Pattern Recognition: Applications in Geosciences, volume three in the Computational Geophysics series, highlights the advantages of pattern recognition and classification methods and examines potential pitfalls and failures that can arise. The handling of data is a typical problem of multivariate analysis in which pattern recognition techniques can offer a suitable key for the processing and extraction of useful information. This book investigates both supervised techniques, such as linear discrimination, and unsupervised techniques, including various concepts of clustering. All approaches are extensively discussed, including example applications and an introduction to available software packages.
This book comprises the formulation of the problem, the choice of suitable features and strategies, and a discussion of the results, thus making it an ideal reference for researchers looking to apply these techniques to their field of geophysics.
- Offers real-world examples of techniques for pattern recognition and handling multivariate data from a variety of geophysical fields
- Includes examples of applications and diagrams to enhance understanding of pattern recognition techniques
- Provides an introduction to relevant software packages, such as TREmOREC and KKAnalysis
Part I: Supervised Techniques 1. A Simple Example 2. Solving the XOR Problem 3. Support Vector Machine (SVM) 4. Hidden Markov Models (HMM)
Part II: Unsupervised Techniques 5. Metric 6. Clustering 7. Clustering Related to Irregular Shapes 8. Self-organizing Map (SOM)
Part III: Example Applications 9. Supervised 10. Unsupervised
Part IV: A Posteriori Analysis 11. Performance 12. Problems with Features 13. Test Issues 14. Generative Approaches 15. What is a Failure?
Part V: Available Software Packages
Appendix: Available Example Data Sets
Horst Langer studied Geology and Geophysics at the University of Stuttgart, where he graduated in 1982. He received his PhD with a seismological thesis at the same University in 1986. From 1983 to 1992 he worked there as a researcher and lecturer. His research activities focused on seismicity, seismotectonics, strong ground motion simulation and volcano seismology. From 1999 on he worked at the "Sistema Poseidon as a researcher in the field of seismology, passing to the Istituto Nazionale di Geofisica e Vulcanologia - Osservatorio Etneo in 2001, where he now holds a position as senior researcher. His current research covers a wide field of seismological topics, such as strong ground motion problems, seismicity and seismotectonics, the identification of the three-dimensional velocity structure, the assessment of significance and stability of seismic inversion.
From 2001 to present, Dr. Falsaperla has been a Senior Researcher at Istituto Nazionale di Geofisica e Vulcanologia (INGV), sezione di Catania, Osservatorio Etneo, Italy. She has been Co-chair of the World Organization of Volcano Observatories (WOVO.org) since October 2011. From 1984 to 2000, she was a researcher at the Istituto Internazionale di Vulcanologia of Catania, department of Consiglio Nazionale delle Ricerche (CNR). She has 31 years' experience in volcano seismology and in seismic monitoring and surveillance activities of southern Italy and has been engaged as an expert volcano seismologist in volcano and seismic crises at Mt. Etna, Stromboli, and Vulcano from 1990s on. Her broad interests lie in the geology and geophysics of active volcanic systems and she is the author of several papers on the application of artificial neural networks and multivariate statistics to seismic data - recorded at Stromboli, Etna and Soufrierè Hills volcanoes - and on integrated geophysical and geochemical time series investigations of volcanic systems undergoing unrest.
Conny Hammer studied Geophysics at the University of Potsdam where she graduated in 2007. In 2012 she received her PhD at the same University with a seismological thesis. From 2012 to 2014 she continued in postdoctoral positions adapting her work to alternative datasets such as seismic events related to ice-movements using data from Antarctica. The generality of the developed pattern recognition approach led to a close collaboration with other researchers over the last years. Currently Conny Hammer holds a research position at the Swiss Seismological Service (SED) at ETH Zurich where she worked on the automatic recognition of different mass movements (e.g. snow avalanches, rockfalls) in the Swiss Alps.