The spatial analysis of the soil compaction risk has been developed at the regional level and applied to Wallonia (Belgium). The methodology is based on the estimation of the probability of exceeding the preconsolidation stress due to the application of loads on the soil. Preconsolidation stresses (Pc) are computed from the pedotransfer functions of Horn and Fleige (2003) at pF 1.8 and 2.5 and classified into 6 categories ranging from very low Pc (< 30 kPa) to extremely high Pc (> 150 kPa). The computation requires the knowledge of pedological (texture, organic content), mechanical (bulk density, cohesion, internal friction angle), and hydraulic variables (water content available, non-available water content, air capacity, saturated hydraulic conductivity). These variables are obtained from databases like HYPRES or AARDEWERK or from pedotransfer functions. The computation of Pc takes into account the spatial structure of the data: in some cases, data are abundant (e.g. texture data) and spatial variability is taken into account through geostatistical methods. In other cases, the data is sparse but uncertainty information can be extracted from the knowledge of the statistical distribution. Maps of the most probable Pc class are produced. Uncertainty is computed as the classification error probability. Implementation of these methods in Wallonia showed that Pc values higher than 120 kPa are reached either on 64 % of the territory at pF 2.5 or on 55 % at pF 1.8. A higher uncertainty was found at pF 2.5 than at pF 1.8. Uncertainty was also found higher for clay and clayed loess than for other textural classes present in Wallonia. The risk of compaction is defined as the probability that Pc is exceeded by the stress created by a load applied to the soil at a depth of 40 cm, the loads being similar to those induced by agricultural or forestry tires. It appeared that subsoil compaction risks exist mainly in loamy forest soils with small coarse fragments supporting loads similar to that existing on logging machines. In the zones where the uncertainty is low, the developed tool could be used as a basis for providing policy measures in order to promote soil-friendly farming and forest practices.