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