The soil moisture of a vineyard soil under two different plantation systems, bench terrace and straight down the slope (German system), was measured weekly from 1991 to 1997 in the Demarcated Region of Douro – the Port wine producing region in Portugal. The soils of the region are extremely stony, located on steep slopes along narrow valleys. These conditions complicate the study of soil water relations with deterministic functions. We fitted a large data set to stochastic models, with the objectives of (a) determining which stochastic model best forecasts soil moisture in various planting systems, (b) based on chosen models, objectively evaluate variation of soil moisture in each planting system, and (c) finding an objective factor on which to base management decisions. Soil moisture variation was best represented by a time frequency model (Fourier series). Soil under straight down the slope plantation had significantly higher water reserves for most of the year than bench terrace plantation but the difference was not high enough to drive a significant shift on production and fundamentally affect the choice of which system to choose for new plantations. The model-generated data suggested a superficial and subsuperficial downflow of water from upper slopes in the straight plantation relative to bench terraces. However, the model failed to shed additional information on the mechanisms related to soil water balance. Data generated by the time frequency model might prove useful for other management decisions such as schedule and intensity of topping and pesticide application.