Transcript Contrasting predictions of experimental and observational

Contrasting predictions of
experimental and observational
studies of the response of plant
communities to changing precipitation
Brody Sandel1 , Leah J. Goldstein2, Nathan Kraft1, Jordan Okie3, Michal
Shuldman1, David D. Ackerly1, Elsa Cleland4 and Katharine N. Suding2
(1)Department of Integrative Biology, UC Berkeley, Berkeley, CA
(2)Ecology and Evolutionary Biology, UC Irvine, Irvine, CA
(3)Department of Biology, University of New Mexico, Albuquerque, NM
(4)Ecology, Behavior & Evolution Section, UC San Diego, La Jolla, CA
Introduction
Plant responses to climate change

How will the composition of plant
assemblages respond to climate change?
Precipitation change Weltzin et al. 2003, Bioscience.
 Plant functional traits Suding et al. 2008, Glob. Change Biol.
 Experimental/observational Rustad 2006, Plant Ecol.

Introduction
Traits and climate change
Traits vary with climate
 Can they predict response to changing climate?
N:P

Wright et al. 2005, Glob. Ecol. Biogeogr.
Reich and Oleksyn 2004, PNAS
Introduction
Traits and climate change

Advantages of trait-based predictions
 Mechanistic
interpretations
 Allows syntheses
 Predictions are generalizable
Introduction
Experimental and observational

Similar predictions?
 Direction

and magnitude of effect
Shifts in functional trait composition are the
bases for comparison
Introduction
Mean trait value
Mean trait value
Similar predictions?
Control
+ Precip
Precipitation
Introduction
Similar predictions?
Mean trait value
Direction
Mean trait value

Control
+ Precip
Precipitation
Introduction
Similar predictions?
Direction
 Magnitude

=? ∆TO
Mean trait value
Mean trait value
 ∆TE
∆TE
Control
+ Precip
∆TO
Equivalent to
experiment
Precipitation
Introduction
Similar predictions?
Combining results
 Same
direction, different magnitude
 (My a priori expectation)
Mean trait value

Observational
gradient
C
T
C
C
T
Precipitation
T
Experimental
studies
Methods
Methods overview




Experimental water additions
Natural precipitation gradient
Match species lists to trait databases
Calculate plot mean trait values
 Test

for effects of increased water
Compare experimental and observational outcomes
 Direction
 Magnitude
Methods
Experimental data

Four water addition experiments
 Konza LTER (1991-2005) Knapp et al. 2001, Ecosystems.
 Shortgrass Steppe LTER (2000)
 Sevilleta LTER (2004-2006) Baez et al. In prep.
 Jasper Ridge Global Change Experiment (1999-2002)
et al. 2003, Ecol. Monogr.




Between 10% and 190% (mean 50%) precip.
increases
Plant community composition data
Grasslands or mixed grass-shrublands
219 species total
Zavaleta
Methods
Observational data
VegBank (vegbank.org)
 21,566 plots from across the country
 Plant assemblage of all plots
 7813 species total


Used PRISM climate data to obtain 30-year
mean precipitation values
Methods
Traits

Matched species lists to trait databases
 USDA Plants
 Kew
Gardens Seed Information Database
 Glopnet leaf traits Wright et al. 2004. Nature.
 More leaf traits Tjoelker et al. 2005, New Phyt.; Reich and Oleksyn
2004, PNAS.
 Height Cleland et al. 2008. Ecology.
Methods
Traits
Exp.
Nat. Grad.
Trait
Coverage
Coverage
LL
30%
21%
SLA
41%
34%
Nmass
42%
43%
Narea
40%
32%
Amass
38%
23%
Aarea
40%
23%
Seed
94%
80%
Form
100%
89%
Lifespan
98%
90%
Height
100%
Methods
Analyses

Abundance-weighted trait means for each plot
Percentage cover by a group
All analyses performed on these plot-level values

Experimental


 ANOVA using

last year of each study
Observational
 Aggregated cells at
 Linear
regression
1 x 1 degree resolution
Results
log(Seed mass (mg))
Seed size example
log(Precip (mm))
Results
log(Seed mass (mg))
Seed size example
log(Precip (mm))
Results
log(Seed mass (mg))
Seed size example
log(Precip (mm))
Results
log(Seed mass (mg))
Seed size example
log(Precip (mm))
Results
log(Seed mass (mg))
Seed size example
log(Precip (mm))
Results
Treatment effect
log(Seed mass (mg)) per log(Precip (mm))
Seed size example
Slopes of line
segments through
time
Year
Results
Summary of all traits
Experimental
Natural Gradient
Trait
Effect
P
Effect
r2
LL
-
0.0129
+
0.154
SLA
+
0.0297
NS
0.006
Nmass
NS †
0.1601
-
0.158
Narea
+
0.0003
-
0.309
Amass
Mixed †
0.0189
-
0.047
Aarea
NS †
0.3116
-
0.101
Seed
-
0.0071
+
0.362
Grass
NS †
0.0717
-
0.373
Forb
+
0.0091
-
0.066
Annual
-
<.00001
-
0.122
Short
-†
<.00001
† indicates a significant site by treatment interaction
Results
Summary of all traits
Experimental
Natural Gradient
Trait
Effect
P
Effect
r2
LL
-
0.0129
+
0.154
SLA
+
0.0297
NS
0.006
Nmass
NS †
0.1601
-
0.158
Narea
+
0.0003
-
0.309
Amass
Mixed †
0.0189
-
0.047
Aarea
NS †
0.3116
-
0.101
Seed
-
0.0071
+
0.362
Grass
NS †
0.0717
-
0.373
Forb
+
0.0091
-
0.066
Annual
-
<.00001
-
0.122
Short
-†
<.00001
† indicates a significant site by treatment interaction
Results
Summary of all traits
Experimental
Natural Gradient
Trait
Effect
P
Effect
r2
LL
-
0.0129
+
0.154
SLA
+
0.0297
NS
0.006
Nmass
NS †
0.1601
-
0.158
Narea
+
0.0003
-
0.309
Amass
Mixed †
0.0189
-
0.047
Aarea
NS †
0.3116
-
0.101
Seed
-
0.0071
+
0.362
Grass
NS †
0.0717
-
0.373
Forb
+
0.0091
-
0.066
Annual
-
<.00001
-
0.122
Short
-†
<.00001
† indicates a significant site by treatment interaction
Results
How will communities change?

Experimental studies
 Tall,
long-lived forbs with short leaf lifespans,
high leaf N concentrations, high specific leaf area,
and small seeds

Observational analysis
 Long-lived
woody species with long leaf
lifespans, low leaf N concentrations and
photosynthetic capacity, and large seeds
Discussion
Why the mismatch?

One is right, the other wrong
 Experimental
artifacts
 Unmeasured covariates

The different responses may reflect a real,
two-phased response to climate change
Discussion
A two-phase model

Response to climate change may occur over
distinct phases
 Why
two phases?
 Why might the responses in each phase differ?
 What determines the time scale of the phases?
Discussion
Two phases

Premise – Abundance changes happen more
quickly than species gain and loss
1 – Changes in local species abundance
 Phase 2 – Changes in species pool
 Phase

Calculating plot trait values not weighted by
abundance revealed fewer treatment effects
 Abundance
shifts were critical in experiments
Discussion
Two phases
Increased water
Phase 1 – Abundance changes
Time
Phase 2 – Species pool changes
Discussion
Phase differences

Why might the trait
responses differ in the
two phases?
 Changing
interactions
among species
 Shifts in the limiting
resource

The traits of local
species that increase are
not the same as those of
immigrating species
Discussion
Phase differences
Increased water
Increasing species are able to
take advantage of increased
resource availability
(tall, high leaf N, short-lived
leaves, small seeds)
Taller stature - light
limitation
Time
Species must cope with
low light environment
(woody, low leaf N,
long-lived leaves, large
seeds)
Discussion
Time scales

Little evidence for phase 2 in the experiments
 No
convergence through time towards observational results
 No treatment affect on species-time relationships
JRG
SEV
KNZ
Discussion
Time scales

What determines the length of phase 1?
 Spatial
extent of climate change
 Life histories of local species (annual/perennial)

At least decades in this case
 Lengthened
by experimental limitations
Discussion
Main messages




Traits useful predictors
Mismatch between experimental and observational
results
Could be due to different time scales captured by
these two types of study
Use the appropriate data to predict for a given time
scale
Acknowledgments






NCEAS, and the coordinators and participants in the distributed graduate
seminar
William Lauenroth
Alan Knapp
William Pockman
Erika Zavaleta
Funding –




NSF grant to NCEAS (EF-0553768)
UC Santa Barbara
LTER network office for cross-site research
NSF LTER program (DEB0218210, BSR 88-11906, DEB9411976, DEB0080529,
DEB0217774, DEB0217631)


David and Lucile Packard Foundation
Morgan Family Foundation
 Jasper Ridge Biological Preserve


The many VegBank contributors
Ian Wright and Peter Reich (Glopnet)
Discussion
A two-phase model