Two experimental plots amended with a co-compost of sewage sludge and green wastes (SGW) or with a municipal solid waste compost (MSW) were compared with a control plot without organic amendment (CONT) in terms of trace metals mobility. These plots were equipped with wick lysimeters, time-domain reflectometry probes and tensiometers for 6 yr (2004–2010). Different soil structures due to tillage and compost incorporation were identified in the tilled layers and reproduced in HYDRUS-2D for simulating water, Cu, and Cd transport. Two sorption estimation approaches were used, either assuming equilibrium between CaCl2 and ethylenediaminetetraacetic acid (EDTA) extractable metals (Kd-1) or using equations based on pedotransfer functions assuming nonlinear sorption for Cu (Kf) and linear sorption for Cd (Kd-2). Lysimeter data on Cu leaching were successfully reproduced with the Kd-1 approach for the SGW and CONT plots (model efficiency coefficient ESGW = 0.97, ECONT = 0.95), while the MSW plot showed better fitting with the Kf approach (EMSW = 0.77), which could be explained by the less stable organic matter of the MSW compost because it takes into account organic matter components (dissolved organic C and soil organic matter). The Cd leaching was reproduced with the Kd-2 approach for the two amended plots (ESGW = 0.12, EMSW = 0.80), while CONT simulation overestimated leaching. The percentage of measured Cu and Cd leached in reference to input mass was 0.6 and 2.7%, respectively, in the SGW plot compared with 5% for both metals in the MSW plot. Trace metal mobility appeared to be limited in the tilled layer by sorption to organic matter.