Testing a soil-andscape model for dry greywacke steeplands on three mountain ranges in the South Island, New Zealand

Abstract

We tested the accuracy of a soil–landscape model, previously developed using a very low sampling density, for predicting soil properties in the dry greywacke steeplands of the South Island, New Zealand. These deeply dissected steeplands have relatively uniform geology and slope form, and in the model development area, elevation and aspect largely controlled the soil pattern. Predicted and sampled values of selected soil properties were compared at randomly selected sites on the Benmore Range (the development area), the adjacent Kirkliston Range, and a similar mountain range on Molesworth Station, 250 km distant. Considering the harsh environmental conditions, parent material variability, management history, and vegetation change, none of which were directly included in the model, model performance was adequate, explaining up to 45% of the variability for 0–7.5 cm pH. The model did not consistently perform better at any one test location, although the mean absolute error increased with distance from the model establishment area for 0–7.5 cm %C, pH, and A-horizon total C and total N. Predictions were more accurate for properties based on one laboratory measurement. Where several field measurements were used to derive values (e.g. nutrient amounts in units of kg/ha), predicted values were less accurate presumably because they incorporate more measurement error. We conclude that at a broad management scale, the model adequately predicts the soil properties on dominant landscape units. The model greatly improved prediction of soil values in national inventories (e.g. A-horizon C) and also has application in environmental risk prediction and economic development. landscape, mountain, soil map, soil carbon.