Clustering has been shown to be an effective architecture to prolong network lifetime in energy-limited wireless sensor networks. In this thesis, we provide analytical frameworks to study clustering performance for two different types of wireless sensor networks. First, using rate distortion theory and random geometry results, we study the impact of partially correlated data on clustering when data sinks are stationary. Our analysis in this case leads to the Non-Uniform Clustering algorithm which substantially enhances network lifetime. Second, we propose to use clustering for wireless sensor networks with mobile agents (C-SENMA) in order to reduce energy consumption and the latency of data collection. Our proposed clustering architecture uses a novel low overhead MAC scheme based on the multi-packet reception capability of the reachback channel. We further verify through analysis and simulation that C-SENMA performs better than the flat topology which has been proposed earlier for sensor networks with mobile agents.