Mapping planted forest age using LandTrendr algorithm and Landsat 5–8 on the Loess Plateau, China

Abstract

Forest age is a key parameter for understanding the water and carbon cycles and carbon sequestration potential of planted forest ecosystems. However, estimating forest age on a large scale is difficult. This study aims to determine whether the distribution of forest age can be mapped over a large area using the LandTrendr (LT) algorithm. With the LT algorithm, the initial year of forestland (breakpoints) can be determined rather accurately through the yearly trajectory of the normalized burn ratio (NBR) index in the revegetated forest. The results show that the LT algorithm is a convenient, efficient, and reliable method for identifying forests age. Moreover, a comparison with ground truth data on the Chinese Loess Plateau (LP) revealed that the overall accuracy of the error confusion matrix was 89 %, with a root-mean-square error (RMSE) of 2.14 years. Using the LT algorithm, we revealed that the forestland on the Chinese LP in 2020 was dominated by planted forests over 30 years old, accounting for approximately 78.27 %. The forestland area increased by approximately 20,386 km2 from 1990 to 2020, and the increase occurs primarily around the original forestland in the eastern and southern parts of the Chinese LP. This study provides an important parameter for assessing and quantifying biomass and carbon sequestration through afforestation on the Chinese LP.