TitelTitel

Gap Filling of COGap Filling of CO22 Fluxes Fluxes

of Frequently Cut Grasslandof Frequently Cut Grassland

Christof Ammann

Agroscope ART Federal Research Station, Zürich

Gap Filling Comparison Workshop, Jena, 2006

Federal Research Station Agroscope Reckenholz-Tänikon ART

MotivationMotivation

Motivation and ContentsMotivation and Contents

Most gap-filling algorithms up to now have been optimized and evaluated based mainly on forest NEE datasets

Managed grassland sites (as other agricultural sites) can experience very rapid changes (discontinuities) in vegetation cover and soil conditions

Not only annual NEE but also management/event-related NEE (over few weeks/months) is of interest.

We applied a specific gap filling algorithm that should be able to reproduce fast changes and yield an adequate seasonal course of NEE.

CONTENT

Short description of Swiss grassland site (with specific problems)

Description of gap filling method

Performance of gap filling for examplary events

Site/RegionSite/Region

Measurement Site near OensingenMeasurement Site near Oensingen

2002 2004

2004 2008

Site plotsSite plots

Oensingen Site: Experimental PlotsOensingen Site: Experimental Plots

Intensively managed grassland• mineral fertilizer and manure

(ca. 200 kg/ha/y total N) • 4-5 cuts per year

Extensively managed grassland• no fertilizer • 3 cuts per year

Various crops (rotation)

Flux measurement systems

(1.2 m above

ground)-200

-150

-100

-50

0

50

100

150

200

-300 -250 -200 -150 -100 -50 0 50 100

local WE scale [m]

loca

l S

N s

cale

[m

]

Highway

0

N

annual distribution of wind directions

CO2 nightCO2 night

Low Wind Conditions during the NightLow Wind Conditions during the Night

intermittent/no turbulence

mostly identified with stationarity and/or integral turbulence criteria

windspeed ca. 1m above ground(Jun-Oct 2002)

0 1 2 3 4 5 6 7

wind speed [m/s]

freq

uenc

y of

occ

urre

nce

[rel

. uni

ts] daytime

nighttime

Data selectionData selection

Quality Control and Data CoverageQuality Control and Data Coverage

 

individual rejection rate data coverage

(combined) effect / rejection criterion

INT EXT INT EXT

power/data acquisition failure 11% 6% 89% 94%

(A) erroneous raw data 15% 13% 76% 82%

(B) integral turbulence (w /u*) 10% 14%

(C) flux stationarity 36% 37% 49% 47%

(D) footprint 35% 38% 32% 30%

ManagementManagement

Vegetation Development and Management EventsVegetation Development and Management Events

0

20

40

60

80

100

Jan 2002 Jan 2003 Jan 2004 Dez 2004

canopy height INT [cm]

canopy height EXT [cm]

-1

-0.5

0

0.5

1

1.5

2

2.5

Jan 2002 Jan 2003 Jan 2004

harvest export EXT [tC/ha]

harvest export INT [tC/ha]

manure import INT [tC/ha]

Gap Method 1Gap Method 1

Applied Gap-Filling MethodApplied Gap-Filling Method

Low data coverage (mostly short gaps: 1 hour...2 days)

Rapidly changing vegetation cover during the entire growing season

Highly adaptive gap-filling 3-day (5-day/7-day) moving window

Best use of available data non-linear regression functions:

NEE = R(Tsoil) – A(QPAR)

To keep the method simple and robust, only R10 and A2000 are fitted with the moving window

T0 and /A2000 are kept konstant (determined by an overall regression)

]t[AQ

2000Q

1

Q)Q(A

2000

PARPAR

PARPAR

0soil010

10soil TT

1

TT

1K309exp]t[R)T(R

Gap Method 2Gap Method 2

Applied Gap Filling MethodApplied Gap Filling Method

0

10

20

30

40

0 500 1000 1500 2000

photosynthetic photon flux densitiy QPAR [E m-2 s-1]C

O2

ass

imila

tion

flu

x A

[ m

ol m

-2 s

-1]

fitted function (Michaelis-Menten)

observed data

0

4

8

12

-5 0 5 10 15 20 25

soil Temperature Tsoil at -5cm [°C]

noc

turn

al C

O2

flux

[m

ol m

-2 s

-1]

measured data

fitted function (Lloyd&Taylor)

Respiration R(Tsoil) was only fitted to nighttime data.

Daytime assimilation was calculated as NEE–R(Tsoil)

Overall fit of A(QPAR) was made with selected dataset

(canopy height > 20 cm)

Gap ResultGap Result

Normalized Assimilation and RespirationNormalized Assimilation and Respiration

... resulting from the gap filling procedure

0

2

4

6

8

Jan 2002 Jan 2003 Jan 2004 Jan 2005 Jan 2006

R10

[

mol

m-2

s-1]

INT

EXT

(b)

0

20

40

60

Jan 2002 Jan 2003 Jan 2004 Jan 2005 Jan 2006

A20

00 [ m

ol m

-2 s

-1]

INT

EXT

NEE topoNEE topo

Seasonal and Diurnal COSeasonal and Diurnal CO22 Exchange (INT 2002 - 2004) Exchange (INT 2002 - 2004)

CO2 flux[mol m-2 s-1]

cumul NEE

Cumulative NEE for Different Years and ManagementCumulative NEE for Different Years and Management

-7

-6

-5

-4

-3

-2

-1

0

1

0 50 100 150 200 250 300 350

julian day

cum

ulat

ive

NE

E [

tC/h

a]

2002

2003

2004

2005

INT

-7

-6

-5

-4

-3

-2

-1

0

1

0 50 100 150 200 250 300 350

julian day

2002

2003

2004

EXT

Examp: CutExamp: Cut

Example of Gap-Filled Time Series: Cutting EventExample of Gap-Filled Time Series: Cutting Event

-30

-20

-10

0

10

09.10.03 11.10.03 13.10.03 15.10.03 17.10.03

CO

2 flu

x [u

mol

m-2

s-1]

gap-filled data

measured datacut

Examp: WinterExamp: Winter

Example of Gap-Filled Time Series: FreezingExample of Gap-Filled Time Series: Freezing

-15

-10

-5

0

5

10

15

16. Dez 02 21. Dez 02 26. Dez 02 31. Dez 02 05. Jan 03 10. Jan 03

CO

2 fl

ux

[um

ol/m

2/s

] ; s

oil

Te

mp

. [°C

]

CO2 flux data

soil resp. (param.)

CO2 flux (param.)

soil Temp. (-5cm)

0

2

4

6

05/2003 06/2003 07/2003 08/2003 09/2003 10/2003 11/2003

norm

. re

spir

atio

n R

10 [

umo

l m-2

s-1

]

R_SWC_1

Respiration during Summer 2003Respiration during Summer 2003

0

5

10

15

-5 0 5 10 15 20 25 30

soil temperature (-5cm) [°C]

noct

urna

l res

pira

tion

[mm

ol m

-2 s-1

]

R_SWC_2

Nocturnal Respiration and Soil MoistureNocturnal Respiration and Soil Moisture

10

20

30

40

05/2003 06/2003 07/2003 08/2003 09/2003 10/2003

soil

wat

er c

onte

nt [

vol.%

] -5 cm

-10 cm

-30 cm

0

2

4

6

05/2003 06/2003 07/2003 08/2003 09/2003 10/2003 11/2003

norm

. re

spir

atio

n R

10 [

umo

l m-2

s-1

]

R_SWC_3

Nocturnal Respiration and Soil MoistureNocturnal Respiration and Soil Moisture

10

20

30

40

05/2003 06/2003 07/2003 08/2003 09/2003 10/2003

soil

wa

ter

cont

en

t [vo

l.%]

0

10

20

30

40

50

60

rain

fall

[mm

d-1

]

-5 cm

-10 cm

-30 cm

rain

0

2

4

6

05/2003 06/2003 07/2003 08/2003 09/2003 10/2003 11/2003

norm

. re

spir

atio

n R

10 [

umo

l m-2

s-1

]

ConclusionsConclusions

ConclusionsConclusions

The applied gap filling method is relatively simple and well suited for rapidly changing conditions and a low data coverage (with short gaps)

Larger gaps can be filled by interpolation of R10 and A2000 or by using

default values (long-term means).

Potential improvement: Time dependent fit for all functional parameters (partly with larger window size?)

Further activities: Comparison with other methods (test performance on discontinuities)

END

Thank You!

Gap Method 3

observed flux NEE[t] (with gaps)

daytime fluxnighttime flux R[t]

assimilation A[t]

3..7-day moving average filter for R10[t]

complete time series NEE[t] (Eq.2)

A[t]R[t]

R10[t] A2000[t]

3..7-day moving average filter for A2000[t]

time-independent fitof param. T0

time-independent fitof ratio /A2000 for

hc>20cm

C-budg avgC-budg avg

Carbon Budget for Intensive and Extensive Management Carbon Budget for Intensive and Extensive Management (2002-2004)(2002-2004)

C/t

CO2

HarvestManure

-6

-4

-2

0

2

4

6

carb

on

exc

han

ge

[tC

ha-1

y-1

]

intensive field

extensive field

NEE + Hexport - Mimport = - NBP

(a)