Geostatistical Modeling of CO2 Flux Spatial Variation

Kwame Awuah-Offei

Fred Baldassare

Moagabo Mathiba

Outline

• Background & motivation

• Objective• Objective

• Methods & materials

• Results & discussions

• Conclusions & future work

BACKGROUND & MOTIVATION

Background

• Elevated levels of CO2 have been found in homes built on/adjacent to reclaimed and abandoned mine land in recent years

– CO2 > 25%

– O2 < 10%

• Stable carbon isotope analysis have shown that AMD-carbonate reactions are responsible in some instances

2

( ) 3( ) ( ) 2 ( ) 2( )2

+ ++ → + +

aq s s l gH CaCO Ca H O CO

Motivation Cont’d

• We are motivated by a desire to develop predictive tools/methods to assess reclaimed mine land prior to reclaimed mine land prior to development

– Apply chamber accumulation trace gas measurement techniques to collect discrete soil CO2 emission rates (fluxes) on reclaimed mine land

– Apply geostatistics to model spatial and/or spatiotemporal variation

OBJECTIVE

Objective

• The objective of this presentation is to use a case study to illustrate the benefits, and future research the benefits, and future research directions, of CO2 flux monitoring and modeling using geostatistics.

METHODS & MATERIALS

Chamber Accumulation Soil Sampling

• LI-8100 automated CO2

flux system (LICOR Biosciences, Lincoln, Biosciences, Lincoln, Nebraska)

• USDA GRACEnet protocl

• 8” Collars

• Samples taken on June 4, 5 & 10

Study Site

• Reclaimed Mine in Somerset County, PA

– Spoil > 70 ft thick– Spoil > 70 ft thick

– 20 tons/acre of agg. Lime (CaCO3) addition to the pit floor was required prior to backfilling

• Stray CO2 in the residence was investigated by the PA-DEP in 2003

Study Site

• Continuous monitoring in the basement recorded:

– >25% CO2– 13% O2

• Isotopic analyses yielded a δ13C of:

– -4.07‰ in the basement

– -4.18‰ in a monitoring well

Geostatistics

• Software: GS+ Version 9

• We used variogram analysis to model the autocorrelation of maximum flux in the autocorrelation of maximum flux in 2-D

• Variogram models considered were :

– Linear

– Exponential

– Spherical

– Gaussian

Geostats Cont’d

• The maximum flux was estimated using ordinary and indicator krigingkriging

– Isotropic search radius 385 ft.

–Minimum samples = 3

–Maximum samples = 15

– Indicator kriging cut-off = 5.5 µmol/m2/sec

RESULTS & DISCUSSIONS

Flux Results

12

14

16

/sec)

6/4/2009 6/5/2009 6/10/2009 Max. Biogenic Mean

0

2

4

6

8

10

12

A

2

A

7

A

8

B

1

B

2

B

3

B

4

B

5

B

6

B

7

B

8

C

0

C

1

C

2

C

3

C

4

C

5

C

6

C

7

C

8

D

0

D

1

D

2

D

3

D

4

D

5

D

6

D

7

D

8

D

9

E

0

E

1

E

2

E

3

E

4

E

5

E

6

E

7

E

8

H

1

H

2

H

3

H

4CO2Flux (µmol/m

2/sec)

Sample Points

293875

294688

295500N

ort

hin

g

Flux: Quartiles

< 4.56

< 6.06

< 9.35

< 14.46 (max)

6

8

10

12

Frequency

292250

293063

1605000 1606200

No

rth

ing

Easting

0

2

4

6

3 5 7 9 11 13 15

Frequency

Maximum flux upper class limit (µmol/m2/sec)

0.75

0.9

0.95

0.99

Pro

ba

bil

ity Max flux: 5.73

Normal probability plot

2 4 6 8 10 12 14 16

0.01

0.05

0.1

0.25

0.5

Max flux (µµµµmol/m2/sec)

Pro

ba

bil

ity

Variogram Analysis

30

35

40

45

50

0.4

0.5

0.6

7.5

10.0

Sem

ivari

an

ce

Flux: Isotropic Variogram

0

5

10

15

20

25

30

0.0

0.1

0.2

0.3

100 150 200 250 300 350 400

Residual

R2

Lag

R2 Residual

0.0

2.5

5.0

0 700 1400 2100

Sem

ivari

an

ce

Separation Distance (h)

Spherical model (Co = 0.25000; Co + C = 8.99000; Ao = 382.00; r2 = 0.543;

RSS = 4.33)

Kriging Results

Cross-Validation

14

16

mol/m2/sec)

Actual Flux Ideal fit (1:1) Linear (Actual Flux)

2

4

6

8

10

12

2 4 6 8 10 12 14 16

Actual Flux (µmol/m2/sec)

Estimated Flux (µmol/m2/sec)

Cross-Validation Cont’d

90%

110%

130%

150%

-50%

-30%

-10%

10%

30%

50%

70%

90%

H

3

H

1

E

7

E

6

E

5

E

4

E

3

E

2

E

0

E

1

D

9

D

7

D

6

D

5

D

4

D

3

D

2

D

1

C

8

C

7

C

6

C

5

C

4

C

2

C

1

B

8

B

7

B

4

B

3

B

2

A

8

A

7

A

2

H

4

H

2

E

8

D

8

D

0

C

0

C

3

B

6

B

5

B

1

Percent Error (E-A)

Samples

Indicator Kriging Results

CONCLUSIONS & FUTURE WORK

Conclusions

• The maximum CO2 fluxes of replicate samples at the site seem to be spatially correlatedspatially correlated

• The grid spacing of 250 ft is too high to accurately quantify the spatial variability

• There is significant temporal variability of the fluxes as well

Conclusions Cont’d

• Geostatistical methods show promise in modeling the spatial and spatiotemporal variabilityand spatiotemporal variability

Limitations/Future Work

• Validation of geostatistical estimates needs to be improved (mean error = 20%)

– The optimal grid spacing needs to be establishedestablished

– More covariance/variogram functions need to be explored and new ones developed if necessary

• Spatiotemporal data collection and modeling may be necessary to model the random field appropriately.

Acknowledgement

• The authors will like to thank

– LICOR Biosciences for supporting this research with a LI-8100 automated CO2 flux research with a LI-8100 automated CO2 flux system

– Several employees of the PA-DEP who assisted in various forms with the field data collection exercise

COMMENTS & QUESTIONS

Isotope Results

Sample ID δ13C CO2

Flux (µmol/m2/sec) CO2 (ppm)

B1 -24.1 7.45 1,299 B1 -24.1 7.45 1,299

B5 -20.3 4.39 792

B6 -19.6 3.36 688

C0 -19.0 4.14 718

C3 -20.3 5.43 920

D0 -16.4 2.12 593

D8 -19.2 4.20 752

E8 -19.4 4.32 749

H2 -21.2 8.40 930

H4 -22.1 7.30 960