Article
On the Accuracy of Factory-Calibrated Low-Cost Soil
Water Content Sensors
Jesús María Domínguez-Niño
1,
* , Heye Reemt Bogena
2
, Johan Alexander Huisman
2
,
Bernd Schilling
2
and Jaume Casadesús
1
1
Program of Efficient Use of Water in Agriculture, Institute of Agrifood Research and Technology (IRTA),
Parc de Gardeny (PCiTAL), Fruitcentre, 25003 Lleida, Spain
2
Institute of Bio- and Geosciences, Agrosphere Institute (IBG-3), Forschungszentrum Jülich GmbH,
52425 Jülich, Germany
* Correspondence: jesus.dominguez@irta.cat; Tel.: +34-973-032-850 (ext. 1598)
Received: 26 June 2019; Accepted: 11 July 2019; Published: 13 July 2019
Abstract:
Soil water content (SWC) monitoring is often used to optimize agricultural irrigation.
Commonly, capacitance sensors are used for this task. However, the factory calibrations have been
often criticized for their limited accuracy. The aim of this paper is to test the degree of improvement
of various sensor- and soil-specific calibration options compared to factory calibrations by taking the
10HS sensor as an example. To this end, a two-step sensor calibration was carried out. In the first step,
the sensor response was related to dielectric permittivity using calibration in media with well-defined
permittivity. The second step involved the establishment of a site-specific relationship between
permittivity and soil water content using undisturbed soil samples and time domain reflectometry
(TDR) measurements. Our results showed that a model, which considered the mean porosity and a
fitted dielectric permittivity of the solid phase for each soil and depth, provided the best fit between
bulk permittivity and SWC. Most importantly, it was found that the two-step calibration approach
(RMSE: 1.03 vol.%) provided more accurate SWC estimates compared to the factory calibration
(RMSE: 5.33 vol.%). Finally, we used these calibrations on data from drip-irrigated almond and apple
orchards and compared the factory calibration with our two-step calibration approach.
Keywords:
soil water content; 10HS sensor; calibration; sensor variability; specific calibration;
CRIM model
1. Introduction
Efficient irrigation management is essential for reducing water consumption. To this end,
real-time monitoring of soil water content (SWC) is essential to optimize the amount and timing
of water irrigation [
1
,
2
]. Electromagnetic (EM) methods, such as time domain reflectometry (TDR)
(e.g., Reference [
3
]) and capacitance sensors [
4
,
5
], are most commonly used for soil water content
measurements at the point scale. Capacitance sensors are often preferred over TDR sensors, as they
provide real-time SWC at a lower cost. In addition, they were shown to be reasonably robust and
precise, and consume less energy compared to TDR sensors [
6
–
8
]. Both TDR and capacitance methods
make use of the strong dependence of the soil dielectric permittivity on volumetric SWC. As the
dielectric permittivity of liquid water is much higher than the dielectric permittivity of the other
soil components, SWC is the principal factor governing the apparent soil permittivity [
9
]. However,
other soil properties such as salinity and texture may cause dielectric losses and disturb the SWC
measurements with EM sensors [
10
]. These dielectric losses depend on the frequency of the electric
field generated by the sensors and are especially important for sensors that work at frequencies between
1 and 200 MHz [
11
]. In addition, capacitance sensors can show substantial sensor-to-sensor variability,
Sensors 2019, 19, 3101; doi:10.3390/s19143101 www.mdpi.com/journal/sensors
