A simplified theoretical model of storm runoff has been combined with analysis of rainfall records to improve understanding of dryland runoff processes in the Guadalentin catchment of SE Spain, and in particular the decrease in runoff fraction with increasing slope length or catchment area. Estimates of storm runoff volumes can be made using a simple runoff threshold, SCS Curve Numbers, or infiltration equations. A modified Green-Ampt equation is applied here to give storm runoff estimates for conditions of constant effective rainfall intensity. This can be seen as both an improvement on estimates made using the simple runoff threshold, although with no additional parameters, and a good approximation to the SCS Curve Number family of rainfall-runoff relationships, providing an improved theoretical basis for the Curve Number approach and an explicit means of linking it to existing soils data on infiltration and surface properties. The frequency distribution of intensities is examined to determine an appropriate effective intensity in relation to slope length, with a duration long enough to allow overland flow to reach the slope base. Hydrologically Similar Surfaces (HYSS) are defined by local runoff characteristics, and the linkage to infiltration allows an explicit comparison with routing over land areas of different area, shape and topography. This provides the basis for up-scaling from at-a-point runoff to runoff from larger areas which are mapped as belonging to the same HYSS class, explicitly allowing for their different topography and size, providing an essential step in up-scaling data on fine-scale surface characteristics to the response of catchment areas. Here we focus on the ways in which catchment area or slope length influence runoff production within a given HYSS class. Characteristics which drive catchment response include (1) variations in intensity over time through a storm (2) areal variations in storm volume. For measured rainfall in the upper Guadalentin catchment, SE Spain, the first of these factors is seen to be the most important for catchments of up to 500 km2, and a procedure is suggested for deriving effective rainfalls from recorded intensity data. The rainfall data can then be combined with a HYSS analysis of the soil/land use response, and analyses of catchment network morphology and connectivity to forecast the heterogeneous distribution of ephemeral stream runoff.