Advanced

Estimating soil solution electrical conductivity from time domain reflectometry measurements using neural networks

Persson, Magnus LU and Bertacchi Uvo, Cintia LU (2003) In Journal of Hydrology 273(1-4). p.249-256
Abstract
Time domain reflectometry (TDR) is a widely used method for measuring the dielectric constant (K-a) and bulk electrical conductivity (sigma(a)) in soils. The TDR measured sigma(a) and K-a can be used to calculate the soil solution electrical conductivity, sigma(w.) The sigma(w), in turn, can be related to the concentration of an ionic tracer. Several models of the sigma(w)-sigma(a)-K-a relationship can be found in the literature. Most of these models require extensive calibration experiments in order to obtaining best-fit parameters. In this paper, we attempt to model the sigma(w)-sigma(a)-K-a relationship using neural networks (NN). We used TDR measured K-a and sigma(a) along with five different soil physical parameters (sand, silt, clay,... (More)
Time domain reflectometry (TDR) is a widely used method for measuring the dielectric constant (K-a) and bulk electrical conductivity (sigma(a)) in soils. The TDR measured sigma(a) and K-a can be used to calculate the soil solution electrical conductivity, sigma(w.) The sigma(w), in turn, can be related to the concentration of an ionic tracer. Several models of the sigma(w)-sigma(a)-K-a relationship can be found in the literature. Most of these models require extensive calibration experiments in order to obtaining best-fit parameters. In this paper, we attempt to model the sigma(w)-sigma(a)-K-a relationship using neural networks (NN). We used TDR measured K-a and sigma(a) along with five different soil physical parameters (sand, silt, clay, and organic matter content and bulk density) measured in nine different soil types using three different sigma(w) levels in each soil type. In total, 2953 K-a and sigma(a) measurements were obtained. The NN estimated sigma(w) was found to have a root mean square error (RMSE) of 0.05-0.13 dS m(-1) for the nine different soil types whereas the RMSE of two traditional sigma(w)-sigma(a)-K-a models was 0.12-0.87 dS m(-1). Furthermore, the traditional models exhibited larger errors for low sigma(a) and K-a, whereas the NN estimated sigma(w) did not show any trend in the errors. A sensitivity analysis showed that the NN model was more sensitive to small changes in sigma(a) compared to K-a. Of the five soil physical parameters, the silt and clay content affected the sigma(w)-sigma(a)-K-a relationship the most. The results presented shows that using NN, the sigma(w)-sigma(a)-K-a relationship can be predicted using soil physical parameters without need for elaborate soil specific calibration experiments. (C) 2003 Elsevier Science B.V. All rights reserved. (Less)
Please use this url to cite or link to this publication:
author
organization
publishing date
type
Contribution to journal
publication status
published
subject
keywords
electrical conductivity, neural networks, time domain reflectometry
in
Journal of Hydrology
volume
273
issue
1-4
pages
249 - 256
publisher
Elsevier
external identifiers
  • wos:000181223900017
  • scopus:0037466130
ISSN
0022-1694
DOI
10.1016/S0022-1694(02)00387-6
language
English
LU publication?
yes
id
d5662f19-1c40-4b3c-900e-4805c12399f4 (old id 317505)
date added to LUP
2007-09-20 09:33:21
date last changed
2018-10-03 11:30:06
@article{d5662f19-1c40-4b3c-900e-4805c12399f4,
  abstract     = {Time domain reflectometry (TDR) is a widely used method for measuring the dielectric constant (K-a) and bulk electrical conductivity (sigma(a)) in soils. The TDR measured sigma(a) and K-a can be used to calculate the soil solution electrical conductivity, sigma(w.) The sigma(w), in turn, can be related to the concentration of an ionic tracer. Several models of the sigma(w)-sigma(a)-K-a relationship can be found in the literature. Most of these models require extensive calibration experiments in order to obtaining best-fit parameters. In this paper, we attempt to model the sigma(w)-sigma(a)-K-a relationship using neural networks (NN). We used TDR measured K-a and sigma(a) along with five different soil physical parameters (sand, silt, clay, and organic matter content and bulk density) measured in nine different soil types using three different sigma(w) levels in each soil type. In total, 2953 K-a and sigma(a) measurements were obtained. The NN estimated sigma(w) was found to have a root mean square error (RMSE) of 0.05-0.13 dS m(-1) for the nine different soil types whereas the RMSE of two traditional sigma(w)-sigma(a)-K-a models was 0.12-0.87 dS m(-1). Furthermore, the traditional models exhibited larger errors for low sigma(a) and K-a, whereas the NN estimated sigma(w) did not show any trend in the errors. A sensitivity analysis showed that the NN model was more sensitive to small changes in sigma(a) compared to K-a. Of the five soil physical parameters, the silt and clay content affected the sigma(w)-sigma(a)-K-a relationship the most. The results presented shows that using NN, the sigma(w)-sigma(a)-K-a relationship can be predicted using soil physical parameters without need for elaborate soil specific calibration experiments. (C) 2003 Elsevier Science B.V. All rights reserved.},
  author       = {Persson, Magnus and Bertacchi Uvo, Cintia},
  issn         = {0022-1694},
  keyword      = {electrical conductivity,neural networks,time domain reflectometry},
  language     = {eng},
  number       = {1-4},
  pages        = {249--256},
  publisher    = {Elsevier},
  series       = {Journal of Hydrology},
  title        = {Estimating soil solution electrical conductivity from time domain reflectometry measurements using neural networks},
  url          = {http://dx.doi.org/10.1016/S0022-1694(02)00387-6},
  volume       = {273},
  year         = {2003},
}