This study investigates the dynamics of glyphosate and AMPA in the soil surface layer of two fields growing glyphosate-resistant crops in the loess Pampas of Córdoba Province, Argentina. Glyphosate decay and AMPA formation/decay were studied after a single application, using decay kinetic models. Furthermore, glyphosate and AMPA concentrations were investigated in runoff to evaluate their off-site risk. During a 2.5-month study, cultivations of glyphosate-resistant soybean and maize received an application of 1.0 and 0.81 kg a.e. ha-1, respectively, of Roundup UltraMax©. Topsoil samples (0-1, 1-2 cm) were collected weekly (including before application) and analysed for glyphosate, AMPA and soil moisture (SM) contents. Runoff was collected from runoff plots (3 m2) and weirs after 2 erosive rainfall events, and analysed for glyphosate and AMPA contents (water, eroded-sediment). Under both cultivations, background residues in soil before application were 0.27-0.42 mg kg-1 for glyphosate and 1.3-1.7 mg kg-1 for AMPA. In the soybean area, the single-first-order (SFO) model performed best for glyphosate decay. In the maize area, the bi-phasic Hockey-Stick (HS) model performed best for glyphosate decay, due to an abrupt change in SM regimes after high rainfall. Glyphosate half-life and DT90 were 6.0 and 19.8 days, respectively, in the soybean area, and 11.1 and 15.4 days, respectively, in the maize area. In the soybean area, 24% of the glyphosate was degraded to AMPA. In the maize area, it was only 5%. AMPA half-life and DT90 were 54.7 and 182 days, respectively, in the soybean area, and 71.0 and 236 days, respectively, in the maize area. Glyphosate and AMPA contents were 1.1-17.5 times higher in water-eroded sediment than in soil. We conclude that AMPA persists and may accumulate in soil, whereas both glyphosate and AMPA are prone to off-site transport with water erosion, representing a contamination risk for surface waters and adjacent fields.