Mekonen, Andsera Adugna (2026) SUSTAINABLE AGROFORESTRY ECOSYSTEM ASSESSMENT THROUGH EARTH OBSERVATION: A CASE STUDY ON ABOVE-GROUND BIOMASS ESTIMATION FROM DRONE AND SATELLITE DATA. [Tesi di dottorato]

Documento PDF MEKONEN_Andsera_Adugna_38.pdf Visibile a [TBR] Amministratori dell'archivio Download (6MB) | Richiedi una copia

Abstract

Agroforestry systems play a critical role in climate‑smart land management, yet their structural complexity and spatial heterogeneity make above‑ground biomass estimation particularly challenging. This dissertation proposes an integrated, multi‑scale methodology for above‑ground biomass estimation by combining drone‑based photogrammetry, Sentinel‑2 satellite imagery, LiDAR‑derived reference datasets, and advanced machine‑learning and deep‑learning approaches. At the fine scale, a complete Unmanned Aircraft System workflow was designed and validated, including mission planning, radiometric calibration, canopy height modelling, vegetation index computation, and extraction of spectral, structural, and textural features for tree‑level biomass prediction. Multiple regression and ensemble learning models were evaluated using both RGB and multispectral imagery to assess sensor performance and feature importance. To enable regional scalability, the research extends to satellite‑based biomass modelling using Sentinel‑2 composites processed in Google Earth Engine, ESA WorldCover masks, and Random Forest regression. To address saturation and reference data limitations, LiDAR-derived above‑ground biomass labels from the BioMassters dataset were used to train state-of-the-art deep learning architectures (U-Net, MA-Net, DeepLabV3, DeepLabV3+), producing pixel-level biomass maps with improved spatial consistency. Results demonstrate that integrating Unmanned Aircraft System‑derived structural metrics with satellite‑based predictors enhances model accuracy and supports cross‑scale calibration. The proposed framework advances a transferable methodology for biomass monitoring in heterogeneous agroforestry landscapes, supporting both precision‑level assessment and regional carbon‑stock estimation. The findings contribute to the development of scalable monitoring tools aligned with climate‑smart agriculture, carbon‑neutrality strategies, and Measurement, Reporting, and Verification requirements for nature‑based climate solutions.

Downloads

Actions (login required)

Modifica documento