Polluted site assessment using Inverse Distance Weighted and Ordinary Kriging: advantages and limitations.

Abstract: Precise and accurate delineation of polluted sites is an important stake
among soil remediation professionals. Information available for estimation
is often incomplete and/or heterogeneous. That usually leads to the
discovery of unexpected polluted volumes during remediation phase and costs
overrun. On the opposite, certain projects may be wrongly judged too
expensive due to an overestimation of the pollution. Different tools are
used today in order to estimate pollution as accurately as possible.
Though, the spatial variability of the pollution, if not reproduced
properly, induces unrealistic contamination delineation. Consequently a
model appropriate to the concerned site is necessary to evaluate the
pollution consistently.
To give solutions to these issues, this master thesis considered the use of
Inverse Distance Weighted (IDW) and Ordinary Kriging (OK) in order to
estimate two polluted sites in Norrbotten, Sweden: Solgårdarna and Tväråns
såg area. Both methods can not be applied without hypotheses verification,
as the second order stationarity of data for OK: the estimation method
should therefore be chosen regarding the specificities of the site and
their adequacy to these hypotheses. It has been shown that some method
parameters such as the power or the search neighbourhood ellipsoid of IDW
have a non negligible impact on results (more than 20%) and should be
chosen carefully. Also transformation of data helps to make them more
stable reducing the impact of outliers but, if not applied properly, may
lead to results different of more than 45%. The different methods have been
compared by cross-validation. Limitations of the cross-validation method
have been examined.