The age of a forest serves as a critical indicator of both its carbon stock and its capacity for carbon volume within forest ecosystems. The high spatial and temporal resolution of forest age improves the accuracy of estimates for forest biomass, storage, and carbon stocks. Obtaining reliable forest age data across various scales remains a significant challenge. This study conducted forest age mapping at a 30 m resolution over long time series, taking into account forest disturbance events, using Xinjiang forest as a case study. In 1991, the age of forests in Xinjiang was estimated utilizing the Random Forest (RF) model, incorporating both forest inventory data and remote sensing data. Utilizing Landsat Time Series Stacked (LTSS) Data from 1991 to 2022, the LandTrendr(LT) algorithm was applied to identify forest disturbance and recovery events from 1992 to 2022 using the Google Earth Engine (GEE) platform. Finally, the forest age distribution in Xinjiang from 1991 to 2022 was mapped in conjunction with the forest disturbance events. The findings revealed that: (1) the R2 values of the RF regression models for estimating forest age exceeded 0.65, with RMSEs below 22 years and RRMSEs ranging from 0.119 to 0.183, demonstrating high credibility of the simulation results; (2) the overall accuracy of the LandTrendr algorithm in identifying forest loss and gain in the study area was 85 % or higher. From 1992 to 2022, Xinjiang experienced a net increase in forest area of 28580.69 km2; (3) the forest age distribution from 1991 to 2022 indicates a predominance of young and middle-aged forests, with the proportion of old forests remaining between 30 % and 40 %, showing a gradual annual increase. This study presents a method to improve the estimation of forest age at high resolution over extended time series. It also supplies essential data for assessing carbon storage in Xinjiang's forests, both historically and prospectively, contributing to a deeper understanding of carbon balance in forest ecosystems within arid regions.