Enhancement of Detecting Permanent Water and Temporary Water in Flood Disasters by Fusing Sentinel-1 and Sentinel-2 Imagery Using Deep Learning Algorithms: Demonstration of Sen1Floods11 Benchmark Datasets

Remote Sensing, Год журнала: 2021, Номер 13(11), С. 2220 - 2220

Опубликована: Июнь 5, 2021

Identifying permanent water and temporary in flood disasters efficiently has mainly relied on change detection method from multi-temporal remote sensing imageries, but estimating the type disaster events only post-flood imageries still remains challenging. Research progress recent years demonstrated excellent potential of multi-source data fusion deep learning algorithms improving detection, while this field been studied initially due to lack large-scale labelled images events. Here, we present new a driven inundation mapping approach by leveraging publicly available Sen1Flood11 dataset consisting roughly 4831 Sentinel-1 SAR Sentinel-2 optical imagery gathered worldwide years. Specifically, proposed an automatic segmentation for surface water, identification, all tasks share same convolutional neural network architecture. We utilize focal loss deal with class (water/non-water) imbalance problem. Thorough ablation experiments analysis confirmed effectiveness various designs. In comparison experiments, paper is superior other classical models. Our model achieves mean Intersection over Union (mIoU) 52.99%, (IoU) 52.30%, Overall Accuracy (OA) 92.81% test set. On Bolivia set, our also very high mIoU (47.88%), IoU (76.74%), OA (95.59%) shows good generalization ability.

Язык: Английский

---

Accounting for uncertainties in compound flood hazard assessment: The value of data assimilation

Coastal Engineering, Год журнала: 2021, Номер 171, С. 104057 - 104057

Опубликована: Ноя. 23, 2021

Язык: Английский

---

An exploratory study of Sentinel-1 SAR for rapid urban flood mapping on Google Earth Engine

International Journal of Applied Earth Observation and Geoinformation, Год журнала: 2022, Номер 113, С. 103002 - 103002

Опубликована: Сен. 1, 2022

Real-time, near-real-time, and accurate flood extent information is critical for emergency response during disaster events such as floods. Accurate extents are management relief efforts. Despite multiple efforts, there still many challenges in automated processing of Sentinel-1 SAR to generate reliable inundation maps. The major advantage compared optical imagery its data collection capability despite any weather conditions even thick cloud situation. Currently, a knowledge gap employing different polarization combinations flooding research. First, ten the two original VH VV polarizations designed rapid urban mapping. To examine significant potentials mapping, four mapping methods namely threshold, change detection, unsupervised supervised classification, combination with zero-depth method, used map extents. Among combinations, multiplication, squared addition, addition have resulted good results In depth estimation approach has been address overestimation flooded areas. all methods, deduction overestimated areas using threshold zero improved overall accuracy on average 7 % methods. show that implemented Google Earth Engine identify but detection method requires little user involvement, this can be applied new study without estimating affected Whereas classification will need more user’s involvement collect sample points. consistently performed well All analysis done platform, strategy environment. finding enhance local governments federal agencies assessment disasters making decisions.

Язык: Английский

---

Burnt-Net: Wildfire burned area mapping with single post-fire Sentinel-2 data and deep learning morphological neural network

Ecological Indicators, Год журнала: 2022, Номер 140, С. 108999 - 108999

Опубликована: Май 23, 2022

Accurate and timely mapping of wildfire burned areas is crucial for post-fire management, planning, next subsequent actions. The monitoring the area by traditional common methods are time-consuming challenging while vital to propose an advanced detection framework achieving reliable results. To this end, study proposed a novel End-to-End based on deep learning Sentinel-2 imagery. known as Burnt-Net combines quadratic morphological operators standard convolution layers. multi-patch multi-level residual (MP-MRM) blocks main part decoder encoder uses transpose evaluate efficiency latest wildfires over different countries was collected then, model trained evaluated them. Furthermore, most learning-based implemented comparing result Burnt-Net. results show robust in provides mean accuracy more than 97% overall (OA). fast can provide map near real-time.

Язык: Английский

---

Integrating remote sensing and social sensing for flood mapping

Remote Sensing Applications Society and Environment, Год журнала: 2022, Номер 25, С. 100697 - 100697

Опубликована: Янв. 1, 2022

Flood events cause substantial damage to infrastructure and disrupt livelihoods. Timely monitoring of flood extent helps authorities identify severe impacts plan relief operations. Remote sensing through satellite imagery is an effective method flooded areas. However, critical contextual information about the severity structural or urgent needs affected population cannot be obtained from remote alone. On other hand, social microblogging sites can potentially provide useful directly eyewitnesses people. Therefore, this paper explores integration data derive informed maps. For purpose, we employ state-of-the-art deep learning methods process heterogeneous four case-study areas, including two urban regions Somalia India coastal Italy The Bahamas. side, observe that models perform generally better than Otsu in water prediction. example, for highly areas India, U-Net achieves F1-scores (0.471 0.310, respectively) (0.297 0.251, respectively). Similarly, FCN yields a F1-score (0.128) (0.083) while on par Bahamas (0.102 0.105, Then, add layers representing relevant tweet text images posted highlight different ways these sources complement each other. Our extensive analyses reveal several valuable insights. In particular, three types signals: (i) confirmatory signals both sources, which puts greater confidence specific region flooded, (ii) complementary requests, disaster impact reports situational information, (iii) novel when do not overlap unique information.

Язык: Английский

---

Perspective on uncertainty quantification and reduction in compound flood modeling and forecasting

iScience, Год журнала: 2022, Номер 25(10), С. 105201 - 105201

Опубликована: Сен. 23, 2022

This perspective discusses the importance of characterizing, quantifying, and accounting for various sources uncertainties involved in different layers hydrometeorological hydrodynamic model simulations as well their complex interactions cascading effects (e.g., uncertainty propagation) forecasting compound flooding (CF). Over past few decades, CF has come to attention across globe this natural hazard results from a combination either concurrent or successive flood drivers with larger economic, societal, environmental impacts than those isolated drivers. A warming climate increased urbanization flood-prone areas are expected contribute an escalation risk near future. Recent advances remote sensing data science can provide wide range possibilities account reduce predictive uncertainties; hence improving predictability events, enabling risk-informed decision-making, ensuring sustainable governance.

Язык: Английский

---

Sentinel-1-Based Water and Flood Mapping: Benchmarking Convolutional Neural Networks Against an Operational Rule-Based Processing Chain

IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Год журнала: 2022, Номер 15, С. 2023 - 2036

Опубликована: Янв. 1, 2022

In this study, the effectiveness of several convolutional neural network architectures (AlbuNet-34/FCN/DeepLabV3+/U-Net/U-Net++) for water and flood mapping using Sentinel-1 amplitude data is compared to an operational rule-based processor (S-1FS). This comparison made a globally distributed dataset scenes corresponding ground truth masks derived from Sentinel-2 evaluate performance classifiers on global scale in various environmental conditions. The impact single versus dual-polarized input segmentation capabilities AlbuNet-34 evaluated. weighted cross entropy loss combined with Lovász augmentation methods are investigated. Furthermore, concept atrous spatial pyramid pooling used DeepLabV3+ multiscale feature fusion inherent U-Net++ assessed. Finally, generalization capacity tested realistic scenario by additional two events Sen1Floods11 dataset. model trained dual polarized outperforms S-1FS significantly increases intersection over union (IoU) score 5%. Using combination IoU another 2%. Geometric degrades while radiometric leads better testing results. FCN/DeepLabV3+/U-Net/U-Net++ perform not different AlbuNet-34. Models showing no distinct inundation very well extent during events, reaching scores 0.96 0.94, respectively, comparatively

Язык: Английский

---

Flood modeling and fluvial dynamics: A scoping review on the role of sediment transport

Earth-Science Reviews, Год журнала: 2024, Номер 253, С. 104775 - 104775

Опубликована: Апрель 15, 2024

Язык: Английский

---

Mapping Compound Flooding Risks for Urban Resilience in Coastal Zones: A Comprehensive Methodological Review

Remote Sensing, Год журнала: 2024, Номер 16(2), С. 350 - 350

Опубликована: Янв. 16, 2024

Coastal regions, increasingly threatened by floods due to climate-change-driven extreme weather, lack a comprehensive study that integrates coastal and riverine flood dynamics. In response this research gap, we conducted bibliometric analysis thorough visualization mapping of studies compound flooding risk in cities over the period 2014–2022, using VOSviewer CiteSpace analyze 407 publications Web Science Core Collection database. The analytical results reveal two persistent topics: way explore return periods or joint probabilities drivers statistical modeling, quantification with different through numerical simulation. This article examines critical causes flooding, outlines principal methodologies, details each method’s features, compares their strengths, limitations, uncertainties. paper advocates for an integrated approach encompassing climate change, ocean–land systems, topography, human activity, land use, hazard chains enhance our understanding mechanisms. includes adopting Earth system modeling framework holistic coupling components, merging process-based data-driven models, enhancing model grid resolution, refining dynamical frameworks, comparing complex physical models more straightforward methods, exploring advanced data assimilation, machine learning, quasi-real-time forecasting researchers emergency responders.

Язык: Английский

---

Forecasting of compound ocean-fluvial floods using machine learning

Journal of Environmental Management, Год журнала: 2024, Номер 364, С. 121295 - 121295

Опубликована: Июнь 13, 2024

Язык: Английский

---