Premium
Calibration of an electromagnetic induction sensor with time‐domain reflectometry data to monitor rootzone electrical conductivity under saline water irrigation
Author(s) -
Coppola A.,
Smettem K.,
Ajeel A.,
Saeed A.,
Dragonetti G.,
Comegna A.,
Lamaddale.,
Vacca A.
Publication year - 2016
Publication title -
european journal of soil science
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.244
H-Index - 111
eISSN - 1365-2389
pISSN - 1351-0754
DOI - 10.1111/ejss.12390
Subject(s) - reflectometry , calibration , environmental science , remote sensing , emi , soil science , soil salinity , sampling (signal processing) , time domain , geology , soil water , electromagnetic interference , computer science , mathematics , telecommunications , statistics , detector , computer vision
Management of saline water irrigation at the field scale requires electrical conductivity to be monitored regularly in the generally shallow soil layer explored by plant roots. Non‐invasive electromagnetic induction ( EMI ) sensors might be valuable for evaluating large‐scale soil salinity. However, obtaining information on soil surface salinity from depth‐integrated EMI measurements requires either an inversion of the electromagnetic signal or an empirical calibration. We opted for an empirical calibration of an EM38 sensor, which requires a reference dataset of local bulk electrical conductivity, σ b , for comparison with EMI readings for estimating regression coefficients. We used time‐domain reflectometry ( TDR ) to replace direct sampling for local σ b measurements. With empirical approaches, the different soil volumes involved with the EMI and TDR sensors become problematic. We resolved this issue by analysing large EMI and TDR datasets recorded at several times along three transects irrigated with water at 1, 3 and 6 dS m −1 degrees of salinity. A Fourier filtering technique was applied to remove the high frequency part (at small spatial scales) of the variation in the original data, which was the main source of dissimilarity between the two datasets. Therefore, calibration focused on the lower frequency information only; that is, information at a spatial scale larger than the observation volume of the sensors. We show that information from the TDR observations derives from a combination of local and larger scale heterogeneities, and how it should be managed for calibration of the EMI sensor. Analysis enabled us to identify characteristics of the calibration data that should be included to improve prediction.HighlightsEMI and TDR sensors have different observation volumes and are not immediately comparable. With F ourier filtering the different observation volumes were accounted for in calibration of the EMI data. We improved the empirical calibration between EMI and TDR datasets. We identified criteria to select data series for EMI sensor calibration.