Exploring the use of learning techniques for relating the site index of radiata pine stands with climate, soil and physiography

•Stepwise, LAR and IFSR approaches do not perform an effective variable selection.•LASSO and PLS show a significant regression to the mean.•MARS approach models SI more effectively than purely linear approaches.•Heat variables, such as the sum of degree-days, have a positive influence on SI.•Frost and hydric stress variables have a negative influence on SI.