Open AccessElectrical Conductivity Imaging of Active Layer and Permafrost in an Arctic Ecosystem, through Advanced Inversion of Electromagnetic Induction Data
Author(s) -
Dafflon Baptiste,
Hubbard Susan S.,
Ulrich Craig,
Peterson John E.
Publication year - 2013
Publication title -
vadose zone journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.036
H-Index - 81
ISSN - 1539-1663
DOI - 10.2136/vzj2012.0161
Subject(s) - permafrost , emi , inversion (geology) , environmental science , arctic , remote sensing , electromagnetic induction , active layer , computer science , prior probability , range (aeronautics) , electromagnetic interference , meteorology , geology , geography , artificial intelligence , engineering , layer (electronics) , bayesian probability , telecommunications , paleontology , oceanography , chemistry , thin film transistor , organic chemistry , structural basin , aerospace engineering , electrical engineering , electromagnetic coil
Characterizing the spatial variability of active layer and permafrost properties is critical for parameterizing process‐rich models that simulate feedbacks from Arctic ecosystem to a changing climate. Because of the sensitivity of electrical conductivity (EC) measurements to moisture content, salinity, and freeze state and the ease of collecting electromagnetic induction (EMI) data with portable tools (e.g., EM38, GEM2, or DUALEM) over large regions, EMI surveys hold great potential for Arctic ecosystem characterization. However, estimation of subsurface EC distribution from such data is challenging because of the insufficient amount of information such data provide towards finding a unique solution. The non‐uniqueness problem is often approached by fixing inversion constraints and initial models without a clear understanding of their possible effects on the obtained results. Here we developed a direct search method, which involves a grid‐based evaluation of one‐dimensional layered model parameters, to estimate EC distribution from EMI data and evaluate the influence of prior constraints, data information content, and solution non‐uniqueness. We applied the new method to EMI data acquired in Barrow, AK, as part of the Department of Energy Next‐Generation Ecosystem Experiments (DOE NGEE–Arctic). Results demonstrate the success of the developed approach for estimating models that reproduce recorded data within a specified range of uncertainty at each measurement location, as well as the value of different types of constraints. Importantly, the method can be used to quickly investigate the need for and effects of different priors at numerous measurement locations, since the time‐consuming simulation of the EMI signals from the multidimensional search grid is performed only once.