Runoff of organic and inorganic contaminants from live firing ranges is a challenging issue because of the extensive size, the variable nature of runoff, the random occurrence of surface contamination, and general inaccessibility. This is particularly true with energetic compounds such as hexahydro-1,3,5-trinitro-1,3,5-triazine (RDX). One potentially promising technology for addressing runoff in a passive, sustainable, and low-cost manner is the use of constructed wetlands to intercept flow. This research effort investigated the ability of down-flow constructed wetlands to remediate RDX-contaminated water at varying loading rates over a period of 2 years. The impact of competing electron acceptors (NO3− and SO42−), the role of plants in the remediation process, and the production of initial daughter products were determined. Significant RDX removal occurred (89–96%) for all loading rates (160–1600 mg/(m2 day)) at a hydraulic retention time of approximately 2 days. RDX degradation occurred in both NO3− and SO42− dominated electron acceptor zones. RDX and NO3− degradation rate constants (kRDX and kN−O3) were also determined using a 1D transport model with dispersion. KRDX declined with the increase of influent RDX (6.2 − 0.8 d−1 for planted mesocosms). Changes in degradation rates may be the result of changing bioavailability of carbon released by the peat moss. kN−O3 followed similar trends but was relatively greater than kRDX. Removal of plants is believed to have a small impact on overall RDX removal. RDX concentration in plant tissue was found to be semi-logarithmically related to RDX concentration in the rhizosphere. MNX, DNX, and TNX were observed as transient products generally proportional to RDX loading and low concentrations of MNX and DNX were detected in plant tissue at high RDX loading rates. Cessation of RDX exposure substantially reduced RDX concentrations plants over 6 weeks. These results support the use of constructed wetlands for the remediation of low-level RDX-contaminated water.