Global Forum on Food Security and Nutrition (FSN Forum)

1. The criteria “no clearly visible indications of human activities…” appearing in the definition is important and adequate when considering the use of Earth Observation data to assess or map ‘primary forests’ (or proxies of primary forests). Indeed remote-sensing based approaches can allow to detect and map disturbances in the forest cover but very old or small-scale disturbances may not be ‘visible’ from such data. Historical assessments of forest cover disturbances can be based on remote sensing time series but are limited by the availability and characteristics (e.g. spatial resolution and temporal coverage) of the satellite data. Currently, for the tropical regions adequate Satellite imagery (Landsat) is not available before the year 1982 for South-America (mostly Brazil) and much later for other tropical countries (the first valid image acquisitions are often not available before 2000 in Congo-Gabon and the Gulf of Guinea). Disturbances due to human activities or natural processes that occurred before the first available images cannot be mapped in regions like the tropics where no other historical wall to wall information is available.

Consequently, we suggest the following operational definition of primary forests that can be consistent with the available historical observation data:

“ Naturally regenerated forest of native tree species, where there are no clearly visible indications of human disturbances in the last 20 years or during the period of existing observation data” similarly to the initial definition of primary forest used by FAO (“not recently disturbed forest”).

3. We have developed an approach to map “undisturbed tropical moist forests” (TMF) that can be considered as an operational proxy to primary forests in the tropical moist domain. The approach is based on the detection of tree cover disturbances visible with Landsat satellite imagery over the full archive available (from Google Earth Engine). We consider a forest as undisturbed when no disturbances are detected over the full period of available historical imagery (varying from 10 to 30 years duration according to the region).

An expert-based system allows processing the full Landsat archive data from 1982 (c. 1 250 000 scenes in total for the tropics). The approach allows detecting tree cover disturbances that are visible in 0.09 ha size pixels - including disturbances from selective logging and fires that can be visible only during a short period -, and to map remaining moist forests without any visible sign of disturbances during the available observation period.  An annual change dataset is also provided depicting the spatial extents of TMF and disturbances for each year.

4. Suggestions to improve the reporting on primary forest area and its changes in humid tropical regions:

• We would need more information on the spatial distribution of historical disturbances. The large geographical and temporal unevenness of the Landsat archive prevents robust monitoring of disturbances before the year 2000 in some countries and consequently consistent monitoring among countries.  However, consistent monitoring is possible during the last 20 years over the full tropical belt.  Expert knowledge could complement the RS-derived maps when historical data are missing, to possibly exclude the forests that have been falsely identified as undisturbed based on ancillary historical data.
• Finer spatial resolution data are needed to capture smaller disturbances (e.g. removal of small single trees). Sentinel 2 data (0.01 ha size pixel) will significantly improve the detection of disturbances but such data exist only since the year 2016 and operational products are not yet existing on the continental/global scale.

5. FAO can help the countries by collecting and providing them access to the most relevant information, in particular, access to historical satellite imagery or existing products (maps of undisturbed forests) and by giving them support for using this information (i) technically, and (ii) by providing some guidelines to ensure consistency among countries.