Skip to main content
EU Science Hub
News article4 June 2019

A comparison of US and EU satellites in monitoring forest disturbances

amazon_rainforest_in_brazil_c_adobestock_by_gustavofrazao_169474397.jpeg
The State of Amazonas holds the largest area of intact forests in the Brazilian Amazon
© Adobe Stock, by gustavofrazo

A recent article co-authored by the JRC finds significant differences between the performance of Landsat 8 and Sentinel-2 data for mapping areas of forest disturbances in the Brazilian Amazon.

The authors analysed a time series of images from both satellites to map selective logging activities in seven legally authorised timber harvesting sites in 2016 and 2017, and compared the results.

While the accuracies of the detection of disturbances were similarly high for both satellites (> 95%), the area of selective logging mapped based on Landsat 8 data was almost 37% larger (area adjusted) than the respective results from Sentinel-2, due to the coarser spatial resolution of Landsat data.

The higher spatial resolution of Sentinel-2 makes it possible to better map smaller forest disturbances, and to map larger selective logging features more precisely.

The authors found that a grid- rather than pixel-based approach led to more accurate results, and facilitates the interoperability of the two sensors.

Monitoring forest disturbances in Amazonian forests

The study was carried out on one of the main timber production zones in the State of Amazonas. The largest state in Brazil, Amazonas holds the largest area of intact forests within the Brazilian Amazon.

Lead author Thais Almeida Lima carrying out field work in Santo Antonio do Matupi, Southern Amazonas State
Lead author Thais Almeida Lima (University of British Columbia, Vancouver, Canada; ex-JRC) carrying out field work in Santo Antonio do Matupi, Southern Amazonas State
© Raimundo Saturnino

Selective logging is a major source of forest degradation in the Brazilian Amazon, and is responsible for a significant amount of greenhouse gas emissions to the atmosphere.

The authors use the term 'forest disturbance' rather than forest degradation when mapping areas that are undergoing selective logging, as not all disturbance events necessarily result in a degradation of the forest.

Currently, satellite imagery offers the only feasible way of accurately mapping changes over large areas of remote tropical forests.

The Landsat program, a series of Earth-observing satellites co-managed by the United States Geological Survey (USGS) and the National Aeronautics and Space Administration (NASA), offers moderate spatial resolution (30 m) satellite imagery for detecting and monitoring deforestation and larger forest disturbances, but that can be too coarse to map accurately small-scale forest disturbances, such as those resulting from selective logging.

The European Space Agency (ESA) recently launched the Sentinel-2A (in 2015) and Sentinel-2B (in 2017) satellites as part of the EU's Copernicus programme. These twin satellites provide multi-spectral imagery at a spatial resolution of 10 to 60 m, with a revisit time frequency of five days.

Performance differences of Landsat 8 and Sentinel-2 sensors

The Sentinel-2 sensor, with its higher spatial resolution, performed slightly better than the Landsat 8 sensor in terms of overall mapping accuracy.

Logging infrastructure was to some extent better detected by Sentinel-2 (43.2%) than Landsat 8 (35.5%) data, confirming the potential of the European satellite for mapping smaller-scale logging.

The higher spatial resolution of Sentinel-2 makes it possible to better map smaller forest disturbances, and to map larger selective logging features more precisely.

Interoperability of both satellites facilitated by a grid-based approach

The authors tested an alternative, grid-based (rather than pixel-based) approach based on collected field data in logging areas identified by satellite. This approach was found to more exhaustively map the forest area affected by selective logging, as smaller logging features (such as skid trails, smaller felling gaps and logging roads, which are either beneath the forest canopy or too small to be mapped via satellite imagery) are also included in the grid cells.

According to the scientific literature these small logging features occur, on average, at a distance of up to about 150 m from larger logging features that can be mapped with satellite imagery, hence the grid cell size of 300 m x 300 m.

The grid approach was found to reduce the difference in the selective logging areas mapped using Landsat 8 and Sentinel-2 data to 5%, and thus facilitates the interoperability of the two sensors. It also significantly increased the area identified as being affected by selective logging from 4.9% (pixel-based) to 50.4% (grid-based) for Sentinel-2 and from 5.7% to 52.8% for Landsat 8, respectively.

The authors recommend that future research should establish some sort of correction method between the (pixel-based) results of two sensors, e.g. by applying the approach described above on larger areas in different types of forests and selective logging techniques, and should look at a possible inclusion of additional Sentinel-2 spectral bands in the analysis of forest disturbances.

Further information

Related Content

Comparing Sentinel-2 MSI and Landsat 8 OLI Imagery for Monitoring Selective Logging in the Brazilian Amazon

Towards Operational Monitoring of Forest Canopy Disturbance in Evergreen Rain Forests: A Test Case in Continental Southeast Asia

Details

Publication date
4 June 2019