Underwater sensor networks have extensive applications ranging from natural ecosystem to security monitoring. The basic task of a sensor network is to facilitate collection of the sensed data. To this end, existing works assume a centralized strategy where the sensed data is immediately forwarded to a central sink from individual sensors. This strategy faces formidable challenges due to the intricacies of underwater acoustic channels. Based on qualitative analyses and quantitative calculations, we introduce the idea of distributed storage and collection for underwater sensor networks. In the ACOustic Underwater Sensor NETwork (ACOUSNET) project, we establish the transforming framework of distributed storage and collection of the sensed data. This transform entails multi-level adaptations to the unique underwater acoustic channels, physical limitations, desirable network features and network heterogeneity. Our foremost objective is to establish distributed data storage and collection mechanisms to optimize the ACOUSNET performance. Due to the very different properties of deep water and shallow water, we also put different emphasis and design requirements on them. The backbone of our ACOUSNET is a suite of physical and data link layer solutions for efficient data transmission. Compared with existing studies on these issues, our endeavor is unique by taking into account the battery nonlinearity and the network heterogeneity. Sea trials are essential for underwater networks since the ocean is very difficult to emulate due to its vast scale and complexity. The proposed research in this project will be significantly informed by the data collected in sea experiments under various conditions.