A catena can be thought of as a sequence of soils along a transect from the top of a hill to the middle of an adjacent stream or local depression. Ideally it will run perpendicular to the contour lines. The term catena was first posited by Milne (1936) who studied tropical semi-arid soils found on the plateaux south of Lake Victoria in eastern Africa. He found that a sequence of seven soils occurred from the top of a residual granite hillock to the end of the slope below it, and that whilst these soils varied in their character, the formation of each soil was linked to those above it on the hillslope by erosive processes. The term catena itself comes from the Latin, and means “chain”. Therefore, not only can individual soils along a catena be thought of as links in a chain, but just as each link has its own position in a length of chain, so each soil has its own position in the catena. The catena concept is a means of describing the variability of soils in space (Sommer and Schlicting, 1997). Field soil surveyors have suggested that sampling methods that more explicitly explored soil-landscape relationships (such as along the catena) are the most appropriate sampling schemes for establishing mapping rules for digital soil mapping. However no formal sampling method for catena has been established. We make a first attempt at such a method here. If we consider soil-landscape relationships to be catenary relationships we can formalise a procedure called random catena sampling. The method is summarized as follows: (1) Using a digital elevation model, define a k-th order stream catchment, with the catchment boundary and streams. The choice of DEM resolution will be dependent on the size of the study area and the scale of the landform features within the study area. (2) Define a point (pixel) at random in the area of interest, trace all points uphill and downhill (to the stream of order k) from this point, this set of points is a random catena. The neighbouring pixel is chosen randomly for uphill or downhill, alternatively using the steepest ascent/descent criteria. (3) Repeat the above processes for the given number of catenas. All points belong to at least one such catena, points higher in the landscape have a larger probability of belonging to two or more catenas. A criterion maybe defined to ensure the catenas are well spread over the area of interest. A subset of all possible catenas can be chosen by simple random sampling, stratified random sampling (strata being e.g., lithology or aspect) or Latin hypercube sampling. Positioning of observation locations along the catena seems more open. It would necessarily seem appropriate to sample the highest and lowest position in the transect. Other points could be sampled using equal intervals of horizontal, vertical or across-the ground distance (or some other metric). An example of a random catena sampling design will be given. References: (i) Milne, G., 1936. Normal erosion as a factor in soil profile development. Nature, 138: 548-549. (ii) Schaetzl, R. and Anderson, S., 2005. Soils: genesis and geomorphology. Cambridge University Press, 817 pp. (iii) Sommer, M. and Schlicting, E., 1997. Archetypes of catenas in respect to matter—a concept for structuring and grouping catenas. Geoderma, 76(1): 1-33.