Marine in-situ data is collected by sensors mounted on fixed or mobile systems deployed into the ocean. This type of data is crucial for understanding climate and forecasting the state of marine ecosystems, and valuable for industry operating at sea, including oil and gas, aquaculture, wind farms, and shipping. In the past decade, several public organizations have collected, managed, and openly shared in-situ marine data. Recently, initiatives such as the Ocean Decade Corporate Data Group have incentivized the sharing of marine data of public interest from private companies aiding in ocean management. The contribution of this paper is to present a software architecture for marine in-situ data streams. Our software architecture is founded on a long-term empirical software engineering process to integrate, process, and share data in a cross-organizational setting composed of stakeholders from public and private organizations working in the marine domain. The research process has been driven by knowledge extracted from the stakeholders’ experience and by challenges identified in the literature pertinent to resource-constrained sensor systems for collecting in-situ marine data. The challenges has been analyzed regarding their impact on the overall software system and associated data quality attributes, and a set of software architectural decisions have been made to address the challenges. The outcome is a proposed data flow-oriented software architecture which we have validated via a prototype implementation.