Decadal scale observations of carbon
exchange
Presented by J. William Munger
Contributions from J. Hadley, S. Urbanski, D. Medvigy, P.
Moorcroft, S. Wofsy and many students, post-docs, and
technicians over the past 20 years
Additional support from DoE Office of Science (TCP, NIGEC, NICCR)
1
Observational approaches
• Tower-based eddy covariance
– EMS tower (~100 yr old mixed oak stand) since 1990,
Hemlock site since 2000 LPH,
younger Oak stand, since 2002
• Plot-based biomass inventories, LAI, litter, CWD,
above ground woody increment (dbh>10 cm)
• Soil respiration
• Setting up mini-rhyzotrons for below-ground
observations
2
Analysis
• Convergence of atmospheric flux and
biometric approaches
• Focus on trends at process level as well as
net fluxes.
3
Long-term record, continuous, consistent NEE
from fall of 1991
HFEMS data
40 Continuous series
of hourly NEE
Missing invalid and
NEE mmole m-2s-1
low u* values filled
based on f(T,PAR),
fit to short intervals
Mean of residual
~0
Note the variability
in magnitude and
width of summer
peaks
-40
1992 1998 2007
4
NEE (gC m-2d-1)
• Conifer stand takes advantage of shoulder seasons
• Small differences between LPH and EMS due to stand age and
5
soil moisture conditions
Cumulative NEE sums 1992-2007
0 Calendar
year
NEE Mg-C ha-1
eco year
Accelerating
40 uptake!
6
Annual NEE sums (for eco years)
Mg-C ha-1y-1
0
1998
Short record looks constant
with some variance
NEE
-5
2001
16 GPP
Reco
10
7
1992-2004, showing magnitude of NEE
confidence intervals
Uncertainty < Anomalies
NEE = -1.28 + -0.146 x (yr-1990); R2 = 0.337
0
-1
NEE (Mg-C/ha/yr)
-2
-3
-4
-5
GEE = 11.1 + 0.363 x (yr-1990); R2 = 0.732
R = 9.82 + 0.217 x (yr-1990); R2 = 0.626
-1 x GEE
16 Resp
Mg-C/ha/yr
14
12
10
1992 1994 1996 1998 2000 2002 2004
Year
8
Consider trends separately for dormant and
active periods
Winter
• Recois similar
from year to year
• Growing season
dominates the
Summer interannual
variability
Empirical fit
Obs
1992 1998 2001 2007
9
Annual cumulative NEE (Mg-C ha-1y-1) illustrates
the seasonality, and differences among years
•Magnitude of
4 summer uptake not
highly variable, but
duration is
•Some fall/winter
periods have
0 enhanced Reco
1998
•Spring onset is
variable
-4 2007
NOV
10
2001
•Variability in timing of spring onset
11
Peak uptake years maintain high
NEE through end of September
NEE ~0 through Oct. instead of
positive values.
12
Comparison between NEE fluxes
and biomass growth Mg-C ha-1 y-1
•Peak AGWI in 2002 follows
high NEE year.
5
•General upward trend in
AGWI
4
3
4
2
1
0
1991 2008
13
Red oak dominates the biomass growth
20% more Above-ground biomass
14
Increasing trend of Annual NEE associated with longer growing season
122 days in 1997; 164 days in 2007
# of days NEE <0
1992 1998 2001
15
April - subcanopy has properties of a
conifer stand
June – Conifers are fully shaded:
Site acts as a deciduous stand
Phenocam images from
below canopy camera
16
Emerging question
• Is extended growing season due to earlier
leafout, or
• Increased contribution from subcanopy
conifers in the spring and fall
– New observation of above and below-canopy
greenness (Phenocam)
– Expanded sampling of understory biomass
– Understory light measurements
17
Light curves show interannual variability
JULY
10
0
-10
1992
-20
-30 2004
-40
0 500 1000 1500
18
Hemlock has reduced uptake at high light relative to deciduous stands 19
Mean GEE in June – Aug
at constant high light level 1200
•Increasing trend
consistent with
0
annual NEE trend
•Reduction in
photosynthetic
-10
capacity in 1998
-20
-30
-30
-40
1992 2006 2000 2004 2007
20
Annual cycles of LAI
• More foliage
• Lasting later in season
• Annual litter input increasing
(Mg-C ha-1y-1)
21
Modeling approaches
• Empirical model including phenology and
light and temperature responses
• Testing NULL hypothesis that ecosystem is
constant (ans., it is not)
– Year partitioned into 8 seasonal blocks
– Fit to all available valid data
– Overall R2 = .80 – weather accounts for hourly
signal
a PAR
NEE a1 + a 2 (T - T ) + 3
a + PAR
4
22
ED2 Ecosystem Model
• Structured canopy model
• Dynamic vegetation with Functional Types
• Physiological Process Parameters optimized
to 2 years of flux data and decadal-scale
biomass inventory data
– Biomass data essential to constrain the long-
time response processes, and estimate
parameters that are difficult to observe directly
(e.g. below-ground allocation)
• 10 years predicted using observed
meteorology and phenology 23
Observations vs Process model and Empirical Fit
Transition seasons have largest residual
1.0
NEE Mg-C ha-1m-1
-1.0
-2.0 ED2 is tracking variability
Empirical fit has R2 =0.8 on hourly
but fails to capture most of variability or trend
24
Flux towers are a focus for other
investigations
• Remote sensing validation
• Canopy structure
• Nitrogen dynamics
25
Conclusions
• Trends in carbon exchange at Harvard Forest are driven
by interspecies competition (Oak vs Maple)
• and successional shifts (rise in conifer subcanopy)
• Climate variability alone does not account for
interannual variability,
(poor skill by empirical fit - assumes constant response)
– Though climate interacting with the ecosystem is surely a
factor
• Photosynthetic efficiency increasing due to higher LAI
and shift in species distribution towards more efficient
oaks
• Ecosystem process model parameterized by a limited
data set is able to represent some of interannual
variability 26
Conclusions, cont’d
• Long-term records are key
– Quantifying the range of variability
– Distinguishing trends
– to capturing impacts of climate change and normal ecosystem
succession
• Capacity for sustained NEE at HFEMS looks promising
• Oak not yet at maximum lifespan
• Barring demise of hemlock
• Reproducing observations with models seems to require
consideration of shifting vegetation and constraint with
observations that account for a range of time scales
27
Key Citations
Medvigy, D., S. C. Wofsy, J. W. Munger, D. Y. Hollinger, and P. R.
Moorcroft (2009), Mechanistic scaling of ecosystem function and dynamics
in space and time: Ecosystem Demography model version 2, Journal of
Geophysical Research-Biogeosciences, 114, G01002, DOI:
10.1029/2008jg000812.
Urbanski, S., C. Barford, S. Wofsy, C. Kucharik, E. Pyle, J. Budney, D.
Fitzjarrald, M. Czikowsky, J. W. Munger, (2007) Factors Controlling CO2
Exchange at Harvard Forest on Hourly to Annual Time Scales, J. Geophys
Res., 112, G02020, doi:10.1029/2006JG000293.
Hadley, J. L., Kuzeja, P.S., Daley, M.J., Phillips, N.G., Mulcahy, T.,
Singh, S. (2008). Water use and carbon exchange of red oak- and eastern
hemlock-dominated forests in the northeastern U.S.: Implications for
ecosystem-level effects of the hemlock woolly adelgid. Tree Physiology 28:
614-627.
Hadley, J. L., Schedlbauer, J. L. (2002). Carbon Exchange of an Old-
Growth Eastern Hemlock (Tsuga canadensis) Forest in Central New
England. Tree Physiology 22: 1079-1092.
28