Publications

Galvao, LS; Petri, CA; Dalagnol, R (2024). Coupled effects of solar illumination and phenology on vegetation index determination: an analysis over the Amazonian forests using the SuperDove satellite constellation. GISCIENCE & REMOTE SENSING, 61(1), 2290354.

Abstract
Despite the importance of the Amazonian rainforests in the global carbon cycle, their phenological responses measured by large field-of-view satellite sensors are still not completely understood. In this scenario, close-to-nadir observations at high spatial and temporal resolutions made by satellite constellations may contribute to improve this knowledge. Here, we investigated the sensitivity of five vegetation indices (VIs) to canopy shadows over 15 protected forests of the Amazon, and the possible existence of coupled effects of solar illumination and vegetation phenology on the VI determination. The VIs are the Enhanced Vegetation Index (EVI), Green-Red Normalized Difference (GRND), Modified Photochemical Reflectance Index (MPRI), Normalized Difference Vegetation Index (NDVI), and RedEdge Normalized Difference (REND). They were calculated from 432 images obtained in 2022 by the Planet's eight-band SuperDove instrument. Few daily images acquired on the same day with distinct Solar Zenith Angle (SZA) were used to disentangle the effects of solar illumination from those of vegetation phenology. The results showed the presence of coupled effects of solar illumination and vegetation phenology on the EVI determination regardless of the site location, especially over dense forests. Such effects were not observed significantly in the GRND, MPRI, NDVI, and REND data. When the vegetation phenology was kept fixed in the analysis, solar illumination generated pseudo-greening effects from the EVI, even for small differences in SZA between daily observations. As the most sensitive VI to illumination conditions, the EVI increased from the beginning to the end of the dry season tracking solar angles and shade fractions. This dry-season trend was not observed for GRND, MPRI, NDVI, and REND, which presented low correlations with SZA and shade fractions. These four VIs were correlated with each other over most sites, which explained the agreement observed between their seasonal profiles. From the analysis of 15 sites distributed throughout the Amazon, our findings did not confirm patterns of large-scale greening at the end of the dry season. Local changes in greening (vegetation productivity) and browning were captured by the VIs over a few sites but in different periods of the dry season (June to September). At the high spatial scale of SuperDove observations, our results highlight the necessity of correcting solar and, in some cases, terrain illumination effects on the EVI before retrieving phenological metrics over the Amazon.

DOI:
1943-7226

ISSN:
10.1080/15481603.2023.2290354