Item

Abstract

Forests play an important role in the global carbon cycle, making accurate assessments of carbon dynamics essential for effective forest management and climate change mitigation strategies. This research examines the spatiotemporal patterns of carbon storage and sequestration (CSS) in forests' aboveground biomass using satellite data, machine learning (Support Vector Machines), carbon modelling and spatial statistics. The methodology follows a two-step classification process: (i) binary forest classification and (ii) forest type classification, mapping seven forest types within two main categories - Broadleaves (Quercus suber, Quercus ilex, Eucalyptus sp., and other species) and Coniferous (Pinus pinaster, Pinus pinea, and other species). We analyzed the relationship between forest type and CSS at the Nomenclature of Territorial Units for Statistics (NUTS) III level and identified spatial clusters, outliers, and hot and cold spots of carbon sequestration at the municipal level across mainland Portugal. The broadleaved category demonstrated the highest classification accuracy in both years, decreasing slightly from 90.3 % in 2018 to 89 % in 2022, while the Coniferous group had the lowest accuracy, declining from 84.1 % in 2018 to 83.6 % in 2022. Anselin's Local Moran's I identified clusters of carbon sequestration, while the Getis-Ord Gi analysis confirmed these findings, revealing statistically significant hotspots of carbon sequestration in the northern and central regions and cold spots in the southern region. By providing insights at the sub-regional and municipal levels, this study offers a robust framework to support sustainable forest management and climate change mitigation strategies. Moreover, it can assist decision-makers in prioritizing natural capital, and developing nature-based solutions to tackle climate change and biodiversity loss.