Forest site productivity maps can be of great help for sustainable forest management. Site productivity maps, commonly representing the site index (SI) of a specific tree species, allow foresters to forecast wood production over the entire area of interest and hence select the most appropriate location for establishing a particular tree species mix. In many situations, forest SI cannot be directly measured from the dominant height and age of a stand and must therefore be directly estimated from relevant local site factors related to climate, topography and/or soil (attribute-based approach). Alternatively, estimations can also be made based on site information available for nearby locations (location-based approach). Also both approaches can be combined (hybrid approach). Since there is no straightforward procedure for selecting the most appropriate approach, the performance of five regionalization techniques was compared for predictive mapping of the SI of two important tree species in the temperate lowland region of Flanders (Belgium): one location-based technique (ordinary kriging), one attribute-based technique (regression) and three hybrid techniques (geomatching, ordinary co-kriging and regression kriging). From the findings of this case study, it cannot be concluded that one technique outperforms the others under all circumstances, but it was possible to build a decision tree providing guidance in selecting an appropriate SI mapping technique depending on the availability and characteristics of the data.