This paper presents a two-stage compression method for accelerating GPU-based volume rendering of time-varying scalar data. Our method aims at reducing transfer time by compressing not only the data transferred from disk to main memory but also that from main memory to video memory. In order to achieve this reduction, the proposed method uses packed volume texture compression (PVTC) and Lempel-Ziv-Oberhumer (LZO) compression as a lossy compression method on the GPU and a lossless compression method on the CPU, respectively. This combination realizes efficient compression exploiting both temporal and spatial coherence in timevarying data. We also present experimental results using scientific and medical datasets. In the best case, our method produces 56% more frames per second, as compared with a single-stage (GPUbased) compression method. With regard to the quality of images, we obtain permissible results ranging from approximately 30 to 50 dB in terms of PSNR (peak signal-to-noise ...