Application of Machine Learning to Classification of Volcanic Deformation in Routinely Generated InSAR Data

Anantrasirichai, Nantheera; Biggs, Juliet; Albino, F.; Hill, Paul; Bull, David

doi:10.1029/2018jb015911

Cited by 154 publications

(159 citation statements)

References 56 publications

Supporting

Mentioning

128

Contrasting

Order By: Relevance

“…For the DCNN models, an advanced type of deep learning models, the features useful for robust classification are dug out from data, instead of being predefined. This idea is repeatedly verified in the computer vision community (Russakovsky et al, ) as well as in the geophysics community (Reichstein et al, ; J. Zhang et al, ; Anantrasirichai et al, ). There are several existing studies for DCNN‐based flooding mapping.…”

Section: Introductionmentioning

confidence: 78%

Coastal Inundation Mapping From Bitemporal and Dual‐Polarization SAR Imagery Based on Deep Convolutional Neural Networks

Liu

Zheng

2019

JGR Oceans

View full text Add to dashboard Cite

This study develops an effective and robust method to mine bitemporal and dual-polarization synthetic aperture radar (SAR) imagery information for coastal inundation mapping, based on deep convolutional neural networks. The specially tailored deep convolutional neural network-based SAR coastal flooding mapping network (SARCFMNet) leverages two modifications to improve the accuracy and robustness: the physics-aware input information design and the regularization. The proposed SARCFMNet is applied to the mapping and impact analysis of the coastal inundation caused by 2017 Hurricane Harvey near Houston, Texas, USA. Six pairs of Sentinel-1 SAR images are analyzed along with corresponding ground truth data from Copernicus Emergency Management Service Rapid Mapping products and land-cover types from Google Earth and OpenStreetMap. Flooded areas of 4,000 km 2 are extracted and analyzed. In an analyzed scene, 76% of the flooded area was agriculture area like pasture and cultivated crops field. A flooding map series shows the inundation shrinking rate is about 1% of the analyzed scene per day after the passage of Harvey. However, there was a delayed inundation in Glen Flora, Texas, after the heavy raining period. The average mapping accuracy and F1 score, that is, the harmonic mean of recall and precision, are 0.98 and 0.88, respectively. The impact of wind and cost-sensitive loss functions on the development of SARCFMNet is also discussed. This study demonstrates the proposed method can accurately map hurricane-induced inundation. The method can also be readily extended to other multitemporal SAR imagery classification applications. Plain Language SummaryCoastal flooding caused by hurricanes has a huge impact on the safety and properties of people in coastal areas. Coastal flooding mapping using synthetic aperture radar (SAR) data is a low-cost and wide-coverage means; moreover, SAR has day-and-night, all-weather observation abilities. High-accuracy and robust coastal flooding mapping from SAR imagery information will help the management to formulate more targeted responses and the researchers to better understand coastal flooding mechanisms and develop more precise forecasting models. Based on deep convolutional neural networks, a specially tailored SAR coastal flooding mapping network method is developed to mine bitemporal and dual-polarization SAR imagery information for coastal flooding mapping. The performance of the proposed SAR coastal flooding mapping network is demonstrated by mapping the coastal flooding near Houston, Texas, USA, after 2017 Hurricane Harvey. The study verifies that the mapping accuracy and robustness of the proposed method are better than those of the classic and widely applied deep convolutional neural network method and also shows the mapping products contribute to a better understanding of the geospatial and temporal characteristics of coastal flooding.

show abstract

Section: Introductionmentioning

confidence: 78%

Coastal Inundation Mapping From Bitemporal and Dual‐Polarization SAR Imagery Based on Deep Convolutional Neural Networks

Liu

Zheng

2019

JGR Oceans

View full text Add to dashboard Cite

show abstract

“…Deep convolutional neural networks (CNN) -a class of neural networks inspired by deeply complex hierarchical structure of neurons that connect in multiple layers via learnable filters [11] -is one of the feasible methods for automatically analysing global datasets. Our previous 'proof-of-concept' study demonstrated the ability of CNNs to detect rapidly deforming systems that generate multiple fringes in wrapped interferograms [12] but could not reliably distinguish between deformation signals and atmospheric artefacts in a small percentage of cases. Our approach is to use machine learning to interrogate the large dataset of wrapped interferograms and identify a subset of images to apply unwrapping algorithms and atmospheric corrections.…”

Section: Introductionmentioning

confidence: 94%

“…al. [12] used a dataset of >30,000 Sentinel-1 interferograms produced by the LICSAR system which covered ∼900 volcanoes globally, but only contains 42 interferograms that show deformation signals. The imbalance in training data can be mitigated by artificially subsampling or upsampling the training set.…”

Section: Introductionmentioning

confidence: 99%

A deep learning approach to detecting volcano deformation from satellite imagery using synthetic datasets

Anantrasirichai

Biggs

Albino

et al. 2019

Remote Sensing of Environment

Self Cite

120

View full text Add to dashboard Cite

Satellites enable widespread, regional or global surveillance of volcanoes and can provide the first indication of volcanic unrest or eruption. Here we consider Interferometric Synthetic Aperture Radar (InSAR), which can be employed to detect surface deformation with a strong statistical link to eruption. Recent developments in technology as well as improved computational power have resulted in unprecedented quantities of monitoring data, which can no longer be inspected manually. The ability of machine learning to automatically identify signals of interest in these large InSAR datasets has already been demonstrated, but data-driven techniques, such as convolutional neutral networks (CNN) require balanced training datasets of positive and negative signals to effectively differentiate between real deformation and noise. As only a small proportion of volcanoes are deforming and atmospheric noise is ubiquitous, the use of machine learning for detecting volcanic unrest is more challenging than many other applications. In this paper, we address this problem using synthetic interferograms to train the AlexNet CNN. The synthetic interferograms are composed of 3 parts: 1) deformation patterns based on a Monte Carlo selection of parameters for analytic forward models, 2) stratified atmospheric effects derived from weather models and 3) turbulent atmospheric effects based on statistical simulations of correlated noise. The AlexNet architecture trained with synthetic data outperforms that trained using real interferograms alone, based on classification accuracy and positive predictive value (PPV). However, the models used to generate the synthetic signals are a simplification of the natural processes, so we retrain the CNN with a combined dataset consisting of synthetic models and selected real examples, achieving a final PPV of 82%. Although applying atmospheric corrections to the entire dataset is computationally expensive, it is relatively simple to apply them to the small subset of positive results. This further improves the detection performance without a significant increase in computational burden (PPV of 100%). Thus, we demonstrate that training with synthetic examples can improve the ability of CNNs to detect volcano deformation in satellite images, and propose an efficient workflow for the development of automated systems.

show abstract

“…In volcanology, an early application of machine learning (in this case, pattern recognition) investigated preeruptive seismic data during 217 episodes of volcanic unrest around the world . To date, however, machine learning algorithms have seen the widest use not directly in eruption forecasting (see 10.1029/2018JB016974 Brancato et al (2016), however, for an early example), but rather in the related problems of detection and classification of seismic signals (Curilem et al, 2009;Esposito et al, 2006Esposito et al, , 2008Ibs-von Seht, 2008;Masotti et al, 2006;Scarpetta et al, 2005), thermal anomalies (from remote sensing data; Piscini & Lombardo, 2014), and satellite-measured ground deformation (Anantrasirichai et al, 2018). These studies demonstrate the importance of machine learning for "data discovery," which can feed directly into forecasts.…”

Section: Pattern Recognition: Quantitative Approachesmentioning

confidence: 99%

Partly Cloudy With a Chance of Lava Flows: Forecasting Volcanic Eruptions in the Twenty‐First Century

Poland

Anderson

2020

JGR Solid Earth

View full text Add to dashboard Cite

A primary goal of volcanology is forecasting hazardous eruptive activity. Despite much progress over the last century, however, volcanoes still erupt with no detected precursors, lives and livelihoods are lost to eruptive activity, and forecasting the onsets of eruptions remains fraught with uncertainty. Long-term forecasts are generally derived from the geological and historical records, from which recurrence intervals and styles of activity can be inferred, while shorter-term forecasts are derived from patterns in monitoring data. Information from geology and monitoring data can be evaluated and combined using statistical analysis, expert elicitation, and conceptual and or mathematical models. Integrative frameworks, such as event trees, combine this diversity of information to produce probabilistic forecasts that can inform the style and scale of the societal response to a potential future eruption. Several developments show promise to revolutionize the utility and accuracy of these forecasts. These include growth in the quantity and quality of multidisciplinary monitoring data, coupled with increases in computing power; machine learning algorithms, which will allow far better utilization of this growing volume of data; and new physiochemical volcano models and data assimilation algorithms, which take advantage of a wide range of monitoring data and realistic physics to better predict the evolution of a given physical state. Although eruption forecasts may never be as generally reliable as weather forecasts, and great caution must be exercised when attempting to predict highly complex volcanic behavior, these and other innovations-particularly when combined in integrative, fully probabilistic forecasting frameworks-should help volcanologists to better issue warnings of volcanic activity on societally relevant time frames.

show abstract

Application of Machine Learning to Classification of Volcanic Deformation in Routinely Generated InSAR Data

Cited by 154 publications

References 56 publications

Coastal Inundation Mapping From Bitemporal and Dual‐Polarization SAR Imagery Based on Deep Convolutional Neural Networks

Coastal Inundation Mapping From Bitemporal and Dual‐Polarization SAR Imagery Based on Deep Convolutional Neural Networks

A deep learning approach to detecting volcano deformation from satellite imagery using synthetic datasets

Partly Cloudy With a Chance of Lava Flows: Forecasting Volcanic Eruptions in the Twenty‐First Century

Contact Info

Product

Resources

About