Retrieve and validate ocean salinity

SAC-D/Aquarius carried NASA’s Aquarius instrument, which combined passive radiometers with an active scatterometer. The radiometers measured microwave emission; the scatterometer supplied information used to correct for ocean-surface roughness. NASA’s data centre describes using sea-surface temperature from other observations as part of the salinity retrieval. The instrument did not take a bottle of seawater or directly count dissolved salt molecules.

This is an inference problem. More than one environmental property can affect the received signal, so a retrieval needs a physical relationship and supporting information. Imagine a fictional instrument whose response changes with both salinity and surface roughness. Assigning every signal change to salt would confuse two causes. Measuring or estimating the roughness helps constrain the solution, while uncertainty in those supporting inputs still affects the result.

Salinity is a concentration. Rainfall and river input can freshen surface water, while evaporation can concentrate salt; currents and mixing also redistribute it. The Aquarius visualization around India uses colour to contrast the Arabian Sea and Bay of Bengal. Its colours encode a scientific quantity, not the visible colour of the ocean, and its low-to-high legend does not supply precise numerical values for an arbitrary pixel.

Use a deliberately simplified mass example: 1,000 grams of solution contain 30 grams of salt. Add 200 grams of freshwater while keeping the salt amount fixed. The concentration changes from 30 grams per kilogram of solution to 30 ÷ 1.2 = 25 grams per kilogram. This toy mass concentration is not a calibration rule for Aquarius or a conversion for every oceanographic salinity scale. It isolates dilution while leaving circulation, mixing and retrieval physics outside the example.

Freshwater dilutes a fixed amount of salt

SAC-D’s operations ended in June 2015 after a platform power-and-attitude-system failure. Aquarius had completed its primary science mission. Ending new observations does not erase the measurements already preserved or prevent later calibration and reprocessing. Distinguish the observation interval, the date a product was processed and the date a user downloaded it.

An original validation exercise compares satellite estimates with three matched reference values. Differences of +0.2, +0.4 and +0.3 in a common salinity unit have a mean bias of +0.3. That does not prove every future pixel has exactly that error. Time, depth and spatial footprint must be comparable, and a small sample may miss other conditions. Retaining reference data, quality flags and processing versions lets later users test the result rather than inherit an unexplained colour map.