Soil maps are being applied in modelling of environmental risk at a regional scale. Analysis of soil maps shows that there can be considerable uncertainty in map unit composition with consequent spatial variability in soil properties within map units. Uncertainty of map unit composition arises from a number of factors including map labels that do not reflect the mix of soil types present and imprecise location of map unit boundaries. Usually little, if any, information is supplied as to the extent and magnitude of these uncertainties. In this paper we develop datasets to quantify soil property variability for map units on the Canterbury Plains and use expert knowledge to quantify uncertainty in map unit composition for a small case-study area. We then apply Monte Carlo sampling to the estimated range of soil properties to input soil data into the GLEAMS simulation model to assess the effect of soil variability on nitrate leaching.

The simulation results show that there is considerable uncertainty in leaching from data derived from current soil map units. Even simple map units (map units characterised by a single soil type) have a wide range of potential leaching related to variation in soil profiles within the soil type. When map-unit inclusions were accounted for, variation in leaching risk can increase by 10–30%. Compound map units encompass the range of variability of the named component soils and there is a smaller increase in leaching variability in these units (compared with simple map units) when inclusions are added to the analysis. We discuss the implications of this study to the use of current soil survey data and make recommendations for future soil survey practice. In particular, greater effort is needed on description of the composition and reliability of soil map units.