ALBERT

Description: Floods can have devastating consequences on people, infrastructure, and the ecosystem. Satellite imagery has proven to be an efficient instrument in supporting disaster management authorities during flood events. In contrast to optical remote sensing technology, Synthetic Aperture Radar (SAR) can penetrate clouds, and authorities can use SAR images even during cloudy circumstances. A challenge with SAR is the accurate classification and segmentation of flooded areas from SAR imagery. Recent advancements in deep learning algorithms have demonstrated the potential of deep learning for image segmentation demonstrated. Our research adopted deep learning algorithms to classify and segment flooded areas in SAR imagery. We used UNet and Feature Pyramid Network (FPN), both based on EfficientNet-B7 implementation, to detect flooded areas in SAR imaginary of Nebraska, North Alabama, Bangladesh, Red River North, and Florence. We evaluated both deep learning methods' predictive accuracy and will present the evaluation results at the conference. In the next step of our research, we develop an XAI toolbox to support the interpretation of detected flooded areas and algorithmic decisions of the deep learning methods through interactive visualizations.

Description: Flood monitoring in arid regions is challenging using Synthetic Aperture Radar (SAR) due to the similar backscatter of water and dry sand in surrounding areas. Since textural information is abundant in SAR images, this study investigates the added value of texture in SAR-based flood detection by providing it as auxiliary information for flood delineation. Results show that texture enhanced SAR images in VH polarization substantially underpredicts the flooded area, so adding texture does not improve the classification accuracy. However, using both polarization (VV and VH) produce ~26% higher overall accuracy for flood detection in arid regions.

Description: Determining outline, volume and effusion rate during an effusive volcanic eruption is crucial as it is a major controlling factor of the lava flow lengths, the prospective duration and hence the associated hazards. We present for the first time a multi-sensor thermal-and-topographic satellite data analysis for estimating lava effusion rates and volume. At the 2021 lava field of Cumbre Vieja, La Palma, we combine VIIRS + MODIS thermal data-based effusion rate estimates with DSMs analysis derived from optical tri-stereo Pléiades and TanDEM-X bi-static SAR-data. This multi-sensor-approach allows to overcome limitations of single-methodology-studies and to achieve both, high-frequent observation of the relative short-term effusion rate trends and precise total volume estimates. We find a final subaerial-lava volume of 212×106±13×106m3 with a MOR of 28.8 ± 1.4 m3/s. We identify an initially sharp eruption-rate-peak, followed by a gradually decreasing trend, interrupted by two short-lived-peaks in mid/end November. High eruption rate accompanied by weak seismicity was observed during the early stages of the eruption, while during later stage the lava effusion trend coincides with seismicity. This article demonstrates the geophysical monitoring of eruption rate fluctuations, that allows to speculate about changes of an underlying pathway during the 2021 Cumbre Vieja eruption.

Description: Up-to-date information on the outline, volume and effusion rate during an effusive volcanic eruption is crucial as it is a major controlling factor of the lava flow lengths, the prospective duration and hence the associated hazards. Here, we present a multi-sensor approach combining thermal and topographic satellite data for estimating lava effusion rates and volume. We investigated the large 2021 volcanic eruption at Cumbre Vieja, La Palma and we combined MODIS and VIIRS thermal data-based effusion rate estimates with digital surface model (DSM) analysis derived from bi-static TanDEM-X synthetic aperture radar (SAR) and optical-tri-stereo Pléiades data. Advantage of this multi-sensor approach is to achieve both, high-frequent observation of the relative short-term effusion rate trends and precise total volume estimates. Our results show a final subaerial lava volume of 212x106 ± 13x106 m³ with a mean output rate (MOR) of 28.8±1.4 m³/s. The 2021 Cumbre Vieja eruption was characterized by an initially sharp eruption rate peak, followed by a gradually decreasing trend, and interrupted by two short-lived peaks in mid and in end November. Comparison with independent seismic observations show a high eruption rate accompanied by weak seismicity during the early stages of the eruption, while during later stage the lava effusion trend coincides with seismicity. We will demonstrate that the geophysical monitoring of eruption rate fluctuations allows to speculate about changes of an underlying pathway during the 2021 Cumbre Vieja eruption.

Description: This study introduces the S1S2-Water dataset—a global reference dataset for training, validation, and testing of convolutional neural networks (CNNs) for semantic segmentation of surface water bodies in publicly available Sentinel-1 and Sentinel-2 satellite images. The dataset consists of 65 triplets of Sentinel-1 and Sentinel-2 images with quality-checked binary water mask. Samples are drawn globally on the basis of the Sentinel-2 tile-grid (100 km × 100 km) under consideration of predominant landcover and availability of water bodies. Each sample is complemented with metadata and digital elevation model (DEM) raster from the Copernicus DEM. On the basis of this dataset, we carry out performance evaluation of CNN architectures to segment surface water bodies from Sentinel-1 and Sentinel-2 images. We specifically evaluate the influence of image bands, elevation features (slope) and data augmentation on the segmentation performance and identify best-performing baseline-models. The model for Sentinel-1 achieves an Intersection over Union (IoU) of 0.845, Precision of 0.932, and Recall of 0.896 on the test data. For Sentinel-2 the best model produces an IoU of 0.965, Precision of 0.989, and Recall of 0.951, respectively. We also evaluate the performance impact when a model is trained on permanent water data and applied to independent test scenes of floods.

Description: During flood events near real-time, synthetic aperture radar (SAR) satellite imagery has proven to be an efficient management tool for disaster management authorities. However, one of the challenges is accurate classification and segmentation of flooded water. A common method of SAR-based flood mapping is binary segmentation by thresholding, but this method is limited due to the effects of backscatter, geographical area, and surface characterstics. Recent advancements in deep learning algorithms for image segmentation have demonstrated excellent potential for improving flood detection. In this paper, we present a deep learning approach with a nested UNet architecture based on a backbone of EfficientNet-B7 by leveraging a publicly available Sentinel‑1 dataset provided jointly by NASA and the IEEE GRSS Committee. The performance of the nested UNet model was compared with several other UNet-based convolutional neural network architectures. The models were trained on flood events from Nebraska and North Alabama in the USA, Bangladesh, and Florence, Italy. Finally, the generalization capacity of the trained nested UNet model was compared to the other architectures by testing on Sentinel‑1 data from flood events of varied geographical regions such as Spain, India, and Vietnam. The impact of using different polarization band combinations of input data on the segmentation capabilities of the nested UNet and other models is also evaluated using Shapley scores. The results of these experiments show that the UNet model architectures perform comparably to the UNet++ with EfficientNet-B7 backbone for both the NASA dataset as well as the other test cases. Therefore, it can be inferred that these models can be trained on certain flood events provided in the dataset and used for flood detection in other geographical areas, thus proving the transferability of these models. However, the effect of polarization still varies across different test cases from around the world in terms of performance; the model trained with the combinations of individual bands, VV and VH, and polarization ratios gives the best results.