Excavation within the root-zone of urban trees is increasing rapidly in compact cities for the installation and repair of belowground infrastructures, but research on excavation damages is mostly limited to the assessment of their effects on growth and dieback, while effects on physiology and long-term stability are still elusive. This experiment was aimed to understand long-term mechanisms of tree response to root severance. Two urban tree species (Aesculus hippocastanum and Tilia x europaea) were subjected to two different root damage intensities by digging either one, or two parallel (either side of the root flare) trenches. An equal number of trees were left undamaged as control. Shoot elongation and DBH growth were measured at the end of each growing season for four years after severance. Root system area was measured after the excavation and four years later. Leaf gas exchange, water relations, and chlorophyll fluorescence were assessed during the growing seasons over a 51-month period after severance. The effects on tree stability were evaluated 2 and 44 months after root severance by conducting pulling-tests and by calculating a theoretical uprooting resistance index based on the size of the root system. Results showed that, even at the most severe root damage treatments, which removed up to 70% of the root system, tree survival, growth and physiology were little affected. CO2 assimilation declined by 16%−25%, mostly due to stomatal limitations, as also supported by a 0.7%−1.9% decrease in the maximum quantum yield of PSII (Fv/Fm). Excavation reduced stem diameter growth rate by 16%−28% and shoot elongation by 30%−41%, compared to control. Although the disturbance to tree physiology induced by excavation had a moderate intensity, the ability to recover from such damage over a 51-month period was scarce, particularly in Aesculus. Results indicate that root damage acts as a predisposing factor which reduces in the long term the capacity of trees to withstand co-occurring stresses by decreasing carbon availability for growth and defence. Similarly, the ability of trees to withstand uprooting was reduced by excavation, and no full recovery occurred in the following 44 months, when uprooting resistance was still 20%–66% lower in damaged plants compared to control, depending on the severity of the damage. A strong correlation (R2 = 0.822) was found between results of the pulling test and the uprooting resistance index. This study highlights the importance of integrating physiological parameters for a better prescription of tree preservation guidelines during excavation work.