A soil-landscape model, comprising 12 land components at a scale of 1 : 5000, has been developed in Neogene close-jointed mudstone in the Gisborne-East Cape region of the North Island, New Zealand. In a validation, soil order was predicted correctly in 81% of observations, soil group in 80%, soil subgroup in 63% and soilform in 60% of observations. A simplified model based on 11 land components for use at a scale of 1 : 50 000 has also been validated. Here soil order was predicted correctly in 71% of observations, soil group in 73% and soil subgroup in 49% of observations. For application with a digital elevation model (1 : 50 000), the number of land components was amalgamated to five. Here the soil order and soil group were predicted correctly in 63% of observations and soil subgroup in 40% of observations during validation. In all trials, the percentage of correct observations increased if a second choice or subdominant soil class was allowed. It took 2 person-weeks to produce a soil map from the 1 :50 000 form of the model over 400 km2 of steep and hilly country by photo interpretation of stereo aerial photographs, compared with 1 day of applying computer algorithms on the digital elevation model (DEM). The soil-landscape model succinctly relates soil class to land component and it enables improved targeting of farm and planning inputs by empowering existing research into soil fertilizer requirements and soil physical properties.
