The development of stem heartwood and the factors that control it play an important role in tree physiology, thereby impacting demographic and ecological processes of woody species. We investigated the relationships of stem heartwood with site- and tree-level variables in Pinus canariensis plantations. A total of 30 plots were sampled in the island of Gran Canaria, Spain, over a large elevation range (995–1875 m) and on terrain with different slopes (4%–70%) and exposures. The 15 pines closest to each plot center were measured and cored to quantify growth rates and the size of heartwood, also known as “tea”. We used generalized linear mixed models (GLMMs) to account for both fixed and random effects while evaluating the best predictors of heartwood presence. Stem diameter was the variable most correlated with heartwood radius, and allowing for a random slope and intercept of this relationship accounted for spatially related variability. Furthermore, the GLMM model became more effective when the relationship between stem diameter and heartwood was modeled using the presence/absence of the “tea” rather than its measured size. Other site- and tree-level variables either were not statistically significant or improved the model relatively little. Because stem heartwood affects both wood quality and the amount of carbon that trees can store, our findings have implications for forest management and carbon-conscious policies.