Remote Sensing of Amazon Functioning: the roles of angular influences and ground validation

by Alfredo Huete (Plant Functional Biology and Climate Change Cluster, UTS)

Earlier this month an interesting and provocative paper was published in Nature entitled “Amazon forests maintain consistent canopy structure and greenness during the dry season”, by Morton et al. (2014).

They report Amazon rainforests to be both structurally and seasonally invariant during the dry season, challenging the recent paradigm of light-limitations on tropical forest primary production.  In effect, this study also appears to overturn a previous paradigm that Amazon forests are water limited, since Morton et al. (2014) neither find greening nor browning during seasonal drought periods nor across major drought years. They present 3 lines of evidence by which they base their conclusions, including the use of Lidar waveform centroid relative height data (WCRH) and GLAS NIR data, a canopy FLIGHT radiative transfer model, and MODIS enhanced vegetation index (EVI) data that has been normalised to remove all angular, bidirectional reflectance distribution function (BRDF) influences. Their results have enormous implications to the carbon balance of tropical forests as well as in predicting the fate of Amazon forests to future climate change.  Their results may also have wide implications on the use of satellite time series data products and hence worth a more detailed look and healthy discussion.

First, the novel aspects of the Morton et al (2014) study include the combined use of Lidar data, canopy modelling, and MODIS data to characterise canopy structure and greenness patterns.  They also make a valid case for a better understanding of BRDF influences and potential sun-induced artefacts that may confuse the interpretation of  landscape seasonality and vegetation phenology.  They demonstrate quite nicely, the seasonal sun-view angle geometry shifts that occur between a June solstice and September equinox, in which the relative sun-view azimuths converge upon the principal plane at the equinox.  The same pattern would be repeated between a December solstice and March equinox (something not considered in the Nature study).  The monthly progression of sensor orbits relative to the sun present two potential artefacts of concern in the interpretation of seasonal time series data; 

(1) The obvious one are local sun angle (LSA) variations present in polar orbiting satellite data with fixed equatorial crossing times. The Terra platform has a crossing time of 10:30am, while the Aqua satellite crosses at 1:30pm. Some MODIS products, such as the nadir-adjusted, NBAR reflectances, combine Aqua and Terra data and output data for local solar noon (LSN), however, both LSN and LSA angles will vary seasonally for a fixed overpass time.  What Morton et al. have done in their BRDF correction scheme is to fix the sun angle, rather than the more common procedure of fixed time of day. 

(2) A second more insidious sun-induced artefact concerns the role of backscatter vs forward scatter observations, that is particularly significant along the principal plane (PP).  Along the PP, one encounters either very bright sunlit canopies (backscatter side) or darker ‘shaded canopies’ (forward scatter) side, as described in the Nature study.  Morton et al. (2014) correctly state that the backscatter side will result in higher NIR reflectances and thereby positively biased EVI values, however, they ignore the forward scatter contributions along the principal plane, in which NIR and EVI become lower, or negatively biased.  

There are many reasons for healthy scepticism of this Nature study and certainly this paper will be strongly debated within and outside the remote sensing community.  I will briefly list a few concerns and points worthy of further discussion on my part.

  • This study essentially sidesteps the vast body of in situ, field-based evidence of leaf flushing in the dry season, field LAI measurements, and a network of eddy-covariance flux tower measurements, all of which show dry season increases in foliage and gross primary production (e.g., Restrepo-Coupe et al 2013).  

  • The “3 lines of evidence” (Lidar data, Flight canopy model, and  the authors own BRDF normalisation technique) appear to be an avoidance of the ground data and a precarious substitute for validation.

  • Amazon greening in the dry season has been reported by Brando et al (2010), using   MODIS NBAR-EVI, normalised to a zero view angle.  Although seasonal sun angle influences are present in this dataset, there are no principal plane backscatter issues (the primary argument of Morton et al). There are also studies that show quite strong Amazon dry season greening using the MODIS LAI product (Myneni et al. 2007), based on radiative transfer theory and explicit correction of sun and view geometries.  There is also an Amazon-specific fully corrected dataset, Multi-Angle Implementation of Atmospheric Correction for MODIS (MAIAC, Lyapustin et al 2013), and the official MODIS BRDF product which enables fixed sun angle corrections (MCD43A1).  All the above products show greening in the dry season, yet were ignored by Morton et al (2014), without explanation. 

  • Morton et al. never present any actual EVI data to support their conclusions.  In fact, they modelled and used the 2-band EVI, or EVI2, which is the backup equation used in the MODIS product for use on poor quality pixels. They also confuse the behaviour of the EVI equation with that of the EVI product, which is composited to avoid such backscatter solar influences

  • There are also many concerns of how the Flight model was parameterised, such that litter had stronger influences than LAI in their EVI simulations; and there will be some debate on the interpretation of seasonally invariant Lidar results vs Lidar sensitivity to leaf area in dense forests. 

Nevertheless, this study raises many important issues of concern to satellite product developers.  Satellite data (Landsat, AVHRR, MODIS, etc) are not normally corrected for sun angle, and hence potential sun angle artefacts may be present in all our commonly used datasets.  Therefore, to what extent should satellite data be routinely corrected for seasonal sun angle influences and what should a standard, fixed sun angle be? an average between summer and winter sun angles, or a continental average between Darwin and Hobart?   This becomes infeasible globally in which there is a midnight sun and periods of no sunlight. There have been orbital drift corrections (e.g., AVHRR) but such corrections restore the data back to their seasonally varying local sun angles.  The further one models the sun position away from actual conditions, the greater will be the (seasonal) uncertainties in the resulting data.  The 30 degree sun angle used in Morton et al doesn’t exist along the equatorial forests.

Lastly, seasonal shifts in sun angles are associated with corresponding changes in illumination conditions acting upon a vegetation canopy, and hence convolve with satellite-derived vegetation seasonality.  One can argue that there are both sun angle-induced artefacts as well a ‘real’ plant physiologic responses that are present, related to the magnitude of, and direct/diffuse composition of the photosynthetically-active radiation (PAR).  Thus, how would one know if a sun angle correction was accurate?   In conclusion, all of the above probably have some truth to it.  In fact, there are most likely multiple causes to observations in a complex world, and hence its not a matter of one explanation vs another, but rather a combination of explanations.  


  1. Brando, P. M. et al. Seasonal and interannual variability of climate and vegetation indices across the Amazon. Proc. Natl. Acad. Sci. U. S. A. 107, 14685–90 (2010).
  2. da Rocha, H. R. et al. Seasonality of water and heat fluxes over a tropical forest in eastern Amazonia. Ecol. Appl. 14, S22–S32 (2004).
  3. Hilker, T. et al. Remote Sensing of Tropical Ecosystems: Atmospheric Correction and Cloud Masking Matter. Rem. Sens. Environ. (2012).
  4. Huete, A. R. et al. Amazon rainforests green-up with sunlight in dry season. Geophys. Res. Lett. 33, L06405, doi:10.1029/2005GL025583 (2006).
  5. Hutyra, L. R. et al. Seasonal controls on the exchange of carbon and water in an Amazonian rain forest. J. Geophys. Res. 112, G03008 (2007).
  6. Jiang, Z. Y., Huete, A. R., Didan, K. & Miura, T. Development of a two-band enhanced vegetation index without a blue band. Remote Sens Environ 112, 3833-3845, doi:Doi 10.1016/J.Rse.2008.06.006 (2008).
  7. Jin, YF, Schaaf, CB, Woodcock, CE, Gao, F, Li, XW, Strahler, AH, Lucht, W, Liang, SL (2003). Consistency of MODIS surface bidirectional reflectance distribution function and albedo retrievals: 2. Validation. JOURNAL OF GEOPHYSICAL RESEARCH-ATMOSPHERES, 108(D5), 4159.
  8. Kim, Y. et al. Seasonal carbon dynamics and water fluxes in an Amazon rainforest. Glob. Chang. Biol. 18, 1322–1334 (2012).
  9. Lyapustin, A. I. et al. Multi-angle implementation of atmospheric correction for MODIS (MAIAC): 3. Atmospheric correction. Remote Sens. Environ. 127, 385-393 (2012).
  10. Morton, D. C. et al. Amazon forests maintain consistent canopy structure and greenness during the dry season. Nature, doi:10.1038/nature13006 (2014).
  11. Myneni, R. B. et al. Large seasonal swings in leaf area of Amazon rainforests. Proc. Natl. Acad. Sci. USA. 104, 4820-4823 (2007).
  12. Nemani, R. R. et al. Climate driven increases in global net primary production from 1981 to 1991. Science, 300, 1560-1563 (2003).
  13. Nepstad, D. C. et al. The role of deep roots in the hydrological and carbon cycles of Amazonian forests and pastures. Nature, 372, 666–669 (1994).
  14. Samanta, A. et al. Seasonal changes in leaf area of Amazon forests from leaf flushing and abscission. J. Geophys. Res. 117, G01015 (2012).
  15. Restrepo-Coupe, N., et al., What drives the seasonality of photosynthesis across the Ama-zon basin? A cross-site analysis of eddy flux tower measurements from the Brasil flux network. Agric. Forest Meteorol. (2013), 
  16. Schaaf, CB, et al. (2002). “First operational BRDF, albedo nadir reflectance products from MODIS”. REMOTE SENSING OF ENVIRONMENT, 83(2-Jan), 135-148.
  17. Vermote, E. F., El Saleous, N. Z. & Justice, C. O. Atmospheric correction of MODIS data in the visible to middle infrared: first results. Remote Sens. Environ. 83, 97-111 (2002). 

Leave a Reply

Your email address will not be published. Required fields are marked *