Insect defoliations are a major natural disturbance in high-latitude ecosystems and are expected to increase in frequency and severity due to current climatic change. Defoliations cause severe reductions in biomass and carbon investments that affect the functioning and productivity of tundra ecosystems. Here we combined dendro-anatomical analysis with chemical imaging to investigate the direct and lagged effects of insect outbreaks on carbon investment. We analysed the content of lignin vs. holocellulose, i.e. unspecified carbohydrates in xylem samples of Salix glauca L. collected at Iffiartarfik, Nuuk fjord, Greenland, featuring two outbreak events of the moth Eurois occulta L. Cross sections of the growth rings corresponding to both outbreaks ±3 years were analysed using confocal Raman imaging to identify possible chemical signatures related to insect defoliation on fibres, vessels, and ray parenchyma cells and to get insight into species-specific defence responses. Outbreak years with narrower rings and thinner fibre cell walls are accompanied by a change in the content of cell-wall polymers but not their underlying chemistry. Indeed, during the outbreaks the ratio between lignin and carbohydrates significantly increased in fibre but not vessel cell walls due to an increase in lignin content coupled with a reduced content of carbohydrates. Parenchyma cell walls and cell corners did not show any significant changes in the cell-wall biopolymer content. The selective adjustment of the cell-wall composition of fibres but not vessels under stressful conditions could be related to the plants priority to maintain an efficient hydraulic system rather than mechanical support. However, the higher lignin content of fibre cell walls formed during the outbreak events could increase mechanical stiffness to the thin walls by optimizing the available resources. Chemical analysis of xylem traits with Raman imaging is a promising approach to highlight hidden effects of defoliation otherwise overlooked with classical dendroecological methods.