Earth Signals and Systems Group (ESSG)

The research goal of the Earth Signals and Systems Group (ESSG) is to develop succinct, data-driven models and computational methods to accurately infer the properties of natural systems. We aim to improve the representation, analysis and observation of geophysical, ecological and environmental phenomena. Toward this end, we study the classics, stochastic processes estimation and control, and the modern, learning information and pattern theories, to combine phenomenology with physics.

Current work suggests new algorithms to overcome the curse of nonlinearity, dimensionality and uncertainty in inference problems characteristic of coherent geophysical fluids (stics.mit.edu). Our findings suggest new low-cost, effective ways for designing observing systems to map such fluids (caos.mit.edu). Contributions to inference problems within the realms of Fluid Imaging (flux.mit.edu), Animal Biometrics (sloop.mit.edu) and Hurricane risk and uncertainty predictions (hurriup.mit.edu) have also been made.

ESSG is strongly interdisciplinary, comfortable developing both methodology and application. At our heart we are a science group motivated to bring systems engineering to science in the context of scientific investigations of the earth and environment. So, you can find us tinkering with "what works in the field" a lot. But we are also a methodologically inclined systems science group comfortable stretching contours of systems theory. Participants come from many pedagogical areas including EAPS, EECS, Mathematics, and Mechanical and Aerospace Engineering. ESSG is funded in part by the AFOSR Data-Driven Dynamic Application Systems (DDDAS) Program, the National Science Foundation (NSF), Lincoln Laboratory, Naval Undersea Warfare Center Division (NUWC), and MIT's MISTI program.

The overarching theme for research is the development and application of Computational Intelligence for Earth, Atmospheric and Planetary Science, interpreted broadly to include both Geophysics, Ecology and Environment. The key aspect of our work is to incorporate physical constraints and reasoning into computational inference methodology.