Table 4.4 Artificial intelligence techniques used to predict arsenic contamination of groundwater
Case
study Predictors
Method
Reference
1
Avalon belt
Logistic regression
Ayotte
et al.
(2006)
Bronson Hill belt
Eugeosyncline sequence
Mesozoic basin
Waits River basin
Pelitic rocks (Bronson Hill)
Peraluminous granite (New Hampshire
Maine sequence)
Mafic rocks (Narragansett Basin)
Concord granite (Dc1m, granite)
Madrid Fm. (DSm, metamorphic)
Rindgemere Fm., lower member (DSrb,
metamorphic)
Berwick Fm., calcareous member (SObc,
Calcpelite)
Eliot Fm., Calef member (SOec,
metamorphic)
Kittery Fm. (SOk, metamorphic)
Perry Mountain Fm. (Sp, metamorphic)
Rangeley Fm., lower part, (Srl,
metamorphic)
Sangerville Fm. (Sspm, metamorphic)
Waterville Fm. (Sw, metamorphic)
Kittery Fm. (SZk, metamorphic)
Massabesic gneiss complex (Zmz, granite)
Stream sediment arsenic, (ln) mg kg
À1
Pleistocene marine inundation intrusive
granitic pluton category (within 3 km of
pluton)
Developed land flag (cut point of 33%)
elevation (1:24,000 scale DEM, m)
Population density (person km
À2
)
Precipitation, mm year
À1
Water bodies (% area in 1000 m radius
buffer)
2
Easting
Regression-kriging
Lado et al.
(2008)
Northing
Depth
Arsenic
NH 4
+ (ammonium cation)
Fe (Iron)
Mn (manganese)
Pb (Lead)
(continued)
4 Application of Artificial Intelligence in Predicting Groundwater Contaminants
85