This work was based on six experiments carried out in the northwestern region of the state of São Paulo, using six common varieties of sugarcane. The soil of each locality was classified and both deformed and undeformed samples of soils were collected in a 150 cm deep profile, by horizon, for chemical, physical and physico-hydric characterization. Data on climate and sugarcane plant-cane yield for the 96/97 harvest season were also collected. To determine which factors were better correlated with productivity, group analysis and multiple linear regressions and correlations were determined for the different groups of data concerning site and variety. The chemical, physical and hydric characteristics of the undersurface (B) layer presented significant correlations with the crop yield; the "stepwise" procedure selected Ca from the B layer as the variable that best correlated with the sugarcane yield; the water regime of the soils were an important parameter to evaluate the yield potential of soils; the eutrofic Podzolic soils showed the highest yield potential capacity in the region.
sugarcane; degrees-day; hydric balance; multiple regression; crop modelling