1. Open-Source Basic Check:
For our "open data" API endpoint we use the Global Forest Watch (GFW) LossYear dataset for deforestation. This is filtered with the JRC EUDR V2 dataset as forest mask, corrected using the GFW Spatial Database of Planted Trees (SPDT) V2 datasets for perennial plantations.
2. Advance Satelligence Check:
Satelligence’s deforestation analysis methodology integrates a robust forest baseline creation process with advanced change detection algorithms to identify forest loss with high accuracy.
The forest baseline is generated using a hybrid approach that merges curated and harmonized open datasets (e.g., JRC Tropical Moist Forest Layers, UMD primary forest maps) with Satelligence’s proprietary data based on forest classification models. These models apply FAO and national forest definitions, categorizing forests into primary, disturbed, regrowth, dry forest, and native vegetation classes. Open data layers are quality-checked, standardized, and adjusted to remove areas deforested before defined cutoff dates (e.g., 31 Dec 2020). Proprietary commodity plot data, based on mapped EUDR commodities and commodities like sugar cane & coconut, are created using Landsat, Sentinel-1, and Sentinel-2 data, and are added as an additional layer to eliminate false positives, especially in agricultural production landscapes.
Satelligence uses two change detection algorithms. One for optical data such as Landsat and Sentinel-2, and another for radar data such as Sentinel-1.
The optical change detection algorithm is called SpatioTemporalAdaptiveBareness (STAB). In short, it uses forest statistics of the entire landscapes and compares these statistics to a pixel of interest. If this pixel has much higher Bareness (level of bare soil) than all the surrounding forest, the pixel is flagged as deforested. The advantage of using an algorithm which uses an adaptive threshold is that it can take seasonality into account, which is the case for many dry tropical forests and savanna such as the Cerrado and Chaco areas in South America.
The radar change detection algorithm is called Bayesian Iterative Updating. This is a statistical method, where current observations are compared to historical observations to determine the probability of change.
Satelligence’s forest detection methodology ensures temporal consistency, minimizes false positives and negatives, and supports multiple reporting frameworks (EUDR, NDPE, RSPO) for near real-time deforestation monitoring.
Data sources used. Besides using and developing commodity maps, that we use a layer to distinguish operational/production areas from forests, we use the following data sources to develop our forest baseline (overview from the resource centre):
Data layer | Spatial coverage | Temporal resolution | Spatial resolution (m) |
EU Forest Observatory Global Forest cover 2020 | Global | 2020 | 10 |
Global | 1990-2023 | 30 | |
Global | 1984-2021 | 30 | |
Global | 2000, 2013, 2016, 2020 | n/a | |
Pantropical region | 2000 | 30 | |
Ecuador | 2020 | 25 | |
Honduras | 2014, 2018 | 25 | |
Ivory Coast | 2020 | 30 | |
Argentina | 1998-2022 | 30 | |
Amazonia | 1985-2022 | 30 | |
Atlantic Forest, Brazil | 1985-2022 | 30 | |
Bolvia | 1985-2022 | 30 | |
Brazil | 1985-2022 | 30 | |
Chaco region, Argentina | 1985-2021 | 30 | |
Chile | 2000-2022 | 30 | |
Colombia | 1985-2022 | 30 | |
Ecuador | 1985-2022 | 30 | |
Pampa Region | 1985-2022 | 30 | |
Paraguay |
|
|
Additional Sources:
Satelligence Methodology Knowledgebase: https://doc.clickup.com/2608438/d/h/2fk9p-14655/0a018ad69e6c32b/2fk9p-11275
Deforestation Detection Methodology: https://doc.clickup.com/2608438/d/h/2fk9p-14655/0a018ad69e6c32b/2fk9p-12995