Electrocoagulation (EC), a disruptive “green” technology, was investigated for the removal of natural organic matter (NOM) from drinking water sources. Three anode materials (aluminum, zinc, and iron) and three NOM sources (Suwannee, Nordic, and a local source) were investigated. After one minute of process time, dissolved organic carbon (DOC) reduction was approximately 70-80%. High performance size exclusion chromatography (HPSEC) fractionation showed reductions mostly in the larger apparent molecular weight (AMW) fraction of NOM, from 76% of NOM > 1450 Da initially to approximately 40% after EC. For iron EC, significant EC design variables were investigated, including:​ current density (i) (2.43-26.8 mA/cm²), and charge loading rate (CLR) (100 to 1000 C/L/min). Optimum NOM removal was found at i ~10 mA/cm² and lower CLR. In-situ identification of iron speciation in EC investigated the impact of i and CLR on speciation and NOM removal from a local natural source. Low i and intermediate CLR increased bulk pH and reduced bulk dissolved oxygen (DO), where green rust (GR) was identified in-situ for the first time in EC by Raman spectroscopy. Further oxidation at higher i and CLR led to magnetite (Fe₃O₄) formation, while all other conditions led to increased DO and/or increased pH, with subsequent identification of only orange lepidocrocite (γ-FeOOH). GR showed the marginally higher NOM DOC and AMW fraction reductions. In synthetic water, differing operating parameters led to differences in φ and iron speciation, characterized by in-situ Raman spectroscopy, aqueous XRD, SEM, and cryo-TEM. High i in the presence of pitting promoters led to φ near unity where a GR intermediate was seen, and an end product of Fe₃O₄. A mechanism scheme summarizing EC speciation is proposed. A general model relating cell potential and current was developed for parallel plate continuous EC, relying only on geometric and tabulated variable inputs. For the model, the anode and cathode were vertically divided into n equipotential segments. Potential and energy balances were simultaneously solved for each vertical segment iteratively. Model results were in good agreement with experimental data, mean relative deviation was 9% for a low flow rate, narrow electrode gap, and polished electrodes.