This paper establishes the possibility of utilizing energy harvesting from mechanical vibrations to estimate extreme value responses of the host structure and demonstrates the calibration of these estimates for excitation spectra typical to natural systems. For illustrative purposes, a cantilever type energy harvester is considered for wind excitation. The extreme value estimates are established through a Generalized Pareto Distribution (GPD). Classically well-known Kaimal and Davenport spectra for wind have been considered in this paper for comparison purposes. The applicability of GPD for processes with short-range dependence is explored in both linear and nonlinear systems. The work also demonstrates how return levels can be mapped using energy harvesting levels and indicates that vibration energy harvesting, in its own right, has the potential to be used for extreme value analysis and estimates. The work has impact on health monitoring and assessment of built infrastructure in various stages of repair or disrepair and exposed to nature throughout their lifetime.