The Thesis.

The problem: a significant ocean, largely unmeasured.

The Indian Ocean covers approximately 70 million square kilometres. It is bounded to the north by the Indian subcontinent and to the west and east by Africa and Australia. It drives the Asian summer monsoon — the rainfall pattern on which roughly 1.5 billion people depend for food security. It has absorbed a disproportionate share of the heat the atmosphere has accumulated since industrialisation. Its fisheries feed hundreds of millions of people across South Asia, East Africa, and Southeast Asia.

It is also among the least observed major ocean basins on Earth. The Argo float network — the most comprehensive automated in-situ observing system currently operating — has fewer instruments per unit area in the Indian Ocean than in the Atlantic or Pacific. Historical hydrographic surveys are concentrated near accessible ports in the northern Indian Ocean. The southern Indian Ocean, in particular, remains largely unconstrained by direct observation. The World Ocean Database shows the Indian Ocean has received roughly one-fifth of the survey attention of the North Atlantic over the past half-century, despite covering a comparable area.WOD, NODC/NCEI, 2023

The gap: consequences that compound.

The consequences of this observation gap are not abstract. Ocean state initialisation errors in the Indian Ocean propagate into monsoon forecast errors in ways that are not fully understood precisely because the observations needed to validate the models are absent. Fishery assessments in the Arabian Sea and Bay of Bengal depend on sea-surface temperature and upwelling indices derived from satellites that see only the skin of the ocean; the thermal structure below — which determines where fish aggregate and where hypoxic zones form — is inferred from models with poorly-constrained initial conditions.

Addressing this gap by conventional means would require significantly more research vessels, more Argo floats, and more moored buoys in some of the world's most logistically demanding waters. No national oceanographic programme operating in the Indian Ocean region currently has the resources to materially change observation density within any credible planning horizon.

The insight: the infrastructure already exists.

The central insight of this work is that the observing infrastructure for the Indian Ocean already exists. It simply has not been used for ocean observation.

India's motorised fishing fleet numbers approximately 200,000 vessels, operating from hundreds of ports along India's coastline, the Andaman and Nicobar Islands, and Lakshadweep.CMFRI Handbook, 2022These vessels make daily or multiday passages across some of the most data-sparse waters in the world ocean. They are equipped with GPS. They return to port on predictable schedules. They are operated by maritime workers with detailed local knowledge of sea conditions, seasonal patterns, and anomalies that no instrument alone can capture.

If even a modest fraction of this fleet — five thousand vessels, say, representing roughly 2.5 per cent — were equipped with low-cost solid-state sensors and enrolled in a voluntary observation programme, the resulting network would provide daily ocean observations across the Arabian Sea and Bay of Bengal at a spatial density that no existing programme approaches, at a cost per observation that is a fraction of any alternative.

The model: data assimilation makes it an intelligence.

Data collected from a distributed fleet network are not, by themselves, an ocean intelligence. A temperature measurement at a point in the ocean says only what the temperature was at that point at that moment. Making use of such data at scale requires a physical model — one that integrates spatially and temporally sparse observations into a coherent, physically consistent representation of the ocean state.

THALWAG proposes a continuously running ocean state estimate for the northern Indian Ocean, initialised from existing global ocean analyses (specifically CMEMS/Copernicus Marine Service products) and locally refined by fleet observations through variational data assimilation. The model produces daily gridded fields of temperature, salinity, and dissolved oxygen at depth. It reports its own uncertainty. When new observations contradict its predictions, it corrects itself. It is designed to be falsifiable and to improve with time, because it must be.

The claim: testable, reported honestly.

The specific, testable claim of this work is as follows: a fleet-based observing network of the design described in the paper can produce a materially more accurate characterisation of upper-ocean state in the Arabian Sea and Bay of Bengal than is achievable with current observing systems, at lower cost per equivalent observation, with greater geographic coverage, and with a partnership model that returns demonstrable value to the fishing communities whose cooperation makes it possible.

This claim is evaluated through observing system simulation experiments using existing model output as synthetic truth. The results are reported honestly, including where the simulated system fails to deliver expected improvements and why. The OSSEs show meaningful improvement in upper-ocean temperature and salinity analysis in the Arabian Sea with fleet densities above approximately 2,000 active vessels; below that threshold, the improvement is marginal. These thresholds are reported as constraints, not aspirations.