Earth & Environmental Science

Share this article:

Earth & Environmental Science

  • Join our comunity:

Making Sense of Earth Science Data: Using Pattern Recognition for Data Analysis

By: , Posted on: December 12, 2019

The digital era has brought an outstanding change in the acquisition and interpretation of information concerning our planet. Working in Earth Sciences one rapidly recognizes the complexity of information collected in these disciplines, the large number of factors and parameters that must be accounted for, and the huge mass of data collected. Continuous monitoring, with data acquisition at a high speed, is now state-of-the-art and steadily blows up the amount of gathered data. All this information provides the background of our decisions and action, which often must be taken in short times, for instance in case of an impending threat. On the other hand, the re-analysis of the large amount of information stored in our archives broadens the understanding of the dynamics controlling our planet. For both aspects – rapid decisions and long-term data analysis –the development of efficient methods of processing and interpretation has become a key issue in the last years.

Pattern recognition offers solutions with respect to the need of efficient data handling and interpretation. The methods are strongly related to data mining and machine learning, revealing structures within data sets and facilitating the set-up of rules for decisions and actions. Pattern recognition techniques are based on mathematically formulated procedures; their results are therefore reproducible. Modern pattern recognition approaches show a high degree of flexibility, such as in supervised techniques, where we can solve discrimination problems of – in principle – arbitrary complexity.  They also allow to “detect” structures in data, even though their description is not straightforward at the first glance. As all these characteristics resemble principles of human cognition, so there is a strong link to the framework of artificial intelligence – the science where we make technology do similar things as human brains.

For a limited time you can read Chapter 1 – Patterns, objects, and features on ScienceDirect

Advantages and Pitfalls of Pattern Recognition: Selected Cases in Geophysics” is addressed to readers interested in the field of Earth Science research and teaching. The book starts with propaedeutic steps regarding the definition of objects, as well as the presentation of strategies of feature selection to identify unique objects and avoid traps like the “curse of dimensionality”. The book covers a wide range of techniques related to supervised and unsupervised learning; their in-depth presentation provides plenty of material for teaching in graduate courses, at the same time offering an intuitive understanding of the mathematical background. The authors present a number of applications, partly taken from the literature, others developed specifically for the purpose of the book. Each case study includes the description of the physical background, the formulation of the problem, the propaedeutic steps of data processing, and a posterior interpretation of the results. Furthermore, the book provides guidelines for the validation of the performance both regarding supervised and unsupervised learning. The authors show that a careful re-analysis of applications may also reveal the source of possible failures, such as flaws in the a priori information, inappropriate feature selection or an erroneous formulation of the problem.    

Real-world example data sets are included with the book, which allow the reader to follow important aspects of feature selection and preprocessing. The data can be used in computer codes – also included with the book – and allowing the users to play with the various tools. Sample sessions are proposed to get started with the data and codes. Students can improve their understanding by trying out the various options, whilst practitioners can adjust the configuration proposed in the examples for their specific needs. The book provides the readers with strategies of data preparation, feature selection, finding the most appropriate recognition technique, and finally interpreting and understanding the output of the performed pattern recognition approach.

For a limited time you can read Chapter 1 – Patterns, objects, and features on ScienceDirect

 The book is available on ScienceDirect now. Want your own copy? Order now on the store, enter code STC320 to save up to 30% at the checkout.

 

About the Authors:

Horst Langer has developed methods for automatic alert systems and early warning on Mount Etna as well as tools that are routinely operated in the monitoring room of the institute and are part of the alert system for Civil Protection. Aside from his documented experience in the application of various pattern recognition techniques, he has also published computer programs for pattern recognition.

Susanna Falsaperla has a long experience in the application of pattern recognition techniques and was among the first seismologists to apply automatic classification to seismic signals on volcanoes. She has made extensive use of pattern recognition in volcanology to relate multidisciplinary data to volcanic unrest and eruptive activity.

Conny Hammer has worked on automatic classification of seismic signals in continuous data streams and has introduced novel concepts and tools into the seismological community from fields of machine learning (e.g., speech processing). Her automatic recognition tools are currently implemented in daily observatory routines. Besides automatic event detection, she has focused on the application of machine learning tools in seismic site characterization.

 

 

Connect with us on social media and stay up to date on new articles

Earth & Environmental Science

The fields of Earth science, planetary sciences, and environmental science encompass disciplines critical to the future of our world and its inhabitants. Our well-being depends on a thorough understanding of air and water resources, soil chemistry, atmospheric dynamics, geology, and geochemistry, along with a myriad of other aspects of the environment we live in. Elsevier supports the efforts of researchers and scholars in these areas with content that meets their cross-disciplinary needs: journals, books, eBooks, and online tools that span computer science, chemistry, energy, engineering, biology, agronomy, ecology, environmental impact and many other topics fundamental to the study of our world. Learn more about our Earth and Environmental Science books here.