Piscis in tabula: Turning Environmental Data into Actionable Reports

Modern mariculture operations generate large volumes of heterogeneous data, ranging from in-situ sensor readings and management logs to satellite imagery and meteorological forecasts. Although this information is essential for monitoring farm conditions and supporting decisions, it is often not readily available to operators. When available, it is typically conveyed through dashboards or raw tables that require substantial expertise and time to interpret. In this work, we introduce a web-based platform that organizes multimodal mariculture data as a vertical digital library, providing unified access, basic metadata organization, and integrated analytical tools. Beyond data aggregation, the system incorporates an AI Reporting Agent (AIRA) module that produces periodic narrative reports summarizing past conditions, highlighting relevant patterns, interpreting potential risks, and offering forward-looking suggestions informed by predictive models. We describe the architecture of the digital library and present early components of the proposed system, with a focus on the forecasting deep learning models and the report generation module. Preliminary quantitative experiments illustrate the feasibility of key analytical modules, while qualitative examples outline the potential of narrative reports to support operator understanding. This work positions digital libraries not only as repositories but as knowledge mediators, translating complex environmental datasets into actionable, human-centered narratives.