Non-invasive measurement of sea ice thickness using low frequency EM waves
Abstract
Measuring sea-ice thickness is of great importance in modern world as for navigation, extracting geological information, monitoring the environment and climate etc. Ice thickness can be retrieved by a variety of methods. Most common of them are drilling, EM sounding, ground penetrating radar (GPR), upward-looking sonar etc. There are several tradeoffs between these methods. Drilling is the most accurate and the most used method but it is destructive in nature and cannot be used for wide spatial coverage. EM induction sounding provides a competitive alternative to drilling. However, this methodology is very much dependent on the conductivity of the water below the ice surface and hence will not work for fresh water or will not be very reliable for seawater with low salt concentration. GPR employs high-frequency EM waves and the method is based on measurements of the travel time of radar pulses that get reflected from the interface medium. However, most sea ice has a groove-shaped dendrite interface. As the wavelength of the EM wave in GPR is comparable to the length of the dendrite layer, it causes significant reduction in the reflected energy from the ice-water interface. Upward-looking sonar gives accurate thickness data but it has limited spatial and temporal coverage as compared to airborne measurements. We propose to employ low-frequency EM waves to measure the thickness of sea ice. The measurement system will consist of two dipole antennas. The transmit dipole antenna will emit low frequency EM waves and the response from different interface boundaries will be recorded by the receiver. Since electric permittivity of floating ice is different from that of water beneath, the electric field response will vary with ice thickness. The proposed methodology is aimed at exploring the quantitative relationship between sea-ice thickness and E-field response. The effects of both anisotropy and environmental noise will be considered and the model will be expanded to n-layer system. Then, our computational results will be compared with simulation data from a high-fidelity finite-element computer code. A data-driven deep learning model will be trained in order to predict the ice thickness from the response received from the sensor data mounted on an airborne system. The advantage of using deep learning method is that the performance of the model can be improved continuously by acquiring more E-field data and adding multiple data types such as surface texture from infrared imagery and freeboard measurement with the E-filed response. Finally, an engineered system will be developed to demonstrate the validity of our methodology for measurement of sea-ice thickness.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 23, 2019
- Source ID
- N629091912013
Entities
People
- Mohammad Ariful Haque
Organizations
- Office of Naval Research
- United States Navy