With the recent reduction in the arsenic maximum contaminant level (MCL), utilities out of compliance must modify current treatment processes. Enhanced coagulation is generally considered to be the least-cost modification option when a ferric coagulation system already exists. Options for implementing enhanced coagulation as a modification to conventional sweep-floc treatment include: increased ferric dose; addition of an acid dosing system; and, a combination of the individual options. This study determines the least-cost enhanced coagulation modification for three model waters (soft, moderate, and hard, based on increasing background inorganic concentration) by incorporating a mathematical representation of arsenic sorption to hydrous ferric oxide, and subsequent removal through sedimentation and filtration within an optimization algorithm for minimizing treatment cost. The optimization framework explicitly includes influent water quality and flow rate variability, and model parameter uncertainty to ensure the arsenic MCL is satisfied with a 95% probability. For the moderate water, incorporating variability and uncertainty increased the minimum required ferric dose to satisfy the MCL by 23.6% - 63.0% relative to a deterministic approach. As for the three modification options, the FERRIC ONLY option was always the least-cost treatment modification. The ACID ONLY option had a relatively small feasible region due to the limited sorption sites from the existing ferric dose. The FERRIC AND ACID option was always slightly more expensive than the FERRIC ONLY option, and, in many cases, the resulting optimal acid dose was at the minimum allowable dose. By including variability and uncertainty, the relative importance of the individual parameter distributions can be determined. The variability of influent arsenic concentration was always critical. The variability of influent pH was critical, or important, at lower pH values for the soft and hard waters, and for all influent pH values for the moderate water. Other important parameter distributions include the uncertainty in the filter removal efficiency of entrapped particles (soft water at higher influent arsenic concentrations), variability in sulfate concentration (low influent pH and arsenic concentrations), and uncertainty in the equilibrium sorption coeficient for the Fe(III)SOHCa²⁺ complex (moderate and hard water at high influent pH). Includes 25 references, tables, figures.