Water flooding occurs at the output region of the anode side, which is one of the main reasons for performance degradation of automotive fuel cell systems. If liquid water saturation ratio in gas diffusion layers can be observed and regulated within a suitable range, water flooding in anode channels can be avoided. For this purpose, a zero dimensional dynamic model which focuses on the internal states (such as liquid water saturation ratio in gas diffusion layer) for the anode side before water flooding is presented. A nonlinear state observer for the anode side is designed using the unscented Kalman filter algorithm. Influences of the relative humidity sensor and initial values of internal states on the observation results are analyzed. The results show (1) four key internal states (e.g., water vapor pressure, hydrogen and nitrogen pressure, average liquid water saturation ratio) can be observed using this method, and the time constant of the sensor has a smaller influence on the observer than the signal noise. (2) With a suitable sensor (noise power <10(-8) bar(2)), the relative errors of the partial gas internal states can be within +/- 5%, while the relative error of the average liquid water saturation ratio in the gas diffusion layer can be within +/- 10%. (3) When the purging valve is controlled based on the liquid water saturation ratio, it is possible to avoid water flooding on the anode side. (c) 2020 Elsevier Ltd. All rights reserved.