Back to search

Boston University

Earth observation data for action: multi-sensor mapping of commodity tree crops in the tropics

Abstract

dc:description.abstract

Commodity-driven deforestation is the primary source of forest loss in the tropics, with negative impacts on biodiversity, ecosystem services, and greenhouse gas emissions. Spatially explicit data on the extent and distribution of tree crop commodities associated with this loss – including cocoa, coffee, rubber, and oil palm – remains incomplete, posing significant challenges for emerging sustainable sourcing regulations. This dissertation explores how commodity tree crop land use is represented using remote sensing, whether Earth observation data can distinguish tree crops from natural forests with sufficient accuracy to support deforestation-free policies, and how such data should be applied in governance contexts. In Ghana and Côte d'Ivoire, high-resolution optical and radar time series analysis reveals that 63% of forest disturbance from 2019-2024 occurred within existing plantations, where routine management was nearly twice as prevalent as natural forest loss. Widely used global forest change products were found to systematically misclassify plantation rotations as deforestation, distorting interpretations of forest dynamics across West Africa. Expanding to pan-tropical scale, a multi-sensor deep learning framework was developed to produce a 10m probability-based dataset for cocoa, coffee, rubber, oil palm, natural forest, and non-forest for 2020 and 2023. Overall map performance was strong in major production regions but regionally uneven, reflecting persistent challenges in modeling and validating diverse tree crop land use systems. A final case study in Indonesia evaluated whether combining publicly available map datasets through convergence-of-evidence approaches improves plot-level land use verification of oil palm for regulatory applications. Across methods of varying complexity, no convergence approach outperformed the best individual dataset, underscoring that methodological sophistication alone does not improve accuracy and highlighting the importance of transparency, error tradeoffs, and evidentiary standards when satellite-derived maps are used for compliance and accountability. Collectively, this research demonstrates that accurately representing commodity tree crop land use is both technically challenging and fundamental to characterizing tropical forest dynamics, advancing both the technical foundation and practical guidance for the application of Earth observation data in deforestation-free policy frameworks and forest monitoring systems.

Degree

thesis:*
Name thesis:degree_name
Doctor of Philosophy
Level thesis:degree_level
doctoral
Discipline thesis:degree_discipline
Earth & Environment
Grantor
Boston University
Year dc:date.issued
2026

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Tarrio, Katelyn R.
Advisor dc:contributor.advisor
  • Friedl, Mark A.
ORCID dc:identifier.orcid

Subjects

dc:subject × 3

Rights

Language dc:language.iso
en_US

Identifiers

dc:identifier.*
Handle dc:identifier.uri
https://hdl.handle.net/2144/53704

Chain of custody

source
Harvested from
Boston University
Base URL
open.bu.edu/oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Tarrio, Katelyn R.. Earth observation data for action: multi-sensor mapping of commodity tree crops in the tropics. doctoral thesis, Boston University, 2026. https://hdl.handle.net/2144/53704