Motion sickness is a common symptom which occurs when the brain receives conflicting sensory information. Although many motion sickness-related biomarkers have been identified, estimating humans' motion sickness level (MSL) remains a challenge in operational environments. Traditionally, questionnaire and physical check are the common ways to passively evaluate subject's sickness level. This study proposes a physiology-based estimation system that can automatically assess subject's motion-sickness level in operational environments. Our previous study showed that increases in self-reported MSL in a Virtual-reality based driving experiment on a motion platform were accompanied by elevated alpha (8-12Hz) power most prominently in the occipital midline electroencephalogram (EEG). This study explores the feasibility of an automatic MSL estimation based on spontaneous EEG spectrum. To this end, this study employed three different estimators: 1) Linear regression (LR), 2) Radial basis function neural network (RBFNN), and 3) Support vector regression (SVR). The results of this study showed that SVR outperformed LR and RBFNN in estimating MSL from EEG spectrum. The averaged accuracy of MSL estimation by SVR was 86.926.09% across 6 subjects. This demonstration could lead to a practical system for noninvasive monitoring of the motion sickness in real-world environments.