Abstract: Current marine navigation radars are capable of high-resolution imagery of marine ice but are not able to classify the marine ice. Classifying marine ice means identifying the ice as first-year ice, multi-year ice or glacier ice. The latter two ice types are as hard as concrete and capable of damaging even ice hardened vessels such as icebreakers. The Canadian Coast Guard has identified the ability of marine navigation radars to classify marine ice as the single greatest improvement to be made in the safety of Arctic navigation. This thesis presents new research that improves our understanding of electromagnetic backscatter from marine ice. The goal of this work was two-fold: to demonstrate the feasibility of using commercial computational electromagnetic modelling software to simulate real-world marine ice targets, and to identify an optimum frequency or range of frequencies at which the marine ice targets can be definitively classified. Engineering models for scattering from electrically large objects made of a highly variable, complex, heterogenous, three-phase mixture of ice, air and brine are developed. To do so, an extensive literature review of the Arctic environment, and the physical and electrical properties of marine ice, is conducted to distill the required geophysical parameters of the three marine ice types of interest in this work. Using well-established dielectric mixing theory, these parameters are applied to homogenize the marine ice and model the target (in the presence of a flat sea halfspace) using a surface integral equation formulation. To reduce the computational resources required to numerically solve the integral equation models using the method of moments, computational electromagnetic modelling studies are conducted to select a suitable seawater halfspace representation and determine if the properties of larger objects can be inferred from scaled down models of the object. A case study is presented for backscatter from marine ice from 6 to 10 GHz, which explores the effects of frequency on the co- and cross-polarized backscatter intensity (and hence the apparent radar cross-section) of the three marine ice types of interest. Good agreement is found between the co- and cross-polarized backscatter intensity responses found from the engineering model computations and some existing experimental data from real-world marine ice targets. This work: (a) proves the feasibility of using computational electromagnetic modelling to simulate real-world marine ice targets, providing a new, cost-effective method for the study of backscatter from marine ice; (b) confirms the viability of using cross-polarization as a method of classification; and (c) identifies 10 to 16 GHz as a potential optimal frequency range for the classification of marine ice using dual-polarization radar.