In the irrigated area of the Ebro valley and in particular in Lleida (NE of Spain), the apple 'Golden Delicious' is marketed after a period of cold storage and the control of losses, during this period, is important. One of the most important physiological losses is due to the occurrence of bitter pit. At harvest, different prediction models using laboratory procedures have already been developed, but because of the high potential costs associated with this physiological disorder, its prevention at field level before orchard establishment is of main interest. The objectives of this work are: i) to identify soil characteristics linked to the incidence of bitter pit that can help farmers in their final decisions on orchard establishment and ii) to develop bitter pit prediction models during storage. For four years, 59 orchards were agronomically monitored including leaf and fruit analyses. They were arranged in different soil groups. The influence of the different parameters recorded on bitter pit incidence (stored under controlled or conventional atmosphere) was evaluated. The model obtained was validated during two years on 112 orchards. The behaviour of orchards according to the different soil groups and related to the incidence of bitter pit was statistically significant. The evaluation of soil characteristics prior to orchard establishment is a reliable preventive measure that would allow a major reduction (∼75%) in the incidence of bitter pit. Productive efficiency, fruit K/Ca relationship and percentage of fruit N concentration were the most important factors in bitter pit prediction during storage for all groups of soils, although the importance of each selected variable was different depending on the soil group.