Estimating carbon losses from disturbances in tropical moist forests (deforestation and forest degradation) since 2011

The need for accurate information to characterize the evolution of forest cover at the tropical scale is widely recognized, particularly to assess carbon losses from processes of disturbances such as deforestation and forest degradation¹. In fact, the contribution of degradation is a key element for REDD+ activities and is presently mostly ignored in national reporting due to the lack of reliable information at such scale.
Recently Vancutsem et al.² produced a dataset at 30m resolution which delineates the tropical moist forest (TMF) cover changes from 1990 to 2019. The use of the Landsat historical time-series at high temporal and spatial resolution allows accurate monitoring of deforestation and degradation, from which the carbon losses from disturbances in TMFs can be estimated. A degradation event is defined here as temporary absence of tree cover (visible within a Landsat pixel during a maximum of three years duration) and includes impacts of fires and logging activities.
We quantify the annual losses in above-ground carbon stock associated to degradation and deforestation in TMF over the period 2011-2019 by combining the annual disturbances in forest cover derived from the Landsat archive the pan-tropical map of aboveground live woody biomass density (AGB) from Santoro et al.³ at 100 m. To reduce the local variability within the estimation of AGB values, we apply a moving average filter under the TMF cover for the year 2010. 
The carbon loss due to degradation is accounted as full carbon loss within a pixel (like a deforestation). The reason is that logging activities usually remove large trees with higher biomass densities than the average value of the disturbed pixel indicated by the pan-tropical maps. To avoid double counting of carbon removal, deforestation happening after degradation is not accounted as carbon loss.
Our results are compared with estimates of previous studies that cover different periods and forest domains: (i) Tyukavina et al.⁴ provide estimates of carbon loss from deforestation for the period 2000-2012 for all forests (evergreen and deciduous) discriminating natural forests from managed forests, and (ii) Baccini et al.⁵ provide estimates of carbon loss from deforestation and degradation for the period 2003-2014 for both evergreen and deciduous forests.

In a further step, we will analyze the sensitivity of the results to the input AGB values by applying the same approach to other AGB maps (e.g. Baccini et al. 2012⁶).
Finally we intend to use Sentinel-2 data (10 m) for monitoring the location and extent of logging activities and burnt areas and further improve the estimates of carbon losses from forest degradation. 

1. Achard F, House JI 2015 doi 10.1088/1748-9326/10/10/101002
2. Vancutsem C. et al. 2019 Submitted to Nat. Geoscience
3. Santoro M et al. 2018 doi 10.1594/PANGAEA.894711
4. Tuykavina A et al 2018 http://iopscience.iop.org/1748-9326/10/7/074002
5. Baccini A et al. 2017 doi 10.1126/science.aam5962
6. Baccini A et al. 2012 doi 10.1038/nclimate1354