Mapping forest age and species dynamics in yeraba plantation forest of northwest Ethiopia using hyperspectral signatures and spectral angle mapper

Abstract Plantation forests provide vital social, economic, and environmental benefits to surrounding communities. However, intensive forest exploitation and harvesting practices, which are often conducted without regard for age or species differences, pose significant setbacks to the sustainability of these ecosystems. There is still a major gap in fully understanding and mapping forest age and species. Using locally developed spectral signatures that are specific to age and species can improve the accuracy of forest monitoring and assessment. This approach is crucial for assessing current forest management practices and their alignment with sustainable forest ecosystem contributions. Thus, this study investigates how the intensive forest exploitation and harvesting bring age and species forest dynamics from 2016 to 2024 in the Yeraba plantation forest. To achieve the intended objective, 15 age and species forest samples were collected from the forest purposively, while high-resolution multispectral imageries for 2016, 2020, and 2024 were freely downloaded from Maxar (via Bing Maps). The collected forest samples spectral signatures were measured using an RS-3500 spectroradiometer, and the spectral libraryies were developed using R3.4. Additionally, the preprocessing and image classification of forest species and age were performed using the SAM algorithm referenced with the locally developed spectral libraries. As a result, the forest species and ag classification maps for 2016, 2020, and 2024 were created. The results showed that harvested areas in the Yeraba plantation forest sharply declined from 74.83 ha (23.94%) in 2016 to 7.49 ha (2.4%) in 2024, reflecting a net reduction of 67.34 ha (21.45%) at an average rate of 7.48 ha (2.39%) per year. The overall accuracy for the forest species- and age-classified maps for the years 2016, 2020, and 2024 ranged between 90.48% and 96.61%, with a 95% confidence interval(CI) ranging between 88.42% and 98.51% and an uncertainty level ranging between ± 1.9% and ± 3.07%. The developed ag and species specific forest spectral libraries will serve as a foundational step toward building a classification system tailored to specific age and species compartments within the forest.