In this paper, we propose an adaptive error concealment method for enhancement layer, which is quantized with a fixed step size, in SNR scalable system. The proposed method consists of two steps. First, we partition the impaired image area into the Not-Coded macroblocks and the Coded macroblocks according to the quantization step sizes of both layers. Since the step sizes of base layer are transmitted through guaranteed channel and the residual data of enhancement layer are quantized with a fixed step size, the decoder can always receive the quantization step sizes of both layers. Second, we choose the best concealment method among spatial-domain and temporal-domain error concealments for the Not-Coded macroblock. For this selection, neighborhood matching is used. Experimental results show that the proposed concealment method gives about 8% improvement in PSNR over the conventional one using base layer.