@InProceedings{DalagnolWYBOFTMGSAASG:2023:InCaEm,
author = "Dalagnol, Ricardo and Wagner, Fabien Hubert and Yang, Yan and
Braga, Daniel and Osborn, Fiona and Favrichon, Samuel and
Takougoum, Le Bienfaiteur Sagang and Mullissa, Adugna and George,
Stephanie and Silva J{\'u}nior, Celso Henrique Leite and
Anderson, Liana O. and Arag{\~a}o, Luiz Eduardo Oliveira e Cruz
de and Saatchi, Sassan and Galv{\~a}o, L{\^e}nio Soares",
affiliation = "{University of California Los Angeles} and {University of
California Los Angeles} and {California Institute of Technology}
and {Instituto Nacional de Pesquisas Espaciais (INPE)} and
CTrees.org and JPL/NASA/Caltech and {University of California} and
{University of California Los Angeles} and CTrees.org and
{Instituto Nacional de Pesquisas Espaciais (INPE)} and {National
Center for Monitoring and Early Warning of Natural Disasters} and
{Instituto Nacional de Pesquisas Espaciais (INPE)} and {NASA Jet
Propulsion Laboratory} and {Instituto Nacional de Pesquisas
Espaciais (INPE)}",
title = "Increasing Carbon Emissions from Amazonian Forest Degradation",
booktitle = "Proceedings...",
year = "2023",
organization = "AGU FAll Meeting",
publisher = "AGU",
abstract = "Selective logging and fire disturbances affect large areas of
tropical forests every year causing forest degradation and the
reduction of biomass and carbon. However, disturbances' true
extent and impacts on carbon emissions are difficult to quantify.
These limitations can be attributed to the fact that conventional
monitoring systems do not accurately map these disturbances or
provide attributions. In this study, we use a deep-learning
approach and high-resolution Planet NICFI imagery (4.77-m) to map
forests degraded by selective logging and fire in the entire
Amazon region from 2017 to 2022 and estimate carbon emissions. To
map degradation, we extended an approach based on the U-Net model,
previously trained over Mato Grosso state (Brazil), to the entire
Amazon basin, obtaining high accuracy (>80%). Carbon emissions
were estimated for areas overlapping our degradation maps using
both Airborne Laser Scanning (ALS) datasets collected by National
Institute for Space Research (INPE/Brazil) between 2016 and 2018,
and multi-temporal regional maps of Aboveground Carbon Density
(ACD) derived from the Global Ecosystem Dynamics Investigation
(GEDI) and remote sensing data. Our maps show that selective
logging and fire degraded an average of 11,452 and 21,745 km2 of
forests per year from 2017 to 2022, respectively. This area has
been steadily increasing for logging and highly varying for fire,
with the largest area found in 2020 (34,702 km2), which was a
drought year. Logging and fire were mostly detected alongside the
Arc of Deforestation. Logging occurred more clustered than fire,
showing hotspots that overlapped known forest concessions such as
Tapaj{\'o}s-Arapiuns/PA, Flona Tapaj{\'o}s/PA,
Sarac{\'a}-Taquera/PA, Flona Jamari/RO, and Itapiranga/AM. We
also found other hotspots in Brazil at Paragominas/PA,
L{\'a}brea/AM, large areas of Mato Grosso state, as well as in
Madre de Dios and west of Ucayali regions (Peru), in Guarayos
(Bolivia), and in Suriname. For the Amazon basin, we estimated
increasing carbon emissions from 2017 to 2022, with similar or
higher magnitudes of carbon emissions from deforestation in some
years, such as 2020. Overall, these new estimates of the extent
and impacts of degradation for forest carbon in the Amazon region
highlight that tackling degradation is key for reducing carbon
emissions.",
conference-location = "San Francisco, CA",
conference-year = "11-15 Dec. 2023",
language = "en",
urlaccessdate = "03 maio 2024"
}