The marine boundary layer (MBL) is the region where energy, momentum and masses are exchanged between the ocean surface and the free troposphere. The lack of observations with high vertical resolution over the ocean significantly restricts the understanding of the complex physical processes that occur inside the MBL. The relatively short vertical extent of the MBL (average about 1~2 km) and the frequent cloudiness at its top make probing the MBL extremely difficult from the space. Several features of the Global Positioning System (GPS) radio occultation (RO) technique suggest that it has a great potential for sensing the MBL. These features include global coverage, high vertical resolution, and the ability of GPS signals to penetrate clouds.Over moist marine areas, a large negative moisture gradient often exists across the thermal inversion capping the MBL, which can cause super-refraction (SR) or ducting. A large number of high-resolution soundings have shown that SR occurs about 90% of the time in a year over the subtropical and tropical oceans and even 50% at high-latitudes during the summer. In the presence of SR, the reconstruction of refractivity from RO data becomes an ill-posed inverse problem, i.e., a given RO bending angle profile is consistent with a continuum (an infinite number) of refractivity profiles. The standard Abel retrieval gives the minimum refractivity solution of the continuum and thus produces the largest negative bias, consistent with a negative bias that is often present in the retrieved refractivity profiles in the moist lower troposphere. Simulation studies indicate a large variation of the negative refractivity biases (could be over -15%). The impact of diffraction effects and the open-loop receiver tracking on the bending angle and refractivity retrievals are assessed. A novel approach is developed and tested to reconstruct the vertical refractivity structure within and below the SR layer, which yields a much-improved retrieval, especially below the SR layer (less than 0.5% error). Such a reconstruction method should greatly enhance our ability to measure the MBL globally using the GPS RO technique as well as to improve the MBL parameterizations used in weather and climate models.