Spatiotemporal Analysis of Surface Water Tetrachloroethene in New Jersey

Spatiotemporal Analysis ofSpatiotemporal Analysis ofSurface Water Surface Water

Tetrachloroethene in New Tetrachloroethene in New JerseyJersey

Presentation of the project of Presentation of the project of Yasuyuki AkitaYasuyuki Akita

Temporal GIS Fall 2004Temporal GIS Fall 2004

AgendaAgenda

• About TetrachloroetheneAbout Tetrachloroethene

• Monitoring DataMonitoring Data

• Details of BME MethodDetails of BME Method

• BME AnalysisBME Analysis

• Results of BME AnalysisResults of BME Analysis

• New CriterionNew Criterion

• Model ComparisonModel Comparison

• ConclusionConclusion

About TetrachloroetheneAbout Tetrachloroethene

About TetrachloroetheneAbout Tetrachloroethene

• Tetrachloroethene: CTetrachloroethene: C22ClCl44• Volatile organic compoundVolatile organic compound

• Nonflammable colorless liquid at Nonflammable colorless liquid at room temperatureroom temperature

• Ether-like odorEther-like odor

• Synonym: Tetrachloroethylene, Synonym: Tetrachloroethylene, Perchloroethylene, and PCEPerchloroethylene, and PCE

Use and ProductionUse and Production

• Mainly Used for dry cleaning, chemical Mainly Used for dry cleaning, chemical intermediates, and industrial solventintermediates, and industrial solvent

• PCE used in dry cleaning industry has PCE used in dry cleaning industry has been declining during 90sbeen declining during 90s

• Recent Demand: 763 million lb (1980) Recent Demand: 763 million lb (1980)

318 million lb 318 million lb (1999)(1999)

End-Use Pattern in 70s and End-Use Pattern in 70s and 90s90s

Exposure pathwayExposure pathway

• Primary routePrimary route– InhalationInhalation– Ingestion of contaminated food and waterIngestion of contaminated food and water

• Widely distributed in environmentWidely distributed in environment– 38% of surface water sampling sites in the 38% of surface water sampling sites in the

U.S.U.S.– 771 of the 1430 National Priorities List sites771 of the 1430 National Priorities List sites– 154 of 174 surface water samples in N.J. 154 of 174 surface water samples in N.J.

(1977~1979)(1977~1979)

Health Effect of Health Effect of TetrachloroetheneTetrachloroethene

• Acute Effect (inhalation exposure)Acute Effect (inhalation exposure)– Dizziness, headache, sleepiness, Dizziness, headache, sleepiness,

confusion, nausea, difficulty in speaking confusion, nausea, difficulty in speaking and walking, unconsciousness, and and walking, unconsciousness, and deathdeath

• Chronic Effect (oral/inhalation Chronic Effect (oral/inhalation exposure)exposure)– Detrimental effect to kidney and liverDetrimental effect to kidney and liver

CarcinogenicityCarcinogenicity

• Reasonably anticipated to be a human Reasonably anticipated to be a human carcinogen (US DHHS)carcinogen (US DHHS)

• Group 2A (Probably carcinogenic to Group 2A (Probably carcinogenic to humans) (IARC)humans) (IARC)

• Animal studies: tumors in Animal studies: tumors in liverliver and and kidneykidney

Quality Standard for Quality Standard for TetrachloroetheneTetrachloroethene

• Maximum Contaminant Level (MCL) Maximum Contaminant Level (MCL) in drinking water - 0.005 mg/Lin drinking water - 0.005 mg/L

• Surface Water Quality Standard in Surface Water Quality Standard in New Jersey - New Jersey - 0.388 μg/L

N.J. adopted more stringent standardN.J. adopted more stringent standard

Monitoring DataMonitoring Data

Monitoring Dataset for New Monitoring Dataset for New JerseyJersey

• Data SourceData Source– NJDEP/USGS Water Quality Network NJDEP/USGS Water Quality Network

WebsiteWebsite– EPA STORET databaseEPA STORET database

• Data used in this studyData used in this study– 369 measured values369 measured values– 171 monitoring stations171 monitoring stations– From 1999 to 2003From 1999 to 2003

Monitoring Data – HistogramMonitoring Data – Histogram

Raw DataRaw Data Log-Transformed Log-Transformed DataData

Monitoring Data – Statistical Monitoring Data – Statistical MomentsMoments

RawRaw

((μμg/L)g/L)Log-Log-

transformedtransformed

(log-(log-μμg/L)g/L)

# # of recordsof records 369369 369369MeanMean 0.1569650.156965 -2.597410-2.597410

Standard Standard DeviationDeviation

0.3042710.304271 1.0437571.043757

Coef. of Coef. of skewnessskewness

5.2128345.212834 1.1191871.119187

Coef. of kurtosisCoef. of kurtosis 39.50591239.505912 4.0806464.080646

Distribution of Data PointsDistribution of Data Points

Distribution of Data PointsDistribution of Data Points

Distribution of Data ValuesDistribution of Data Values

What we want to know is …What we want to know is …

• Challenge of our researchChallenge of our research– Assess all river reachesAssess all river reaches– Taking into account the space/time Taking into account the space/time

variabilityvariability

Framework for the space/time Framework for the space/time estimationestimation

Bayesian Maximum Entropy (BME) analysis of Bayesian Maximum Entropy (BME) analysis of TGISTGIS

Details of BME MethodDetails of BME Method

Space/Time Random FieldSpace/Time Random Field• The concentration field is modeled in terms The concentration field is modeled in terms

of Space/Time Random Field (S/TRF)of Space/Time Random Field (S/TRF)– Collection of random variablesCollection of random variables

mapmapmapmapmapmap )(][Prob χχχχχχ dfd

),p,,p(pxxx vv ･･･　･･･ 21map21map at ),,,( px),,,( 21map v ･･･χ

S/TRF: Collection of all possible realizationS/TRF: Collection of all possible realization

• Stochastic characterization of S/TRF is Stochastic characterization of S/TRF is provided by multivariate PDFprovided by multivariate PDF)( mapχf

Knowledge BaseKnowledge Base

• General Knowledge Base: General Knowledge Base: GG– Describe global characteristics of the Describe global characteristics of the

random field of interestrandom field of interest– Expressed as statistical momentsExpressed as statistical moments

• Site-specific knowledge Base: Site-specific knowledge Base: SS– Available monitoring data over the Available monitoring data over the

space/time domain of interestspace/time domain of interest

• Total Knowledge Base: Total Knowledge Base: KK– KK = = GG ∪ ∪ SS

General Knowledge Base General Knowledge Base GGMean TrendMean Trend

– Global trend of the S/TRF of interestGlobal trend of the S/TRF of interest

• CovarianceCovariance

– Measure of dependency between two pointsMeasure of dependency between two points– Sill = variance = covariance(r=0)Sill = variance = covariance(r=0)– Range shows the extent that co-variability Range shows the extent that co-variability

existsexists

)()( pp XEmx

))'()'())(()(()',( pppppp xxx mXmXEc

BME analysisBME analysis of Temporal of Temporal GISGIS

• Prior stagePrior stage– Examine all general knowledge base Examine all general knowledge base GG and and

calculate Prior PDFcalculate Prior PDF

• Integration stageIntegration stage– Update Prior PDF using Bayesian Update Prior PDF using Bayesian

conditionalization on the site-specific conditionalization on the site-specific knowledge base knowledge base S S and obtain posterior PDFand obtain posterior PDF

• Interpretive stageInterpretive stage– Obtain estimation value from Posterior PDFObtain estimation value from Posterior PDF

BME analysis of Temporal BME analysis of Temporal GISGIS• General KB Prior PDFGeneral KB Prior PDF

})()(exp{)(1

0 mapmapmap

N

gf

χpχ

)(

),(

)(

)()(

hard

khard

hard

map

k

χ

χ

χ

χ

f

f

f

ff

• Update prior PDF with Site-specific KBUpdate prior PDF with Site-specific KB– Bayesian conditionalizationBayesian conditionalization– Posterior PDF is given by conditional Posterior PDF is given by conditional

probability probability

Summary of BME analysis of Summary of BME analysis of TGISTGIS

Estimation Point

Data Point

longlati

t

• General KB– Mean

trend– Covarianc

e

• Site-Specific KB

– Hard Data

BME

fK(k)

Posterior PDF at estimation point

hard|km hard|kCEstimation Value

longlati

t

BME AnalysisBME Analysis

S/TRF for Log-transformed S/TRF for Log-transformed PCE concentrationPCE concentration

• S/TRF representing Log-tranformed S/TRF representing Log-tranformed concentration: concentration:

• Residual field describes purely Residual field describes purely stochastic aspect of the stochastic aspect of the concentration fieldconcentration field

)()()( ppp XmY Y

)(pY

Mean Mean TrendTrend

Residual Residual FieldField

Mean Trend of Log-Mean Trend of Log-transformed concentration transformed concentration fieldfield• Mean trend consist of two Mean trend consist of two

componentscomponents– Purely spatial componentPurely spatial component– Purely temporal componentPurely temporal component

• Each component is calculated by Each component is calculated by exponential smoothingexponential smoothing

)()()( ,, tmmm tYYY sp s

)(, ssYm)(, tm tY

Mean Trend – Temporal Mean Trend – Temporal ComponentComponent

• Increase from Jan. 1999 to Jan. 2003Increase from Jan. 1999 to Jan. 2003

• Decrease from Jan. 2003~Decrease from Jan. 2003~

Mean Trend – Spatial ComponentMean Trend – Spatial Component

• Contaminated AreaContaminated Area– Northeastern Northeastern

regionregion– Southwestern Southwestern

regionregion

Homogeneous/Stationary Homogeneous/Stationary S/TRFS/TRF

Log-transformed dataLog-transformed data

Residual data for S/TRF: Residual data for S/TRF:

Removing the mean Removing the mean trendtrend

• Homogeneous/Stationary Random Homogeneous/Stationary Random FieldField– Its mean trend is constantIts mean trend is constant– Its covariance is only function of the Its covariance is only function of the

spatial lag and temporal lagspatial lag and temporal lag'ss r 'tt

)(pX

Covariance for Residual Covariance for Residual S/TRFS/TRF

22

22

11

1

3exp

3exp

3exp

3exp),(

2

aa

rc

aa

rcrc

rr

days 2000

deg. 25.0

1766.0

days 1700

deg. 30.0

8832.0

2

2

2

1

1

1

a

a

c

a

a

c

r

r

Covariance for Residual Covariance for Residual S/TRFS/TRF

Covariance SurfaceCovariance Surface

Experimental DataExperimental Data Covariance ModelCovariance Model

Results of BME AnalysisResults of BME Analysis

BME Estimation – Temporal BME Estimation – Temporal FluctuationFluctuation

BME Estimation – Spatial BME Estimation – Spatial DistributionDistribution



(Apr. 15, (Apr. 15, 2002)2002)

BME Estimation – Contaminated BME Estimation – Contaminated AreaAreaArea above the quality standard: 0.388μg/LArea above the quality standard: 0.388μg/L

BME mean estimateBME mean estimate

Upper bound of the BME 68% confidence Upper bound of the BME 68% confidence interval interval

Upper bound of the BME 95% confidence Upper bound of the BME 95% confidence intervalinterval

(Apr. 15, (Apr. 15, 2002)2002)

BME Estimation – Along River BME Estimation – Along River StreamStream

• Equidistance points Equidistance points along river streamalong river stream– More accurate More accurate

estimation for estimation for surface water surface water

BME Estimation – Along River BME Estimation – Along River StreamStream

MeanMean 68% CI68% CI 95% 95% CICI

Feb. 5, 2000Feb. 5, 2000 0.79%0.79% 1.48%1.48% 15.03%15.03%Mar. 11, Mar. 11, 20012001

0.98%0.98% 6.86%6.86% 66.96%66.96%

Apr. 15, Apr. 15, 20022002

1.50%1.50% 9.04%9.04% 69.63%69.63%

May 20, May 20, 20032003

0.59%0.59% 3.24%3.24% 46.08%46.08%

• Fraction of river miles that does not Fraction of river miles that does not attain the quality standardattain the quality standard

New CriterionNew Criterion

Assessment CriterionAssessment Criterion

• S/TRF is characterized by Posterior PDFS/TRF is characterized by Posterior PDF

• Area under the curve = ProbabilityArea under the curve = Probability

Prob[PCE>QSTD]=Area under the Prob[PCE>QSTD]=Area under the curvecurve (QSTD<PCE<(QSTD<PCE<∞)∞)

Assessment CriterionAssessment Criterion

Prob[Non-Attainment]=Prob[PCE>0.388μProb[Non-Attainment]=Prob[PCE>0.388μｇｇ ]]

• Highly Likely in AttainmentHighly Likely in Attainment– Prob[Non-Attainment]<10%Prob[Non-Attainment]<10%

• Highly Likely in Non-AttainmentHighly Likely in Non-Attainment– Prob[Non-Attainment]>90%Prob[Non-Attainment]>90%

• Non-AssessmentNon-Assessment– 10%10%≦≦Prob[Non-Attainment]Prob[Non-Attainment]≦90%≦90%

• More Likely Than Not in Non-AttainmentMore Likely Than Not in Non-Attainment– Prob[Non-Attainment]>50%Prob[Non-Attainment]>50%

Fraction of River MilesFraction of River Miles

Identifying Contaminated Identifying Contaminated WMAsWMAs• The state of New Jersey is divided into The state of New Jersey is divided into

20 20 Watershed Management Area Watershed Management Area (WMA)(WMA)

• Assess which part of the state is Assess which part of the state is contaminatedcontaminated

• Contribution of each WMA to the Contribution of each WMA to the fraction of river miles assessed as fraction of river miles assessed as – Highly Likely in Non-AttainmentHighly Likely in Non-Attainment– More Likely Than Not in Non-AttainmentMore Likely Than Not in Non-Attainment

Contribution of WMAsContribution of WMAs

• Highly Likely in Non-AttainmentHighly Likely in Non-Attainment

Contribution of WMAsContribution of WMAs

• More Likely Than Not in Non-More Likely Than Not in Non-AttainmentAttainment

Fraction of River Miles in Fraction of River Miles in WMAsWMAs

Model ComparisonModel Comparison

Model Comparison – Error Model Comparison – Error VarianceVariance

Space/Time Space/Time AnalysisAnalysis

Purely Spatial Purely Spatial AnalysisAnalysis

(Feb. 5, (Feb. 5, 2000)2000)

Model Comparison – Cross-Model Comparison – Cross-ValidationValidation



Model Comparison – Cross-Model Comparison – Cross-ValidationValidation

Purely Purely Spatial Spatial

AnalysisAnalysis


ImprovemeImprovementnt

(st-s)/s*100(st-s)/s*100

Mean ErrorMean Error -0.10728-0.10728 -0.039363-0.039363 -64.3%-64.3%

Mean Absolute ErrorMean Absolute Error 0.621330.62133 0.332750.33275 -46.4%-46.4%

Mean Square ErrorMean Square Error 0.835080.83508 0.36760.3676 -56.0%-56.0%

Model Comparison – Fraction of Model Comparison – Fraction of River MilesRiver Miles



ConclusionConclusion


• About Monitoring DataAbout Monitoring Data– Some high concentration values are Some high concentration values are

observed in New Jersey between 1999 to observed in New Jersey between 1999 to 2003.2003.

– Monitoring data shows high Space/Time Monitoring data shows high Space/Time variability in terms of location of the variability in terms of location of the monitoring point and monitoring valuemonitoring point and monitoring value

• Application of BME method of TGISApplication of BME method of TGIS– It enables us to take into account high It enables us to take into account high

space/time variability and to estimate space/time variability and to estimate the concentration all river reachesthe concentration all river reaches


• New CriterionNew Criterion– New criterion takes into account the New criterion takes into account the

uncertainty information of posterior PDFuncertainty information of posterior PDF– It is used to complementary criterion for It is used to complementary criterion for

the conventional onethe conventional one– Fraction of the river miles assessed as Fraction of the river miles assessed as

“Highly Likely in Non-Attainment” “Highly Likely in Non-Attainment” reached about 0.45% in 2000reached about 0.45% in 2000

– Fraction of the river miles assessed by Fraction of the river miles assessed by the conventional criterion (More Likely the conventional criterion (More Likely Than~) reached about 1.8% in 2002Than~) reached about 1.8% in 2002


• Model ComparisonModel Comparison– Space/Time analysis produces more Space/Time analysis produces more

accurate estimation than the conventional accurate estimation than the conventional purely spatial analysispurely spatial analysis

– Space/Time analysis produced very Space/Time analysis produced very different estimatedifferent estimate

– In purely spatial analysis, non-assessment In purely spatial analysis, non-assessment river miles reach about 99%river miles reach about 99%

• NJ DEP will be able to better assess NJ DEP will be able to better assess PCE concentration in all river reaches PCE concentration in all river reaches by using this method and new by using this method and new criterioncriterion

Spatiotemporal Analysis of Surface Water Tetrachloroethene in New Jersey

Documents