Continental, Landscape, and Ecosystem Scale Fluxes of CO2, CO, and other Greenhouse Gases: Constraining Ecosystem Processes from Leaf to Continent

1
Continental, Landscape, and Ecosystem Scale
Fluxes of CO2, CO, and other Greenhouse Gases
Constraining Ecosystem Processes from Leaf to
Continent (a.k.a. CO2 Budget Regional Aircraft
experiment-Maine COBRA-ME) . PIs Steven C.
Wofsy (Division of Engineering and Applied
Science and Department of Earth and Planetary
Science) Paul R. Moorcroft (Department of
Organismic and Evolutionary Biology), Harvard
University CO-Is David Hollinger (USFS) John
C. Lin (Harvard currently at Colorado State)
Christoph Gerbig (Harvard currently at MPI
Jena) Arlyn E. Andrews (NOAA CMDL) Collaborators
Prof. Maria Assuncaõ de Silva Dias, Saulo
Freitas,and Marcos Longo (Universidade de Saõ
Paulo and Centro de Previsão de Tempo e Estudos
Climáticos ( CPTEC), Brazil)
Synopsis Future concentrations of atmospheric
greenhouse gases will be strongly affected by the
rates at which terrestrial ecosystems add or
remove CO2,CH4, and CO from the atmosphere. Our
BE project is developing a framework to link
process-level biological knowledge that describes
individual plants and ecosystems for short time
scales, with observations and models that
characterize atmosphere-biosphere exchange for
large spatial domains and long time scales. Our
case study attempts to determine the sources and
sinks for CO2 and CO in Northern New England and
Quebec (4 million sq. km), using data that we
acquired from satellites, forest inventories, an
extensive campaign of aircraft observations of
trace gas concentrations in summer of 2004, and
tall and short flux towers. In summer, 2004, we
flew an instrumented aircraft (Wyoming King Air)
for 200 hours, spanning the growing season. To
synthesize these data across space and time
scales, we developed an integrated
ecosystem-atmosphere model by coupling the
Ecosystem Demography model and the BRAMS
mesoscale atmospheric simulation system
Moorcroft, 2003 Medvigy et al., 2005, to
capture slow and fast ecosystem processes with
accurate environmental forcing. Our goal is to
assimilate biological knowledge and diverse
atmospheric and ecological data generating an
integrate model satisfying complementary
constraints from atmospheric concentration and
flux data, and structure and growth data form
inventories, satellite data, and more. The
product, a model with realistic structure and
responses to environmental forcing, meeting all
relevant biotic and atmospheric constraints,
quantitatively links emergent properties of the
terrestrial biosphere-atmosphere system with the
underlying fundamental biological and physical
processes, for length scales from individual
trees to 2000 km and time scales from hours to
centuries. This new type of model will provide
new ways to estimate the carbon budget for areas
as large as continents, using observations from
remote sensing and atmospheric data, and, using
the Bayesian framework, these estimates will be
bracketed within defensible error bars. Moreover,
the model will also tell us why the budget
behaves as we infer, and the model can readily be
incorporated into an analysis of future carbon
budgets in a changing environment.
Aircraft and tower measurements of CO2
concentrations and fluxes
b)
PM
Altitude km ASL
Cumulative Distance km
Atmospheric adjoint (receptor-oriented inverse)
model STILT
11 June 2004
(above) CO2 data from aircraft on 11 June 2004.
The King Air picked up air measured on June 10,
executing two cross sections that moved with the
mean flow (see below). Mixed layer heights (left)
validated well, a key success for BRAMS. The
data quantitatively determine CO2 uptake by Maine
forests in a Lagrangian framework. (below) Fluxes
(upper panel) and concentrations (lower panel) of
CO2 on the 30m tower in Howland Forest and the
100m tower at Argyle, 10 miles away. Note the
increasing uptake (negative flux) between mid-
and late June (left, right respectively) seen in
both fluxes and in daytime gradients, also note
the close agreement in forest uptake of CO2 at
these sites. The tall tower data can be used to
determine CO2 fluxes over a large area (see STILT
panel of this poster).
STILT Stochastic Time-Reversed Lagrangian
Transport Model STILT uses the same assimilated
meteorological fields as the biosphere models
(ED-LSM, VPRM) to compute the influence
function (response ppm at the aircraft or
tower caused by unit flux at a given time
upstream. STILT resembles a trajectory model,
but accounts for mixing in the planetary boundary
layer, convective storms, etc.. STILT used
forecast winds to predict where to fly to best
observe atmosphere-biosphere exchange. It uses
analyzed winds to link atmosphere-biosphere
exchange from ED-LSM or VPRM to our aircraft and
tower data. The figures in this panel show
reconstruction of CO and CO2 data observed at
Harvard Forest, based on constrained flux models
from COBRA flights in 2000. Note the changing
influence area during the synoptic progression
from northerly to westerly flow.
Lin et al., 2003, 2004 Gerbig et al., 2003,
2004
TOP DOWN VPRM (Satellite) model of CO2 exchange
with the Biosphere
Weather and Climate Drivers of the Biosphere
Respiration ar (ßr (EVI-EVImin)/(EVImax-
EVImin))
BOTTOM UP ED-LSM Size and Age-Structured
Ecosystem Model (SEM)
Distribution of clouds in the study region(dashed
box) on two flight days, 10 and 11 June 2004. A
particularly favorable combination of fair
weather and stable flow was present.
Wscalar (1 LSWI)/(1 LSWImax) Pscalar (1
LSWI)/2 ltbudbreak to full canopygt
Adaptation of the Vegetation Photosynthesis Model
(VPM Xiao et al., 2004) and Respiration models.
(above) Model equations a, ar, br are fit
parameters, adjusted initially using tower flux
data for each vegetation type. These values are
used as priors for the atmospheric inverse
(Bayesian) problem. The same tower flux data are
are used to constrain the parameters of the ED
model. (below) VPM compared to Howland flux data
for summer, 2004. The left panel shows the fit
using parameters developed from 2003 data,
indicating that parameters are reproducible and
conservative. The right panel shows daily data
variability is driven by solar input.
Brazilian Regional Atmospheric Modelling System
(BRAMS) High-resolution, nested, mesoscale
meteorological model for data assimilation and
forecasting
Ecosystem Demography Model ED-- Ecological
Statistical Mechanics -- accurately captures the
behavior, including competition, mortality,
vegetation change, and carbon budget, of
individual-based ecosystem models (e.g. patch
models).
Tracks the dynamic, sub-grid scale heterogeneity
in canopy structure within grid cells
statistically (computes the PDF, p, ?pda1).
Mass Flux in the study region(left, from our RAMS
assimilation run), flight track center, and
surface temperatures right on 10 June 2004.
The flow traversed the study areas along a NW
track, the weather was cool and sunny over much
of the study area, but there were dense clouds
and rain over the southern boundary. The flights
sampled along the core of the main regional flow
on the first of the two flight days (center, 10th
of June), then examined cross sections moving
with the mean flow on the 11th of June.
ED model equations
Citations Gerbig, C., J. C. Lin, S. C. Wofsy,
B. C. Daube, A. E. Andrews, B. B. Stephens, P. S.
Bakwin, and C. A. Grainger, Towards constraining
regional scale fluxes of CO2 with atmospheric
observations over a continent 1. Observed
Spatial Variability, J. Geophys. Res. 108,. D24,
4756 (14 pp) , 2003 2. Analysis of COBRA-2000
data using a receptor oriented framework, J.
Geophys. Res. 108,. D24, 4757 (27 pp) ,
2003. Lin, J. C. , C. Gerbig, S.C. Wofsy, A.E.
Andrews, B.C. Daube, K.J. Davis, A. Grainger,
The Stochastic Time-Inverted Lagrangian Transport
Model (STILT) Quantitative analysis of surface
sources from atmospheric concentration data using
particle ensembles in a turbulent atmosphere, J.
Geophys. Res. 108, No. D16, 4493,
10.1029/2002JD003161, 2003. Lin, J. C., C.
Gerbig, S.C. Wofsy, B.C. Daube, An Empirical
Analysis of the Spatial Variability of
Atmospheric CO2 Implications for Space-borne
Sensors and Inverse Analyses, Geophys. Res. Lett.
31 (23) Art. No. L23104 DEC 2 2004. D. Medvigy,
P.R. Moorcroft, R. Avissar R.L. Walko (2005.
Mass conservation and atmospheric dynamics in the
Regional Atmospheric Modeling System (RAMS).
Environmental Fluid Mechanics 4 (in
press). P.R. Moorcroft (2003). Recent advances
in ecosystem-atmosphere interactions an
ecological perspective. Proceedings of the Royal
Society Series B, 2701215-1227. Xiao, XM
Zhang, QY Braswell, B Urbanski, S Boles, S
Wofsy, S Berrien, M Ojima, D. 2004. Modeling
gross primary production of temperate deciduous
broadleaf forest using satellite images and
climate data. REMOTE SENSING OF ENVIRONMENT 91
(2) 256-270.
-NEE gC m-2 d-1 Howland d Forest
Initial 2-Year (1995-1996) Model test using
Harvard Forest data and the Ecosystem
Demography-Land Surface model (ED-LSM). Harvard
Forest flux tower data were used to optimize
ED-LSM behavior in temperate mixed hardwood
forests -- an important forest type with the
region. The forest structure was initialized with
FIA-style ecological plot data. Then 3 parameters
were optimized m (relates stomatal conductance
to CO2 flux per unit leaf area), a leaf
respiration parameter, and a growth respiration
parameter (partitions fixed C to growth),
constrained against hourly/ monthly/yearly NEE,
hourly ET, and above-ground woody increment.
ECMWF ERA-40 provided environmental forcing, and
soil respiration was matched to night-time CO2
-flux. The model provides excellent simulation of
Harvard Forest data for time scales from hours to
a decade.

x