region is presented. This inventory (HKH-PK-2020) has been prepared using high resolution PlanetScope imagery acquired in 2020 over northern Pakistan. It contains a total of 8808 lakes. We compare our database with the High Mountain Asia (HMA) glacial lakes inventory over northern Pakistan, prepared in 2018 using Landsat imagery. The new inventory contains 6537 more glacial lakes than the HMA inventory. Furthermore, we have prepared an annotated dataset containing 3525 images (of high resolution PlanetScope imagery over a selected number of lakes from the inventory). Each image comprises 4 bands, namely red, green, blue, and near infrared. The annotations are binary: lake or background. Finally, we have performed an ablation study with two encoder-decoder based convolutional neural networks (CNNs) trained on this dataset for pixel-based classification. Our results show an intersection over union (IoU) score of 72.81% for the lake class, which is a promising first result indicating a use of deep learning for automated inventory updates in future.

Glacial lake outburst floods (GLOFs) are a major threat to the local communities and important infrastructures in the high mountain regions. This paper focuses on the development of a benchmark dataset for glacial lakes classification in Sentinel 2 multi-spectral data and subsequent detection of glacial lakes prior to a glacial lake outburst flood (GLOF). Towards this end, we collected Sentinel 2 true color scenes of High-Mountain Asia (HMA) region using glacial lakes inventory of this region. It covers an area of 2080.12 km 2 with nearly 30,121 glacial lakes. After data collection, we retained 1200 cloud free true color images and manually generated their ground truth masks. The dataset covers lakes with different shapes, sizes and radiometric signatures. For detection of glacial lakes, we used an encoder-decoder based convolutional neural network (CNN). The model is trained on the labelled dataset of glacial lakes for semantic segmentation of true color images into two relevant classes: lake and no lake. The performance of the proposed model is evaluated using intersection over union (IoU) score. It classifies glacial lakes correctly with an IoU score of 79.90%, which is quite good as far as complexity of the problem is concerned.

Oil spills cause a significant threat to marine and coastal ecosystems. It is one of the major causes of water pollution. This research focuses on the use of deep learning for oil spills detection and classification. UNet is a convolutional neural network, originally proposed for biomedical image segmentation and modified for the discrimination of oil spills and look-alikes. The model is trained on a publicly available benchmark oil spill detection dataset of Sentinel-1 synthetic aperture radar (SAR) images. The images have been semantically segmented into multiple regions of interest such as sea surface, oil spills, look-alikes, ships and land. The proposed UNet-based model achieves intersection over union (IoU) value of 95.69% for sea surface, 60.85% for oil spills, 54.90% for look-alikes, 70.27% for ships and 96.79% for land class. The mean intersection over union (mIoU) value for all the classes is 75.70% which consitutes a nearly 10% increase compared to state of the art for this dataset.

This paper reports the first results of a comparative study of tropospheric delays retrieved by means of PSI processing of an interferometric stack of SAR images against those derived independently from a permanent GNSS network. The stack comprises 33 Cosmo-SkyMed stripmap images acquired in the summers between 2008-13 over the Matter Valley in the Swiss Alps. The long-term objective of the study is to explore whether GNSS-derived delays from existing networks (i.e., not deployed specifically for a test site) in Swiss Alpine regions can aid in tropospheric phase corrections in SAR data, or rather the phase corrections derived within the PSI processing being at a higher spatial resolution might be appropriate to build upon the GNSS products by improving their resolution.

Synthetic aperture radar (SAR) tomography with repeat-pass acquisitions generally requires a priori phase calibration of the interferometric data stack by compensating for the atmosphere-induced phase delay variations. These variations act as a disturbance in tomographic focusing. In mountainous regions, the mitigation of these disturbances is particularly challenging due to strong spatial variations of the local atmospheric conditions and propagation paths through the troposphere. In this paper, we assess a data-driven approach to estimate these phase variations under a regression-kriging framework. The vertical stratification of the troposphere is modeled functionally, while the impact of the spatial turbulence is considered in a stochastic sense. The methodology entails an initial persistent scatterer interferometry (PSI) analysis. The atmospheric phases isolated for the persistent scatterers (PS) within the PSI processing are considered as samples of the 3-D distribution of the phase delay variations over the scene. These atmospheric phases are regressed against the spatial coordinates in map geometry at PS locations. In turn, kriging predictions are obtained at each location along the elevation profile, where tomographic focusing is intended. A key point of this approach is that the requisite atmospheric corrections are incorporated within the tomographic focusing model. A case study has been performed on a data stack comprising 32 COSMO-SkyMed stripmap images acquired over the Matter Valley in the Swiss Alps, in the summers of 2008-2013. The results show locally improved deformation sampling with tomographic methods compared to the initial PSI solution, primarily due to the improved phase calibration. In general, this paper underscores the indispensability of height-dependent correction of atmospheric phases for SAR tomography.

Differential synthetic aperture radar (SAR) tomography allows separation of multiple coherent scatterers interfering in the same range-azimuth resolution cell as well as the estimation of the deformation parameters of each scatterer. In this way, the spatio-temporal tomographic inversion serves as a means to resolve the layover and simultaneously improve deformation sampling. Compared to metropolitan regions with several man-made structures, the prevalence of coherent scatterers in the villages of alpine regions is generally low, while at the same time layovers are widespread due to the ruggedness of the terrain. Moreover, the drastic height variations in the imaged scene necessitate height-dependent compensation of the atmospheric phase delay variations within the tomographic inversion. This paper addresses these concerns while performing experiments on an interferometric stack comprising 33 Cosmo-SkyMed strimap images acquired in the summers between 2008-13 over Matter Valley in the Swiss Alps. The results show improved deformation sampling along the layover-affected mountainside.

Differential SAR tomography is a means to resolve layover of temporally coherent scatterers while simultaneously estimating their elevation and average deformation. In alpine regions, drastic height variations result in frequent layovers which are rejected during typical persistent scatterer interferometric (PSI) analyses. In this paper, we explore the potential of tomographic techniques to improve deformation sampling in an alpine region of interest relative to a PSI-based deformation assessment. The mitigation of the atmospheric phase contributions, as required for both tomography and PSI, is often more involved in alpine regions due to strong spatial variations of the local atmospheric conditions and propagation paths through the troposphere. We assume a linear multivariate dependence of atmospheric phase on the spatial location and height of the scatterers, estimate it using universal/regression kriging and subsequently incorporate it within the tomographic focusing. Experiments are performed on an interferometric stack comprising of 32 Cosmo-SkyMed strimap images acquired in the summers of 2008-2013 over Mattervalley in the Swiss Alps.

An acknowledgement provided by the authors was lost during production for the above paper[1]. It is provided here as follows:

SAR tomography can be used as an add-on to persistent scatterer interferometry (PSI) to increase deformation sampling in urban areas by resolving the frequently occurring layovers that are by definition rejected in the PSI processing. This paper, while focusing on the case of a typical high-rise building in layover, quantitatively assesses the potential gain in deformation sampling achieved by the added use of an advanced SAR tomographic technique relative to a PSI approach. At the same time, the quantity of the detected scatterers is weighed against their quality, as assessed on the basis of root-mean-square (RMS) phase deviation between the measurements and the model fit. The quality of the scatterers is also compared with the quality of the persistent scatterers as identified with a PSI approach. The experiments are performed on an interferometric stack of 50 TerraSAR-X stripmap mode images.

Persistent scatterer interferometry (PSI) is in operational use for spaceborne synthetic aperture radar (SAR)-based deformation analysis. A limitation inherently associated with PSI is that, by definition, a persistent scatterer (PS) is a single dominant scatterer. Therefore, pixels containing signal contributions from multiple scatterers, as in the case of a layover, are typically rejected in the PSI processing, which in turn limits deformation retrieval. SAR tomography has the ability to resolve layovers. This paper investigates the added value that can be achieved by operationally combining SAR tomography with a PSI approach toward the objective of improving deformation sampling in layover-affected urban areas. Different tomographic phase models are implemented and compared as regards their suitability in resolving layovers. Single-look beamforming-based tomographic inversion and a generalized likelihood ratio test (GLRT)-based detection strategy are used to detect single and double scatterers. The quantity of the detected scatterers is weighed against their quality as defined in terms of the phase deviation between the single-look complex (SLC) measurements and the tomographic model fit. The gain in deformation sampling that can be derived with tomography relative to a PSI-based analysis is quantitatively assessed, and alongside the quality of the scatterers obtained with tomography is compared with the quality of the PSs identified with a PSI approach. The experiments are performed on an interferometric stack of 50 TerraSAR-X stripmap images. The results obtained show that, although there is a tradeoff between the quantity and the quality of the detected scatterers, the tested SAR tomography approach leads to an improvement in deformation sampling in layover-affected areas.

Persistent scatterer interferometry (PSI) assumes the presence of a single temporally coherent scatterer in a range-azimuth pixel. Multiple scatterers interfering in the same pixel, as for the case of a layover, are typically rejected. Conventional SAR tomography (3D SAR) is a means to separate the individual scatterers in layover. Advanced tomographic inversion approaches employing extended phase models additionally allow simultaneous retrieval of scatterer elevation and deformation parameters. In this way, SAR tomography can increase deformation sampling and thereby complement a PSI-based analysis. This paper investigates the use of tomography as an add-on to PSI for spatio-temporal inversion of single and double scatterers in urban areas. Results are provided on an interferometric stack of 50 stripmap TerraSAR-X images acquired over the city of Barcelona.

The development of Geographical Information Systems applications involving fusion of data from different space-borne imaging sensors inevitably requires a preliminary registration of the images. In case of Synthetic Aperture Radar (SAR) and optical sensors, the registration is particularly challenging due to the vast radiometric differences in the data. In this paper, we present a novel method to register SAR and optical images automatically. It provides an accurate registration despite the radiometric differences in the images. Moreover, this paper introduces a Monte Carlo formulation of the image registration problem.

A processor's performance and power consumption are tied; an increased performance demands more power, and vice versa. An optimal tradeoff can only be achieved by an improved prediction of the task execution times, prior to an efficient scheduling. Moreover, since the processor's soft error rate is a function of its operating voltage, it is also linked to the performance-power trade-off. The situation is further complicated for the case of multicore architectures where the tasks are to be mapped on separate cores (processing elements). This paper proposes a joint State-Space model to achieve improved task execution time estimation, leading to better scheduling for optimizing the trade-off, particularly in the context of multicore soft real-time systems. It does not assume any `a priori' knowledge about the task graph or its properties, and is independent of the underlying architecture. It learns the system dynamics over time. The state-space solution is formulated using a recursive implementation of the online Monte Carlo Method. Having obtained the estimates of the execution times, they are compensated for the soft error according to a given soft error rate. At the beginning of each scheduling interval, the low power EDF scheduling decision is carried out to execute the tasks. The proposed method (SEAL) achieves 29% better energy savings compared to state-of-the-art, while the deadline misses are under 7% without the loss of system failure probability. The results obtained clearly show the advantage in terms of energy savings.

This paper addresses the problem of execution time estimation for tasks in a software pipeline independent of the application structure or the underlying architecture. A regression model is developed to obtain the estimates from previously observed data. To improve the quality of the estimates execution times of predecessor task in a software pipeline is exploited. Since the Model order (number of past observations required to obtain optimal estimate) cannot be determined at design time and to circumvent this, we propose means to dynamically update the order and hence obtain a critical-fit model without resorting to analytical benchmarking or calibration runs. The estimation scheme comprises of two estimation methods, namely `Wiener-Hopf' and Order-recursive estimation. The selection of the estimation method is automatic and depends on the required quality of the estimate against a user selectable threshold. In order recursion, new model order is obtained in conjunction to estimates, so order recursion solve the system both for order and estimate simultaneously. We experimented on two multicore platforms using H.264 decoder, a control dominant, computationally demanding application. Results show that estimates obtained by our method are up to 39% better in case of the first task in the software pipeline. The estimate quality improves significantly for the task with predecessor(s) in pipeline and comparison shows up to 54% improvement in estimation results.

This paper addresses the problem of stochastic task execution time estimation agnostic to the process distributions. The proposed method is orthogonal to the application structure and underlying architecture. We build the time varying state space model of the task execution time. In the case of software pipelined tasks, to refine the estimate quality, the state-space is modeled as Multiple Input Single Output (MISO) system by taking into account the current execution time of the predecessor task. To obtain nearly Bayesian estimates, irrespective of the process distribution, the sequential Monte Carlo method is applied which form the recursive solution to reduce the overheads and comprises of time update and correction steps. We experimented on three different platforms, including multicore, using the time parallelized H.264 decoder: a control dominant computationally demanding application and AES encoder: a pure data flow application. Results show that estimates obtained by our method are superior in quality and are up to 68% better in comparison to others.

The paper presents the development of a ground based bistatic receiver using TerraSAR-X as a transmitter. The receiver subsystems like antennas, low-noise amplifiers, mixers, filters, synthesizers, etc. have been developed using low-cost monolithic devices in order to allow affordable deployment and at the same time offer final year students a challenging SAR engineering project. First raw data have been acquired on the Barcelona harbor area that has been focused producing geocoded images well matched with existing maps. A preliminary interferogram have been also produced.