Urea-based selective catalytic reduction (SCR) system is a promising way to obtain high NOx reduction and commonly adopted in diesel engine aftertreatment systems. The ammonia storage ratio is critical for SCR feedback control but it is difficult to be directly measured by sensors. This paper aims to effectively estimate the ammonia storage ratio on line and reduce the cost of using ammonia sensors. In the proposed method, the ammonia storage ratio is treated as an external disturbance in the NOx dynamic model and estimated by the nonlinear disturbance observer (NDO) methods. Furthermore, to reduce estimation errors of ammonia storage ratio caused by the high-frequency measurement noises, a novel robust nonlinear disturbance observer (robust NDO) is proposed and compared with a typical design method (regular NDO). Both the NDOs are developed based on part of the three-state SCR model and cost-effective, since NOx sensors are only used. The stability and noise attenuation properties of both estimations were also analyzed in the paper. The simulation results based on the full-vehicle simulator of FTP-75 test demonstrate that the regular NDO and the robust NDO can effectively estimate the ammonia storage ratio even in cases where ammonia cross-sensitivity affects the response. Among the two observers, the robust NDO has better noise attenuation properties.