Remote sensing bio-optical models for aquatic systems can estimate optically active components in the water column, In your case I think you could try to estimate the nutrients through an empirical relation using Remote Sensing chlorophyll-a products. It means that you will be using chl-a as a bio-indicator of your nutrients concentration.
Useful information, in my opinion i hope a can measure the nutrient not from numerical model or biogeochemical model, but directly from satellite data by using an algorithm, or something like that...
I think, you should look for spectra of backscattering of dissolved substance groups, which you want to distinguish, in vawelengths 380-700 nm (only these shortvawes are usable to sound water bodies). But I am sceptic about the possibility to detect different substances directly because of complex signal from different components and also water may mitigate spectral signatures from different components.
Maybe, as Igor suggests, you should use some model of empirical data about nutrient species contribution into the whole mass, knowing total nutrients, particulate matter, chlorofyll, cianobacteries (they are distinguishable) load in the akvatory.