Performance Evaluation of Three Newly Developed Soil Moisture Sensors

Abstract

An automated and robust measurement of soil moisture content is vital for the effective use of increasingly scarce water resources in irrigated agriculture. The various factors affecting the performance of dielectric soil moisture sensors include soil texture, bulk density, salinity and temperature variations. It is therefore important to take these factors into account when deploying these sensors. This study evaluated the performance of three newly developed soil moisture sensors; GS 1 (Decagon Devices), Stevens Hydraprobe II (Stevens Water) and TDR-315 (Accilma, Inc.). Measured soil moisture contents on three different soil types were compared with corresponding values derived from gravimetric samples. The sensors were also evaluated under conditions of varying bulk density, salinity and temperature. The calibration equations derived in the laboratory for the 3 soil moisture sensors performed satisfactorily in the three soil types. Variations in bulk density, salinity and temperature were found to introduce slight errors in the volumetric moisture content estimate by the three sensors. These sensors can be useful in monitoring soil moisture fluxes and in irrigation scheduling with laboratory derived calibration functions remarkably improving their accuracy.

Full text

CIGR-AgEng conference Jun. 26–29, 2016, Aarhus, Denmark
∙ 1 ∙
Performance Evaluation of Three Newly Developed Soil Moisture Sensors
Olutobi Adeyemi, Tomas Norton, Ivan Grove, Sven Peets
Engineering Department, Harper Adams University, Newport, Shropshire, TF10 8NB, United Kingdom
Corresponding author. Email: oadeyemi@harper-adams.ac.uk
Abstract
An automated and robust measurement of soil moisture content is vital for the effective use of increasingly scarce
water resources in irrigated agriculture. The various factors affecting the performance of dielectric soil moisture sensors
include soil texture, bulk density, salinity and temperature variations. It is therefore important to take these factors into
account when deploying these sensors.
This study evaluated the performance of three newly developed soil moisture sensors; GS 1 (Decagon Devices),
Stevens Hydraprobe II (Stevens Water) and TDR-315 (Accilma, Inc.). Measured soil moisture contents on three different
soil types were compared with corresponding values derived from gravimetric samples. The sensors were also evaluated
under conditions of varying bulk density, salinity and temperature.
The calibration equations derived in the laboratory for the 3 soil moisture sensors performed satisfactorily in the three
soil types. Variations in bulk density, salinity and temperature were found to introduce slight errors in the volumetric
moisture content estimate by the three sensors.
These sensors can be useful in monitoring soil moisture fluxes and in irrigation scheduling with laboratory derived
calibration functions remarkably improving their accuracy.
Keywords: Dielectric, Salinity, Bulk density, Temperature, Calibration
1. Introduction
Due to the reduced global availability of fresh water, it has become imperative to develop methods that improve water
use in irrigated agriculture. A common approach is the use of soil moisture sensors that monitor the field scale volumetric
water content (VWC). This enables growers to schedule irrigation when the soil moisture is depleted to a defined
threshold which results in improved irrigation timing and application depths (Varble & Chávez, 2011). Dielectric soil
moisture sensors provide a suitable means of continuously monitoring field scale soil moisture status. They take
advantage of the high dielectric permittivity of water relative to other soil constituents to infer soil moisture content. The
dielectric permittivity of soil is however influenced by other factors including soil texture, bulk density, salinity and
temperature, therefore a careful consideration of these factors is essential for the accurate determination of soil moisture
content (Paige & Keefer, 2008).
The variability in the dielectric properties of different soil types and the influence of dry plant tissues make it
necessary to calibrate dielectric sensors for every soil type (Polyakov et al. 2005). A number of researchers have
conducted studies with dielectric sensors in various soil types and generally conclude that a soil specific calibration
developed either in the field or laboratory will generally improve sensor accuracy (Mittelbach et al. 2012; Francesca et al.
2009; Kammerer et al. 2014; Varble & Chávez, 2011; Fares et al. 2011).
The dielectric measurement of electromagnetic soil moisture sensors is widely affected by salinity which is closely
linked to the soil bulk electrical conductivity especially at low operating frequencies less than 50 MHz. The effect of
salinity on the operation of dielectric soil moisture sensors is a function of the dielectric losses in the imaginary part of
the complex permittivity and it is positively dependent on the soil’s ionic conductivity (Saito et al. 2008). The effect of
salinity on dielectric measurements is usually masked at low temperatures (Bogena et al. 2007). Dielectric losses in soils
due to salinity can however be ignored in soils with conductivities less than 0.5 at low operating frequencies
(Bosch, 2004). Thompson et al. (2007) reported an increase in volumetric moisture measurements of 7.5% for every
increase of 1 increase in pore water electrical conductivity in a sandy loam soil and an increase of 4% in VWC
measurement for every increase of 1 increase in pore water electrical conductivity in a clay soil when using a
capacitance sensor.
Temperature variations in the field affect the performance of electromagnetic soil moisture sensors. The dielectric
permittivity of free water decreases with an increase in temperature which usually leads to errors in soil moisture content
estimation in sandy soils at high temperatures. At high temperatures, an increase in dielectric measurements is observed
in soils with high clay content due to the release of bound water at these high temperatures. The errors related to
dielectric measurements is saline soils is also more pronounced at high temperatures.
Czarnomski et al. (2005) reported an underestimation of VWC of about 0.1% for every 1°C increase in temperature
by a capacitance probe installed in a sandy loam soil. Polyakov et al. (2005) evaluated a capacitance sensor in a clay soil
and reported a 15% overestimation of VWC over a 45°C range. Theoretical approaches based on effective frequency and
complex permittivity model for compensating for the effect of temperature and salinity on the accuracy of dielectric
sensors have been successfully applied by (Schwartz et al. 2009; Evett et al. 2006).

CIGR-AgEng conference Jun. 26–29, 2016, Aarhus, Denmark
∙ 2 ∙
These methods improve the accuracy of the sensors but they require an extensive knowledge of electromagnetics and
high cost spectrum analyzing equipment limiting their use to research.
We consider three soil moisture sensors; the GS 1 volumetric soil moisture sensor (Decagon Devices), the Stevens
Hydraprobe II (Stevens Water) and the TDR 315 (Acclima Inc.). The three sensors are newly developed and we have no
knowledge of any peer reviewed articles on their evaluation. In the reviewed literature, all the authors have suggested that
a soil specific calibration equation improves the accuracy of dielectric soil moisture sensors, the influence of salinity,
bulk density and temperature on sensor performance is also emphasized. The objective of this study is therefore to
develop soil specific calibrations for the three soil moisture sensors to predict soil moisture in three different soil types.
We also investigate the effect of soil texture, bulk density, salinity and temperature on the performance of the sensors.
2. Materials and Methods
2.1. Sensors
GS 1 volumetric soil moisture sensor
The GS 1 volumetric soil moisture sensor is a capacitance sensor operating at a frequency of 70 MHz. The sensor
applies an oscillating wave at the stated frequency to the soil to form a complete capacitor. The charge stored in the
sensor probes after a predetermined time is directly proportional to the soil’s apparent dielectric permittivity which can
be related empirically to the volumetric water content of the soil. According to the manufacturer, the sensor’s output is
unaffected by variations in soil texture and with an accuracy of ±3% VWC in soils with EC of less than 8 ds and
temperatures less than 50°C. The manufacturer also states that the accuracy can be increased to ±1% VWC using soil
specific calibrations. The GS 1 sensor has a measurement region with a diameter of 11 cm. The GS 1 sensor outputs an
analog voltage of between 0 - 5 V which can be related to the VWC of the soil using the manufacturer supplied
calibration equation for mineral soils.
Stevens Hydraprobe II
The Stevens Hydraprobe II operates at a frequency of 50 MHz. The sensor calculates the amplitude ratio of reflected
waves within its probes when installed in soil and applies a numerical solution of the Maxwell’s equation to calculate the
real dielectric permittivity of the surrounding soil based on this. The real dielectric permittivity is then related empirically
to the volumetric water content of the soil. According to the manufacturer this procedure makes the probe immune to
variations in soil texture, salinity and temperature. The sensor has a measurement region with a diameter of 3 cm. The
stated accuracy of the sensor is ±3% VWC in all soil types. The Hydraprobe is an SDI-12 sensor which outputs the raw
VWC of the soil in water fraction by volume (wfv; .
Acclima TDR 315
The TDR 315 operates at a wave propagation bandwidth of 3500 MHz. It measures the time taken by a reflected
wave to travel through its probes which can be related to the apparent dielectric permittivity of the sensed soil medium.
The calculated apparent dielectric permittivity is related to the volumetric soil water content using propriety equation
similar to the Topp’s model. According to the manufacturer this procedure makes the probe immune to variations in soil
texture, salinity and temperature. The stated accuracy of the sensor is ±2% VWC in all soil types up to a maximum bulk
EC of 5 . A temperature accuracy of ±1% VWC is reported by the manufacturer for a temperature range of 0-50°C.
The TDR 315 is an SDI-12 sensor which outputs the raw VWC of the soil in % water content.
Data Acquisition
The output of the GS 1 sensor was logged using a Campbell scientific CR 1000 datalogger (Campbell Scientific)
programmed with the manufacturers default calibration equation for mineral soils. The output of the Hydraprobe and
TDR 315 were logged using an Acclima datasnap SDI-12 datalogger (Acclima Inc.). For the purpose of uniformity the
output of all sensors are presented in units of water fraction by volume, .

CIGR-AgEng conference Jun. 26–29, 2016, Aarhus, Denmark
∙ 3 ∙
2.2. Soils
Soils were collected from the top 30 cm in three different sites in Harper Adams University to represent a range of
soil types typical of the Midlands of United Kingdom. The table below shows the physical properties of the soils.
Table 1. Properties of soils tested
Site
Soil
type
Sand
%
Silt
%
Clay
%
Dry bulk
density
( )
Field
capacity
(
Permane
nt wilting
point
(
Crabtree
(CT)
Sandy
loam, light
textured
79
9
12
1.15
0.114
0.06
Back of
CERC
(BOC)
Sandy
loam,
medium
textured
72
15
13
1.28
0.144
0.071
Blackbirtch
(BB)
Sandy
loam,
heavy
textured
67
16
17
1.32
0.18
0.082
2.3. Tests
Three tests were performed. The first test established the relationships between the output of the sensors and the
gravimetrically derived volumetric soil water content for the various soils. The second test examined the relationships
between the VWC estimate of the sensors and the gravimetrically derived VWC estimates in a soil subjected to both a
medium and high level of compaction similar to that experienced in the field. The third test examined the effect of soil
temperatures on the output of the three sensors. Soil temperatures were varied within the range typically experienced in
the field.
Test 1
The laboratory calibrations were performed using the soils from the three sites at a room temperature of 22 ± 2°C.
The laboratory calibration was based on the procedure proposed by Campbell et al. (2009). Soils collected from each
field were air dried and passed through a 5mm sieve. They were then packed into 4 L containers (diameter 16 cm, height
19 cm) at the approximate field bulk density by adding equal volumes of soil in three layers. The water content of each
container was altered by adding deionized water in increments of 400 ml to represent soil moisture contents from air dry
to saturation. The containers were wrapped with polythene to prevent surface evaporation and left for 48 hours in order
for the soil moisture to equilibrate. The sensors were randomly assigned to the containers at each moisture level and the
readings over 10 mins intervals were averaged. After each reading gravimetric samples were taken from the containers
and oven dried a 105°C for 24 h.
Test 2
The medium textured soil from BOC was used in this test. Soils collected from the field were air dried and passed
through a 5mm sieve. Adapting the methodology outlined by John et al. (1986) the soil was packed into 4 L calibration
containers and compacted in three layers to a medium and high level of compaction similar to that experienced in the
field by imposing a load of 2.1 KN and 3.5KN respectively using a tensile testing machine (Samuel Denison and Son
Ltd). This produced an average bulk density of 1.37 in the moderately compacted soil and 1.42 in the
highly compacted soil. The Laboratory calibration equations were then developed for the soils at both compaction levels
following the methodology outlined in test 1.
Test 3
For this test, the light textured soil from CT and the heavy textured soil from BB were used. The soils from each of
the fields were packed into 4 L containers at the approximate field bulk density by adding equal volumes of soil in layers.
Following the methodology proposed by Benson and Wang (2006) the soils were brought to five moisture levels by
adding deionized water in increments of 400 ml. The containers were then wrapped in polythene to prevent surface
evaporation and left for 48 h for soil moisture to equilibrate. Each of the soil container/ water content combination was
then subjected to temperatures of 5, 15, 25 and 35°C in an incubator (Model ICI 180, Sanyo Electric Co). The
temperature of the soil was monitored with a thermocouple and at each temperature step, time was allowed for the soil
temperature to equilibrate. At each temperature step the averaged sensor readings were then logged over 10 mins

CIGR-AgEng conference Jun. 26–29, 2016, Aarhus, Denmark
∙ 4 ∙
intervals by randomly assigning sensors to each container. After the temperature variation procedure gravimetric samples
were taken from the containers and oven dried at 105°C for 24 h.
To investigate the effect of variable salinity on the performance of the sensors a one-time addition of salts to soils
from both sites was performed. The aim was to increase the bulk EC at saturation of each soil to values less than 12
which is the recommended limit for agricultural soils (Kizito et al. 2008). To achieve this 100 g of calcium
chloride di hydrate was dissolved in 400 ml of deionized water and mixed thoroughly with air dried soils from CT and
BB. Water was then added incrementally to produce five different moisture levels in the soils. This produced bulk EC
readings in the range of 0.4 to 5 in the soil from CT and 0.8 to 8.3 in the soils from BB. The
bulk EC was measured using the TDR 315 sensor. Tests on the soils were then conducted following the procedure
outlined in the paragraph above.
2.4. Statistical Analysis
Calibration equations using linear least squares regression were developed to relate laboratory derived gravimetric
water content to the values measured by the sensors. Following recommendations by Varble and Chávez (2011), two
statistical tests were used to evaluate the default manufacturers calibration equations and the laboratory derived
laboratory equations. They include the mean bias error (MBE) and root mean square error (RMSE). A calibration
equation with a MBE value of ±0.02 and RMSE value less than 0.035 was considered accurate. These
values are chosen to reflect the measurement accuracy of 0.01 – 0.02 required in agricultural applications (Iaea,
2008).
In the temperature changing experiment the performance of the sensors were evaluated by relating the sensor output
to the temperature range investigated using linear least square regression.
3. Results
A family of linear calibration equations was developed for each sensor and soil type combination.
3.1. Factory Calibration Evaluation
Table 2 shows that under laboratory conditions the factory based calibration of the three sensors achieved the required
accuracy within the air dry to saturation range in the light textured soil (CT). Varble and Chávez (2011) reported similar
results for a Decagon 5TE sensor evaluated in a sandy soil. The MBE values for the TDR 315’s factory calibration in
Table 2 show that under laboratory conditions, the sensor underestimated VWC by an average of 0.052 in the
heavy textured soil (BB) and an average of 0.032 in the medium textured soil (BOC). The highest errors in VWC
estimates by TDR 315 were recorded in the heavy textured soil from low moisture content to high moisture content range
(P=0.0001). The Hydraprobe sensor’s factory calibration recorded the highest errors in VWC estimates in the heavy
textured soil (P<0.001). The sensor underestimated VWC by an average of 0.08 in the heavy textured soil and an
average of 0.023 in the medium textured soil. The factory calibration of the GS 1 sensor was accurate in the
medium textured soil. However, the calibration was not accurate in the heavy textured soil with an underestimation of
VWC by an average of 0.035 and an RMSE of 0.05 . During the laboratory evaluation, we recorded a
maximum EC of 0.1 in the soil from CT, 0.19 in the soil from BOC and 0.35 in the soil from BB.
3.2. Laboratory Calibration Evaluation
The soil specific linear calibration equations developed for the three sensors in the laboratory improved the accuracies
of the sensors as shown in Table 3. These calibration equations yielded lower levels of errors in all soil types (P<0.001)
in comparison to when the factory calibrations were used. The RMSE and MBE were within statistical targets in all tests
except in the GS 1 and BB combination where the RMSE value recorded was 0.04 .

CIGR-AgEng conference Jun. 26–29, 2016, Aarhus, Denmark
∙ 5 ∙
Table 2. Comparison of factory calibration based VWC ( ) with laboratory measurements of VWC ( ) for
the different sensors and soils.
Sensor and soil type
MBE ( )
RMSE ( )
TDR 315
BB
0.94
-0.052
0.02
BOC
0.94
-0.032
0.02
CT
0.93
-0.02
0.03
Hydraprobe
BB
0.85
-0.08
0.03
BOC
0.94
-0.023
0.03
CT
0.96
-0.02
0.03
GS 1
BB
0.74
-0.035
0.05
BOC
0.92
-0.007
0.03
CT
0.93
0.014
0.03
Table 3. Comparison of laboratory calibration based VWC ( ) with laboratory measurements of VWC ( )
for the different sensors and soils.
Sensor and soil type
MBE ( )
RMSE ( )
TDR 315
BB
0.94
0
0.02
BOC
0.94
0
0.02
CT
0.93
0
0.03
Hydraprobe
BB
0.85
0
0.03
BOC
0.94
0
0.03
CT
0.96
0
0.02
GS 1
BB
0.74
0
0.04
BOC
0.92
0
0.03
CT
0.93
0
0.03

CIGR-AgEng conference Jun. 26–29, 2016, Aarhus, Denmark
∙ 6 ∙
3.3. Sensor Sensitivity to Compaction
The three sensors tested in the compacted medium textured soil recorded the highest errors in VWC estimation in the
highly compacted medium textured soil (P<0.05). The statistical parameters for the linear regression between factory
calibrated sensor output and gravimetric moisture content have been inserted in Table 4. Table 4 shows that at medium
compaction level, the TDR 315 sensor underestimated soil moisture by an average of 0.034 . The Hydraprobe
underestimated soil moisture at medium compaction by an average of 0.024 . Table 4 also shows that magnitude
of soil moisture underestimation by the TDR 315 and the Hydraprobe sensor increased in the highly compacted soil. An
average underestimation of soil moisture by 0.05 was recorded for the TDR-315 and an average underestimation
of soil moisture by 0.027 was recorded for the Hydraprobe sensor. The GS 1 sensor performed within statistical
targets at both medium and high levels of compaction in the medium textured soil as indicated in Table 4.
Table 4. Comparison of factory calibration based VWC ( ) with laboratory measurements of VWC ( ) for
the different sensors in the compacted medium textured soil.
Sensor and Soil
MBE (
RMSE (
TDR 315
BOC Medium
Compaction
0.96
-0.034
0.03
BOC High
Compaction
0.93
-0.05
0.03
Hydraprobe
BOC Medium
Compaction
0.94
-0.024
0.03
BOC High
Compaction
0.92
-0.027
0.03
GS 1
BOC Medium
Compaction
0.94
0.008
0.03
BOC High
Compaction
0.93
-0.01
0.03
3.4. Sensor Sensitivity to Soil Temperature and Salinity Variations
The factory calibrated sensor output showed a significant linear response to increase in temperature for all the soil-
sensor combinations (the lowest =0.73).The parameters of the linear regression between sensor output and temperature
are not presented here due to restrictions in space. The output of the TDR 315 probe in the light textured soil decreased
with increasing temperature. The rate of temperature effect increased with an increase in moisture, thus the highest effect
0.00207 was observed at a moisture content of 0.3516 . The output of the GS 1 sensor in the light
textured soil exhibited a positive response to increasing temperature. The rate of temperature effect increased with
increasing moisture content, thus the highest effect 0.0016 was observed at a water content of 0.3516 .
The Hydraprobe sensor’s output also exhibited a positive response to increasing temperature in the light textured soil, the
response was however similar at all moisture content values with an increase in sensor output of between 0.0003 - 0.0005
observed.
At low to medium moisture content values, the output of the TDR 315 sensor exhibited a positive response to an
increase in temperature (highest response 0.00048 ) in the heavy textured soil. The sensor output however
exhibited a negative correlation to temperature at higher moisture content values. The output of the Hydraprobe sensor
exhibited a positive response to an increase in soil temperature in the heavy textured soil with the increase being more
pronounced at low to medium moisture contents (highest response 0.00141 ) and less pronounced with
increase in VWC. The GS 1 exhibited a positive dependence of sensor output on temperature at low to medium moisture
content (highest response 0.001 ). The sensor response to increasing temperature however became negative at
the higher moisture content range with the effect being less significant with an increase in VWC.

CIGR-AgEng conference Jun. 26–29, 2016, Aarhus, Denmark
∙ 7 ∙
An increase in salinity in the light textured soil had a significant effect (P<0.001) on the average error in VWC
estimates by the three sensor types when compared with the non -saline soil. The output of the TDR 315 in the saline
light textured soil exhibited a positive correlation to increasing temperature at low moisture content. The sensor output
however exhibited a negative correlation to temperature at medium to higher moisture content values with the rate of
temperature effect decreasing with increase in VWC. The highest decrease, 0.00079 was recorded at a
moisture content value of 0.3513 while the lowest decrease, 0.00048 was recorded at a moisture
value of 0.3967 . The result also indicated that at high moisture content in the saline soil, the TDR 315 sensor’s
output was less affected by temperature. The output of the Hydraprobe sensor exhibited a strong positive relationship
with increasing soil temperature at all moisture content values (the lowest in the saline light textured soil. The
rate of increase in sensor output increased with an increase in moisture content, thus the highest response, 0.00409
was observed at a moisture content of 0.3967 . The output of the GS 1 sensor also exhibited a similar
positive response to increase in temperature in the saline light textured soil. The highest response, 0.0024
was recorded at a moisture content value of 0.3967 .
An increase in salinity in the heavy textured soil also had a significant effect (P<0.001) on the average error in VWC
estimates by the three sensor types when compared with the errors observed in the non-saline soil. The error in VWC
estimates by the three sensors in the saline heavy textured soil was positively dependent on temperature (P<0.001).
Results also indicated that there was a significant increase in the error in VWC estimates by the three sensors at low
temperatures when compared with the errors observed in the non-saline soil (P<0.001). The TDR 315 sensor output
exhibited a positive correlation to soil temperature. The rate of increase in sensor output increased with an increase in
moisture content, thus the highest response, 0.00473 was observed at a moisture content value of
0.3857 . The response of the Hydraprobe sensor was also positively correlated with temperature in the heavy
textured saline soil. The increase in sensor output was positively dependent on moisture content. The highest increase
0.0105 was observed at a moisture content value of 0.3857 . A similar response of sensor output to
increase in temperature was also observed in the GS 1 sensor tested in the heavy textured saline soil. The rate of increase
in sensor output increased with increase in moisture content and the highest increase in sensor output observed was
0.0042 at a moisture content value of 0.3857 .
4. Discussions
The factory based calibration of the sensors evaluated achieved the required accuracy only in the light textured soil
and the GS 1 sensor tested in the medium textured soil. The factory based calibration of the three sensors consistently
underestimated soil moisture in the heavy textured soil. This may be due to the large amount of bound water present in
soils with high clay content. The dielectric permittivity of bound water is lower than that of free water leading to an
underestimation of soil water content by dielectric sensors. The factory calibration of the three sensors can be applied in
light texture soils but for heavier soils a laboratory calibration procedure is recommended.
The linear calibration developed in the laboratory improved the performance of the three sensors in all the soils. This
indicates that a laboratory calibration process is important for soil moisture sensors deployed in irrigation scheduling
applications where slight inaccuracies in estimated soil moisture content may lead to the onset of plant water stress.
An increase in compaction level in the medium textured soil increased the magnitude of soil moisture underestimation
by both the TDR 315 and Hydraprobe sensors. The underestimation of soil moisture content by the TDR 315 sensor and
the Hydraprobe sensor can be explained by the increase in the volume ratio of solid particles to air with an increase in
soil bulk density. This causes an increase in the dielectric permittivity of the solid particles accompanied by a decrease in
the dielectric permittivity of the soil water. This mechanism consequently leads to an underestimation of soil moisture
content by dielectric sensors. An increase in the level of compaction had a negligible effect on the accuracy of the GS 1
sensor in the medium textured soil. The GS 1 sensor performed within statistical targets at all levels of compaction.
Changes in soil temperature had an effect on the output of the three sensors in all the soils tested. The TDR 315
sensor tested in the light textured soil exhibited a negative response to an increase in soil temperature. The most
pronounced effect of soil temperature was observed at high moisture content. This is because the dielectric permittivity of
free water reduces with increasing temperature leading to an underestimation of soil moisture by the TDR sensor. Light
textured soils such as the one evaluated in our study hold a large amount of free water. The highest decrease in sensor
output observed in our study corresponds to an underestimation of 0.027 over a 13°C increase in temperature
relative to a reference temperature of 22°C used in our study. The GS 1 and Hydraprobe sensors exhibited a positive
response to increase in temperature in the light textured soil. The highest response to temperature observed for the GS 1
sensor corresponds to an overestimation of soil moisture content by 0.02 while the highest response to
temperature observed for the Hydraprobe sensor corresponds to an overestimation of soil moisture by 0.007 over
a 13°C increase in soil temperature. These errors in soil moisture estimates by the GS 1 and Hydraprobe sensor are
considered negligible in most applications.
The three sensors tested in the heavy textured soil exhibited a strong positive response to an increase in temperature
especially at low moisture contents. This may be due to the release of bound water in soils high in clay content with an

CIGR-AgEng conference Jun. 26–29, 2016, Aarhus, Denmark
∙ 8 ∙
increase in temperature. The bound water effect is however less pronounced at higher moisture contents as the ratio of
free water to bound water is more at this level often leading to a slight underestimation of soil moisture with an increase
in temperature. The temperature response of the TDR 315 sensor indicated an overestimation of soil moisture by
0.02 , an overestimation of 0.018 by the Hydraprobe sensor and an overestimation of 0.013 by the
GS 1 sensor over a 13°C increase in soil temperature. The range of soil moisture overestimation by the TDR 315,
Hydraprobe and GS 1 sensors tested in the heavy textured soils may be considered negligible in most applications.
The performance of the TDR 315 sensor was slightly affected in the saline light textured soil. The decrease in the
dielectric permittivity of free water with an increase in temperature still seemed to have a predominant effect on the
output of the TDR 315 sensor. The highest temperature effect observed corresponds to a soil moisture underestimation of
0.01 which may be considered negligible for most applications. The Hydraprobe and GS 1 sensor exhibited a
positive response to increasing temperature in the saline light textured soil. The temperature effect was positively
dependent on soil moisture for both sensors suggesting the greater contribution of pore water to the bulk electrical
conductivity of the light textured soil. Results also indicate that both sensors are more sensitive to increasing temperature
in the saline light textured soil. This may be due to increased signal attenuation resulting from increase in salinity and the
low operating frequencies of both sensors. The highest temperature effect observed for the Hydraprobe correspond to a
soil moisture overestimation of 0.053 while the highest effect observed for the GS 1 sensor corresponds to a soil
moisture overestimation of 0.031 . This suggests that sensors operating at low frequencies should be deployed with
caution under conditions of variable salinity in light textured soils.
The three sensors evaluated exhibited a positive response to increasing temperature when tested in the saline heavy
textured soil. Results indicated that the three soil moisture sensors significantly overestimated soil moisture at low
temperatures with the magnitude of overestimation larger at higher temperatures. This is explained by the contribution of
the highly charged clay particles to the bulk EC of the soil and the positive dependence of the bulk soil EC on
temperature. The magnitude of soil moisture overestimation by the three sensors was remarkably larger than statistical
targets over a 13°C increase in temperature. The highest temperature effect observed corresponds to a soil moisture
overestimation of 0.06 by TDR 315, 0.137 by Hydraprobe and 0.052 by GS 1. This emphasizes
the need for a temperature compensation procedure when deploying these sensors under conditions of variable salinity in
a heavy textured soil.
5. Conclusions
This research evaluated the performance of TDR 315, Hydraprobe II and GS 1 soil moisture sensors, within air dry to
saturation range of soil moisture contents, under laboratory conditions for soils typically found in the Midlands of United
Kingdom. Acceptable statistical targets for this test were set as an MBE value of ±0.02 and a RMSE value less
than 0.035 . Linear calibration equations were developed for the three sensors in all soils tested. The factory based
calibration of the three sensors performed within required accuracy in the light textured soil and the GS 1 sensor tested in
the medium textured soil. It however failed to achieve the required accuracy in all the other soil-sensor combinations
evaluated.
The linear calibration equation developed in the laboratory reduced the error in soil moisture estimates by the three
sensors in all the soils tested. The laboratory calibration did not however achieve the required accuracy with the GS 1
sensor tested in the heavy textured soil.
The TDR 315, Hydraprobe II and GS 1 sensors experienced errors in reporting soil moisture content with increase in
soil compaction. The magnitude of soil moisture underestimation by the TDR 315 sensor and the Hydraprobe sensor
increased with increasing soil bulk density. The GS 1 sensor was however slightly sensitive to increasing soil bulk
density due to compaction.
The output of the three soil moisture sensors exhibited a significant linear response to increasing temperature when
tested in both the light and heavy textured soil. The TDR 315 sensor underestimated soil moisture with an increase in
temperature while the Hydraprobe and GS 1 sensors overestimated soil moisture with an increase in temperature in the
light textured soil. At low moisture content, the three sensors overestimated soil moisture with increasing temperature in
the heavy textured soil. The magnitude of errors in soil moisture estimate by the three sensors increased with an increase
in salinity level in the heavy textured soil. A similar result was recorded for the Hydraprobe and GS 1 sensors tested in
the saline light textured soil. A saturation effect was however observed for the TDR 315 sensor tested in the saline light
textured soil with the temperature dependence of the sensor output exhibiting less dependence on increasing temperature
at high moisture contents.
In summary, this study has demonstrated that laboratory developed calibration equations improved the accuracy of
the evaluated soil moisture sensors. A careful consideration of the influence of varying compaction levels, temperature
and salinity when deploying dielectric sensors is also emphasized.

CIGR-AgEng conference Jun. 26–29, 2016, Aarhus, Denmark
∙ 9 ∙
References
Benson, C. H., & Wang, X. (2006). Temperature-compensating calibration procedure for Water Content Reflectometers.
Tdr 2006, Paper ID 50, 16 p. Retrieved from https://engineering.purdue.edu/TDR/Papers
Bogena, H. R., Huisman, J. a., Oberdörster, C., & Vereecken, H. (2007). Evaluation of a low-cost soil water content
sensor for wireless network applications. Journal of Hydrology, 344(1-2), 32–42.
http://doi.org/10.1016/j.jhydrol.2007.06.032
Bosch, D. D. (2004). Comparison of Capacitance-Based Soil Water Probes in Coastal Plain Soils. Vadose Zone Journal,
3(4), 1380–1389. http://doi.org/10.2136/vzj2004.1380
Campbell, C. S., Campbell, G. S., Cobos, D. R., & Bissey, L. L. (2009). Calibration and evaluation of an improved low-
cost soil moisture sensor. Decagon Devices Application Note.
Czarnomski, N. M., Moore, G. W., Pypker, T. G., Licata, J., & Bond, B. J. (2005). Precision and accuracy of three
alternative instruments for measuring soil water content in two forest soils of the Pacific Northwest. Canadian
Journal of Forest Research, 35(8), 1867–1876. http://doi.org/10.1139/x05-121
Evett, S. R., Tolk, J. A, & Howell, T. A. (2006). TDR Laboratory Calibration in Travel Time, Bulk Electrical
Conductivity, and Effective Frequency. Vadose Zone Journal, 5(4), 1020–1029.
http://doi.org/10.2136/vzj2006.0062
Iaea. (2008). Field estimation of soil water content: A practical guide to methods, instrumentation and sensor technology.
Atomic Energy, 131 p. http://doi.org/10.2489/jswc.64.4.116A
John, O. O., Raghavan, G. S. V, McKyes, E., & Mehuys, G. (1986). Shear Strength Prediction of Compacted Soils with
Varying Added Organic Matter Contents. American Society of Agricultural Engineering ASAE, 29(2), 351–355.
Retrieved from http://asae.frymulti.com/abstract.asp?adid=30151&t=3
Kizito, F., Campbell, C. S., Campbell, G. S., Cobos, D. R., Teare, B. L., Carter, B., & Hopmans, J. W. (2008). Frequency,
electrical conductivity and temperature analysis of a low-cost capacitance soil moisture sensor. Journal of
Hydrology, 352(3-4), 367–378. http://doi.org/10.1016/j.jhydrol.2008.01.021
Mittelbach, H., Lehner, I., & Seneviratne, S. I. (2012). Comparison of four soil moisture sensor types under field
conditions in Switzerland. Journal of Hydrology, 430-431, 39–49. http://doi.org/10.1016/j.jhydrol.2012.01.041
Paige, G. B., & Keefer, T. O. (2008). Comparison of field performance of multiple soil moisture sensors in a semi-arid
rangeland. Journal of the American Water Resources Association, 44(1), 121–135. http://doi.org/10.1111/j.1752-
1688.2007.00142.x
Polyakov, V., Fares, a., & Ryder, M. H. (2005). Calibration of a Capacitance System for Measuring Water Content of
Tropical Soil. Vadose Zone Journal, 4(4), 1004. http://doi.org/10.2136/vzj2005.0028
Saito, T., Fujimaki, H., & Inoue, M. (2008). Calibration and Simultaneous Monitoring of Soil Water Content and Salinity
with Capacitance and Four-electrode Probes. American Journal of Environmental Sciences, 4(6), 683–692.
http://doi.org/10.3844/ajessp.2008.683.692
Schwartz, R. C., Evett, S. R., Pelletier, M. G., & Bell, J. M. (2009). Complex Permittivity Model for Time Domain
Reflectometry Soil Water Content Sensing: I. Theory. Soil Science Society of America Journal, 73(3), 886 - 897.
http://doi.org/10.2136/sssaj2008.0194
Thompson, R. B., Gallardo, M., Fernandez, M. D., Valdez, L. C., & Martinez-Gaitan, C. (2007). Salinity Effects on Soil
Moisture Measurement Made with a Capacitance Sensor. Soil Scienc3 Society of Americe Journal, 71(6), 1647–
1657. http://doi.org/10.2136/sssaj2006.0309

CIGR-AgEng conference Jun. 26–29, 2016, Aarhus, Denmark
∙ 10 ∙
Varble, J. L., & Chávez, J. L. (2011). Performance evaluation and calibration of soil water content and potential sensors
for agricultural soils in eastern Colorado. Agricultural Water Management, 101(1), 93–106.
http://doi.org/10.1016/j.agwat.2011.09.007