Ideally, dosimetric models for pine trees will accurately reflect the doses to its critical organs. However, existing models (e.g. that in ICRP 108) for the pine tree are over simplified or solely focused on estimating doses to one specific organ, and none of the existing models attempt dosimetry to the root system of the tree. The focus of this work was to develop four geometrically accurate sectional models for the pine tree in order to determine if the composition simplifications used for the ICRP 108 pine tree ellipsoid model are conservative, e.g. provide bounding estimates of dose, for internal dose estimates compared to more accurate compositions. To accomplish this, geometrically accurate models were created for a section of roots, small branch, large branch, and the trunk using tree samples and medical imaging software to develop surface mesh representations of all the tree organs. The surface mesh models were used to define the geometric boundaries for the monoenergetic photon and electron simulations in GEANT4 with a setup of simplified composition assumptions and more accurate compositions from literature. The resulting absorbed fractions for each internal source target pairing for each sectional model were compared as well as some simplified ellipsoidal model absorbed fractions. The results show that ICRP composition assumptions for the model materials do not always yield conservative absorbed fractions compared to more accurate compositions. ICRP compositions similarly do not always yield conservative absorbed fractions when considering physiologically likely scenarios such as the whole trunk as a source and the living layer as a target. Though there are situations noted in this work when the ICRP composition leads to conservative absorbed fractions for the pine tree sectional models, it was not true for all situations and so the ICRP 108 compositions should not be considered universally conservative for pine tree dosimetry. For dose response relationships for pine trees, accurate composition and geometric models should be used rather than simplified models.