Paul R. MoorcroftDavid Medvigy, Stephen Wofsy, J. William Munger, M. Dietze

Harvard University

Developing a predictive science of the biosphere

Overview

• theoretical considerations: capturing the long-term, large scale response of heterogeneous plant canopies

• parameterization: connecting terrestrial biosphere models to field-based measurements of ecosystem composition, structure & function.

The rationale for structured biosphere models

Predicted collapse of Amazon ecosystems in response to rising CO2

(Cox et al. 2000)

carbon flux: land-air global mean temperature

- we now have models that make predictions for the long-term responses of terrestrial ecosystems to climate change.

- but are they predictive?

How close are we to a predictive science of the biosphere?(Moorcroft Tnds Ecol. & Evol. 2006)

- current generation terrestrial biosphere models are interesting, but essentially untested hypotheses for the effects of climate change on terrestrial ecosystems.

- developing accurate terrestrial biosphere models is challenging because the predictions of interest are at scales larger than those at which most measurements are made.

atm

osph

eric

CO

2 fla

sk s

ampl

es

satellite observations(leaf phenology, soil

moisture)

Can

opy

CO

2&

H2O

flu

xes

+ ec

osys

tem

exp

ts.

forest inventories(vegetation dynamics)

spatial scale

1m2 1000km2100km210km21km2 earth

decades

years

months

hours

time

scal

e

- as a result, scaling is a key issue

Aircraft measurementsof CO2 & H2O fluxes

(Moorcroft Tnds Ecol. & Evol. 2006)

Current terrestrial biosphere models are “big leaf” models

scale: 1o x 1o (~104 km2)

Time (yrs)

AG

B (k

gCm

2 )

xxxxx x

x xxx

x

xx

xxx

x

x

xxxx

xxxx

Comparison of above-ground biomass dynamics to observations at San Carlos (tropical forest) 2oN,68oW --->

- predict unrealistic long-term ecosystem dynamics

canopy physiology ecosystem dynamicse.g.: above-ground biomass dynamics of evergreen tree spp. in IBiS big leaf model

San Carlos successional dynamics

Stand Age (0-200yrs) -->

Species --->

AG

B (k

gCm

2 )

Time (yrs)

(Saldarriaga. et al 1988)

growth

height-structured competition

mortality

dispersal + recruitment

Individual-based vegetation models (gap models)(Moorcroft et al. 2001)

mortalitygrowth

~ 15 m

evapo-transpiration

leaf carbon fluxes

EcosystemDemographyModel (ED) dispersal+recruitment

waternitrogencarbon

ha (~10-2 km2)

ha (~10-2 km2)

ED dynamics at San Carlos Tropical forest (2oN,68oW): trajectory of above-ground biomass:

(Moorcroft et al. 2001)

ha (~10-2 km2)

“Ecological Statistical Mechanics”

of plant type i

(Moorcroft et al. 2001)

- accurately captures the behavior of corresponding individual-based model by tracking the dynamic horizontal & vertical sub-grid scaleheterogeneity in canopy structure.

- a size & age-structured terrestrial biosphere model

(recruitment is a lower boundary condition in z)

ED Model: Regional pattern of above-ground biomass (AGB) after 200 year simulation (kgCm-2)

(Moorcroft et al. 2001)

Time (yrs)

AG

B (k

gCm

-2)

AG

B (k

gCm

-2)

Incorporating land-use changeAlbani et al. 2006

historical fraction of agricultural land in each county 1800-2100

regional historical patterns of forest harvesting (USFS)

PrimaryVegetation

NaturalDisturbance(λ)

Agricultural

PrimaryVegetation

SecondaryVegetation

Land clearing

(λSA)

ForestHarvesting

(λPS)

NaturalDisturbance(λ)

NaturalDisturbance(λ)

ForestHarvesting(λSS)

(λPA)

(λAS)

Land abandonment

= [ ]

ED model: predicted impacts of land-use history on the carbon dynamics of the Eastern US

Albani et al. 2006

above gnd. biomass (tC ha-1) carbon uptake (NEP, tC ha-1 y-1) land use

in contrast to traditional ‘big-leaf’ models, structured biosphere models such as ED scale formally between fast-timescale plant-level physiological responses to climate and long-term, large-scale, ecosystem dynamics.

- have both realistic short-term and long-term vegetation dynamics.

(Moorcroft et al. 2001, Moorcroft 2003)

Answer: since structured biosphere models such as ED are formulated at scale individual plants, it should be possible to parameterize & test them against field-based measurements of ecosystem performance.

- incorporate the effects of anthropogenic sub-grid scale disturbances (land clearing, land abandonment and forest harvesting) on ecosystem composition, structure & function.

Enables them to:

Question: How do we know that the underlying model formulation & parameters are correct?

carb

on u

ptak

e(N

EE

tC h

a-1 y

-1)

Harvard Forest LTER ecosystem measurements

Initial model: traditional approach used in big-leaf models, in which model parameters are specified from values pulled from the literature.

Optimized model: a subset of the observations are used to estimate a number of important, but relatively unconstrained, model parameters.

Compare 2 formulations:

ED2 model fitting at Harvard Forest (42oN, -72oW)

> Can we demonstrate that the ED model’s ability to connect to field measurements improves its ability to predict long-term, large-scale ecosystem dynamics?

- initialize with observed stand structure

- model forced with climatology and radiation observed at Harvard Forest meteorological station.

ED2 biosphere model

Atmospheric Grid Cell

ED-2 model fitting at Harvard Forest (42oN, -72oW)

- 2 year model fit (1995 & 1996), in which model was constrained against:

- hourly, monthly and yearly GPP and Rtotal- hourly ET - above-ground growth & mortality of deciduous & coniferous trees

optimizedinitialobserved

= optimization period

Improved predictability at Harvard Forest: 10-yr simulations (1992-2001)

Net Carbon Fluxes (NEP)

Improved predictability at Harvard Forest: 10-yr patterns of tree growth and mortality (1992-2001)

observedinitial optimized

growth

mortality

= optimization period

GPP

respiration (ra + rh )

Improved predictability at Harvard Forest: 10-yr simulations (1992-2001)

observedinitial optimized = optimization period

conifers hardwoods m

orta

lity

grow

th

Vegetation model optimization: results

model parameters are generally well-constrained: average coefficient of variation: 17%

(= 95% confidence interval)

(-85, +160)

Change in goodness of fit: 450 log-likelihood (Δl) units (sig level: Δl= 20)

Harvard Forest

Summary: Harvard Forest: 10-yr simulations (1992-2001)

Demonstrated improved predictability in time.But what about in space?

HowlandForest

Harvard Forest

Howland Forest (45oN, -68o W)

Howland forest Composition:

growth

observedinitial optimized

net carbon fluxes (NEP)

(no changes in any of the model parameters)

Gross Primary Productivity (tC ha-1 mo-1 )

conifer basal area increment (tC ha-1 mo-1 )

hardwood basal area increment (tC ha-1 mo-1 )

Improved predictability at Howland Forest: 5-yr simulations (1996-2000)

=> model improvement is general, not site-specific

Conclusions: Developing a predictive science of the biosphere

capture observed regional scale variation in ecosystem dynamics without the need for site-specific parameters or tuning (scale accurately in space).

capture short-term & long-term vegetation dynamics of heterogeneous ecosystems (scale accurately in time).

The formal approach to scaling vegetation dynamics embodied in structured biosphere models such as ED2 them to:

shown that it is possible to develop terrestrial biosphere models that not only make predictions about the future of ecosystems but are also trulypredictive.

Future Directions

• Coupled modeling studies:- successfully coupled ED2 to a regional-scale atmospheric model (RAMS)- collaborations to implement structured terrestrial biosphere models in earth-system models used to predict future global climate change

• Applying this approach in other ecosystems:

- expanding from the regional to continental scale for N. Am ecosystems

- ecosystem dynamics and ecosystem-atmosphere feedbacks in Amazonia

Biosphere-atmosphere feedbacks Amazonia(Cox et al 2000)

In reality, deforestation is happening at sub-grid scales.

Amazonian deforestation predicted to change South American climate

(Shukla et al 1990)

Change in Annual Precipitation (mm)

Santarem Flux tower(3oS, -55oW)

Forest Inventory:

Collaborators: Steve Wofsy, Bill Munger, Dan Matross, Christoph Gerbig, John Lin, Roni Avissar, Bob Walko

Lab: Marco Albani, David Medvigy, Heather Lynch, Takeshi Ise, Daniel Lipsitt,Oliver Soong

Acknowledgements

Funding: National Science Foundation, Department of Energy (NIGEC/NICCR).

References:Moorcroft et a l. 2001. Ecological Monographs, 74:557-586. Hurtt et al. 2002. PNAS, 99:1389-1394.Moorcroft 2003. Proc. Roy. Soc. Ser. B, 270:1215-1227Medvigy et al. 2005. Env. Fluid Mechanics 4 Albani & Moorcroft (2005) Global Change Biology (accepted).Moorcroft (2006) Trends in Ecology and Evolution (in press)Medvigy & Moorcroft (2006) (in prep).