Publicação:
An ecologically-constrained deep learning model for tropical leaf phenology monitoring using PlanetScope satellites

dc.contributor.authorWang, Jing
dc.contributor.authorSong, Guangqin
dc.contributor.authorLiddell, Michael
dc.contributor.authorMorellato, Patricia [UNESP]
dc.contributor.authorLee, Calvin K.F.
dc.contributor.authorYang, Dedi
dc.contributor.authorAlberton, Bruna [UNESP]
dc.contributor.authorDetto, Matteo
dc.contributor.authorMa, Xuanlong
dc.contributor.authorZhao, Yingyi
dc.contributor.authorYeung, Henry C.H.
dc.contributor.authorZhang, Hongsheng
dc.contributor.authorNg, Michael
dc.contributor.authorNelson, Bruce W.
dc.contributor.authorHuete, Alfredo
dc.contributor.authorWu, Jin
dc.contributor.institutionShenzhen Campus of Sun Yat-sen University
dc.contributor.institutionThe University of Hong Kong
dc.contributor.institutionJames Cook University
dc.contributor.institutionUniversidade Estadual Paulista (UNESP)
dc.contributor.institutionBrookhaven National Laboratory
dc.contributor.institutionInstituto Tecnológico Vale
dc.contributor.institutionPrinceton University
dc.contributor.institutionLanzhou University
dc.contributor.institutionInternational Research Center of Big Data for Sustainable Development Goals
dc.contributor.institutionEast China Normal University
dc.contributor.institutionNational Institute for Amazon Research (INPA)
dc.contributor.institutionUniversity of Technology Sydney
dc.date.accessioned2023-07-29T12:45:40Z
dc.date.available2023-07-29T12:45:40Z
dc.date.issued2023-03-01
dc.description.abstractIn tropical forests, leaf phenology signals leaf-on/off status and exhibits considerable variability across scales from a single tree-crown to the entire forest ecosystem. Such phenology signals importantly regulate large-scale biogeochemical cycles and regional climate. PlanetScope CubeSats data with a 3-m resolution and near-daily global coverage provide an unprecedented opportunity to monitor both fine- and ecosystem-scale phenology variability along large environmental gradients. However, a scalable method that accurately characterizes leaf phenology from PlanetScope with biophysically meaningful metrics remains lacking. We developed an index-guided, ecologically constrained autoencoder (IG-ECAE) method to automatically derive a deciduousness metric (percentage of upper tree canopies with leaf-off status within an image pixel) from PlanetScope. The IG-ECAE first estimated the reflectance spectra of leafy/leafless canopies based on their spectral indices characteristics, then used the derived reflectance spectra to guide an autoencoder deep learning method with additional ecological constraints to refine the reflectance spectra, and finally used linear spectral unmixing to estimate the relative abundance of leafless canopies (or deciduousness) per PlanetScope image pixel. We tested the IG-ECAE method at 16 tropical forest sites spanning multiple continents and a large precipitation gradient (1470–2819 mm year−1). Among these sites, we evaluated the PlanetScope-derived deciduousness against corresponding measures derived from WorldView-2 (n = 9 sites) and local phenocams (n = 9 sites). Our results show that PlanetScope-derived deciduousness agrees: 1) with that derived from WorldView-2 at the patch level (90 m × 90 m) with r2 = 0.89 across all sites; and 2) with that derived from phenocams to quantify ecosystem-scale seasonality with r2 ranging from 0.62 to 0.96. These results demonstrate the effectiveness and scalability of IG-ECAE in characterizing the wide variability in deciduousness across scales from pixels to forest ecosystems, and from a single date to the full annual cycle, indicating the potential for using high-resolution satellites to track the large-scale phenological patterns and response of tropical forests to climate change.en
dc.description.affiliationSchool of Ecology Shenzhen Campus of Sun Yat-sen University, Guangdong
dc.description.affiliationResearch Area of Ecology and Biodiversity School for Biological Sciences The University of Hong Kong
dc.description.affiliationCentre for Tropical Environmental and Sustainability Science College of Science and Engineering James Cook University
dc.description.affiliationDepartment of Biodiversity Bioscience Institute São Paulo State University UNESP, São Paulo
dc.description.affiliationDepartment of Environmental and Climate Sciences Brookhaven National Laboratory
dc.description.affiliationInstituto Tecnológico Vale, Pará
dc.description.affiliationDepartment of Ecology and Evolutionary Biology Princeton University
dc.description.affiliationCollege of Earth and Environmental Sciences Lanzhou University
dc.description.affiliationInternational Research Center of Big Data for Sustainable Development Goals
dc.description.affiliationKey Laboratory of Geographic Information Science (Ministry of Education) East China Normal University
dc.description.affiliationDepartment of Geography The University of Hong Kong
dc.description.affiliationInstitute for Climate and Carbon Neutrality The University of Hong Kong
dc.description.affiliationInstitute of Data Science and Department of Mathematics The University of Hong Kong
dc.description.affiliationNational Institute for Amazon Research (INPA)
dc.description.affiliationSchool of Life Sciences University of Technology Sydney
dc.description.affiliationUnespDepartment of Biodiversity Bioscience Institute São Paulo State University UNESP, São Paulo
dc.description.sponsorshipCoordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)
dc.description.sponsorshipFundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)
dc.description.sponsorshipUniversity of Hong Kong
dc.description.sponsorshipNational Natural Science Foundation of China
dc.description.sponsorshipU.S. Department of Energy
dc.identifierhttp://dx.doi.org/10.1016/j.rse.2022.113429
dc.identifier.citationRemote Sensing of Environment, v. 286.
dc.identifier.doi10.1016/j.rse.2022.113429
dc.identifier.issn0034-4257
dc.identifier.scopus2-s2.0-85145773613
dc.identifier.urihttp://hdl.handle.net/11449/246610
dc.language.isoeng
dc.relation.ispartofRemote Sensing of Environment
dc.sourceScopus
dc.subjectCarbon cycles
dc.subjectDeciduousness
dc.subjectEnvironmental gradient
dc.subjectMachine learning
dc.subjectMulti-scale remote sensing
dc.subjectTropical forests
dc.titleAn ecologically-constrained deep learning model for tropical leaf phenology monitoring using PlanetScope satellitesen
dc.typeArtigo
dspace.entity.typePublication

Arquivos

Coleções