Current methods of soil survey and land evaluation are based on transfer by analogy and results can be limited by the quality of the classification system used for mapping. Parametric soil survey is an alternative where predictions of individual soil characteristics and qualities, rather than soil types, are provided for the complete survey area. Predictions can be derived from either local soil classifications or interpolation and surface fitting procedures. The feasibility for a third approach to parametric survey has been investigated using data from the lower Macquarie Valley, NSW, Australia. Soil characteristics were predicted using generalized linear models with more readily observed environmental variables as predictors. These variables (e.g. geomorphic unit, local relief, etc.) are related to pedologic factors controlling soil distribution. A large percentage of variation for most soil characteristics was accounted for by pedoderms which were mapped using stratigraphic relationships determined in the field and by air photo interpretation. (Variation accounted for in the A horizon: clay content = 63.9%, CEC = 66.5%, EC = 26.5%, pH = 43.0%, bulk density = 33.3%, COLE = 61.2%, −10 kPa gravimetric water content = 66.4%, −1.5 MPa gravimetric water content = 64.0%; B horizon: clay content = 34.9%, CEC = 58.2%, ESP = 41.5%, EC = 17.2%, pH = 33.3%, bulk density = 32.1%, COLE = 59.9%, −10 kPa gravimetric water content = 55.5%, −1.5 MPa gravimetric water content = 39.8%). Within-pedoderm statistical models were developed with landform as an explanatory for several units and these further improved predictions. The potential of the approach in routine soil survey is discussed and the advantages of generalized linear models for analysing survey data are noted.