So, I went ahead and attempted to do what I just suggested in the previous post. I have a Class A 100 mL volumetric flask, which has an uncertainty of +/- 0.08 mL, or 0.08% when used at 20 degrees C. I also have a scale that can measure up to 300 grams with a precision of 0.01 g. I also have a NIST-traceable thermometer with an uncertainty of +/- 0.5 degrees C. I also have a refractometer that I've calibrated with a DIY standard made according to Randy's article.
I took about 250 mL of my tank water in a sampling flask, and let it sit on my lab bench together with the clean, dry volumetric flask, the scale, and the refractometer at 20 degrees C for a couple of hours for the temperature to stabilize. I then measured the salinity using the refractometer as being 34.9. I then carefully weighed the empty flask. Next, I added 100 mL of 20C tank water to the volumetric flask, using a black/white card to correctly read the meniscus. I then weighed the full flask, and determined the weight of the 100 mL of tank water, which was 102.41 g, meaning a density of 1.0241.
There are various seawater density calculators, most of which take temperature and salinity as inputs, and give density as an output, such as the one found here:
http://www.csgnetwork.com/h2odenscalc.html. I'm not aware of any that take temperature and density as inputs, and give salinity as an output, but you can play around with the salinity input until you get the right density output. For the purposes of this discussion, though, the chart found here is also very helpful:
http://www.msc.ucla.edu/oceanglobe/pdf/temp_densal_chart.pdf.
Now, according to either the calculator or the chart, a density of 1.0241 at 20C indicates a salinity of right at 34.1. But, how certain am I of that number? What is the measurement uncertainty of what I just did? Well, fortunately, I can readily calculate that from the uncertainties of the various measurements taken to arrive at that 34.1 number, and determine the uncertainty of the salinity result.
The uncertainty in the density value is the sum of the uncertainties of the volumetric flask and the scale, which are 0.08% and 2 * 0.01% (because we used the scale twice), or a total of 0.10%. That means that the actual density could be within plus/minus 0.10%, or 0.001, meaning the real density value could easily be anywhere between 1.0231 and 1.0251. Plugging these values into either the calculator or the chart, that shows us that the actual salinity could be anywhere from 32.8 up to 35.3. And that isn't even taking into account the uncertainty in temperature, which widens the uncertainty even greater, from 32.6 to 35.6!
My point is that even with quite reasonably accurate equipment, used correctly to perform a very straightforward analysis, the ability to measure salinity gravimetrically is fraught with challenges. Using this method I can say with good confidence that my salinity is somewhere between 32.6 and 35.6.
Randy is right. Why not just use conductivity?
