mountain forest ecology seminar; jan. 22 nd 2004 niklaus e. zimmermann biome-bgc: calibration and...

Mountain Forest Ecology Seminar; Jan. 22nd 2004 Niklaus E. ZimmermannBiome-BGC: Calibration and Sensitivity Analyses

Adjusting a global ecosystem model for local performance

– calibration and sensitivity analysis of BIOME-BGC in

Switzerland

N.E. ZimmermannWSL


Niklaus E. Zimmermann1, Peter. E. Thornton2, Matt Jolly3, Paolo Cherubini4, Felix Kienast1, Norbert Kräuchi4, Otto Wildi1, Matthias Dobbertin4, Werner Schoch1, Marcus

Schaub4, Luzi Bernhard4

1Swiss Federal Research Institute WSL, Division of Landscape, CH-8903 Birmensdorf, Switzerland;

2Climate & Global Dynamics Division, National Center for Atmospheric Research, Boulder, CO, USA; 3NTSG lab, School of Forestry, University of Montana, Missoula, MT, USA;

4Swiss Federal Research Institute WSL, Division of Forest, CH-8903 Birmensdorf, Switzerland

Scaling carbon fluxes from stands to landscapes: calibrating and testing BIOME-BGC along multiple

environmental gradients


The Biome-BGC Terrestrial Ecosystem Process Model

Biome = an area characterized by its flora, fauna, and climateBGC = BioGeochemical Cycles

The model uses a daily time-step. Flux (and pools) are estimated for a one-day period. Between days, the program updates its memory of the mass stored in different components of the vegetation, litter, and soil.

Weather is the most important control on vegetation processes. Flux estimates in Biome-BGC depend strongly on daily weather conditions. Model behavior over time depends on the history of these weather conditions, the climate.

• New leaf growth and old leaf litterfall • Sunlight interception by leaves, and penetration to the ground• Precipitation routing to leaves and soil• Snow accumulation and melting• Drainage and runoff of soil water • Evaporation of water from soil and wet leaves• Transpiration of soil water through leaf stomata• Photosynthetic fixation of carbon from CO2 in the air• Uptake of nitrogen from the soil • Distribution of carbon and nitrogen to growing plant parts• Decomposition of fresh plant litter and old soil organic matter• Plant mortality• Fire

Biome-BGC is a computer program that estimates fluxes and storage of energy, water, carbon, and nitrogen for the vegetation and soil components of terrestrial ecosystems. It is a process model because its algorithms represent physical and biological processes that control fluxes of energy and mass. These processes include:


Biome: required meteorological inputs

• Daily maximum temperature (°C) • Daily minimum temperature (°C)• Daylight average temperature (°C)• Daily total precipitation (cm)• Daylight average partial pressure of water vapor (Pa)• Daylight average shortwave radiant flux density (W/m2)• Daylength (s)

In many cases, the only data available for a particular site are daily Tmin/Tmax and Prec. The model MT-CLIM can then be used to derive estimates of the other required meteorological parameters.

Biome: required soils inputs: % clay, silt, sand


Site Characteristics• Meteorological inputs

• Maximum temperature• Minimum temperature• Precipitation• Radiation• Humidity/ VPD

• Soil characteristics

Ecophysiological Parameters• Biome properties• Leaf/Litter/Wood chemistry & characteristics (C:N, SLA, lignin,

cellulose, etc.)• Photosynthetic characteristics• Allocation rules• Turnover rates• etc.

BIOME-BGC

BIOME-BGC inputBIOME-BGC input


1:(flag) 1 = WOODY 0 = NON-WOODY 2:(flag) 1 = EVERGREEN 0 = DECIDUOUS 3:(flag) 1 = C3 PSN 0 = C4 PSN 4:(flag) 1 = MODEL PHENOLOGY 0 = USER-SPECIFIED PHENOLOGY 5:(yday) yearday to start new growth (when phenology flag = 0) 6:(yday) yearday to end litterfall (when phenology flag = 0) 7:(prop.) transfer growth period as fraction of growing season 8:(prop.) litterfall as fraction of growing season 9:(1/yr) annual leaf and fine root turnover fraction 10:(1/yr) annual live wood turnover fraction 11:(1/yr) annual whole-plant mortality fraction 12:(1/yr) annual fire mortality fraction 13:(ratio) (ALLOCATION) new fine root C : new leaf C 14:(ratio) (ALLOCATION) new stem C : new leaf C 15:(ratio) (ALLOCATION) new live wood C : new total wood C 16:(ratio) (ALLOCATION) new croot C : new stem C 17:(prop.) (ALLOCATION) current growth proportion 18:(kgC/kgN) C:N of leaves 19:(kgC/kgN) C:N of leaf litter, after retranslocation 20:(kgC/kgN) C:N of fine roots 21:(kgC/kgN) C:N of live wood 22:(kgC/kgN) C:N of dead wood 23:(DIM) leaf litter labile proportion 24:(DIM) leaf litter cellulose proportion 25:(DIM) leaf litter lignin proportion 26:(DIM) fine root labile proportion 27:(DIM) fine root cellulose proportion 28:(DIM) fine root lignin proportion 29:(DIM) dead wood cellulose proportion 30:(DIM) dead wood lignin proportion 31:(1/LAI/d) canopy water interception coefficient 32:(DIM) canopy light extinction coefficient 33:(DIM) all-sided to projected leaf area ratio 34:(m2/kgC) canopy average specific leaf area (projected area basis) 35:(DIM) ratio of shaded SLA:sunlit SLA 36:(DIM) fraction of leaf N in Rubisco 37:(m/s) maximum stomatal conductance (projected area basis) 38:(m/s) cuticular conductance (projected area basis) 39:(m/s) boundary layer conductance (projected area basis) 40:(MPa) leaf water potential: start of conductance reduction 41:(MPa) leaf water potential: complete conductance reduction 42:(Pa) vapor pressure deficit: start of conductance reduction 43:(Pa) vapor pressure deficit: complete conductance reduction

Biome propertiesPhenology info

Turnover infoMortality info

Allocation rules

C:N ratios

Decomposition info

Canopy/leaf info

Leaf physiology


BIOME-BGC outputBIOME-BGC output


NP

P

[kgC

/m2 ]

Net Primary Productivity for Othmarsingen, 1931-2001

<



NP

P

[kgC

/m2]

Net Primary Productivity for Othmarsingen, 1931-2001

Net

Pri

mar

y P

rodu

ctiv

ity (

NP

P)

[kg

C/m

2 ]

Net Primary Productivity for Missoula, Montana 1994



<



Net Primary Productivity for Othmarsingen, 1998

kgC

/m2

<



Net Ecosystem Exchange for Othmarsingen, 1998

kgC

/m2

Concept & state of the Swiss Biome-BGC projectConcept & state of the Swiss Biome-BGC project

Calibrate and thoroughly test the Biome-BGC model for Swiss forests along multiple environmental gradients

A

Test the hypothesis of increased tree growth in Central European forests in the decade of 1991 to 2000 compared to earlier decades.

B

Testing MODIS-MOD17 (NPP/GPP) data on the same test sites (calculated using the Biome-BGC logic, but based on RS inputs, partly)

C


Calibration and test sitesCalibration and test sites


Project work flow and progressProject work flow and progress

Test dataLWF sites

• stand level data, structure, mass• leaf chemistry, mass• intra-annual growth rates• soils & roots data

Calibrationdata

• C:N of leafs, roots, wood• leaf-level physiology• C-partitioning• decay rates, etc.

Test model • testing against field data• sensitivity analyses

Run simulations

• past 4 decades, test w| tree rings• test set of hypotheses• test effect of scale & resolution


prog

ress

Project work flow and progressProject work flow and progress

Test dataLWF sites

• stand level data, structure, mass• leaf chemistry, mass• intra-annual growth rates• soils & roots data

Calibrationdata

• C:N of leafs, roots, wood• leaf-level physiology• C-partitioning• decay rates, etc.

Test model • testing against field data• sensitivity analyses

Run simulations

• past 4 decades, test w| tree rings• test set of hypotheses• test effect of scale & resolution


Calibration & test data preparationCalibration & test data preparation

The following data were primarily collected in the field

• C:N of sunlit/shaded leafs & stem wood• A/Ci curves of sunlit (some shaded) leafs• Sunlit and Shaded leafs SLA• Leaf phenology (flush, drop) of deciduous trees• Number of needle years (needle turnover fraction)• Wood/tree cores for inter-/intraannual growth and density

People involved: Matt Jolly, Theo Forster, Stéphanie Schmid, Markus Schaub, Dani Nievergelt, Paolo Cherubini, Peter Suter, Werner Schoch, Matthias Dobbertin, Norbert Kräuchi, Felix Kienast, Dmitry Golikov, Lorenz Waltert, Stefan Zimmermann, Gustav Schneiter, Peter Jakob, etc. etc.


Vcmaxactfnrflnrlnc

sm

CO mol

sRubg

CO mol

N g

Rub g

N g

N g

m

N g2

222

Rubleaf

Rubleaf

Vcmax: Maximum carboxylation rate (=PS-rate)lnc: Leaf nitrogen content (from C:N and SLA)flnr: Fraction of leaf nitrogen in Rubiscofnr: Rubisco nitrogen fraction (enzyme structure)act: enzyme activity of Rubisco


valid if stomatal resistance is minimal


Leaf physiology

(lnc) x (flnr) x (fnr) x (act) = (Vcmax)

Area-based leaf N concentration = 1/ (SLA C:Nleaf)

Rubisco activity = f(Tleaf)

Maximum rate of photosynthesis(determined from fit to ACi curves)

Determined from protein structure

Leaf physiology

sm

CO mol

sRubg

CO mol

N g

Rub g

N g

N g

m

N g2

222

Rubleaf

Rubleaf



leafRubRub NfractNfractact

VcFLNR

..max


Leaf physiology


constant constant 1/ (SLA C:Nleaf)

ACi curves

Beauty and pain of field work - Vcmax



Beauty and pain of field work – SLA / C:N



Leaf physiologyA/Ci curve Pinus sylvestris

-20

-10

0

10

20

30

40

50

60

70

80

0 200 400 600 800 1000 1200 1400

Internal CO2 concentration

Ass

imil

atio

n

LapseRate

Vcmax

Plateau Jmax &/or

TPU



A-Ci curve Quercus pubescens

-1

-0.5

0

0.5

1

1.5

2

2.5

3

0 50 100 150 200 250 300 350 400

Internal CO2 concentration

Ass

imil

atio

n

uselessdata ....



Leaf physiology

Wc = Vcmax; Maximum rate of carboxylation by RuBisCO (CO2 limitation)

Wj = Jmax; RuBP-Regeneration (electron transport limitation)Wp = Regeneration of inorganic phosphate; (TPU-limitation)

A/Ci curves are analyzed using the Photosyn Assistant software [Dundee Scientific, UK]

Param Est. SEResp 1.53 3.38E-01Vcmax 57.5 1.12E+00Jmax 189 4.38E+00TPU 12.2 1.83E-01CO2 comp. est. (from Wc) = 5.13Sy.x = 6.10E-01 DF= 8Iteration count 2014SSqs 2.98

Example output:



Leaf physiology

Modified Farquhar (1980) photosynthesis eqn.:

dayi

RWpWjWcC

OA

,,min

5.01

Where, e.g.:

KoOKcCi

CiVcWc

1

max



Leaf physiology

internal CO2

concentrationinternal O2

concentration

Michaelis-Menten constants of Rubisco for CO2 and O2

respiration other than

photorespiration

Calibration of C:N for Fagus sylvatica



Calibration of SLA for Fagus sylvatica



Calibration of FLNR for Fagus sylvatica



Vcmax seasonal pattern: Fagus sylvatica




Model runs and sensitivity analysesModel runs and sensitivity analyses

How much do simulations differ with locally optimized parameters compared to the global DBF parameter set?

1

How sensitive is the model with respect to the accuracy of the measured parameters?

2

These questions are mostly analyzed by evaluating effects on carbon and nitrogen pools (not so much fluxes).

Most variables varied 5-15% between sites.



Set up for sensitivity analyses

A annual whole plant mortality fractionB C:N of leavesC C:N of live woodD C:N of dead woodE dead wood lignin/cellulose proportionF canopy light extinction coefficientG canopy average SLAH ratio of shd:sun SLAI fraction of leaf N in RubiscoJ B & G & H at the same timeK B & G & H & I at the same timeL soil depthM sand proportion in soil

each variable was increased by 5% and by 15% respectively



Simulation results Othmarsingen (C-Pool)



Simulation results Othmarsingen (N-Pool)



Simulation results Othmarsingen (C-Pool, change)



Simulation results Othmarsingen (N-Pool, change)



Simulation results Lausanne (C-Pool, change)



Summary-1

The accuracy of the measured parameter has a clear impact on the model simulations.

1

C:N ratio’s have a higher impact in Othmarsingen, where soils are shallow(er) and less rich.

2

Mortality and soil parameters are additionally important.

3



How does the sensitivity of parameters change across environmental gradients?

1

Do different parameters behave differently?2



Distribution of F. sylvatica in the environmental space



Set up for 2nd sensitivity analyses

A annual whole plant mortality fractionB C:N of leavesG canopy average SLAI fraction of leaf N in Rubisco

each variable was increased by 15% on each plotand compared to the standard calibration

Soil parameters (sand, silt, clay, soil depth) are kept constant

The climate of Othmarsingen is linearly adjusted for temperature and/or precipitation for each lattice point in order to exclude the effect of climate seasonality



BBGC Sensitivity G07085

-15.00%

-10.00%

-5.00%

0.00%

5.00%

+15

%M

ort

+15

%le

afC

:N

+15

%ca

nopy

avg.

SLA

+15

%F

LNR

Tested variables

Res

ult

ing

C-P

ool

size

VEG-C

LITTER-C

SOIL-C

Simulation results for Tave=7.0; Prcp=850mm



Climate effects on Veg-C

Simulated Fagus Veg-C across env. gradients

Prcp [cm]

Tave

[de

g. C

]

100 120 140 160

89

10

11

46 48

50 52 54

54

56



Simulated Fagus Soil-C across env. gradients

Prcp [cm]

Tave

[de

g. C

]

100 120 140 160

89

10

11

12

12

12

12

14161818

Climate effects on Soil-C



Effects of 15% C:N parameter error on Veg-C

Prcp [cm]

Tave

[de

g. C

]

100 120 140 160

89

10

11

-3.5-3 -3

-2.5

-2.5

-2

Sensitivity along env. gradients (C:NVeg-C)



Sensitivity along env. gradients (SLAVeg-C)

Effects of 15% SLA parameter error on Veg-C

Prcp [cm]

Tave

[de

g. C

]

100 120 140 160

89

10

11

-2 -1.5

-1

-1

-0.5

-0.5



Sensitivity along env. gradients (FLNRVeg-C)

Effects of 15% FLNR parameter error on Veg-C

Prcp [cm]

Tave

[de

g. C

]

100 120 140 160

89

10

11

0.450.45

0.450.45

0.5

0.5

0.5

0.550.60.65



Sensitivity along env. gradients (MortVeg-C)

Effects of 15% Mortality parameter error on Veg-C

Prcp [cm]

Tave

[de

g. C

]

100 120 140 160

89

10

11

-13.4 -13.4-13.4 -13.4-13.3 -13.3-13.3 -13.3



Summary-2

The model reacts non-linear to changes in the model parameters along environmental gradients.

1

Different parameters show differing patterns along environmental gradients

2

Mortality and leaf level chemistry parameters are highly sensitive to ecosystem simulations.

3



Outlook – next steps

Finish this sensitivity analysis for enlarged gradients.

1

Include other major tree species that are now calibrated.

2

Test model on LWF and WSI sites.3



Evaluate hypotheses of increased/decrease NPP along environmental gradients using single point and spatial Biome-BGC simulations.

1

Different parameters show differing patterns along environmental gradients

2

Mortality and leaf level chemistry parameters are highly sensitive to ecosystem simulations.

3

Outlook – and thereafter



Requirements – 1: daily climate maps

Daily Tmin simulations (L. Bernhard & N.Zimmermann)



Requirements – 1: daily climate maps

Daily Tmax simulations (L. Bernhard & N.Zimmermann)



Requirements – 2: fractional covers

Daily Tmax simulations (M. Schwarz, L. Mathys & N.Zimmermann)


Thank you for your attention


Some more stuff / results

Progress report on test data & -simulationsProgress report on test data & -simulations

Examples of preliminary LWF site simulations


Simulated NPP vs. Ring width

400

500

600

700

800

900

1000

1100

1200

1300

1400

1932

1934

1936

1938

1940

1942

1944

1946

1948

1950

1952

1954

1956

1958

1960

1962

1964

1966

1968

1970

1972

1974

1976

1978

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

Year

NP

P (

gC

/m^2

/yr)

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

Sta

nd

ard

ized

Rin

g w

idth

BGC_NPP

Ringwidth

Progress report on test data & -simulationsProgress report on test data & -simulations

Examples of preliminary LWF site simulations

Othmarsingen


Simulated Max. LAI and Evapotranspiration

3456789

1930 1940 1950 1960 1970 1980 1990 2000Year

max

. L

AI

500

700

900

1100

1300

1500

ET

max LAIann ET

Simulated NPP

0

200

400

600

800

1000

1200

1400

1930 1940 1950 1960 1970 1980 1990 2000

Year

% d

aily

wo

od

fo

rmat

ion

Avg

. d

aily

NP

P (kg C/m

2 /day)

simulated NPP (mass+storage) vs. stemwood production (mass)OthmarsingenOthmarsingen (Fagus sylvatica)

Leaves


-0.200

0.000

0.200

0.400

0.600

0.800

1.000

1.200

1.400

1.600

1.800

80 100 120 140 160 180 200 220 240 260 280 300 3200.000

0.001

0.002

0.003

0.004

0.005

0.006

0.007

0.008

0.009

0.010

simulated NPP (mass+storage) vs. stemwood production (mass)VordemwaldVordemwald (Abies alba & Picea abies)

Needles

% d

aily

wo

od

fo

rmat

ion

Avg

. d

aily

NP

P (kg C/m

2 /day)


0.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

1.60

80 100 120 140 160 180 200 220 240 260 280 300 320

-0.002

-0.001

0

0.001

0.002

0.003

0.004

0.005

0.006

0.007

ΔDBH

Counts

Avg_core

BBGC_NPP

Next tasks (loose list)Next tasks (loose list)

• Stand structural data (3-4 year colume increments per size class)

• Densitometry and Ring width of individual years, sizes• Calibrate for 3-4 year period per LWF plot• Reconstruct NPPa for last few decades



Finish additional calibration data (roots, LWF, …)• Mostly root turnover, C:N, allocation is not well

included yet• Lots of C:N (leafs, needles, wood) is in the analyses lab• Leaf litter data from LWF is currently under analysis

Finish preparation for spatial simulations• Prepare input maps for dominant forest types• Using a combination of RS and GIS/Statistical modelling



Finish preparation for spatial simulations (ctnd.)

Developed by M. Schwarz & N. ZimmermannMountain Forest Ecology Seminar; Jan. 22nd 2004 Niklaus E. ZimmermannBiome-BGC: Calibration and Sensitivity Analyses

Progress report on calibrationProgress report on calibration

Intra-annual Growth

Sampling:• every 2 weeks (later 4 weeks)• 6 sites (LWF)• dominant spp., 3 size classes, 2 cores per tree

Analysis• # of cells added since last date• diameter increment (1/100mm) since last date• % daily average growth per period• average per species (size).

Intraannual growth dynamics

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

100 120 140 160 180 200 220 240 260 280 300

Yearday

% d

aily

gro

wth

Fagus sylvatica



Intra-annual Growth some results

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

100 120 140 160 180 200 220 240 260 280 300

Yearday

% d

aily

gro

wth

Fagus sylvatica

Climate and Soils

-500

-450

-400

-350

-300

-250

-200

-150

-100

-50

0

100 120 140 160 180 200 220 240 260 280 300

Yearday

Wat

er t

ensi

on

(h

Pa)

0

5

10

15

20

25

Tem

par

atu

re (

°C)

15cm30cm45cm80cm130cmTave

Othmarsingen



Intra-annual Growth some results

Vordemwald

Climate and Soils

-100.0

-50.0

0.0

50.0

100.0

150.0

200.0

250.0

300.0

100 120 140 160 180 200 220 240 260 280 300

Yearday

Glo

bal

Rad

& W

ater

Ten

sio

n

0.0

2.0

4.0

6.0

8.0

10.0

12.0

14.0

16.0

18.0

20.0

Tav

e

GlobRadSW15SW30SW45SW80SW130Tave

0.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

100 120 140 160 180 200 220 240 260 280 300

Yearday

% d

ail

y g

row

th

Counts

Increment

Conifers


mountain forest ecology seminar; jan. 22 nd 2004 niklaus e. zimmermann biome-bgc: calibration and...

Documents

new leaf c

ratio allocation new

ratio allocation new

new total wood c

new growth

testing biomebgc

phenology flag

model phenology