An Algorithm of Forest Age Estimation Based on the Forest Disturbance and Recovery Detection

This is the authors' abstract. We don't add key points for this paper.

Forest age is a crucial parameter for evaluating the state and potential of carbon sequestration in forest ecosystems. However, the lack of a time-series forest age will lead to an inability to capture forest disturbance and restoration history, resulting in increased uncertainty in estimating forest carbon sinks. To address this issue, we aimed to propose an integrated algorithm of forest age estimation based on forest disturbance and recovery history, using Zhejiang Province forests as an example. Based on the Google Earth Engine (GEE) platform, we first used random forest (RF) to estimate the forest age in 2004, and then integrated the LandTrendr algorithm with RF to detect the forest loss and gain year during 2005-2019, and finally derived and mapped the time series forest age distribution from 2004 to 2019. The results show that: (1) TheR2(≥0.6) and theRMSE(≤5 years) revealed that the constructed RF models could estimate forest age with high reliability; (2) The integrated algorithm of LandTrendr and RF effectively extracted the forest disturbance and recovery regions with overall accuracies (OAs) above 0.7, while the extracted forest area in Zhejiang Province has a net increase of 2206.86 km2during 2005-2019; (3) The forest age from 2004 to 2019 was still dominated by young and middle-aged forests, with a shift in age dominance from 20-30 to 30-40 years old after 2013. This study provided an effective methodological idea for time-series forest age estimation and reliable basic data for estimating past and future forest carbon sinks.