Plenary 1
Hyperspectral remote sensing of land cover characteristics and dynamics in versatile landscapes of East Africa
Petri Pellikka, Janne Heiskanen, Ilja Vuorinne, Zhaozhi Luo, Hengwei Zhao, Yanfei Zhong, Rami Piiroinen, Mark Boitt, Ian Ocholla, Ashfak Mahmud, Elisa Schäfer, Niklas Sädekoski, Linda Pesonen, Saana Järvinen
East Africa encompasses diverse landscapes ranging from tropical montane forests to heterogeneous agricultural areas and dry savannahs, which present unique challenges and opportunities for remote sensing applications. Taita Hills in Southern Kenya is a miniature of East Africa, featuring all these landscapes within a compact area, alongside varied land use, human activities, and conservation efforts. This presentation summarizes the past and current hyperspectral remote sensing research in the Taita Hills exploring land cover characteristics and dynamics within this region.
Over the years, multiple airborne hyperspectral datasets have been acquired using Specim AisaEagle and AisaKestrel sensors, along with airborne laser scanning data. These data have been used for developing tree species classification methods for species-rich landscapes, with a particular emphasis on invasive tree species, such as Acacia mearnsii, Eucalyptus spp., and Prosopis spp. using one-class classification methods. Related applications include mapping of tree species that produce nectar and pollen for beekeeping, and crop type classification. An unsupervised classification approach has been developed to characterize tree species diversity over tropical montane forests. The feasibility of hyperspectral data for modelling grass biomass has been studied in livestock herding and conservation areas. Furthermore, the integration of airborne laser scanning and hyperspectral data has been examined for tree species classification and developing aboveground biomass prediction models. In addition, to complement the airborne datasets, proximal hyperspectral sensing has been studied to model soil attributes along altitudinal gradients. Recently, the first drone-based hyperspectral datasets have been acquired to support agricultural applications.
These studies advance the understanding of land cover dynamics, biodiversity, and soil properties in East Africa through the application of hyperspectral remote sensing technologies. The findings emphasize the critical role of high-resolution spectral data in addressing ecological and agricultural challenges in diverse tropical landscapes.