Impacts of including trait variation on predictions of global carbon fluxes and vegetation distribution
Peter van BodegomDepartment of Systems Ecology
VU University AmsterdamThe Netherlands
Traits vary considerably within and between communities
Kattge et al. 2011 GCB
Traits also tend to respond to climate manipulations
Cornelissen et al. 2007 EcolLett; Aerts, van Bodegom and Cornelissen 2012 New Phytol
incr
ease
d C
O2
incr
ease
d U
V
incr
ease
d ra
infa
ll
war
min
g
defo
liatio
n
shad
ing
ferti
lizat
ion
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
low altitude incubation
High altitude incubation
global change treatment
rel.
incr
ease
in d
ecom
posi
tion
rate
sup
on c
limat
e m
anip
ulat
ion *
** *
*
*
This trait variation is not well-captured by classification in biomes/PFTs
Van Bodegom et al. 2012 GEB
Keeping trait values PFT-constant within DGVMs has various disadvantages
Van Bodegom et al. 2012 GEB
Constant trait values in modelling hampers:1. including acclimation and adaptation processes2. Accounting for non-random species turnover3. Quantifying vegetation-environment feedbacks
For these reasons, trait variation/responses are increasingly incorporated into the DGVMs
A partial solution: incorporation of observation-driven trait/process estimates
Brovkin et al. 2012 BiogeosciencesGlobal litter stocks
Cornwell et al. 2008 EcolLett
Potential solutions for further incorporating trait responses/ranges into DGVMs
1. more PFTS2. Incorporating variation within PFTs:
a. Based on habitat filtering principlesb. Based on evolutionary principles
3. Fully traits-based approach
1. More PFTs may not be a fruitful approach given functional redundancy
observedpr
edic
ted
2a. Incorporation of trait variation within PFTs: habitat filtering principles
Based on assembly theory: environment acts a ‘filter’
Filtering by environment
Potential range of trait values
Trait range in habitat 1
Trait range in habitat 2
Ordonez et al. 2009 GEB
2a. Incorporation of habitat filtering principles into JSBACH
For PFT 1: trait X = a * temperature + b * radiation + CO2-acclimation
1 2 3 4 534
38
42
46
YearSL
A (m
2 / k
g C)
1 2 3 4 534
38
42
46
Year
SLA
(m2
/ kg
C)
C3-grasses
Default: fixed traits variable traits responses
C3-grasses
Verheijen et al. 2012 Biogeosci.Disc.
2a. JSBACH-simulated trait variation based on habitat filtering
Red dots: fixed values from default setting
Verheijen et al. 2012 Biogeosci.Disc.
2a. Impacts of JSBACH-simulated trait variation on productivity
Verheijen et al. 2012 Biogeosci.Disc.
2a. Impacts of JSBACH-simulated trait variation on vegetation distribution
default
variable traits
2a. Impacts of JSBACH-simulated trait variation on future carbon sink
Verheijen et al. 2012 in prep.
2b. Incorporation of trait variation based on evolutionary principles
Van Bodegom & Franklin in prep.
GPP
(gC
m-2
year
-1)
Latitude
N:C & allocation (no root. comp)N:C & allocation (with root. comp.)Allocation (with root. comp.)N:C
Effects included
- Forest stand model- No water limitation- Maximizing net growth & reproduction
2b. Incorporation of trait variation based on evolutionary principles: on-site evaluation of
variable allocation
Van Bodegom & Franklin in prep.
Prod
uctiv
ity (k
g C ha
-1ye
ar-1 )
0.06 0.08 0.1 0.12 0.140
100
200
300
400
500
All xx C0 5 1 5All2 xx C0 1.09 5 1 5AffR2 xx C0 1.03 1 5
xx
Soil N availability (max gN g root C -1year-1)
0.06 0.08 0.1 0.12 0.140
50
100
150
200
All xx C0 5 1 0All xx C0 5 1 1All xx C0 5 1 2AffR2 xx C0 1.03 1 0AffR2 xx C0 1 1 1AffR2 xx C0 1 1 2
xx
stem
fine-roots
foliage
a b
Fixed allocationOptimal allocation
3. A fully traits-based approach: separating trait predictions from vegetation distribution
predictions
Douma et al 2012 EcographyVan Bodegom et al. in revision
3. Trait predictions based on trait-environment relationships
Van Bodegom et al. in revision
3. Predicting vegetation probabilities from traits: kernel density fitting
Douma et al 2012 Ecography Van Bodegom et al. in revision
SSD
LMASeed mass
Biome A
Biome B
Biome C
Seed massLMA
SSD
For each position in trait space, multiple plant functional types may in principle be possible. The probability of each is described by Gaussian kernels
3. Predicting vegetation probabilities from traits: global vegetation distribution
Van Bodegom et al. in revision
Conclusions
Trait responses to climate (manipulations) are important and strong impact (predictions of ) vegetation distribution and functioning.
There are multiple ways to continue refining DGVMs.
Exchange of ideas between modellers and experimentalists will remain essential for more reliable predictions of our future climate.
Douma et al 2012b Ecography