At the height of an extreme rainfall event, an officer managing a reservoir in Keralam may have to decide whether, when, and how much water to release. The reservoir level is displayed on the screen before him. But much of the other information relevant to that decision may be held in different systems and by different institutions.That, in a single image, captures an important part of Keralam’s flood-management challenge. The question is increasingly not whether the State of Keralam has data, but whether the right pieces of information can be brought together quickly enough to support timely and informed decisions.Vast improvements in the StateKeralam has considerably greater monitoring capability today than it did in 2018 (a year of severe flooding in the State). The Kerala State Disaster Management Authority (KSDMA) reported that 100 Automated Weather Stations were operational by December 2025. The Centre for Water Resources Development and Management (CWRDM) has developed the Reservoir Assessment Tool-Kerala, which uses satellite-derived storage estimates and modelled inflows to provide near-real-time information on reservoir conditions. The Kerala State Electricity Board (KSEB) and the Water Resources Department maintain reservoir-level and discharge data; the India Meteorological Department (IMD) provides weather and rainfall forecasts; and the Indian National Centre for Ocean Information Services (INCOIS) forecasts waves, currents, swells and tides.The information exists. The larger challenge is interoperability.There are lessons from 2018. The floods in 2018 demonstrated why a river basin must be understood as a connected system. Extreme rainfall was the principal driver, but what happened downstream depended on the volume of runoff, reservoir conditions, tributary flows and the capacity of rivers, lakes and outlets to carry the accumulated water.The Central Water Commission (CWC)’s study of the 2018 floods provides a striking example. Between August 15 and 17, it estimated that approximately 1.63 billion cubic metres (BCM) of runoff was generated by the Pamba, Manimala, Achenkovil and Meenachil river systems, against an estimated carrying capacity of about 0.6 BCM in Vembanad Lake. The CWC also identified the then discharge capacity of the Thottappally spillway — about 630 cubic metres per second — as a major constraint. As a result, around 1 BCM of runoff remained within the system, contributing to rising water levels in the lake and surrounding areas.Keralam’s Irrigation Department notes that Thottappally was designed for a discharge of about 1,800 cubic metres per second, but siltation in the downstream channel can reduce its effective capacity substantially.The lesson is important. A release that appears manageable at a reservoir cannot be assessed in isolation from conditions many kilometres downstream. Water does not recognise departmental boundaries.From databases to a decision networkKeralam may therefore not need another stand-alone portal or database, but rather a secure interoperability layer that connects the systems it already has.Such an architecture need not take ownership of the underlying data. Individual agencies could continue to own and operate their sensors, models and databases. Instead, a read-only integration layer could map the relationships among them in a machine-readable format. Which rainfall stations feed which catchments? Which catchments feed which reservoirs? Where will a release travel? What other flows will join it? Which settlements and critical infrastructure lie downstream? What is the river’s current carrying capacity, including that of its outlets? And what will the coastal conditions be when the water reaches the sea?Every input should retain its source, timestamp and confidence level. Hydrological and hydraulic models must remain scientifically validated, and institutional responsibility clearly assigned.Modern data architectures, knowledge graphs and carefully governed Artificial Intelligence (AI) make such integration increasingly feasible. But AI should come last, not first. Without trusted data and validated relationships, AI merely processes uncertainty faster.Technology must also address an institutional problem. Sharing operational information across departments raises legitimate questions about ownership, reliability, accountability and liability. A properly designed system can help address these concerns by maintaining an auditable decision trail: what information was available at a particular moment, which models were consulted, what warnings were generated, and why a particular recommendation emerged.