We developed a wideband, low-power, lightweight, prototype ground-penetrating radar (GPR) for subsurface exploration for Mars. The transmitter and receiver subsystems were constructed using a commercially available radio frequency (RF) and digital integrated circuits, connectorized components, and evaluation boards. The transmit/receive system, antennas, laptop controller, and batteries are all accommodated on a 2.5-m sled. Field experiments were conducted in Kansas and Alaska. The experiments in Kansas tested the operation of the system. The primary objective of the Alaska experiments was to investigate the ability of a GPR to detect and distinguish between subsurface deposits of ice and ice-cemented soils. The investigation depths of these experiments ranged from 1 m to 30 m, and the subsurface geology included near-surface thaw, discontinuous permafrost, water-saturated soils, and lenses of pure ice. Dielectric contrasts within the ground were detected with near-meter resolution; however, identifying the geologic context of an interface was difficult due to ambiguities associated with reflection data. These investigations demonstrate some of the difficulties associated with inverting reflection data to obtain dielectric properties of the subsurface. Fortunately, previous geophysical investigations such as drilled cores and seismic surveys helped to constrain the geology, and numerical simulations provided an additional resource for the interpretations. On Mars, additional geological context will be limited and numerical simulations will become extremely important for data interpretation. Due to these difficulties, we demonstrate in one experiment how dielectric information can be obtained directly from bistatic measurements with a fixed transmitter and mobile receiver. Finally, a comparison with a commercially available Geophysical Survey Systems, Inc. (GSSI) radar system is presented, and we discuss how the system can be modified and improved for future exploration on Mars.