New Zealand has a large set of soil resource data gathered from more than 80 years of soil survey activity, which include survey reports, soil maps, and a laboratory analysis database. There is also a cadre of pedologists with expert knowledge of soil–landscape relationships. However, current heritage soil survey data in digital and analogue form can not be easily utilised in pedometric analyses because it is either hidden in the grey literature of soil survey reports, inconsistent due to poor correlation or varying standards, or because of concerns about sampling bias or the accuracy and relevance of attributes. The rich information content of these data needs to be captured in a form suitable for integration with other digital datasets. Since 2003 New Zealand has undertaken a revitalised effort in soil survey and soil database development called S-map (Lilburne et al., 2004). S-map is designed to capture knowledge from existing soil surveys as well as to fill gaps with new data using appropriate modelling techniques. The S-map project will provide a consistent national coverage of soil class and attribute data, with confidence estimates being recorded for each attribute entry. Attributes chosen for an S-map minimum dataset were those known to be useful for prediction of target outputs, and which could be readily extracted using expert interpretation from heritage data. Emphasis was given to soil physical attributes, particularly those that control water storage, and water dynamics. It was expected that most soil chemical and biological attributes can be modelled from these physical attributes, with inclusion of appropriate non-soil attributes, such as effective rainfall, geology and vegetation. Attributes for taxonomic classes were entered and correlated to a developing national taxonomic legend of soil families and soil siblings (Lilburne et al. 2004, http://www.landcareresearch.co.nz) within the framework of the New Zealand Soil Classification (Hewitt 1998). Soil profile variation was summarised by definition of 43 functional horizons. Functional horizons are defined by classes of stone content, texture group, ped size, and soil strength. These are intended as building blocks for modelling soil physical attributes for all New Zealand soils. Inferences from functional horizons are strongest for saturated hydraulic conductivity, macro-porosity and bulk density (Webb 2003, 3004). Variation within the soil taxonomic units was recorded as probability distribution functions within functional horizons, from available data and expert knowledge. The database allowed preservation of unique ranges for local areas within soil families that are found throughout New Zealand. We describe S-map with reference to its application in the Otago Region of New Zealand. In this region the original data comprised 13 soil surveys carried out over a 50-year period. Survey reports are of varying standards, many with poor representative pedon descriptions, and poor correlation of soil taxonomic classes. There is only a small set of soils with detailed profile descriptions and comprehensive laboratory analysis. The outputs of the Otago region S-map include data layers for modelling input, and soil fact sheets. Fact sheets for each soil class were derived dynamically from the S-map database, and accessed by a public web site (http://www.orc.govt.nz). Fact sheet attributes were derived by pedotransfer models based on the S-map minimum dataset.