Natural and anthropogenic disasters can cause significant damage to urban infrastructure, landscape, and loss of human life. Satellite based remote sensing plays a key role in rapid damage assessment, post-disaster reconnaissance and recovery. In this study, we aim to assess the performance of Sentinel-1 and Sentinel-2 data for building damage assessment in Kyiv, the capital city of Ukraine, due to the ongoing war with Russia. For damage assessment, we employ a simple and robust SAR log ratio of intensity for the Sentinel-1, and a texture analysis for the Sentinel-2. To suppress changes from other features and landcover types not related to urban areas, we construct a mask of the built-up area using the OpenStreetMap building footprints and World Settlement Footprint (WSF), respectively. As it is difficult to get ground truth data in the ongoing war zone, a qualitative accuracy assessment with the very high-resolution optical images and a quantitative assessment with the United Nations Satellite Center (UNOSAT) damage assessment map was conducted. The results indicated that the damaged buildings are mainly concentrated in the northwestern part of the study area, wherein Irpin, and the neighboring towns of Bucha and Hostomel are located. The detected building damages show a good match with the reference WorldView images. Compared with the damage assessment map by UNOSAT, 58% of the damaged buildings were correctly classified. The results of this study highlight the potential offered by publicly available medium resolution satellite imagery for rapid mapping damage to provide initial reference data immediately after a disaster.
A global geospatial liquefaction model (GGLM‐2017) was previously developed (Zhu et al., 2017) using logistic regression (LR) and is currently used by the U.S. Geological Survey as the preferred liquefaction model to map liquefaction probability immediately after the occurrence of earthquake events. This research proposes an ensemble modeling approach to improve the performance of the GGLM‐2017 for geospatial liquefaction modeling of the 2023 Türkiye earthquakes using an updated inventory of liquefaction occurrence locations in Europe (the OpenLIQ database, which includes prior events occurring in Türkiye) and a new inventory from the 2023 Türkiye earthquakes (gathered from multiple sources). Using the same geospatial proxies for soil saturation, soil density, and earthquake loading, and the same non‐liquefaction sampling strategy used to develop GGLM‐2017, the proposed ensemble method is validated on the data of the 2023 Türkiye earthquakes by integrating four models, including global (GGLM‐2017), continental (LR model trained on eight European events), regional (LR model trained on three historical events in Türkiye), and event‐specific (LR model trained on partially available data from the 2023 Türkiye earthquakes) models. The inventory from the 2023 Türkiye earthquakes is split into two batches, in which the first batch (163 liquefaction occurrences) resulted from the preliminary reconnaissance and is used for training the event‐specific model, and the second batch (284 liquefaction occurrences) resulted from a more complete reconnaissance (which was made available later) and is used for validating all models. The rationale for using the first batch for training the event‐specific model is to exploit the information as they become available to optimize the performance of global model in liquefaction prediction. The final ensemble probability assignment is done by averaging the probabilities derived by the four individual models, and a 50% threshold is used for classification accuracy evaluations. Comparative analysis of the ensemble model’s performance with the GGLM‐2017 showed improved predictive accuracy, resulting in higher liquefaction detection for the specific event under study (the 2023 Türkiye earthquakes). The ensemble model also provides an estimate of model uncertainty.
Soil liquefaction often occurs as a secondary hazard during earthquakes and can lead to significant structural and infrastructure damage. Liquefaction is most often documented through field reconnaissance and recorded as point locations. Complete liquefaction inventories across the impacted area are rare but valuable for developing empirical liquefaction prediction models. Remote sensing analysis can be used to rapidly produce the full spatial extent of liquefaction ejecta after an event to inform and supplement field investigations. Visually labeling liquefaction ejecta from remotely sensed imagery is time-consuming and prone to human error and inconsistency. This study uses a partially labeled liquefaction inventory created from visual annotations by experts and proposes a pixel-based approach to detecting unlabeled liquefaction using advanced machine learning and image processing techniques, and to generating an augmented inventory of liquefaction ejecta with high spatial completeness. The proposed methodology is applied to aerial imagery taken from the 2011 Christchurch earthquake and considers the available partial liquefaction labels as high-certainty liquefaction features. This study consists of two specific comparative analyses. (1) To tackle the limited availability of labeled data and their spatial incompleteness, a semi-supervised self-training classification via Linear Discriminant Analysis is presented, and the performance of the semi-supervised learning approach is compared with supervised learning classification. (2) A post-event aerial image with RGB (red-green-blue) channels is used to extract color transformation bands, statistical indices, texture components, and dimensionality reduction outputs, and performances of the classification model with different combinations of selected features from these four groups are compared. Building footprints are also used as the only non-imagery geospatial information to improve classification accuracy by masking out building roofs from the classification process. To prepare the multi-class labeled data, regions of interest (ROIs) were drawn to collect samples of seven land cover and land use classes. The labeled samples of liquefaction were also clustered into two groups (dark and light) using the Fuzzy C-Means clustering algorithm to split the liquefaction pixels into two classes. A comparison of the generated maps with fully and manually labeled liquefaction data showed that the proposed semi-supervised method performs best when selected high-ranked features of the two groups of statistical indices (gradient weight and sum of the band squares) and dimensionality reduction outputs (first and second principal components) are used. It also outperforms supervised learning and can better augment the liquefaction labels across the image in terms of spatial completeness.