Comparison of the accuracy of kriging and IDW interpolations in estimating groundwater arsenic concentrations in Texas

Gordon Gong, Sravan Mattevada, Sidney O'Bryant

Research output: Contribution to journalArticle

55 Citations (Scopus)

Abstract

Exposure to arsenic causes many diseases. Most Americans in rural areas use groundwater for drinking, which may contain arsenic above the currently allowable level, 10. μg/L. It is cost-effective to estimate groundwater arsenic levels based on data from wells with known arsenic concentrations. We compared the accuracy of several commonly used interpolation methods in estimating arsenic concentrations in >8000 wells in Texas by the leave-one-out-cross-validation technique. Correlation coefficient between measured and estimated arsenic levels was greater with inverse distance weighted (IDW) than kriging Gaussian, kriging spherical or cokriging interpolations when analyzing data from wells in the entire Texas (p<0.0001). Correlation coefficient was significantly lower with cokriging than any other methods (p<0.006) for wells in Texas, east Texas or the Edwards aquifer. Correlation coefficient was significantly greater for wells in southwestern Texas Panhandle than in east Texas, and was higher for wells in Ogallala aquifer than in Edwards aquifer (p<0.0001) regardless of interpolation methods. In regression analysis, the best models are when well depth and/or elevation were entered into the model as covariates regardless of area/aquifer or interpolation methods, and models with IDW are better than kriging in any area/aquifer. In conclusion, the accuracy in estimating groundwater arsenic level depends on both interpolation methods and wells' geographic distributions and characteristics in Texas. Taking well depth and elevation into regression analysis as covariates significantly increases the accuracy in estimating groundwater arsenic level in Texas with IDW in particular.

Original languageEnglish
Pages (from-to)59-69
Number of pages11
JournalEnvironmental Research
Volume130
DOIs
StatePublished - 1 Apr 2014

Fingerprint

Spatial Analysis
Groundwater
Arsenic
kriging
interpolation
arsenic
Interpolation
well
Aquifers
groundwater
aquifer
Regression analysis
regression analysis
Regression Analysis
comparison
drinking
Drinking
rural area
method
Costs and Cost Analysis

Keywords

  • Arsenic
  • Groundwater
  • IDW
  • Kriging
  • Texas

Cite this

@article{6034db1aa19a4d77840a62259561d474,
title = "Comparison of the accuracy of kriging and IDW interpolations in estimating groundwater arsenic concentrations in Texas",
abstract = "Exposure to arsenic causes many diseases. Most Americans in rural areas use groundwater for drinking, which may contain arsenic above the currently allowable level, 10. μg/L. It is cost-effective to estimate groundwater arsenic levels based on data from wells with known arsenic concentrations. We compared the accuracy of several commonly used interpolation methods in estimating arsenic concentrations in >8000 wells in Texas by the leave-one-out-cross-validation technique. Correlation coefficient between measured and estimated arsenic levels was greater with inverse distance weighted (IDW) than kriging Gaussian, kriging spherical or cokriging interpolations when analyzing data from wells in the entire Texas (p<0.0001). Correlation coefficient was significantly lower with cokriging than any other methods (p<0.006) for wells in Texas, east Texas or the Edwards aquifer. Correlation coefficient was significantly greater for wells in southwestern Texas Panhandle than in east Texas, and was higher for wells in Ogallala aquifer than in Edwards aquifer (p<0.0001) regardless of interpolation methods. In regression analysis, the best models are when well depth and/or elevation were entered into the model as covariates regardless of area/aquifer or interpolation methods, and models with IDW are better than kriging in any area/aquifer. In conclusion, the accuracy in estimating groundwater arsenic level depends on both interpolation methods and wells' geographic distributions and characteristics in Texas. Taking well depth and elevation into regression analysis as covariates significantly increases the accuracy in estimating groundwater arsenic level in Texas with IDW in particular.",
keywords = "Arsenic, Groundwater, IDW, Kriging, Texas",
author = "Gordon Gong and Sravan Mattevada and Sidney O'Bryant",
year = "2014",
month = "4",
day = "1",
doi = "10.1016/j.envres.2013.12.005",
language = "English",
volume = "130",
pages = "59--69",
journal = "Environmental Research",
issn = "0013-9351",
publisher = "Academic Press Inc.",

}

Comparison of the accuracy of kriging and IDW interpolations in estimating groundwater arsenic concentrations in Texas. / Gong, Gordon; Mattevada, Sravan; O'Bryant, Sidney.

In: Environmental Research, Vol. 130, 01.04.2014, p. 59-69.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Comparison of the accuracy of kriging and IDW interpolations in estimating groundwater arsenic concentrations in Texas

AU - Gong, Gordon

AU - Mattevada, Sravan

AU - O'Bryant, Sidney

PY - 2014/4/1

Y1 - 2014/4/1

N2 - Exposure to arsenic causes many diseases. Most Americans in rural areas use groundwater for drinking, which may contain arsenic above the currently allowable level, 10. μg/L. It is cost-effective to estimate groundwater arsenic levels based on data from wells with known arsenic concentrations. We compared the accuracy of several commonly used interpolation methods in estimating arsenic concentrations in >8000 wells in Texas by the leave-one-out-cross-validation technique. Correlation coefficient between measured and estimated arsenic levels was greater with inverse distance weighted (IDW) than kriging Gaussian, kriging spherical or cokriging interpolations when analyzing data from wells in the entire Texas (p<0.0001). Correlation coefficient was significantly lower with cokriging than any other methods (p<0.006) for wells in Texas, east Texas or the Edwards aquifer. Correlation coefficient was significantly greater for wells in southwestern Texas Panhandle than in east Texas, and was higher for wells in Ogallala aquifer than in Edwards aquifer (p<0.0001) regardless of interpolation methods. In regression analysis, the best models are when well depth and/or elevation were entered into the model as covariates regardless of area/aquifer or interpolation methods, and models with IDW are better than kriging in any area/aquifer. In conclusion, the accuracy in estimating groundwater arsenic level depends on both interpolation methods and wells' geographic distributions and characteristics in Texas. Taking well depth and elevation into regression analysis as covariates significantly increases the accuracy in estimating groundwater arsenic level in Texas with IDW in particular.

AB - Exposure to arsenic causes many diseases. Most Americans in rural areas use groundwater for drinking, which may contain arsenic above the currently allowable level, 10. μg/L. It is cost-effective to estimate groundwater arsenic levels based on data from wells with known arsenic concentrations. We compared the accuracy of several commonly used interpolation methods in estimating arsenic concentrations in >8000 wells in Texas by the leave-one-out-cross-validation technique. Correlation coefficient between measured and estimated arsenic levels was greater with inverse distance weighted (IDW) than kriging Gaussian, kriging spherical or cokriging interpolations when analyzing data from wells in the entire Texas (p<0.0001). Correlation coefficient was significantly lower with cokriging than any other methods (p<0.006) for wells in Texas, east Texas or the Edwards aquifer. Correlation coefficient was significantly greater for wells in southwestern Texas Panhandle than in east Texas, and was higher for wells in Ogallala aquifer than in Edwards aquifer (p<0.0001) regardless of interpolation methods. In regression analysis, the best models are when well depth and/or elevation were entered into the model as covariates regardless of area/aquifer or interpolation methods, and models with IDW are better than kriging in any area/aquifer. In conclusion, the accuracy in estimating groundwater arsenic level depends on both interpolation methods and wells' geographic distributions and characteristics in Texas. Taking well depth and elevation into regression analysis as covariates significantly increases the accuracy in estimating groundwater arsenic level in Texas with IDW in particular.

KW - Arsenic

KW - Groundwater

KW - IDW

KW - Kriging

KW - Texas

UR - http://www.scopus.com/inward/record.url?scp=84896737644&partnerID=8YFLogxK

U2 - 10.1016/j.envres.2013.12.005

DO - 10.1016/j.envres.2013.12.005

M3 - Article

C2 - 24559533

AN - SCOPUS:84896737644

VL - 130

SP - 59

EP - 69

JO - Environmental Research

JF - Environmental Research

SN - 0013-9351

ER -