Semi-automated filtering of data outliers to improve spatial analysis of piezometric data

Authors: Tremblay, YohannLemieux, Jean-MichelFortier, RichardMolson, JohnTherrien, RenéTherrien, PierreComeau, GuillaumeTalbot Poulin, Marie-Catherine
Abstract: The identification and removal of data outliers remains a major challenge for spatial analysis of piezometric data. In this context, a simple semi-automated procedure for filtering outliers of depth to static water level was developed and used as a part of a regional groundwater-mapping project in the Québec Metropolitan Community, Québec, Canada. Following a few basic steps of data control, potential outliers were detected using two simple automated steps: (1) identifying water levels that are deeper than the 99th percentile of a high-reliability dataset compiled by groundwater professionals and assumed to adequately represent depths to static water level, and (2) using moving averages within a search radius of 250 m calculated around each well. All detected potential outliers were visually examined in a geographic information system and compared to neighbouring data before being kept or discarded. To evaluate the efficiency of the procedure, exploratory statistics, histograms and semi-variograms of the initial, intermediate and filtered datasets were compared to the high-reliability dataset. Objective interpolation was then performed using ordinary kriging. A cross-validation analysis showed a less biased and more accurate interpolation after applying the proposed outlier filtering procedure. Qualitative knowledge of the hydrogeological settings is an important component of this procedure which combines advantages of both manual and automated processing, making the procedure adaptive and easy to use. The final outcome of the proposed procedure is an improved interpolation map of depth to static water level along with minimised and low squared estimation errors.
Document Type: Article de recherche
Issue Date: 19 April 2015
Open Access Date: Restricted access
Document version: VoR
This document was published in: Hydrogeology Journal, Vol. 23 (5), 851-868 (2015)
Verlag Heinz Heise
Alternative version: 10.1007/s10040-015-1257-y
Collection:Articles publiés dans des revues avec comité de lecture

Files in this item:
Description SizeFormat 
Tremblay et al. 2015 - HJ.pdf
21.51 MBAdobe PDF    Request a copy
All documents in CorpusUL are protected by Copyright Act of Canada.