The invention is directed towards a method and system for optically-based monitoring of chemicals of potential concern (COPCs), including mercury (Hg), methyl mercury (MeHg), and polychlorinated biphenyls (PCBs). It employs a multi-parameter statistical analysis model, such as partial least square regression (PLS regression), to identify combinations of predictors (e.g., optical, water quality, and physical properties) that have large covariance with the response values. The predictive model is validated against analytic samples from laboratories. The steps in the method are repeated until predictors, or parameters are selected that avoid over-fitting and which yield physically-based results.