One of the central claims of the THALWAG thesis is that cheap sensing makes a fleet-scale ocean observation network economically viable in a way it was not ten years ago. This deserves more precision than the claim usually receives.
What has actually become cheap
Temperature sensors have become very cheap. A solid-state thermistor with the accuracy and stability required for oceanographic data assimilation now costs a small fraction of what it cost fifteen years ago. Conductivity sensors - which you need to derive salinity - have followed, though more slowly and with more variation in quality across price points.
Dissolved oxygen sensors remain more expensive per unit than temperature sensors, but the price has fallen significantly with the expansion of environmental monitoring markets. GPS is effectively free. Satellite data transmission has become substantially cheaper with the expansion of low-earth-orbit constellations.
What has not become cheap
Pressure sensors with the resolution and stability needed for accurate depth profiling remain expensive relative to the other parameters. Calibration - both factory calibration and field calibration - is a significant and often underestimated cost component. Battery systems that can operate sensors reliably in a marine environment for an extended season are not free.
These are not prohibitive costs for a network that is operating on existing vessels with existing crew. But they are real costs, and they need to be in the model honestly.
What this means for the design
The THALWAG sensor suite for initial deployments prioritises temperature and salinity - the parameters where the cost-per-observation economics are clearest and where the data assimilation value is best established. Dissolved oxygen is included where the deployment class supports it. Pressure profiling is included on vessels capable of deploying a sensor on a line; surface-only sensors are deployed on others.
This creates a heterogeneous network, which is actually an advantage: it mirrors the heterogeneity of real ocean conditions and produces a dataset that is more spatially diverse than a uniform deployment of any single sensor type would be.
The post on sensor calibration - the part of this that gets the least attention and causes the most problems - is coming next.