Dr. Suteanu is an Associate Professor cross-appointed in the Department of Geography & Environmental Studies and the Department of Environmental Science at Saint Mary’s University, Halifax, Canada. His research focuses on nonlinear analysis and modeling of natural systems. Applications include climate variability, renewable energy, natural hazards (e.g. space-time patterns associated with seismicity and volcanism, landslide dynamics), and structural aspects of geosystems from small to large scale. In addition, he studies epistemological aspects of our interactions with the environment. His courses include Environmental pattern analysis and modeling, Environmental information management, as well as Statistics, Natural hazards, and graduate and post-doc courses on nonlinear approaches to natural complex systems. Results of his research are published in journals such as Pure and Applied Geophysics, Geomorphology, Fractals, Surveys in Geophysics, Meteorology and Atmospheric Physics, Journal of Environmental Informatics, Quaternary International, etc.
Environmental big data and nonlinear systems with high output rate: persistent challenges and novel solutions
Valuable environmental big data are often associated with major challenges regarding their appropriate handling. Moreover, their effective interpretation and their integration in evolving patterns with strong potential for application raise barriers that are even more difficult to overcome. This presentation argues that for challenges of such proportions, while a methodological arsenal focusing on aspects of fast handling of large amounts of environmental data of a wide diversity is beneficial, it is insufficient and sometimes inadequate. The effectiveness of efforts related to environmental big data can be notably enhanced by a comprehensive and suitably targeted assessment of the dynamic system characteristics to which the acquired data are associated. In the case of complex dynamic systems, understanding the nature of the system and monitoring its spatio-temporal change should be seen as more than an optional set of operations.
Considered in isolation, in-depth system-identifying operations can be relatively expensive in terms of time and computing resources, but they may provide insights capable of considerably improving the performance of the tools to be subsequently applied to the data. Not only can the processing time be slashed significantly, and not only can proper resources be more realistically allocated to each task: new, initially unforeseen value may be identified in the data without any additional acquisition effort. Moreover, most natural systems are complex and nonlinear. Their output is usually non-stationary and often strongly irregular. How can we make sense of their output in order to better understand them, when fast data streaming is added to their non-stationary and wild irregularity?
The paper presented addresses the above-mentioned difficulties with the help of methodologies developed by the author, which build on and incorporate some powerful approaches to nonlinear pattern analysis. A key step in this context is the transformation of time series into transition structures, which can be two-dimensional or multi-dimensional matrices. Properties of the resulting matrices are analyzed using rapid procedures and then followed in time in order to characterize systems with fast streaming output.
The methods are introduced systematically, step by step, and illustrated with practical examples. Key aspects addressed include scaling, universality, the detection of long-range correlations and their possibly spurious nature, as well as multifaceted aspects of pattern change identification. Applications concern renewable energy resources and natural hazards.