The establishment of patterns of radial and longitudinal variations and the development of models to predict the wood anatomical properties, especially from juvenile wood, are of interest for both wood industry and researchers. Linear regressions were used to predict whole-tree, breast height and mature tracheid length and diameter in white spruce (Picea glauca (Moench) Voss) and the WBE model was used to predict the variation of tracheid diameter. Tracheid length and diameter increased from pith to bark. Tracheid length decreased, while tracheid diameter increased from apex to lower heights. Cambial age was the most important predictor of tracheid length. The final tracheid length models with either a log transformation or a third-order polynomial of cambial age explained 82% of the variation in the whole-tree tracheid length. At breast height, 83% of the variation in the whole tracheid length was explained using the juvenile value at a cambial age of 3 years. Up to 87% of the variation was explained by the model, including the average value of juvenile wood. However, mature wood tracheid length at breast height could not be predicted from juvenile wood. Distance from the apex predicted the tracheid widening in outer rings but failed to predict tracheid expansion of samples collected at fixed cambial ages. The WBE explained 86% of conduit widening in the outer rings. The sampling strategy, i.e. collecting samples longitudinally at a fixed cambial age vs. at a fixed calendar year is important in predicting tracheid diameter.