calculation of contaminated soil volumes : geostatistics ... · better understand the...

Calculation of contaminated soil volumes :Geostatistics applied to a hydrocarbons spill

Lac Megantic Case

www.envisol-canada.ca

1

Sara Godoy del Olmo

CONTEXT OF THE STUDYThis study was conducted for pedagogical purposes

In June 2014 a geostatistical training was performed for the EnvironmentalMinistry of Quebec and for other public and private companies in order tobetter understand the Geostatistical method and its applications to thecontaminated sites

The data set obtained in the Lac Megantic first characterizations was usedfor performing the study case during this training

Geostatistics: study of a PH contamination

2

PetroleumHydrocarbons


3

OBJECTIVES OF THE GEOSTATISTICAL STUDYThe geostatistical treatment has been performed in order to:

• Accurately understand the contaminants’ behaviour in soil (correlations betweencontamination and depth/topography/geology…)

• Once the behaviour of the contamination is established, we perform acontamination mapping in order to visualize the pollution migration

• Delineate, if possible, the pollution extent and the volumes of contaminated soilaffected by the PH spill


4

1. Data field integration

2. Exploratory Data Analysis

3. Variogram fitting

4. Interpolation methods:Krigging / Simulations

5. 3D mapping / Estimate ofcontaminated soil volumes

Remember:

A geostatistical study is conducted as follows:


5

1. Integrating the field data

Laboratory data

- In total, 450 PH laboratory analyses- Preferential sampling performed in surface fills- Maximum sampling depth = 8 m

Field Data: geology

- Fills layer (0-2m) and natural soil (silt and clay) until 8 m depth

Remediation threshold

Field data must be geo-referenced in a coordinate X Y Z system


6

700 ppm for the PH


Objectives:

Geological study,

Contamination base map,

Quick statistics, histograms

Correlations between concentration and depth/geology,

Experimental variogram construction


7


Contamination base map

0 10000 20000 30000 40000 50000 60000 70000 80000Contamin C10-C50

0.00.10.20.30.40.50.60.70.80.9

Frequencies

Nb Samples: 450Minimum: 100.0000Maximum: 79000.0000Mean: 2556.0222Std. Dev.: 7816.4959


8


Correlation between concentration and depth

719,5

5291,51

1513,911935,5

1016,81900

137,72115,38533,33

133,33 0 0 0 0 0 1000

1000

2000

3000

4000

5000

6000

Conc

entr

atio

n m

oyen

ne e

n C1

0-C5

0(p

pm)

The highest concentrations aremeasured between 0.5 and 1 m depth


9


Sampling

fills Sand, gravel, Silt-F

Natural soil Silt-NS, clay, organic matter


10


Correlation between concentration and geology

Lithology Code Frequency EffectiveMean

(PH ppm)

Minimum

concentrationMaximum concentration

2. Gravel 2 11,8 53 7584,53 100 42018,9

3. Sand 3 50,2 225 2539,87 100 79019,9

4. Silt_F 4 7,6 34 1904,41 100 30005,4

5. Silt_NS 5 28,1 126 870,95 100 25004,1

1. Clay 1 0,7 3 100 100 100

6. Org Matter 6 1,6 7 137,14 100 281,07



Experimental Variogram Construction

B.1/ Isotrope Directional Variogram– NO/D90

N0

19

13822349271029123501

34243760

399041663929

D-901

1

543

0 100 200 300Distance (m)

-1.0e+0070.0e+0001.0e+0072.0e+0073.0e+0074.0e+0075.0e+0076.0e+0077.0e+0078.0e+007

Variogram:ContaminC10-C50

Contamin C10-C50D-90

85

160

16 3

0 1 2 3 4 5Distance (m)-1

.0e+0070.0e+0001.0e+0072.0e+0073.0e+0074.0e+0075.0e+0076.0e+007


Contamin C10-C50


12

3. Variogram Fitting

Géostatistique: étude d’une contamination aux HP et au Ni

N0

19

13822349271029123501

34243760

399041663929

D-903

3

9

0 100 200 300Distance (m)

-1.0e+0070.0e+0001.0e+0072.0e+0073.0e+0074.0e+0075.0e+0076.0e+0077.0e+0078.0e+007


0 1 2 3Distance (m)0.

0e+0001.0e+0072.0e+0073.0e+0074.0e+0075.0e+0076.0e+007


B.1/ Isotrope Directional Variogram– NO/D90

13

3. Variogram Fitting

Cross Validation

Mean VariancePercentage of robust

dataCorrelation

Coefficient Z/Z*

3D VariogramError -44.55817 29886681.8726

98.6 0.687Standard error -0.00355 0.96465


14

4. Interpolation method: ordinary krigging

Petroleum Hydrocarbons 0-2 m


15


Petroleum Hydrocarbons 2-2.5 m


16

4. 4. Interpolation method: ordinary krigging

Petroleum Hydrocarbons 2.5-5 m


17


Petroleum Hydrocarbons 5-8 m


18


Petroleum Hydrocarbons – 3D view


19


Is the whole site contaminated? Why?


20

Petroleum Hydrocarbons Concentrations > 700 ppm - 3D view


Krigging reproduces the expected value of the distribution: estimates are stronglyinfluenced by the mean value of the distribution

0 10000 20000 30000 40000 50000 60000Contam HP estimated Sara 3D omni

0.0

0.1

0.2

0.3

0.4

0.5

0.6

Frequencies

Nb Samples: 56410Minimum: -3105.0647Maximum: 62710.9023Mean: 1685.3650Std. Dev.: 2420.8222

0 10000 20000 30000 40000 50000 60000 70000 80000Contamin C10-C50

0.00

0.25

0.50

0.75

1.00

Frequencies

Nb Samples: 361Minimum: 100.0000Maximum: 79000.0000Mean: 1986.1087Std. Dev.: 7518.6112

Original data histogramme - PH Krigging histogramme- PH


21

5. Interpolation Methods: Conditional simulations

Krigging Conditional Simulations

Krigging allows reproducing the expectedvalue of the distribution: estimates arestrongly influenced by the mean value

Each simulation respects the valuesobserved in the sample points. Mean andvariance are preserved in eachrealization

Several random fonctions whit the samemean and variance can have the samevariogram

Simulations can reproduce the statisticsof order 1 (histogram) and 2 (variogram)

Allows only one mapping of thecontamination

Performing a large number ofsimulations allows to better reproducethe real diversity of possible cases,especially if we are considering a randomvariable.

Quantifying the uncertainty is notpossible

The uncertainty is quantified so it ispossible to asses the risk of the project


22

5. Interpolation Method: Conditional Simulations

Conditional simulations allow us to perform several charts (with equal probability) to betterreflect the diversity of real possible cases

Simulation 1 : grid 5*5*0.5 m


23


Simulation 3 : grid 5*5*0.5 m


24


Simulation 250: grid 5*5*0.5 metc…


25


Performing a post-treatment of conditional simulations allows to ask the followingquestions:

• Where are the contaminated areas? This question is used to locally estimatethe risk of exceeding the remediation threshold

• What is the volume of contaminated soil ? This question is used to estimate theoverall level of contamination in the study area

How to answer these questions ???


26


Where are the contaminated areas? This question is used to locally estimate the riskof exceeding the remediation threshold

Setting a remediation threshold probability map for the occurrence of PH in soil (foreach bloc) in concentrations higher than 750 ppm


27


Blocks showing a probability of exceeding the remediation threshold greater than 0.6%


28


Which risk level are we willing to accept? How to manage blocks showing anuncertain estimate?

Blocks having a probability of exceeding the remediation threshold (750 ppm)between 0.3 and 0.6%


29


What is the volume of contaminated soil ? This question is used to estimate theoverall level of contamination in the study area

P 5 ( 57470P10 ( 554100.

P15 ( 539700.0

P50 ( 486937.50)

P85 ( 432987.50)P90 ( 414237.50)

P95 ( 395637.50)

350000 400000 450000 500000 550000 600000 650000Volume (m3)

0

10

20

30

40

50

60

70

80

90100

Frequencies

Risk curve

Most probable volume

Optimistic volume

Pessimistic Volume


30


Performing conditional simulations shows that:

• The estimated volume of contaminated soil is between 574 700 m3 (pessimisticvolume) and 395 637 m3 (optimistic volume)

• The most probable volume is 486 937 m3, which corresponds to 25% probabilityof exceeding the remediation threshold (low risk)

• The uncertain volume (probability between 30% and 60%) represents a totalvolume of 237 787 m3

• This volume represents 48% of the most probable contaminated volumeannounced by the risk curve


31


The calculations show a great uncertainty of the contaminated soil estimate

900 1000 1100 1200 1300 1400 1500X (m)

-8

-7

-6

-5

-4

-3

-2

-1

0

Z(m)

Sampling barelyperformed from 5 m

depth


32


If the estimate is stopped at 5 m depth:

P 5 ( 298400P10 ( 283475.00

P15 ( 278175.00)

P50 ( 248250.00)

P85 ( 219700.00)P90 ( 214100.00)

P95 ( 208575.00)

200000 250000 300000 350000Volume (m3)

0

10

20

30

40

50

60

70

80

90100

Frequencies

• The difference between the pessimistic and optimisticvolumes is less important,

• The most probable volume is 248,250 m3, which ishalf of the uncertain volume calculated above, andcorresponds to approximately 25% probability ofexceeding the remediation threshold

• The volume of contaminated soil showing uncertainestimation (30% to 60% probability of exceeding thethreshold) is 66 375 m3

• This volume represents 25% of the most probablecontaminated volume (P50) announced by the riskcurve


33


Mapping : Soil having 25% probability or more of being above the remediation threshold and correspondingto the most probable contaminated volume


34


Blocks having a probability greater than 0.6% of exceeding 750 ppm

Changing the grid: blocks size = 25*25*1 m


35


Blocks with 30 to 40% probability ofexceeding the criterion

Blocks with 40 to 60% probability ofexceeding the criterion

Blocks with more than 60% probability ofexceeding the criterion

It is possible to produce maps suitable for decontamination which will be based on the risk weare willing to take in order to define a real strategy for the remediation project


Changing the grid: blocks size = 25*25*1 m

36

Comparison between different methods: Polygons vs Geostatistics

Polygons method performed according to the observed geology (thickness of the fills layer)


37

Total volume103 000 m3

to 126 300 m3

Risk curve

Risk curve performed on the fills forthe north area, along the railways

P 5 ( 92050.0P10 ( 88675.00)

P15 ( 85950.00)

P50 ( 77975.00)

P85 ( 70275.00)P90 ( 68500.00)

P95 ( 65275.00)

60000 70000 80000 90000 100000 110000Volume (m3)

0

10

20

30

40

50

60

70

80

90100

Frequencies


Comparison between different methods: Polygons vs Geostatistics

38

Conclusions

Performing the geostatistical study highligths the following information :

• Geostatistics provide a coherent analysis of data collected at each step of a contaminated sitecharacterization by performing statistical processing of spatial data.

• Geostatistics’ ultimate goal is to determine the uncertainties associated to interpolationmethods in order to reduce the financial risks of remediation sites projects.

• The choice of the geostatistical method used for each project has a large influence whenmaking estimates, especially if the original data does not show a clear spatial structure. Tomake relevant calculations, it is necessary to conduct a comprehensive analysis of the inputdata before integrating it into the geostatistical treatment. Each choice must be justified bythe geostatistical theories but also judged by the experts working on contaminated sites.


39

Thank you for your attention

40

calculation of contaminated soil volumes : geostatistics ... · better understand the...

Documents