Coupling carbon, water, energy and nutrients in ecosystem models Physical climate system
Climate
change
Stratospheric
chemistry/dynamics
Volcanoes
External forcing
Sun
Atmospheric physics/dynamics
Ocean dynamics
Terrestrial
energy/moisture
Soil
Global moisture
Marine
biogeochemistry
Terrestrial
ecosystems
Tropospheric chemistry
Biogeochemical cycles
Human
activities
CO2
Land
use
CO2
Pollutants
Shameless Commerce Moment: Climate and Ecosystems David Schimel 2013 Princeton University Press Structure of an Earth System Model Atmospheric general circulation model
CO2
Land model
Vegetation and soil
Land-ice model
Atmospheric chemistry
aerosols
Ocean
circulation model
Marine
ecosystems
Sea-ice
model
Climate model components
Earth-system model components
ElevaCon and Depth Structure Ecosystems Tundra
Alpine forest
High-elevation forest
Mid-elevation forest
Woodland
Grassland
Coastal/
intertidal
Pelagic
Euphotic
Bathyal
Abyssal
Hadal
Ecosystem structure and climate A central paradox •  Water and energy variables parsimoniously explain global paGerns of terrestrial vegetaCon •  Most ecosystems respond to nutrient addiCons, suggesCng nutrients as a primary constraint •  Does water/energy or nutrient availability beGer explain terrestrial vegetaCon paGerns and producCvity? Water, energy and nitrogen coupling Global steady-­‐state paGerns of carbon, water and energy Similar paGerns across models: a blast from the past Schimel, D.S., VEMAP ParCcipants, and B.H. Braswell. 1997. SpaCal variability in ecosystem processes at the conCnental scale: models, data, and the role of disturbance. Ecological Monographs 67: 251-­‐271. Steady state to transient: nutrients are key controls ProducCvity will change at the rate of the slowest-­‐
adjusCng variable (this is a pointer to Ben’s talk). Del Grosso et al. 23
1
Figure 1
2
3
2000
a)
4
1500
Tree
Non-tree
Miami model
Schuur model
NCEAS model tree
NCEAS model non-tree
2000
5
1500
500
0
0
2000
4000
6000
8
-2
7
TNPP gC m yr
-1
6
1000
9
1000
500
10
0
11
0
500
12
b)
13
14
2000
Tree
Miami model
Schuur model
1000
MAP mm
1500
2000
2500
Biological N fixaCon tracks water and energy availability 639
CLEVELANDEI'AL.:GLOBALPAITERNSOF1ERRESlRIALNFIXATION
70
,-..
......,
60
• Central N Fixation 2
[f(x)=0.234x-0.172; R =0.63]
I-<
•
;>-
......,
ro
50
Z
40
OJ:)
30
•
20
•
,.!od
'-"
I:::::
.....
.....
0
•
Century-derived N Fixation
0
,.J:::l
•
•
••
•
@
Z
••
10
0
0
0
0
0
0
o
0
20
40
60
0
0
0
80
0
100
o
OJ
120
0
140
160
ET (em yr- 1)
Figure 1. Conservative, central, and upper-bound data-based and modeled estimates of terrestrial biological
nitrogen fixation by ecosystem plotted versus modeled ecosystem ET.
abundant in some parts and/or successional stages of all
ecosystems, we believe that it is extremely unlikely that this
value represents a realistic average spatial and temporal
coverage. The contribution suggested by the conservative and
central data-based estimates (100 Tg N yr'! and 195 Tg N yr'! ,
central value is based on an average of explicitly reported
coverages of N fixers, our best estimate of potential N fixation
in natural ecosystems globally is 195 Tg N yr'!, or an average
of -15 kg N yr'! for each hectare of the Earth's land surface.
Our failure to include an error term on this central value reflects
___
210
Mapped limiCng factors from Churkina and Running (1998) G. Churkina and S. W. Running
Dominant Environmental Controls on Net Primary Productivity
Climatic Controls and Vegetation Productivity
Annual NPP in Climate Space
0
3)
>
Figure 3. Annual global net
primary productivity (NPP)
simulated by BIOME-BGE
io climate space represented
by mean annual tempera
ture and water balance coef
ficient. Each data point rep
resents one 0.5° X 0.5° grid
cell.
0
010
,01.”0
E
211
E
C
3)
C.)
)
U)
0
C)
3)
C)
C
0
0
:3
9
C
-o
=
U)
a
:3
S
.E
9
ts
a
0
0)
,
low
high
water limitation
—20
—10
p
‘
Figure 2. Map of
weighted climatic
trols on net prima
ductivity determin
from water availa
average temperatu
cloudiness. Each d
point represents 3
of the membershi
tions based on ann
mean temperature
balance coefficien
percentage of sun
hours per year in
0.5°)< 0.5° grid ce
9
‘0
-j
low
high
watar limitation
0
10 Plants may
20 experience30water stress at a
processes.
annuci mean temperature (dec 0)
a
at
low
hi.,..
radiation limitation
water limitation on NPP (Figure 1) clearly
site with sufficient precipitation because a high
eated sites with available water deficiencies
fraction of incoming water has been intercepted and
excesses.
then evaporated by canopies. Consequently, NPP is
Radiation is another important environ
Table 1. Biome Types and Environmental Controls on Their Productivity
controlled not by the amount of precipitation, but
control on NPP, because photosynthesis occu
by the water available to plants.
in environments with a sufficient amount
No
To estimate the amount of available water, we
clouds can dramatically reduce the
Although
Water
Radiation Climate
Environmental
coefficient (WBC) as a
water balance
computed aReduction
of incoming photosynthetically active ra
Biome Type
Temperature Availability
Limitation Controls
difference between mean annual precipitation and
plants still photosynthesize on a cloudy day b
3 grassland
C
76%
21.5%
1.5%
1%
Churkina, S. water
W.
(G. temperature,
potential evapotranspiration
diffuse radiation, but at lower rates. Th
grassland
4
C
<1%
99%
<1%
watermanuscript),
unpublished
Running, and0%A. Schloss
assumed that cloudiness considerably redu
Deciduous broadleaf forest
1 l%
64%
8% evapotranspiration
17%
water,
controls than climate
was other
a function
where potential
coming solar radiation and NIP in areas w
Deciduous needle leaf forest 80%
4.5%
8.5%
7%
water
net solartemperature,
radiation (Priest
and
of mean temperature
of sunshine hours per year. Ve
percentages
Evergreen broad leaf forest
4%
31%
12%1972). WBC
53% computation
other controls
than climate, water
was based
ley and Taylor
was not limited by radiation i
productivity
Evergreen needle leaf forest 74%
4%
16% of annual
6%precipitation,
temperature,
radiation limitation
solar radia
on global means
or with negligible cloud
clouds
without
Shrubland/Desert
4.5%
<1%
95%
<1%
water database.
from the CLIMATE
tion, and temperature
Cloudiness data, expressed in sunshine ho
This water balance computation has the advantage
year, were derived from the CLIMATE datab
of being independent of any models and can be
According to the aforementioned assum
derived purely from climate data. To develop the
mapped
weighted climatic controls (te
we
To analyze the relationships between environmen
limiting
productivity
the
majority
of
of
biome
grid
relationship between NPP limitation and WBCs, we
and radiation) on N
availability,
tal controls and productivity of different vegetation
water
ture,
cells
was
defined
dominant
as
a
environmental
followed a logic similar to the one suggested by
biomes, we determined the relative importance of
of
the
globe (Figure 2). Th
surfaces
the
land
control
on
biome
productivity
(Table
1).
The
grid
Stephenson (1990). Stephenson showed the impor
each climatic factor to biome productivity. First, we
of each 0.5° >< 0.5° grid cell on this map is a
cells where the values of all 3 membership functions
Extending Churkina and Running: Global nutrient limita>on in terrestrial vegeta>on Global Biogeochemical Cycles Volume 26, Issue 3, GB3007, 3 AUG 2012 DOI: 10.1029/2011GB004252 hGp://onlinelibrary.wiley.com/doi/10.1029/2011GB004252/full#gbc1905-­‐fig-­‐0001 Global nutrient limita>on in terrestrial vegeta>on ValidaCon? Global Biogeochemical Cycles Volume 26, Issue 3, GB3007, 3 AUG 2012 DOI: 10.1029/2011GB004252 hGp://onlinelibrary.wiley.com/doi/10.1029/2011GB004252/full#gbc1905-­‐fig-­‐0003 Global nutrient limita>on in terrestrial vegeta>on Global Biogeochemical Cycles Volume 26, Issue 3, GB3007, 3 AUG 2012 DOI: 10.1029/2011GB004252 hGp://onlinelibrary.wiley.com/doi/10.1029/2011GB004252/full#gbc1905-­‐fig-­‐0002 Theory for nutrient limited adjustment rate The rate of change of carbon depends on the rate of change in available nutrients (not total but related). What controls available nitrogen (nutrients)? First, the fracCon of total nutrient that is available: this fracCon may change as water and energy availability change: fNavail Second, the total amount of nutrient, noCng water and energy affect inputs and losses: Ntotal dfNa/dt =F(Ntotal) where F, inputs and losses all depend on energy and water, anf F is the fracConal availability of a total nutrient Coupled carbon-­‐nutrient modeling •  CriCcal to simulate nutrient cycles (mineralizaCon, immobilizaCon, uptake). •  But, equally important to simulate budgets (inputs, losses, mineral phases). •  The moCvaCon for modeling budgets has ohen been climate, air or water quality issues, but emerging as important for the carbon cycle. Evidence for dynamic adjustment of total and available nutrients. •  When producCvity decreases, N losses increase. •  When plant uptake declines, precipitaCon of P into insoluble mineral forms increases. •  Long-­‐term changes in nutrient budgets correlate with carbon stocks (eg, Hawai’i) and nutrient parCConing, resource use efficiency. •  Microbial communiCes respond to changes in the relaCve availability of energy and nutrients, affecCng the acCve/total parCConing. Modeling nutrient inputs and losses •  A game changer for modeling… •  Process-­‐level modeling of A and R requires modest site-­‐ and landscape-­‐specific processes. •  Process modeling of local nutrient budgets requires detailed soil physics and local hydrology. •  Scale becomes a major issue, with dependence on soil pores, wetlands, hillslopes, even herbivores. Approaches •  Simple (Schimel et al 1997) parameterizaCon of inputs and losses and environmental funcCons. •  Complex models of each flux process as in DayCent or DNDC… •  Different biomes and regions have different dominant processes (BNF vs deposiCon, fire versus leaching). PARTONET AL.: MODEL FORNOx AND N20 EMISSIONS
17,407
a)1_
;
0.8
Example control funcCons: -,
ß
/ •
,,
_,•
Tmax=20
',, - Tmax=27
0.6 / / / •'!• ',,:
.....
Trnax=35
.;
0.4
"x
0.2
\',.,
0 -,-::: .............
0
• .......
10
20
',
• :-
,-- ._._:-_:--.._.•_:•__•__:.._•_____•
30
Soil Ternp
40
50
60
b)1
0.8
complex but these are the easy part 0.6
##
0.4
coarse
*
,,
0.2
,'
fine
......
volcanic
i
0.2
I
i
0.4 WFPS O.6
_
0.2
3
i
i
i
i
i
i
1
4
5
6
7
8
9
10
pH
Figure3. Effectof (a) soilwater-filled
porespace,(b) soiltemperature,
and(c) soilpH onnitrification.
Thereare generalpatternsof decreasing
NOx to N20 ratios
To considerbothWFPS and the constraints
on soil gas
with increasing WFPS and decreasingD/Do values but diffusivityrepresented
by D/Do,we grouped
thedatafor each
NOx:N20for the aggregated
datafrom all of the soilsis quite site into classes based on WFPS, and then derived the
17,410
Modeling soil water PARTON
ETAL.:MODEL
FORNOx ANDN20EMISSIONS
a)
b)
Soil Water Apr-Oct
r2=0.64
Soil Water
Nov-Mar
r2=0.27
ß
ß
60
ß
ßß
ß
E
'ooe._e
cA 40 -
coo ß
ß
ß
ß ß
eoee!A
• ee e
q•
ßt
ß
•e ß ee ß
e ß ee ß
Oebee.•.e
•
ß
ß
20-
0
0
20
40
WFPSobs
c)60_
60
80 0
Soil WFPS
20
40
WFPSobs
60
80
1992
ß
50-
observed
simulated
30 ß ßß
2O
10
ß ee
ß
ß
0
1/1
,
I
I
I
3/31
6/29
9/27
12/26
Figure5. Comparison
ofsimulated
versus
measured
soilwater-filled
porespace
(WFPS)
forallthesites
(a)during
thegrowingseason
and(b) duringthewintermonths.
(c) Dailycomparison
of simulated
andobserved
WFPS for
the sandyloamsiteduring1992.
1998]. Overall, with the exceptionof the winter soil water
observed
andsimulated
totalmineralN (NO3'+ NH4+).For
this analysiswe defined seasonsas winter (Decemberand temperature
parameters
that are usedas inputsinto the February),
spring(March-May),
summer
(June-August),
and
dynamics,the DAYCENT model simulateswell the soil water
tracegasflux submodel(Figures5 and6).
autumn (September-November).The model underestimated
5.2. SoilNH4+and NO3'Concentrations
mineralN concentrations
in all of the soilsbut correctly
simulated
highermineralN valuesin theSLFsoilcompared
to thenativeSL soil.However,themodelsimulated
higher
The comparisonof observedversussimulatedsoil NH4+
andNO3-concentrations
is limitedby the representativenessvaluesin the SCLF comparedto thenativeSCL, but the data
of the samplesand concentrationsnear the detectionlimit. showed no significant differences in the mineral N
Measurements
of soil NO3- and NH4+ are spatiallyand concentrations of these soils. The model and data both
temporallyvariable.Long-termexperiments
requirea small showedlower mineral N values in the native coarse-textured
PARTON
ETAL.-MODELFORNOx ANDN20EMISSIONS
17,415
_
ß
•
_
Model evaluaCon is event-­‐scale since there is no integral constraint. observed
simulated
_
o
-90
Apr-90
dul-90
Oct-90
dan-91
91d•l-91
Oct-9•
'd•n-;•
'Ap;-92
dul'Oct-9•
Jar
-1
_
5
0
Apr-g4
Jut-g4
O0t-94
Jan-g5
Apr-g5
Jul-g5
O0t-95
Jar
-1 -
7
6
0
Ja•
-1
-
,.•-
' '11i?TT ......... t'l• •
' + lTt
' Apr
-c
.)7
-96 Apr-96 Jul-•)6 Oct_96•
l + Jan-97
-
Jul-97 Oct-97
Figure
9. Comparison
ofobserved
and
simulated
daily
N20fluxes
forthenative
sandy
loam
site.
Error
bars
represent
standard
deviations
ofmeasurement
repetitions
(n=4);
standard
deviations
were
notavailable
forall
measurements.
•  N budgets are controlled by different factors in different environments (leaching, fire, trace gases, BNF, deposiCon), most of which remain incomplete in global models •  Diagnosis is needed to determine a “minimal” model for coupled C-­‐nutrient cycles in the climate system (Wait for Ben’s talks) to determine the nutrient-­‐limited rate of carbon change.