Shifting cultivation is by definition a temporally dynamic system in which fields are shifted to exploit the energy and nutrient capital of the natural vegetation–soil complex of the new plot. The current needs for improved soil productivity under this system call for a development of methods and strategies for maintaining soil fertility, based on a good understanding of soil behaviour in time under each phase of this land use practice, with quantified rates of changes in soil properties. This research was designed (i) to determine and quantify critical soil factors that are the most affected by these practices, and from these select a Minimum Data Set (MDS); and (ii) to develop empirical models of soil dynamics within this agricultural system as a function of land use time series based on the MDS. An analysis of this farming system led to the development of a conceptual model of the spatio-temporal dynamics of shifting agriculture. Soil samples were collected using mixed synchronic and diachronic approaches along a chronosequence (from zero to more than 30 years) of shifting agricultural land use systems. A robust quantitative multi-criteria procedure was developed for selecting soil variables, which are the most sensitive to this land use system. Application of the method to our dataset selected five soil properties (pH in water, calcium, available phosphorous, bulk density and organic carbon) as a Minimum Data Set (MDS) that can be used individually or in combination to assess the effect of this practice on soil conditions. These variables were easily interpretable in terms of their relation to land management practices and land use changes. Empirical models of linear/quadratic fractional rational functions were fitted to the time series data of the MDS using non-linear least squares. The fitted functions on four variables explained 50 to 80% of soil dynamics with time in the 0-20 cm layer, but only 25% for organic carbon. The functions showed a very quick reaction to forest conversion for calcium, available P and organic carbon which maxima are reached at the end of the first year. Soil reaction and bulk density showed significant changes a bit later (2.5 to 3.5 years). The relaxation times of soil chemical properties were much shorter than those of bulk density. These curves are useful to derive quantitative measures on temporal changes in soil with land use as input data to develop improved agricultural strategies and new research orientation.