In a tracer experiment TDR transect measurements were made to study percolation behaviour in a 120-year-old pine stand (Pinus sylvestris) on a water-repellent sandy soil (Haplic Arenosol). The experiment (with potassium iodide) showed an 80% labelling of the total flow in organic layers, whereas the area of transport in the mineral soil was sharply reduced to 12–30%. The average diameters of these preferential flow paths were about 8–15 cm. The TDR measurements indicate a homogeneous flow only for a short period from February until April. At this time of the year preferential flow is insignificant, because the soil is at approximately field capacity and not repellent to water. During summer (May to September) the soil dries out, and most precipitation results in preferential flow during this period. For any daily rainfall exceeding 10 mm, water infiltrates down to 1 m depth in the soil, which nevertheless, is still within the root zone. This kind of deep percolation results in the subsoil’s wetting to field capacity (pF 1.8) earlier than the topsoil. A one-dimensional numerical model (SWAP) was used to simulate mean water balance with hydraulic functions with and without a water-repellency term. From the results of our tracer experiment we showed that the de-watering process in spring could be simulated well using the traditional piston flow concept, while the rewetting behaviour could be described more realistically using the mobile–immobile concept for water repellency.