Aeration and rapid sand filtration are common practice for the treatment of iron, manganese and ammonium containing ground water for more than a century. Nevertheless, this process has not become obsolete. On the contrary, this thesis assumes that enhanced rapid sand filters remain standard treatment technology in the 21st-century. Iron can be removed by homogeneous, heterogeneous and biological oxidation. The potential advantages of heterogeneous iron oxidation over homogeneous iron removal are high filtration rates, excellent iron removal and reduced losses of backwash water as demonstrated by the excellent performance of the pre-filters of water treatment plant Grobbendonk. Heterogeneous iron oxidation dominates at Grobbendonk because the supernatant is slightly anoxic and the pH is low. Thus adsorption and oxidation occur simultaneously. The target feed water in this thesis is deep anoxic ground water, characterized by high iron and methane concentrations. The typical high methane values demand intensive aeration to prevent excessive formation of mucus or slime by methanotrophic bacteria. This research focuses therefore on heterogeneous iron oxidation that can be applied by running an adsorptive filter under anoxic conditions and regenerate it intermittently under oxic conditions. In other words, adsorption and oxidation are separated in time. This work consists of three main parts: characterization of filter medium, batch experiments and pilot plant experiments. Sequential extraction provided useful information about the accumulation on filter medium grains. The accumulation on sand/anthracite grains of the two investigated production sites contained predominantly iron(hydr)oxides e.g. ferrihydrite, which was in agreement with literature. A method was developed to determine Freundlich isotherms constants that describe the adsorption of iron or manganese on filter medium of water treatment plants Holten and Spannenburg. These Freundlich isotherms allowed a rough estimation of the maximum operating time of a rapid sand filter in intermittent regeneration mode, which amounted to 189 bed volumes. Pilot plant research was initiated following up on these positive results. The pilot plant consisted of two columns. The first column was filled with iron-coated sand of Holten as an adsorbent. The aim of the experiments was to determine the optimal regeneration time and the influence of the Empty-Bed Contact Time (EBCT). An increase of the EBCT, e.g. a reduction of flow rate, resulted in an increase of the iron removal. Nevertheless, the performance was lower than estimated from the batch experiments. The explanations are: 1) there was no sharp-edged adsorption front, 2) calcium adsorption competed with iron adsorption. The absence of a sharp-edged adsorption indicates that kinetic limitations play an important role. Furthermore, a salt spiking experiment showed that the filter column did not show ideal plug flow reactor behavior. The anoxic process requires a considerable amount of oxic water for its regeneration. Oxic water beyond that to oxidize adsorbed iron(II) was necessary to expel methane. The slow start-up with virgin sand is an additional drawback, because it results in high water losses. However, a start-up with iron-coated sand is, of course, an option to be further investigated. The second filter column was filled with virgin sand and then used to determine the time that the process requires to get going, i.e. the startup time. It took more than 123 cycles, which implies a water loss of 3295 bed volumes, before iron became significantly removed. The research proved that a conventional oxic filter with aerators performs better than an anoxic filter in intermittent mode. With equal iron removal efficiencies, the number of bed volumes that can be treated and the accumulated iron load of a conventional filter are much higher. It is plausible that the anoxic pilot plant performs better at an even lower EBCT. However, this comes at the expense of one of the potential advantages of heterogeneous iron oxidation, a high filtration rate. Therefore, the outcome of this thesis does not yet warrant a change of the process configuration. When we furthermore compare the outcome of the pilot plant experiments with the performance of the pre-filters of Grobbendonk, we have to conclude that heterogeneous iron removal performs better when adsorption and oxidation occur simultaneously, instead of separated in time.