Identifying whether an area is a plantation is a critical step in understanding land use, deforestation risk, and supply-chain impacts. At Orbify, this identification is achieved by combining multiple upstream datasets with advanced machine-learning approaches to detect commodity crops at a global scale.
Orbify does not rely on a single dataset to classify land as plantation. Instead, the platform fuses several open-source, peer-reviewed commodity datasets. This multi-source approach improves accuracy, expands geographic coverage, and helps account for regional differences in crop type, canopy structure, and management practices.
Where dataset coverage allows, Orbify’s system provides global commodity crop detection, flagging areas that are likely to be plantations based on spatial overlap and probabilistic modeling.
Currently, Orbify integrates the following key datasets into its plantation detection pipeline:
Spatial Database of Planted Trees (Richter et al., 2024)
This dataset captures approximately 264 million hectares of planted forests across 140 countries, representing around 90% of the world’s total planted forest area. It is essential for distinguishing planted tree systems from natural forests.
Global Map of Oil Palm Plantations (Descals et al., 2021)
Focused on closed-canopy oil palm plantations, this dataset covers 49 countries and identifies roughly 19.6 million hectares of oil palm worldwide.
Canopy Top Height and Indicative High Carbon Stock Map (Lang et al., 2021)
Using the High Carbon Stock (HCS) approach, this dataset enables detection of palm oil and coconut plantations in Indonesia, Malaysia, and the Philippines by analyzing canopy height and carbon density patterns.
Soy Planted Area (Song et al., 2021)
A high-resolution map of soy cultivation in South America’s major soy-producing regions, supporting precise identification of soy plantations.
MapBiomas Brazil (Souza et al., 2020)
A detailed land-use and land-cover classification for Brazil that includes a wide range of commodity crops commonly grown in the country.
In addition to static datasets, Orbify employs machine-learning probability models to strengthen plantation identification. These models help infer plantation presence in areas where direct mapping is incomplete or uncertain, improving robustness across different geographies.
This work is supported by the Forest Data Partnership, which provides methodological guidance and validation frameworks for forest and land-use data.
By combining authoritative open-source datasets with probabilistic machine-learning approaches, Orbify can reliably identify areas that are likely plantations. This layered methodology ensures higher confidence in results and enables users to assess land-use patterns consistently across regions, commodities, and supply chains.
