Relating Optical Indices to Carbon and Water Fluxes in a Chaparral Ecosystem

1
Relating Optical Indices to Carbon and Water
Fluxes in a Chaparral Ecosystem
Helen C. Claudio1 (hclaudi_at_calstatela.edu), John
A. Gamon1 (jgamon_at_calstatela.edu), Yufu Cheng1
(ycheng5_at_calstatela.edu), Daniel A. Sims2
(dsims_at_bsu.edu), Hongyan Luo3 (luo_at_sunstroke.sdsu.
edu), Walter Oechel3 (oechel_at_sunstroke.sdsu.edu) 1
California State Univeristy, Los Angeles,
Department of Biological Sciences 5151 State
University Drive, Los Angeles, CA 90032 2 Ball
State Univeristy, Department of Geography CL425
Ball State University, Muncie IN 47306 3 San
Diego State University, Global Change Research
Group Department of Biology, PS-240 San Diego
State University 5500 Campanile Drive, San Diego,
CA 92182
Abstract Between 2001 and 2003, spectral
reflectance coupled with CO2 and H2O flux data
were collected at Sky Oaks Biological Field
Station, a chaparral-dominated ecosystem in
southern California. The reflectance data were
collected by walking along a transect (early 2001
and after July 2003) and semi-automated 100 meter
tram system installed at the site (mid 2001 until
July 2003), while CO2 and H2O flux data were
gathered with an eddy covariance flux tower.
Over the study, which included a normal (2001),
an extremely dry (2002), and a recovery year
(2003), the water band index (WBI) was found to
be more closely correlated with ecosystem H2O
flux than with the CO2 flux. In the wet year,
WBI was more closely correlated with both the H2O
and CO2 fluxes, but when a record drought struck
in 2002, the correlation between WBI and CO2
disappeared as vegetation died. Also, WBI is
dynamic over time, precipitation conditions, and
between species in the region. These results
suggest that the ecosystem average WBI is an
overall more robust estimator of the H2O flux
than of the CO2 flux at the ecosystem level, and
water fluxes can be directly estimated from
optical remote sensing. However, the WBI region
is noise-prone and needs to be examined carefully.

• Discussion
• There was a significant statistical correlation
  between water band index (WBI) and the water
  flux. This is an indication that the ecosystem
  is indeed driven by the presence or lack of
  water.
• Sky Oaks is a particularly good site for
  correlating WBI and water flux because the data
  cover wide variations in precipitation, ranging
  from wet years to record drought years. However,
  there is limited vegetation in the region and the
  low productivity in the region makes the
  amplitude of the readings relatively small. This
  contrasts with the types of readings expected in
  a wetter environment readings with larger
  amplitudes but with less variation due to the
  lack of extremes.
• Canopy structure plays a significant role in
  interspecies WBI differences. Manzanita, for
  example, has larger, broader leaves than chamise
  or redshank, which both have small, needle-like
  leaves.
• WBI and CO2 flux do not always have a clear
  correlation because of the lead and lag effects.
  This model does not account for atmospheric
  effects or seasonal changes that affect the CO2
  exchange rates.
• The results indicate that WBI is an indicator of
  the ecosystem evapotranspiration rate. In
  addition to temporal dynamics and canopy
  structure effects, environmental factors such as
  relative humidity and temperature, and
  physiological effects such as stomatal
  conductance are also key factors in modelling
  evapotranspiration and CO2 flux.
• The water band region at 970nm is particularly
  sensitive to noise and background moisture.

Figure 4 Typical reflectance spectra for each
species (Af Adenostoma fasciculatum, Ap
Arctostaphylos pungens, As Adenostoma
sparsifolium). To the left are the pre-drought
reflectance spectra and to the right are the
drought spectra, taken December 2001 and July
2002, respectively. The wavelengths used for the
two reflectance indices, NDVI (normalized
difference reflectance index) and WBI (water band
index), are also shown.
Figure 5 Precipitation, water flux, and water
band index (WBI) over time from January 2001
until July 2003. Note the seasonal variations in
WBI, NDVI, and the water flux. Also note the low
NDVI and WBI during the drought period (summer
2002) compared to previous and following years.
There are also lead-lag effects between the
optical indices, evapotranspiration, and
precipitation. Also, note the difference in WBI
in 2001 before and after quality control checks
for poor data points.
Figure 1 A contour map of Sky Oaks (left) with
the location of the transect, flux ttower, and
footprint marked.
Figure 2 Eddy flux tower for taking CO2 and H2O
flux data.
Introduction The use of remote sensing allows for
ecosystem sampling that are normally
inaccessible. Remote sensing can range anywhere
from global scale such as from satellites or
large regions such as through aircrafts or small
(ecosystem level) regions through localized
manual and automatic sampling devices. This
particular study focuses on the ecosystem level
measurements in the 400-1000nm wavelength range.
Some commonly used indices include normalized
difference vegetation index (NDVI), photochemical
reflectance index (PRI), and water band index
(WBI). This study focuses on the 970 nm WBI,
which is a measure of a water absorption
feature. One of the primary purposes in this
study are to correlate water and carbon fluxes
with the various reflectance indices and use
these data to model fluxes. In addition to local
fluxes, this study is a part of a larger effort
to run cross-ecosystem analyses by a network
called SpecNet, a collaboration of investigators
combining optical and flux measurements to
understand ecosystem flux controls.
Figure 3 Instrumentation of the tram system
including the cart, white calibration panel,
track, and spectrometers equipped to take
reflectance while correcting for changing sky
conditions.
Figure 9 Transformed AVIRIS image to show WBI
over a larger space. Image was taken on October
3, 2002. Darker areas indicate areas of high
relative WBI, while lighter areas indicate low
WBI. Here, the riparian cooridor has been
highlighted to show the contrast in WBI.
Figure 10 A sample reflectance spectra (June 7,
2001) with an oxygen absorption feature at
760-765 nm and noise in the 930-1000 nm region.
The water absorption region is prone to noise and
this noise (zig-zags) confound WBI.
Figure 7 WBI of the overall transect average and
the three individual dominant species
(Adenostoma. fasciculatum, A. sparsifolium,
Arctostaphylos pungens) at Sky Oaks over time
from January 2001 until July 2003. Note the
inter-species differences in response,
particularly in 2003 A. fasciculatum and A.
sparsifolium show a greater extent of recovery
than does A. pungens.

• Conclusion
• The statistical correlation between the water
  band index (WBI) and water flux is significant,
  indicating optical remote sensing can provide
  direct information on water status and
  evapotranspiration. The correlation between WBI
  and CO2 flux, while weaker than between WBI and
  water flux, is still significant.
• While WBI can yield some information on
  evapotranspiration and CO2 exchange, there needs
  to be incorporation of other factors, such as
  structural, physiological, meteorological, and
  environmental factors, to generate more dynamic
  models.
• Ongoing studies at this and other sites in the
  SpecNet network will enable us to further
  evaluate the general applicability of WBI to
  evapotranspiration and carbon exchange.

Figure 6 Transect NDVI and WBI under wet and
drought conditions. Note that different species
have different peaks at different times due to
varying greenness (NDVI) and water status (WBI).
Individual shrubs appear as distinct, individual
peaks, especially in the case of Arctostaphylos.
Methodology The Field Site Sky Oaks is a
chaparral vegetated site located in northern San
Diego County and is part of a developing network
known as Spec-Net. The site of interest is a
site that had gone unburned for nearly a century
and then burned again in July 2003. Some
dominant plant species of particular interest are
chamise (Adenostoma fasciculatu), redshank (A.
sparsifolium) and manzanita (Arctostaphylos
pungens).. Methods CO2 and water flux data
were gathered using an eddy tower with an IRGA
CO2 detector and meteorlogical analyzers.
Reflectance data were taken by using a dual
detector spectrometer (UnispecDC, PP Systems,
Haverhill MA) capable of correcting for changing
sky conditions. The water band index (WBI), a
measure of vegetation water content, was
calculated using with the reflectance data the
formula is R900/R970, where Rx is the reflectance
at a given wavelength. Simultaneous sampling of
CO2 and H2O fluxes and reflectance from 2001 to
2003 allows for comparisons to observe the
dynamics of water status in the ecosystem.
Fluxes and indices were correlated with each
other, and the effects of precipitation and
species were also investigated. In addition, the
data were quality checked, particularly the NIR
(near-infrared region) for noise and other
difficulties.
Figure 8 Correlations between H2O (left) and CO2
(right) fluxes with transect average WBI
respectively. Note the stronger correlation
between WBI and H2O flux than with CO2 flux. For
the CO2 flux, negative values indicate uptake
into the ecosystem, while positive values
indicate loss into atmosphere.
Acknowledgements This study was funded by the
NCEAS, CEA-CREST and NSF Ecosystems grants. The
precipitation and flux data were provided by Dr.
Walter Oechels Global Change Research Group at
San Diego State University. I also want to thank
everyone in the VCSARS lab for giving me helpful
insights along the way and keeping my morale up
during difficult times.