A MODIS-Derived Photochemical Reflectance Index to Detect Inter-Annual Variations in the Photosynthetic Light-Use Efficiency of a Boreal Deciduous Forest

1
A MODIS-Derived Photochemical Reflectance Index
to Detect Inter-Annual Variations in the
Photosynthetic Light-Use Efficiency of a Boreal
Deciduous Forest
G.G. Drolet1, K.F. Huemmrich2, F.G. Hall2, E.M.
Middleton2, T.A. Black3, A.G. Barr4 H.A.
Margolis1 1Departement des sciences du bois et
de la forêt, Université Laval, Quebec, QC,
Canada, G1K 7P4 2NASA Goddard Space Flight
Center, Code 923, Greenbelt, MD 20771,
USA 3Department of Soil Science, University of
British Columbia, Vancouver, BC, Canada, V6T
1Z4 4Meteorological Service of Canada, Climate
Research Branch, 11 Innovation Blvd, Saskatoon,
SK, Canada, S7N 3H5
Biochemical changes in leaves during stress
events have been shown to result in changes in
spectral reflectance at 531 nm. These changes are
correlated with variations in canopy light-use
efficiency. The Photochemical Reflectance Index
(PRI) was developed to measure this phenomenon.
We calculated PRI from MODerate resolution
Imaging Spectroradiometer (MODIS) reflectance
data for cloud-free days between 2001 and 2003
for the Fluxnet-Canada Research Network (FCRN)
Old Aspen (OA) flux tower in Saskatchewan. The
flux and meteorological data from the tower
allowed us to calculate photosynthetic light-use
efficiency (LUE) at the time of MODIS overpasses.
A linear relationship was found between PRI and
LUE only when backscatter spectral data (minimal
shadowing) was used. The relationship was
stronger for top of the atmosphere reflectance
data (R20.76) than for data that had been
atmospherically corrected with MODIS-derived
aerosol optical depth values and the 6S
atmospheric correction model (R20.53). While
our analysis of MODIS-derived PRI did not capture
seasonal variations in LUE, it seemed to detect
inter-annual variations. An ability to reliably
estimate LUE from satellites would significantly
improve large-scale modeling of the carbon cycle.
NDVI from tower sensors at Old Aspen site (1200
PM local time)
Daily LUE at Old Aspen site (1200 PM local time)
Time series of NDVI for the 2001-2003 growing
seasons at the OA site showed that once the
canopy was fully-leaved, around day 160 (June 9),
NDVI stayed relatively stable throughout the
growing season. This reflects the absence of
major changes in the structural characteristics
of the canopy during this period. However, a time
series of instantaneous LUE, computed as Gross
Ecosystem Photosynthesis (GEP) / Absorbed
Photosynthetically Active Radiation (APAR)
derived from tower measurements for the same
period, showed that LUE was very dynamic and
responded to changes in environmental variables.
In the most extreme cases, four fold variations
in instantaneous LUE were observed. This makes it
difficult to adequately estimate ecosystem
productivity using LUE models driven by remotely
sensed inputs.
2001
LUE (µmol C µmol-1 APAR)
NDVI
2003
LUE (µmol C µmol-1 APAR)
Day of year
Day of year
For the period between June and August of 2001,
2002 and 2003, 52 clear days were identified
using radiation data from the Old Aspen tower.
For these days, MODIS data were downloaded and
processed (top-of-atmosphere reflectance, aerosol
optical thickness (AOT), geolocation, cloud mask,
and sun/sensor geometry). 30-minute periods of
GEP and PAR, corresponding to the MODIS
overpasses over the tower, were also downloaded
from the FCRN DIS. An atmospheric correction
model (6S) was used to obtain the surface
reflectance values for bands 11 (526-536 nm), 12
(546-556 nm), and 13 (662-672 nm). The PRI was
then calculated as where r11 is the reflectance
in MODIS band 11, containing the PRI signal at
531 nm, and rref is the reflectance in a
reference band unaffected by stress events, in
this case bands 12 or 13. To obtain positive
values only, PRI were scaled using Even though
the range of LUE values for cloud-free days is
narrow, a positive linear relationship (R20.76)
between LUE and sPRI was found when using only
backscatter reflectance and MODIS band 13 as the
reference band. Moreover, the LUE-sPRI
relationship was weakened when using
atmospherically corrected data (R20.53). This is
probably due, in part, to the uncertainty in the
MODIS-derived AOT.
Forward scatter
Backscatter
July 6, 2001
R2 0.02
R2 0.08
August 14, 2003
R2 0.76
R2 0.53
To verify whether the results from the
regression analysis were consistent with
information from a larger spatial scale, a
regional analysis of sPRI was performed over a
9900 km2 area. Two MODIS images from the
backscatter LUE-sPRI relationship (July 6, 2001
and August 14, 2001 see above) were
co-registered and overlaid on a 1994 Landsat 5 TM
physically based land cover classification. The
deciduous and medium-age deciduous regeneration
classes were then aggregated into a larger
deciduous class. MODIS pixels containing more
than 60 deciduous Landsat pixels were identified
and sPRI values for these pixels were extracted.
Frequency histograms of the sPRI values extracted
from the regional analysis show that, for MODIS
deciduous pixels in the 9900 km2 area, sPRI
values differed between the two dates. For July
6, 2001, the mean sPRI was 0.60 (s.d. 0.02)
while it was 0.55 (s.d. 0.01) for August 14,
2003. The red arrows on the histograms indicate
the classes where the sPRIs from the tower data
were located, for these respective days. Both
tower sPRIs fell inside the mean 1 s.d.
interval. These results indicate that sPRI
calculated for the OA site at a specific date was
representative of the sPRI of deciduous stands
across the landscape. Conclusions sPRI
calculated from MODIS reflectance in bands 11 and
13 (backscatter), without atmospheric correction,
showed a strong relationship with LUE from tower
measurements. sPRI calculated with band 12 as a
reference band did not show a relationship with
LUE. For this site, the sPRI failed to track
daily or seasonal variations in LUE, although it
captured inter-annual variations. To adequately
estimate LUE from space-borne remote sensing
platforms and use it for modeling of ecosystem
photosynthesis, further research is needed to
determine whether the sPRI is responding to
changes in canopy physiological function or
changes in structural properties and the extent
to which it is confounded with NDVI. The degree
to which such a relationship holds across the
Fluxnet-Canada transect needs to be determined as
well.
This research was supported by NSERC, CFCAS,
Biocap and NASA.