Mesospheric radars are used to investigate the characteristics of a Kelvin wave from two equatorial sites: Jakarta, Indonesia, in the western Pacific and Christmas Island in the central Pacific. Our study focuses on the time span from mid-January through mid-October 1993. A Kelvin wave with a period near 3 days was detected throughout this 9-month duration, although it underwent deep amplitude modulations on a ~20-day timescale. A fitting procedure is applied to study the phase/amplitude behavior of the wave. The vertical wavenumber was measured by the radars and found to be small, wandering around zero with only a weak bias toward downward phase progression. The long vertical wavelength suggests that the wave was predominantly zonal wavenumber 1. The amplitude of the wave measured by the Jakarta meteor scatter radar was much larger than the amplitude measured by the MF partial reflection radar at Christmas Island. The smaller wave amplitude at Christmas Island could at least partially be due to a measurement bias associated with MF radars. The radar at Jakarta is a VHF meteor scatter radar and is not susceptible to this bias. However, the mean velocities and the amplitudes of the tidal and quasi 2-day wave components were in good agreement at the two sites. The estimated 9-month averaged zonal acceleration was ~0.67 m s-1 day-1 over Jakarta at 94--98 km and only about half as large over Christmas Island. The magnitude of the zonal acceleration occasionally showed large enhancements which suggest the importance of refractive effects associated with vertical and temporal variations in the mean winds. The larger 3-day wave amplitudes and inferred acceleration at Jakarta may reflect its location in the western Pacific, a region of high convection, and hence an excitation region for equatorial waves. The relative phase of the wave between the two radar sites gradually shifted over a timescale of weeks. These smooth variations in relative phase are suggestive of a superposition of waves with different zonal wavenumbers, perhaps radiating preferentially from one longitude. The phase of the wave as a function of altitude and time was much more disordered at Jakarta than at Christmas Island. The conjecture can be made that the more chaotic phase structure observed over Jakarta is due to higher-order zonal wavenumber components which weaken as they propagate eastward. ¿ 1997 American Geophysical Union