To develop a method for the selection of suitable predictive indicators for the assessment of soil quality, we used a general approach for choosing the most representative indicators from large existing data sets, for mountainous rangeland in northern Iran. The approach involves identifying a suite of soil indicators and landscape attributes for an area of relatively uniform climate. The interrelationships between soil properties and plant growth in various landscape units were investigated and interpreted based on statistical analysis and expert knowledge. Multivariate statistical techniques were used to determine the smallest set of chemical, physical, and biological indicators that account for at least 70% of the variability in the whole soil data set at each site. We defined this set as the minimum data set (MDS) for evaluating soil quality. Using investments of time and budget considerations, two minimum data sets were selected. The MDSs were selected for their ability to predict soil stability and productivity, as components of site potential assessment for extensive grazing. The efficacy of the two chosen MDSs were evaluated in terms of their capacity to assess range capability by performing multiple regressions of each MDS against the plant growth characteristics: total yield, herbaceous plant production, and utilizable forage as iterative dependent variables. Variations in the plant response variables were best predicted by the variables, soil profile effective thickness, followed by nutrient cycling index, which is a landscape function indicator: total nitrogen; slake test; first layer thickness; and water retention at wilting point. The relationships between soil properties and plant growth showed that plant variables were more sensitive to soil physical properties than to soil chemical properties. (c) 2006 Elsevier B.V. All rights reserved.