Compressive sensing (CS)-based data aggregation has become an increasingly important research topic for large-scale wireless sensor networks since conventional data aggregations are shown to be inefficient and unstable in handling huge data traffic. However, for CS-based techniques, the discrete cosine transform, which is the most widely adopted sparsification basis, cannot sufficiently sparsify real-world signals, which are unordered due to random sensor distribution, thus weakening advantages of CS. In this paper, an energy-efficient CS-based scheme, which is called “treelet-based clustered compressive data aggregation” (T-CCDA), is proposed. Specifically, as a first step, treelet transform is adopted as a sparsification tool to mine sparsity from signals for CS recovery. This approach not only enhances the performance of CS recovery but reveals localized correlation structures among sensor nodes as well. Then, a novel clustered routing algorithm is proposed to further facilitate energy saving by taking advantage of the correlation structures, thus giving our T-CCDA scheme. Simulation results show that the proposed scheme outperforms other reference approaches in terms of communication overhead per reconstruction error for adopted data sets.