Spectral analysis of amazon canopy phenology during the dry season using a tower hyperspectral camera and modis observations

Author(s):Show less -

de Moura, Yhasmin Mendes ^[1] ; Galvao, Lenio Soares ^[1] ; Hilker, Thomas ^[2] ; Wu, Jin ^{[3, 4]} ; Saleska, Scott ^[3] ; do Amaral, Cibele Hummel ^[5] ; Nelson, Bruce Walker ^[6] ; Lopes, Aline Pontes ^[1] ; Wiedeman, Kenia K. ^{[3, 7]} ; Prohaska, Neill ^[3] ; de Oliveira, Raimundo Cosme ^[8] ; Machado, Carolyne Bueno ^[1] ; Aragao, Luiz E. O. C. ^{[9, 1]}

Affiliation

^[1] INPE, Ave Astronautas 1758, BR-12245970 Sao Jose Dos Campos, SP - Brazil

^[2] Univ Southampton, Dept Geog & Environm, Southampton SO17 1BJ, Hants - England

^[3] Univ Arizona, Dept Ecol & Evolutionary Biol, Tucson, AZ 85721 - USA

^[4] Brookhaven Natl Lab, Environm & Climate Sci Dept, Upton, NY 11973 - USA

^[5] Univ Fed Vicosa, BR-36570900 Vicosa, MG - Brazil

^[6] INPA, Dept Dinam Ambiental, Av Andre Araujo 2936, BR-69067375 Manaus, Amazonas - Brazil

^[7] Harvard Univ, John A Paulson Sch Engn & Appl Sci, Cambridge, MA 02138 - USA

^[8] Empresa Brasileira Pesquisa Agr EMBRAPA, Ctr Pesquisa Agroflorestal Amazonia Oriental, BR-66095100 Belem, Para - Brazil

^[9] Univ Exeter, Coll Life & Environm Sci, Exeter EX4 4RJ, Devon - England

• Document type: Journal article
• Source: ISPRS JOURNAL OF PHOTOGRAMMETRY AND REMOTE SENSING; v. 131, p. 52-64, SEP 2017.
• Web of Science Citations: 18

Abstract

The association between spectral reflectance and canopy processes remains challenging for quantifying large-scale canopy phenological cycles in tropical forests. In this study, we used a tower-mounted hyper spectral camera in an eastern Amazon forest to assess how canopy spectral signals of three species are linked with phenological processes in the 2012 dry season. We explored different approaches to disentangle the spectral components of canopy phenology processes and analyze their variations over time using 17 images acquired by the camera. The methods included linear spectral mixture analysis (SMA); principal component analysis (PCA); continuum removal (CR); and first-order derivative analysis. In addition, three vegetation indices potentially sensitive to leaf flushing, leaf loss and leaf area index (LAI) were calculated: the Enhanced Vegetation Index (EVI), Normalized Difference Vegetation Index (NDVI) and the entitled Green-Red Normalized Difference (GRND) index. We inspected also the consistency of the camera observations using Moderate Resolution Imaging Spectroradiometer (MODIS) and available phenological data on new leaf production and LAI of young, mature and old leaves simulated by a leaf demography-ontogeny model. The results showed a diversity of phenological responses during the 2012 dry season with related changes in canopy structure and greenness values. Because of the differences in timing and intensity of leaf flushing and leaf shedding, Erisma uncinatum, Manilkara huberi and Chamaecrista xinguensis presented different green vegetation (GV) and non-photosynthetic vegetation (NPV) SMA fractions; distinct PCA scores; changes in depth, width and area of the 681-nm chlorophyll absorption band; and variations over time in the EVI, GRND and NDVI. At the end of dry season, GV increased for Erisma uncinatum, while NPV increased for Chamaecrista xinguensis. For Manilkara huberi, the NPV first increased in the beginning of August and then decreased toward September with new foliage. Variations in red-edge position were not statistically significant between the species and across dates at the 95% confidence level. The camera data were affected by view-illumination effects, which reduced the SMA shade fraction over time. When MODIS data were corrected for these effects using the Multi-Angle Implementation of Atmospheric Correction Algorithm (MAIAC), we observed an EVI increase toward September that closely tracked the modeled LAI of mature leaves (3-5 months). Compared to the EVI, the GRND was a better indicator of leaf flushing because the modeled production of new leaves peaked in August and then declined in September following the GRND closely. While the EVI was more related to changes in mature leaf area, the GRND was more associated with new leaf flushing. (C) 2017 International Society for Photogrammetry and Remote Sensing, Inc. (ISPRS). Published by Elsevier B.V. All rights reserved. (AU)

• FAPESP's process: 13/50533-5 - Understanding the response of photosynthetic metabolism in tropical forests to seasonal climate variations
• Grantee: Luiz Eduardo Oliveira e Cruz de Aragão
• Support Opportunities: Regular Research Grants