There are correlations of data in adjacent sensor nodes in wireless sensor networks (WSNs). Distributed source coding (DSC) is an idea to improve the energy efficiency in WSNs by compressing the sensor data with correlations to others. When utilizing the DSC, the network architecture that, deciding which nodes to transmit the side information and which nodes to compress according to the correlations, influences the compression efficiency significantly. Comparing with former schemes that have no adaptations, a dynamic clustering scheme is presented in this article, with which the network is partitioned to clusters adaptive to the topology and the degree of correlations. The simulation indicates that the proposed scheme has higher efficiency than static clustering schemes. There are correlations of data in adjacent sensor nodes in wireless sensor networks (WSNs). Distributed source coding (DSC) is an idea to improve the energy efficiency in WSNs by compressing the sensor data with correlations to others. When utilizing the DSC, the network architecture that, deciding which nodes to transmit the side information and which nodes to compress according to the correlations, influences the compression efficiency significantly. Comparing with former schemes that have no adaptations, a dynamic clustering scheme is presented in this article, with which the network is partitioned to clusters adaptive to the topology and the degree of correlations. The simulation indicates that the proposed scheme has higher efficiency than static clustering schemes.