Short answer: I would look to find iso-surfaces (or surface values) of temperature/salinity/potential density propagating downwards into different water masses.
It all depends on what kind of data you have. In case of a vertical profile time-series, you could make a contour plot of one of these properties, with depth in the y-axis, and time in the x-axis. Then, downwelling would be indicated by spikes of surface properties propagating downward.
In case you have several spatially distributed vertical profiles, at a single moment in time, you could look for downward "dents" on the iso-surfaces.
using the temperature and salinity profile, you can compute the potential density at each points and draw the isopycnal, the downward movement of isopycnals indicate the downwelling phenomina