Fechar
Metadados

1. Identificação
Tipo de ReferênciaArtigo em Evento (Conference Proceedings)
Sitemtc-m21d.sid.inpe.br
Código do Detentorisadg {BR SPINPE} ibi 8JMKD3MGPCW/3DT298S
Identificador8JMKD3MGP3W34T/4878HEL
Repositóriosid.inpe.br/mtc-m21d/2022/12.13.18.30
Repositório de Metadadossid.inpe.br/mtc-m21d/2022/12.13.18.30.29
Última Atualização dos Metadados2023:01.03.16.46.27 (UTC) administrator
Chave SecundáriaINPE--PRE/
Chave de CitaçãoDalagnolWBBPSSFYCCRAAS:2022:MaTrFo
TítuloMapping Tropical Forest Degradation using High-Resolution Planet NICFI Satellite Imagery and Deep Learning
Ano2022
Data de Acesso09 maio 2024
Tipo SecundárioPRE CI
2. Contextualização
Autor 1 Dalagnol, Ricardo
 2 Wagner, Fabien Hubert
 3 Braga, Daniel
 4 Bispo, Polyanna da Conceição
 5 Payne, Matt
 6 Silgueiro, Vinicius
 7 Silva Júnior, Celso
 8 Favrichon, Samuel
 9 Yang, Yan
10 Cushman, Katherine
11 Carter, Griffin
12 Ritz, Alison L.
13 Anderson, Liana O.
14 Aragão, Luiz Eduardo Oliveira e Cruz de
15 Saatchi, Sassan
Grupo 1
 2
 3 SER-SRE-DIPGR-INPE-MCTI-GOV-BR
 4
 5
 6
 7
 8
 9
10
11
12
13
14 DIOTG-CGCT-INPE-MCTI-GOV-BR
Afiliação 1 University of California Los Angeles
 2 University of California Los Angeles
 3 Instituto Nacional de Pesquisas Espaciais (INPE)
 4 Centre for Landscape and Climate Research (CLCR)
 5 University of Leicester
 6 Instituto Centro de Vida (ICV)
 7 University of California Los Angeles
 8 JPL/NASA/Caltech
 9 Jet Propulsion Laboratory
10 JPL/NASA/Caltech
11 CTREES.org
12 Virginia Polytechnic Institute and State University
13 University of Oxford
14 Instituto Nacional de Pesquisas Espaciais (INPE)
15 NASA Jet Propulsion Laboratory
Endereço de e-Mail do Autor 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14 luiz.aragao@inpe.br
Nome do EventoAGU Fall Meeting
Localização do EventoChicago, IL
Data12-16 Dec. 2022
Editora (Publisher)AGU
Histórico (UTC)2022-12-13 18:30:29 :: simone -> administrator ::
2023-01-03 16:46:27 :: administrator -> simone :: 2022
3. Conteúdo e estrutura
É a matriz ou uma cópia?é a matriz
Estágio do Conteúdoconcluido
Transferível1
Tipo do ConteúdoExternal Contribution
ResumoForest degradation caused by logging and fire disturbances affect large areas of tropical forests every year. However, their true extent is not quantified, because conventional monitoring systems do not accurately map these disturbances or provide correct attributions. Having a monitoring system that can map areas and attributions of forest disturbances is important for calculating emissions from deforestation and degradation and enforcing climate mitigation policies. Here, we present a novel approach to map and monitor tropical forest degradation from logging and fire using the state-of-the-art deep-learning models and high-resolution Planet NICFI imagery. By focusing on forests across Amazonia, we develop training data for the deep-learning model by visually interpreting logging, fire, and roads in the Planet imagery (4.77 m spatial resolution). The model was based on the U-Net architecture, a convolutional neural network (CNN) that understands spatial patterns in the imagery and produce pixel-by-pixel image classification. The inputs were image patches of 256x256 pixels, and the outputs were maps of logging, fire, and roads. The model learned to detect the disturbance and its types at the same time and produced independent outputs for each degradation type. The trained model was applied to predict over large regions and every 6-month from 2016 to 2022, producing cumulative degradation maps. We found that the deep-learning model was able to detect degradation with an overall accuracy above 98%, and showed F1-Scores of 0.82 for logging, 0.84 for fire, and 0.65 for roads. When compared to other forest disturbance products from Landsat data (TMF from JRC and GLAD from UMD), our products from the Planet data showed a better detection of all pixels affected by degradation and provided a more accurate attributions of logging, fire, and roads/trails. For example, in the Mato Grosso rainforests, we found ~4600 km² of new burned forests and ~4000 km² of new logged forests in 2020, covering a much larger area than the official PRODES/INPE deforestation rate of 1,779 km². Our findings over Amazon forests indicate that the proposed approach can be used for operational large-scale and near real-time monitoring of tropical forest degradation from Planet data.
ÁreaSRE
Arranjo 1urlib.net > BDMCI > Fonds > Produção pgr ATUAIS > SER > Mapping Tropical Forest...
Arranjo 2urlib.net > BDMCI > Fonds > Produção a partir de 2021 > CGCT > Mapping Tropical Forest...
Conteúdo da Pasta docnão têm arquivos
Conteúdo da Pasta sourcenão têm arquivos
Conteúdo da Pasta agreement
agreement.html 13/12/2022 15:30 1.0 KiB 
4. Condições de acesso e uso
Grupo de Usuáriossimone
Visibilidadeshown
Permissão de Atualizaçãonão transferida
5. Fontes relacionadas
Unidades Imediatamente Superiores8JMKD3MGPCW/3F3NU5S
8JMKD3MGPCW/46KUATE
Acervo Hospedeirourlib.net/www/2021/06.04.03.40
6. Notas
Campos Vaziosarchivingpolicy archivist booktitle callnumber copyholder copyright creatorhistory descriptionlevel dissemination doi e-mailaddress edition editor format isbn issn keywords label language lineage mark mirrorrepository nextedition notes numberoffiles numberofvolumes orcid organization pages parameterlist parentrepositories previousedition previouslowerunit progress project publisheraddress readergroup readpermission resumeid rightsholder schedulinginformation secondarydate secondarymark serieseditor session shorttitle size sponsor subject targetfile tertiarymark tertiarytype type url versiontype volume
7. Controle da descrição
e-Mail (login)simone
atualizar 


Fechar