The soil hydraulic conductivity (<i>K</i> function) of three layered soils cultivated at Paradys Experimental Farm, near Bloemfontein (South Africa), was determined from in situ drainage experiments and analytical models. Pre-ponded monoliths, isolated from weather and lateral drainage, were prepared in triplicate on representative sites of the Tukulu, Sepane and Swartland soil forms. The first two soils are also referred to as Cutanic Luvisols and the third as Cutanic Cambisol. Soil water content (SWC) was measured during a 1200 h drainage experiment. In addition soil physical and textural data as well as saturated hydraulic conductivity (<i>K</i><sub>s</sub>) were derived. Undisturbed soil core samples of 105 mm with a height of 77 mm from soil horizons were used to measure soil water retention curves (SWRCs). Parameterization of SWRC was through the Brooks and Corey model. Kosugi and van Genuchten models were used to determine SWRC parameters and fitted with a RMSE of less 2%. The SWRC was also used to estimate matric suctions for in situ drainage SWC following observations that laboratory and in situ SWRCs were similar at near saturation. In situ <i>K</i> function for horizons and the equivalent homogeneous profiles were determined. Model predictions based on SWRC overestimated horizons <i>K</i> function by more than three orders of magnitude. The van Genuchten–Mualem model was an exception for certain soil horizons. Overestimates were reduced by one or more orders of magnitude when inverse parameter estimation was applied directly to drainage SWC with HYDRUS-1D code. Best fits (<i>R</i><sup>2</sup> ≥ 0.90) were from Brooks and Corey, and van Genuchten–Mualem models. The latter also predicted the profiles' effective <i>K</i> function for the three soils, and the in situ based function was fitted with <i>R</i><sup>2</sup> ≥ 0.98 irrespective of soil type. It was concluded that the inverse parameter estimation with HYDRUS-1D improved models' <i>K</i> function estimates for the studied layered soils.