Walnut agroforestry systems require regular tree pruning, generating a large volume of biomass residues which are mainly valorized as wood-energy, Ramial Chipped Wood (RCW) or animal litter. However, walnut is recognized as a rich source of different chemical compounds, which could be recovered as valuable chemicals. This study aims to improve the knowledge on the composition of the water and ethanol extractive contents of wood, knot and bark fractions from walnut branches, harvested in agroforestry (AF) and traditional forestry control (FC) systems. LC-MS analyses were carried out to identify the chemical composition of extracts. Additionally, all samples were analyzed by near-infrared (NIR) spectroscopy with the aim of developing a fast tool to distinguish walnut branches, knots and bark wood fractions from trees grown under agroforestry and plantation management. Extractive contents and chemical composition of AF and FC wood samples were slightly different among branch, knot and bark. Despite these small chemical differences, PLS-DA models based on NIRS measurements can distinguishing among wood samples from walnut trees grown under different silvicultural regimes. In addition, in the both forestry systems, branch and knot extractive contents were significantly lower than those of bark specimens. Principal Component Analyses (PCA) based on NIR-spectrometry of raw samples revealed that the chemical composition of branch and knot woods are similar to each other and are very different compared to those of bark samples. This study provides new knowledge on branch woods from agroforestry systems, which are still very under-studied at present.