Journal of Hydrology 430–431 (2012) 39–49

Contents lists available at SciVerse ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Comparison of four soil moisture sensor types under field conditions in Switzerland
Heidi Mittelbach ⇑, Irene Lehner, Sonia I. Seneviratne ⇑
Institute for Atmospheric and Climate Science, ETH Zurich, Universitätstrasse 16, CHN N12.2, 8092 Zurich, Switzerland

a r t i c l e i n f o s u m m a r y

Article history: Many environmental and hydrological applications require knowledge about soil moisture. Its measure-
Received 9 September 2011 ment accuracy is known to depend on the sensor technique, which is sensitive to soil characteristics such
Received in revised form 27 January 2012 as texture, temperature, bulk density and salinity. However, the calibration functions provided by instru-
Accepted 30 January 2012
ment manufacturers are generally developed under laboratory conditions, and their accuracy for field
Available online 8 February 2012
This manuscript was handled by
applications is rarely investigated, in particular over long time periods and in comparison with other sen-
Konstantine P. Georgakakos, Editor-in-Chief, sors types. In this paper, four side-by-side profile soil moisture measurements down to 110 cm using
with the assistance of V. Lakshmi, Associate three low-cost sensors and one high-accuracy and high-cost time domain reflectometry (TDR) sensor
Editor are compared over a 2-year period at a clay loam site in Switzerland. The low-cost instruments include
the (1) 10HS (Decagon Devices, United States), (2) CS616 (Campbell Scientific, United States), and (3)
Keywords: SISOMOP (SMG University of Karlsruhe, Germany) sensors, which are evaluated against the (4) TDR-
Soil moisture based TRIME-IT/-EZ (IMKO GmbH, Germany) sensors. For the comparison, the calibration functions pro-
Sensor comparison vided by the manufacturers are applied for each sensor type. The sensors are evaluated based on daily
Low-cost data regarding their representation of the volumetric water content (VWC) and its anomalies, as well
Evapotranspiration as the respective temperature dependency of the measurements. Furthermore, for each sensor type the
Measurement networks
actual evapotranspiration is estimated using the soil water balance approach and compared with mea-
surements from a weighing lysimeter. It is shown that the root mean square difference (RMSD) of
VWC for the low-cost sensors compared to the TDR measurements are up to 0.3 m3/m3, with highest val-
ues in near-surface layers. However, the RMSD for the VWC anomalies are lower compared to those for
absolute values. We conclude that under the studied conditions none of the evaluated low-cost sensors
has a level of performance consistent with the respective manufacturer specifications. Hence the deriva-
tion of site-specific calibration functions is vital for the interpretation of measurements with low-cost soil
moisture sensors. Furthermore, some weaknesses of the tested low-cost sensors such as the lack of sen-
sitivity in certain soil moisture regimes or spurious dependency on soil temperature, imply intrinsic
issues with the measurements derived with this type of instruments. This is particularly critical for a
number of environmental and hydrological applications, including the assessment of remote sensing
measurements.
Ó 2012 Elsevier B.V. All rights reserved.

1. Introduction tial role for climate variability and extremes, including heat waves
(Seneviratne et al., 2006; Diffenbaugh et al., 2007; Fischer et al.,
Soil moisture is an essential environmental, hydrological and 2007; Vautard et al., 2007; Lorenz et al., 2010; Teuling et al.,
climate variable. In particular, it strongly affects the land surface 2010; Hirschi et al., 2011; Jaeger and Seneviratne, 2011). Moreover,
fluxes of the water and energy balances with consequent impacts potential (positive or negative) feedbacks with precipitation have
on temperature, evapotranspiration, planetary boundary layer sta- also been suggested in various regions (Koster et al., 2004; Hohe-
bility or runoff generation (see Seneviratne et al., 2010 for a re- negger et al., 2009; Seneviratne et al., 2010; van den Hurk and
view). In recent years, soil moisture–atmosphere interactions van Meijgaard, 2010; Findell et al., 2011; Taylor et al., 2011). Global
received increasing attention in climate research. In particular, effects of soil moisture variability on climate have been proposed as
both numerical and observational studies highlighted their poten- well (Jung et al., 2010), and recent studies have further highlighted
the potential role of soil moisture for sub-seasonal to seasonal fore-
⇑ Corresponding authors. Addresses: Universitätsstrasse 16, CHN N12.3, 8092 casting (e.g. Koster et al., 2010a,b; Weisheimer et al., 2011).
Zurich, Switzerland. Tel.: +41 44 632 80 77; fax: +41 44 633 10 58 (H. Mittelbach), Ground truth data for soil moisture are essential to analyze the
Universitätsstrasse 16, CHN N11, 8092 Zurich, Switzerland. Tel.: +41 44 632 80 76; processes underlying land–atmosphere interactions and to evaluate
fax: +41 44 633 10 58 (S.I. Seneviratne).
their role in land surface and climate models (e.g. Vinnikov et al.,
E-mail addresses: [email protected] (H. Mittelbach), sonia.senevir-
[email protected] (S.I. Seneviratne). 1996; Dirmeyer et al., 2006). Thereby, information about root-zone

0022-1694/$ - see front matter Ó 2012 Elsevier B.V. All rights reserved.
doi:10.1016/j.jhydrol.2012.01.041
40 H. Mittelbach et al. / Journal of Hydrology 430–431 (2012) 39–49

soil moisture is particularly critical, given that feedbacks of soil The present study is designed to compare the performance of
moisture on climate are mostly mediated by vegetation (e.g. Sene- two FDR- and one capacitance-based sensor types, when applying
viratne et al., 2010). Comprehensive ground-based soil moisture the calibration function supplied by the manufacturers. For their
observational networks are nonetheless still very scarce, although evaluation TDR measurements are used as reference. The analysis
some networks have been recently established, such as the is based on 2 years of field measurements at a site of the SwissS-
Oklahoma Mesonet in United States (Basara and Crawford, 2000), MEX network using parallel measurements down to 110 cm. The
OzNet in Australia (Rüdiger et al., 2007) and in Europe the networks focus of the investigation is on the uncertainties in measured
RHEMEDUS in Spain (Martínez-Fernández and Ceballos, 2005), VWC and its anomalies, as well as on the temperature dependency
SMOSMANIA in France (Albergel et al., 2008), SwissSMEX in Switzer- of the measurements. Furthermore, the ability of the four sensor
land (http://www.iac.ethz.ch/url/SwissSMEX; e.g. Mittelbach et al., types to represent changes in absolute soil moisture storage is
2011), and the TERENO observatories in Germany (http://teodoor. evaluated using the soil water balance approach and a direct com-
icg.kfa-juelich.de/). These measurements are also essential for the parison with evapotranspiration measurements from a weighing
development and evaluation of new remote sensing-based soil lysimeter. This study does not provide new calibration functions
moisture estimates (Wagner et al., 2007; De Jeu et al., 2008). For in- for the investigated sensors. Our emphasis is rather on assessing is-
stance, Jackson et al. (2010) recently emphasized the need for more sues with the quality of soil moisture measurements when the
ground-truth observations networks over wide spatial scales and manufacturers’ calibration function is applied without correction.
with increased temporal frequency to improve respective satellite In addition, we provide an extensive cross-evaluation of several
algorithms. Many of the newly available but also already existing commonly used sensors under field conditions, which is useful
long-term networks, as well as remote sensing observations have when assessing measurements of distinct soil moisture networks,
been recently collected and harmonized as part of the International as they often rely on the use of a single sensor type.
Soil Moisture Network (ISMN; Dorigo et al., 2011). The article is structured as follows. First, data and methods,
Different techniques for soil moisture measurements have been including instruments, field measurements and applied methods
developed over the last decades (see e.g. Robinson et al., 2008 for a for the sensor comparison are described in Section 2. The results
review). Most commonly, electromagnetic sensors are used to are presented in Section 3. The discussion of the results and con-
establish continuous in situ soil moisture networks. These sensors clusions with respect to climate, hydrology and remote sensing
make use of the high permittivity of water to estimate the volu- applications are provided in Sections 4 and 5, respectively.
metric water content (VWC) in the soil (Topp, 2003) and are gen-
erally based either on time domain reflectometry (TDR),
frequency domain reflectometry (FDR), or capacitance techniques. 2. Data and methods
The TDR-based sensors are known to be of higher accuracy, but are
also of significantly higher cost than the FDR and capacitance 2.1. Field site and measurements
probes. Due to the accuracy of TDR measurements, these sensors
are used as reference sensors in previous studies (e.g. Plauborg Field measurements from the research catchment Rietholzbach
et al., 2005; Bogena et al., 2007; Mittelbach et al., 2011). Neverthe- (http://www.iac.ethz.ch/url/rietholzbach) are used in this study.
less, because of the spatial and temporal variability of soil moisture This pre-alpine catchment is located in northeastern Switzerland
(e.g. Western et al., 1999, 2004; Famiglietti et al., 1999; Albertson (47.37°N, 8.99°E) and has been in operation since 1975. For the
and Montaldo, 2003; Teuling and Troch, 2005; Brocca et al., 2010; period 1976–2006 the catchment is characterized by a mean an-
Vereecken et al., 2010) it has been suggested to use less accurate nual precipitation of 1459 mm, an actual yearly evapotranspira-
but cheaper sensors in order to increase the density of measure- tion, estimated using a weighing lysimeter, of 560 mm and a
ments within soil moisture networks (e.g. Bogena et al., 2007; Rob- mean annual 2 m air temperature of 7.1 °C. The weighing lysimeter
inson et al., 2008; Vereecken et al., 2010). In addition to their cost, of the Rietholzbach site is a backfilled lysimeter with a surface of
the high power consumption of TDR instruments (Veldkamp and 3.14 m2 and a depth of 2.5 m. The container is positioned on a scale
O’Brien, 2000) may be another dissuasive argument for their use which has a resolution of 100 g. For further information about the
if a site has to be e.g. operated with stand-alone power supply. catchment and respective observations see Seneviratne et al. (sub-
As for any types of measurements, the sensors’ performance mitted for publication). A comparison of evapotranspiration mea-
and accuracy are important. Many sensor types for soil moisture surements and estimates for the catchment are also provided in
measurements have been evaluated in previous studies (e.g. Roth Lehner et al. (2010).
et al., 1990; Walker et al., 2004; Czarnomski et al., 2005; Robinson Within the SwissSMEX project (http://www.iac.ethz.ch/url/
et al., 2008; Blonquist et al., 2005). These studies generally con- SwissSMEX) the grassland site Rietholzbach was enhanced in
cluded that the accuracy of the measured VWC when using univer- 2009 with additional soil moisture measurement profiles. The set-
sal calibration functions provided by the manufacturers, needs to up of a SwissSMEX site is carried out in several steps: First, a hole is
be carefully evaluated using laboratory and field measurements. dug taking into account the original sequence of the soil horizons,
Further studies (e.g. Seyfried and Murdock, 2004; Evett et al., and separating the respectively extracted soil amounts. Second,
2006; Logsdon, 2009; Rüdiger et al., 2010; Mittelbach et al., disturbed and undisturbed soil samples are taken from each soil
2011) also recommended more specifically to establish soil- or horizon for subsequent soil analysis. In a third step, all sensor types
site-specific calibration functions in such applications. It should are installed horizontally in the undisturbed soil to provide similar
be noted that the accuracy of the respective sensors may also de- conditions. Finally, the hole is systematically refilled, ensuring that
pend on the site characteristics for various reasons (e.g. soil mois- the soil horizons are arranged in the original order and close to the
ture regime, soil type, soil homogeneity, presence of stones and original density using compaction. For the Rietholzbach site, the
roots). Some of these site-specific effects may not only be related TRIME-IT/EZ, 10HS, and CS616 sensors as well as the soil temper-
to the measurement technique, but also to the sensor design. For ature sensors were installed side-by-side at seven depths: 5, 15,
instance, soil moisture sensors with long rods provide a more rep- 25, 35, 55, 80, and 110 cm. The SISOMOP sensors were installed
resentative soil moisture measurement due to the integration over at five depths: 5, 15, 35, 55, and 80 cm.
a larger volume. But their installation in stony and also clayey soils In this study, measurements of soil moisture and soil tempera-
with low water content can be more difficult compared to that ture down to 110 cm, precipitation and 2 m air temperature as well
with smaller and more compact sensors. as data from the weighing lysimeter of the time period 1 June 2009
H. Mittelbach et al. / Journal of Hydrology 430–431 (2012) 39–49 41

to 31 May 2011 were considered. The temporal evolution of these types make use of electromagnetic characteristics to estimate the
variables is shown in Fig. 1. Basic soil characteristics for the site permittivity of a medium and relate this information to VWC. De-
and each soil horizon are listed in Table 1. tails of the instruments are provided in the next subsections.
Parallel to the soil moisture measurements, soil temperature
2.2. Instruments (Tsoil) was measured in all depths using the temperature sensor
107-L (Campbell Scientific, United States). All measurements, ex-
The considered soil moisture instruments are the TRIME-IT/-EZ cept those from the SISOMOP sensor, are logged with a Campbell
(IMKO GmbH, Germany), the 10HS (Decagon Devices, United Scientific CR1000 data logger. The data from the SISOMOP sensor
States), the CS616 (Campbell Scientific, United States), and the are logged with the SISOMOP logging unit (Schlaeger, 2007a). All
SISOMOP (SMG University of Karlsruhe, Germany). All four sensor data are measured and recorded at a 10-min interval. For the

1 2 3
a 60
P Tair
P (mm)

Tair (°C )
40 20
20 10
0 0

b 5 cm
VWC (m3/m3)

0.8 20

Tsoil (°C )
0.6
0.4 10
0.2
0
c 15 cm
VWC (m3/m3)

0.8 20

Tsoil (°C )
0.6
0.4 10
0.2
0

d 25 cm
VWC (m3/m3)

0.8 20

Tsoil (°C )
0.6
0.4 10
0.2
0
e 35 cm
VWC (m3/m3)

0.8 20

Tsoil (°C )
0.6
0.4 10
0.2
0
f 0.6
55 cm
VWC (m3/m3)

20
Tsoil (°C )

0.4
10

0.2 0
g 0.6
VWC (m3/m3)

80 cm 20
Tsoil (°C )

0.4
10

0.2 0
h 0.6
VWC (m3/m3)

110 cm 20
Tsoil (°C )

0.4
10

0.2
01/06/09 01/09/09 01/12/09 01/03/10 01/06/10 01/09/10 01/04/11 01/06/11
Ts oil TRIME 10HS CS616 wT CS616 wo T SISOMOP

Fig. 1. Temporal evolution of (a) precipitation and 2-m air temperature and of (b–h) soil moisture measurements of the different sensors and soil temperature measurements
at the different depths at daily resolution. Dry down periods are indicated in gray and correspond to the following time frames: 8–13 September 2009 (1), 24 September–05
October 2009 (2), and 25 June–04 July 2010 (3). Note the different scale for the y-axis in plots f–h.
42 H. Mittelbach et al. / Journal of Hydrology 430–431 (2012) 39–49

Table 1
Basic soil characteristics of the Rietholzbach site with texture according to USDA soil taxonomy.

Depth (cm) Particle size distributiona (%) Texture Bulk density (g/cm3) Organic fractionb (%) pH
Clay (<2 lm) Silt (2–63 lm) Sand (>63 lm)
0–15 30.6 35.9 33.5 Clay loam 1.08 4.7 6.9
15–23 30.8 31.0 38.2 Clay loam 1.37 2.5 7.0
23–70 25.6 32.7 41.7 Loam 1.50 1.3 7.1
70–120 26.9 34.4 38.7 Loam 1.50 1.7 7.1
a
Using pipette method (Scott, 2000) after organic matter was removed.
b
Using dichromate method (Margesin and Schinner, 2005).

present analysis they are averaged and analyzed on a daily time inversely related to the number of reflected pulses, to the VWC
scale. (Blonquist et al., 2005). The frequency of pulsing in free air is about
70 MHz. The CS616 measures in a range of 0–1 m3/m3 and operates
2.2.1. TDR based sensor TRIME-IT/-EZ between 25 and +50 °C. The accuracy between 0 and 0.50 m3/m3
The TRIME-IT and TRIME-EZ sensors are based on the TDR tech- is reported to be ±0.025 m3/m3 using standard calibration with a
nique. This technique makes use of the travel time of an electro- bulk electrical conductivity 60.5 dS/m and a bulk density of
magnetic impulse propagating along the rods of a sensor. The 1.55 g/cm3 (Campbell Scientific, 2006). The sensor consists of two
signal is reflected at the end of the rods and the returned signal parallel round rods of 300 mm length, 3.2 mm diameter, and a
is sampled. The travel time of the electromagnetic impulse is re- spacing of 32 mm.
lated to the permittivity of the medium, which can itself be related The sensor output rawCS616 (ls) was transformed to VWCCS616
to the VWC (Blonquist et al., 2005). The TRIME-IT/-EZ operate at a (m3/m3) using the standard calibration provided by the manufac-
frequency of 1 GHz. They measure in a range of 0–1 m3/m3 and turer (Campbell Scientific, 2006):
operate at temperatures ranging from 15 to +50 °C. The accuracy
VWCCS616 ¼ 0:0663 0:0063rawCS616 þ 0:0007raw2CS616 : ð3Þ
is reported to be of ±0.01 m3/m3 for 0–0.40 m3/m3 and ±0.02 m3/
m3 for 0.40–0.70 m3/m3 in soils with bulk electrical conductivity The manufacturer reports an error in measured VWC caused by
of up to 2 dS/m (IMKO, 2006). The TRIME-IT and TRIME-EZ sensors the temperature dependency of the CS616 sensor, and provides the
are similar sensors that have only different dimensions. The following correction equation to be applied to the uncorrected sen-
TRIME-IT instrument consists of two parallel round rods of sor output rawCS616woT (ls) using the soil temperature Tsoil (°C) at
110 mm length with 3.5 mm diameter, and a spacing of 20 mm. the specific installation depth (Campbell Scientific, 2006):
It has a smaller measurement volume than the TRIME-EZ, which
also has two round metal rods but of 160 mm length with 6 mm rawCS616withT ¼ rawCS616woT þ ð20 T soil Þð0:526
diameter, and a spacing of 40 mm. Due to its smaller measurement 0:052rawCS616woT þ 0:00136raw2CS616woT Þ: ð4Þ
volume, the TRIME-IT was installed at 5 and 15 cm depth, whereas
the TRIME-EZ was installed at deeper depths down to 110 cm. The In this study both the rawCS616withT and the rawCS616woT sensor
sensor output rawTRIME (mV) range between 0 and 1000 mV and is output were transformed to VWC. The respective CS616 estimates
linearly related to the VWCTRIME (m3/m3): are distinguished as CS616wT (with temperature correction) and
CS616woT (without temperature correction).
VWCTRIME ¼ rawTRIME =1000: ð1Þ The SISOMOP sensor is made up of a ring oscillator based on a
digital inverter, driving a transmission line, whose end is fed into
2.2.2. Capacitance based sensor 10HS its input (Schlaeger, 2007b). The resulting oscillation frequency is
The 10HS sensor is based on the capacitance technique. The expressed as moisture counts (MC) and has a relative accuracy of
charging time of an electromagnetic field is related to the capaci- the permittivity of ±4%. The sensor measures in a permittivity
tance of the soil, which is related to the permittivity of the medium. range of 5–25 and operates at 5 to 60 °C (Schlaeger, 2007b).
The permittivity can then be related to the VWC of the soil. The 10HS The measurement range of VWC is not explicitly provided by the
is the successor of the commonly used EC-5 soil moisture sensor, manufacturer. The sensor consists of one flat pronged plastic rod
but integrates over a larger soil volume (rod length is 10 cm instead with 100 mm length and 30 mm width. The exponential relation-
of 5 cm). It operates at a frequency of 70 MHz with a measurement ship between the moisture counts (MC) and the VWCSISOMOP re-
range indicated by the manufacturer between 0 and 0.57 m3/m3. quires a material dependent calibration according to Krauss et al.
The instrument operates between 0 and +50 °C. The accuracy using (2010):
the standard calibration is reported to be ±0.03 m3/m3 in mineral VWCSISOMOP ¼ a expðbMCÞ=100: ð5Þ
soils that have a solution electrical conductivity <10 dS/m. Using a
soil specific calibration the accuracy is reported to be ±0.02 m3/m3 The manufacturer provides a sensor and soil specific calibration
(Decagon Devices, 2009). The sensor consists of two parallel- based on Krauss et al. (2010).
pronged plastic rods of 100 mm length and 9.8 mm width, and a
spacing of 12.1 mm. The sensor reading raw10HS (mV) was trans- 2.3. Sensor comparison
formed to VWC10HS (m3/m3) using the standard calibration function
provided by the manufacturer (Decagon Devices, 2009): All investigations were performed based on daily-averaged
values from 1 June 2009 to 31 May 2011 focusing on the VWC
VWC10HS ¼ 2:97 109 raw310HS 7:37 106 raw210HS þ 6:69 obtained by applying the respective calibration function by the
103 raw10HS 1:92: ð2Þ manufacturer. Because of their reported accuracy and the tested
accuracy under laboratory conditions (e.g. Mittelbach et al.,
2011) the TRIME-IT/-EZ sensors were taken as reference for the
2.2.3. FDR based sensors CS616 and SISOMOP following analysis. As a quality check for the TRIME-IT/-EZ field
Both the CS616 and SISOMOP sensors are based on the FDR measurements the relation to the lysimeter weight was taken into
technique. The CS616 sensor directly relates the period, which is account. Given by a divergent behavior of the relation the time
H. Mittelbach et al. / Journal of Hydrology 430–431 (2012) 39–49 43

period 23 September 2010–15 April 2011 was excluded from the significant body of literature exists on the evaluation of different
analysis. The measurements of the soil moisture sensors 10HS, measurement techniques and their dependency on different fac-
CS616 (CS616wT and CS616woT), and SISOMOP were compared with tors, such as measurement frequency, temperature, soil texture,
respect to the absolute daily VWC (m3/m3) and its anomalies rela- salinity, as well as VWC and its consequent impact on measure-
tive to the average VWC over the investigated period. The RMSD ment accuracy (e.g. Wraith and Or, 1999; Kelleners et al., 2005;
and correlations with respect to VWCTRIME were calculated for each Blonquist et al., 2005; Evett et al., 2006; Escorihuela et al., 2007).
measurement depth over the four seasons, winter (DJF), spring
(MAM), summer (JJA), and fall (SON), as well as for the entire con- 3.1. Volumetric water content and its anomalies
sidered period. In addition, the absolute error of daily VWC depen-
dent on the frequency distributed VWCTRIME was analyzed. In the following analysis, daily VWCTRIME measurements are
Therefore, the VWCTRIME was merged over all depths and binned used as reference. As displayed in Fig. 1 VWCTRIME is found to be
in 0.05 m3/m3 intervals and the difference between the tested sen- most variable at 5 and 15 cm depth over the entire measurement
sor and the TRIME-IT/-EZ was calculated for each bin. period. VWCTRIME has minimum values (in spring 2011) of about
Furthermore, the effect of Tsoil on the obtained daily VWC was 0.25 and 0.32 m3/m3 with a measurement range of about 0.42
assessed for each installation depth. By assuming TRIME-IT/-EZ and 0.47 m3/m3 at these two depths. By contrast, VWCTRIME at
as physically correct the test criterion was the difference in VWC 25 cm depth shows a similar minimum at 0.31 m3/m3 but a
to the other sensor types as used in e.g. Verhoef et al. (2006): strongly reduced measurement range of only about 0.09 m3/m3.
At 25 cm depth and below, the VWCTRIME measurements show a
DVWC ¼ ðVWCtest VWCTRIME Þ ðVWCtest VWCTRIME Þ: ð6Þ clear continuous decrease in variability with a nearly steady
behavior for the whole observation period and similar VWC for
all depths. The relatively shallow extent of soil moisture dynamics
2.4. Comparison of change in integrated column soil moisture with
in Rietholzbach was also identified from long-term TDR measure-
evapotranspiration measurements from a lysimeter
ments within the lysimeter (Seneviratne et al., 2012).
Fig. 1 shows further that the 10HS measurements do not exceed
As last criterion, the estimation of evapotranspiration (ET) using
0.40 m3/m3 independently of depth. In addition, the 10HS sensors
the soil water balance approach was evaluated using the lysimeter
do not capture the daily VWC fluctuations for moist conditions.
measurements available at the Rietholzbach site (Section 2.1). For
This effect has the highest influence at 5 and 15 cm depth, where
precipitation-free periods and with the assumption of no drainage,
the VWCTRIME shows highest variations. At 25 cm depth and below,
ET over the considered time period can be approximately assumed
the 10HS measurements are close to the VWCTRIME measurements,
to be equal to the change in weight of the lysimeter as well as to
but the soil moisture fluctuations are almost negligible at these
the change in absolute integrated column soil moisture storage S
depths. The CS616wT measurements display a higher variability
(mm) using soil moisture measurements. To minimize the effect
than the 10HS measurements but tend to overestimate high water
of drainage, only drying periods starting on the fourth day after a
contents. Moreover, the CS616wT estimates, which use the temper-
precipitation event were taken into account. As reference, hourly
ature correction provided by the manufacturer (Section 2.2.3),
data of weight and outflow from the weighing lysimeter at the Rie-
present significant artifacts at 25 cm depth and below, with appar-
tholzbach site were available. Three dry periods were investigated
ent variability in estimated VWC related to the measured soil tem-
and are highlighted in Fig. 1: 8 September–13 September 2009
perature. These features appear erroneous compared to the
(6 days), 24 September–5 October 2009 (12 days), and 25 June–4
VWCTRIME measurements. In contrast, the CS616woT estimates,
July 2010 (10 days). For each sensor type, S was calculated at an
which do not use the temperature correction, reach values of up
hourly time step by integrating the VWC measurements over the
to 0.47 m3/m3 and show a lower VWC range. Nevertheless, the
whole soil column z (from the surface down to 110 cm for
CS616woT performs reasonably well at 5 and 15 cm depths. The
TRIME-IT/-EZ, 10HS, CS616 and down to 80 cm for SISOMOP). As
SISOMOP sensor shows an underestimation of VWC with an upper
integration method, we used the trapezoidal method (e.g.
limit around 0.53 m3/m3, but indicates acceptable VWC variability
Hupet et al., 2004) including an additional value VWC at the
at all depths.
surface which is assumed to be equal to the measurement VWC
The RMSD for the measured absolute and relative VWC and the
in 5 cm:
correlation of the absolute VWC with respect to the VWCTRIME val-
Z N ues are shown in Fig. 2 at seasonal scale and for the overall period.
SðtÞ ¼ VWCðt; zÞdz For all sensor types, the RMSD of the absolute VWC (Fig. 2a) at the
0
XN different measurement depths are overall lowest in JJA, when the
ðVWCðt; zi Þ þ VWCðt; ziþ1 Þ lowest VWCTRIME occur. For each sensor type, the highest RMSD
ffi ðziþ1 zi Þ ð7Þ
i¼1
2 are found at different depths with similar ranking for all seasons.
The 10HS displays highest RMSD of about 0.18–0.27 m3/m3 at 5
where z indicates the depth, N the deepest measurement depth, the
and 15 cm depth, where the VWCTRIME is highest and most vari-
subscript i indicates the respective measurement level and t indi-
able. The lower and more steady VWCTRIME at depths from 25 to
cates the time. The change in absolute integrated column soil mois-
110 cm and similar behavior of the 10HS measurements leads to
ture storage S for these precipitation-free periods was estimated by
clearly smaller RMSD in absolute VWC in all seasons. The cumula-
the difference of the last (S(T)) and the first value (S(0)) over the
tive distribution function (cdf) for all measurement depths and
considered time period:
over the whole measurement period (Fig. 3a) confirms that the
DS ¼ SðTÞ Sð0Þ: ð8Þ RMSD of the 10HS in 5 and 15 cm depth is mainly related to its lim-
itation to measure VWC above 0.4 m3/m3. The probability of low
VWCTRIME agrees well with the probability of the TRIME-IT/-EZ.
3. Results For VWCTRIME above 0.3 m3/m3 the cdfs shows almost no flattening
of the curve, which results in a underestimation. The CS616wT esti-
While this paper intends to compare the VWC obtained by the mates show for all seasons the highest RMSD at 25 and 35 cm
sensor types and does not intend to characterize the permittivity depth with values of up to 0.32 and 0.27 m3/m3 (Fig. 2a), respec-
and its relation to the VWC, it has to be acknowledged that a tively. The cdfs (Fig. 3a) indicate that the RMSD in 15–110 cm
44 H. Mittelbach et al. / Journal of Hydrology 430–431 (2012) 39–49

all DJF MAM JJA SON

a
0.3 0.3 0.3 0.3 0.3
RMSD (m /m )
3
3

0.2 0.2 0.2 0.2 0.2

0.1 0.1 0.1 0.1 0.1

0 0 0 0 0

b
0.1 0.1 0.1 0.1 0.1
RMSDanom(m /m )
3

0.08 0.08 0.08 0.08 0.08

3

0.06 0.06 0.06 0.06 0.06

0.04 0.04 0.04 0.04 0.04

0.02 0.02 0.02 0.02 0.02

0 0 0 0 0

c
1 1 1 1 1
correlation (−)

0.8 0.8 0.8 0.8 0.8

0.6 0.6 0.6 0.6 0.6

0.4 0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2 0.2

0 0 0 0 0
10HS CS616wT CS616 woT SISOMOP 10HS CS616wT CS616woT SISOMOP 10HS CS616wT CS616woT SISOMOP 10HS CS616wT CS616woT SISOMOP 10HS CS616wT CS616woT SISOMOP

5 cm 15 cm 25 cm 35 cm 55 cm 80 cm 110 cm

3 3
Fig. 2. RMSD (m /m ) of (a) VWC measurements, (b) anomalies of VWC, and (c) the correlation between the sensor to be tested and the reference sensor (TRIME-IT/-EZ) in
daily resolution for the seasons DJF, MAM, JJA and SON as well as for the whole period (1 June 2009–31 May 2011).

represent an overestimation of the VWC. Furthermore, from 25 cm anomalies, indicating dry conditions, are captured well. In contrast,
on the distribution results in a different shape displaying a larger the 10HS does not capture extreme dry and wet conditions of the
measurement range of the CS616wT. By contrast, the CS616woT esti- cdf at the uppermost two depths. As for the cdfs for the absolute
mates display a marked different behavior: The RMSD values for VWC (Fig. 3a), the CS616wT display a different shape than the
the absolute VWC are highest at 5 to 25 cm depth with values of VWCTRIME.
up to 0.16 m3/m3 (Fig. 2a) resulting in over- and underestimation The correlations between the TRIME-IT/-EZ measurements and
of the VWCTRIME (Fig. 3a). The measurements at the other depths the estimates derived from the three tested sensor types (Fig. 2c)
result in clearly smaller RMSD and display a good agreement with show high values (r > 0.8) for the near surface depths. A clear linear
distribution of the reference sensors. Similar to the 10HS sensor, relationship (r P 0.6) is found for 10HS and SISOMOP in all depths
the SISOMOP sensor shows highest RMSD for the first two mea- for all seasons. The correlation for CS616wT is generally above 0.5
surement depths with values of up to 0.17–0.31 m3/m3 (Fig. 2a). but displays more variations between the single depths and sea-
However, its cdfs display a similar shape as the reference sensor, sons. If the temperature correction is not applied (CS616woT), the
but shifted to lower VWCs (Fig. 3a). correlation is lower and more variable for the different depths
The RMSD for the measured VWC anomalies (Fig. 2b) display far and seasons.
smaller values compared to the RMSD of the absolute VWC for all The absolute error of VWC for the tested sensors as a function of
tested sensor types. Note that the anomalies are computed with re- the measured VWCTRIME as well as the frequency histogram of the
spect to the average VWC of the given analyzed time frames (i.e. respective VWCTRIME values are displayed in Fig. 4. A first striking
whole time period for the annual values and respective seasonal feature is that the underestimation of actual VWC by the 10HS
values of the 2 years for the seasonal analyses). Furthermore, the and SISOMOP sensors increases with increasing VWC. Nonetheless,
ranking of depths with smallest RMSD is nearly the same for all for 74% of the data (VWC range of 0.30–0.45 m3/m3) the 10HS mea-
four estimates and three sensor types. All show the highest error surements underestimate the VWCTRIME with a maximum error of
in the first two measurement depths where highest and most var- about 0.06 m3/m3. Within the same measurement range, the
iable VWCTRIME is found. Consequently, lowest RMSD are indicated absolute error of the SISOMOP sensor is around 0.16 m3/m3. By
in DJF with lowest VWC variability (temporal evolution of anoma- contrast the CS616wT and CS616woT measurements overestimate
lies are visualized in the Supplementary material). Largest differ- the VWCTRIME within this VWC range by up to 0.15 m3/m3, and
ences can be identified with the CS616wT in all depths as well as 0.06 m3/m3, respectively. Their absolute error decreases with
for all of the tested sensor types in shallow depths (5–15 cm). increasing VWC and shifts to an underestimation for higher VWC
Fig. 3b displays the cdfs for the anomalies. Consistent with the values, which reaches at maximum around 0.08 m3/m3 and
Fig. 2a and b it shows a better agreement between the measure- 0.18 m3/m3 for the CS616wT and CS616woT measurements,
ments for the anomalies than for the absolute values (Fig. 2a). respectively. For higher VWC values (>0.45 m3/m3), the 10HS and
The cdf of VWCTRIME in 5 cm is well represented by CS616wT, SISOMOP sensors underestimate the VWCTRIME by up to
CS616woT, and SISOMOP. Moreover, the probability of negative 0.42 m3/m3.
H. Mittelbach et al. / Journal of Hydrology 430–431 (2012) 39–49 45

a 1 1 1 1

0.8 0.8 0.8 0.8

probability
0.6 0.6 0.6 0.6

0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2

5 cm 15 cm 25 cm 35 cm
0 0 0 0
0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8
3 3
1 1 1 VWC (m /m )

0.8 0.8 0.8

probability

0.6 0.6 0.6

0.4 0.4 0.4

0.2 0.2 0.2

35 cm 80 cm 110 cm
0 0 0
0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8 0.2 0.4 0.6 0.8
3 3 3 3 3 3
VWC (m /m ) VWC (m /m ) VWC (m /m )

b 1 1 1 1

0.8 0.8 0.8 0.8

probability

0.6 0.6 0.6 0.6

0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2

5 cm 15 cm 25 cm 35 cm
0 0 0 0
−0.2 0 0.2 −0.2 0 0.2 −0.2 0 0.2 −0.2 0 0.2
3 3
1 1 1 anomV WC (m /m )

0.8 0.8 0.8

probability

0.6 0.6 0.6 TRIME

10HS
0.4 0.4 0.4 CS616wT
CS616 woT
0.2 0.2 0.2
35 cm 80 cm 110 cm SISOMOP
0 0 0
−0.2 0 0.2 −0.2 0 0.2 −0.2 0 0.2
anom (m3/m3) anom (m3/m3) anom (m3/m3)
V WC V WC V WC

Fig. 3. Cumulative distribution function of the different sensors measurements at all depths of (a) VWC measurements and (b) anomalies of VWC.

3.2. Temperature dependency

In this section, we assess the temperature dependency of the

0.1
soil moisture measurements performed with the different sensor
types. Assuming the VWC measurements of TRIME-IT/-EZ to be
abs. Error (m3 /m3 )

0
physically correct we concentrate on the scatter plots of the differ-
−0.1 50
ences of the tested sensor types to VWCTRIME (DVWC) and soil tem-
−0.2 40 perature (Tsoil). Fig. 5 provides the respective analysis by the single
% of data

measurement depths. In general, the lowest errors are found at Tsoil

−0.3 30
of 10 °C, and all low-cost sensor types switch at this temperature
−0.4 20 either from an over- to an underestimation or vice versa. The
10HS, CS616woT, and SISOMOP display a positive relation between
10
Tsoil and DVWC at 5 and 15 cm depth, while at depths from 25 cm
downwards the DVWC is close to zero with only a small slope. It is
0.30 0.40 0.50 0.60 0.70 0.80
important to note that for the 10HS and SISOMOP sensors the rela-
VWC (m3 /m3 ) tion between DVWC and Tsoil at 5 and 15 cm should be interpreted
10HS CS616wT CS616 woT SISOMOP rel histogram in sight of their limitations in measuring actual VWC values rather
than be attributed to an effective temperature dependency. Both
Fig. 4. Absolute errors of VWC measurements for the different sensor types and
relative frequency distribution of VWC measured with TRIME-IT/-EZ. The error is
sensor types have highest RMSD in these depths (Fig. 2a). Under
calculated as the difference between the measurements of the respective sensor the given conditions it is difficult to assess a temperature depen-
and those from the TRIME-IT/-EZ. dency for the measurements. Concerning the CS616wT a contrast-
46 H. Mittelbach et al. / Journal of Hydrology 430–431 (2012) 39–49

0.2 0.2 0.2 0.2

35 cm
Δ VWC (m3/m3)
5 cm 15 cm 25 cm
0.1 0.1 0.1 0.1

0 0 0 0

−0.1 −0.1 −0.1 −0.1

−0.2 −0.2 −0.2 −0.2

0 10 20 0 10 20 0 10 20 0 10 20

0.2 0.2 0.2 T (°C)

s oil
55 cm 80 cm 110 cm
Δ VWC (m /m3)

0.1 0.1 0.1 10HS10

3

CS616 wT
0 0 0 CS616 woT
SISOMOP
−0.1 −0.1 −0.1

−0.2 −0.2 −0.2

0 10 20 0 10 20 0 10 20
T (°C) T (°C) T (°C)
s oil s oil s oil

Fig. 5. Relation between soil temperature and difference in VWC (DVWC) for different installation depths. The data represent the mean and standard deviation using bins of
2 °C.

ing behavior compared to the previous mentioned sensors is iden- This induces an enhanced effect of the temperature dependency on
tified, which is characterized by a strong negative relation with a the estimated change in soil moisture. By contrast the CS616woT
similar slope at all depths. estimates display an overestimation of 4.8 mm (42%) and 0.8 mm
(5%), respectively for the two events. In the case of the SISOMOP
sensor, one should take into account that the integration of VWC
3.3. Soil water storage estimation and comparison to lysimeter was only possible down to 80 cm with no measurement at 25 cm
measurements depth. Since the Rietholzbach site shows low variation in VWC
for depths below 25 cm (Fig. 1), the measurements of the SISOMOP
Fig. 6 displays the changes in soil moisture over the whole soil sensor should nonetheless be comparable to those of the TRIME-
column (mm) for three precipitation free periods (see Fig. 1) by IT/-EZ sensor. The analysis suggests that the SISOMOP sensor rep-
applying the soil water balance approach for each sensor type resents both events reasonably well with an overestimation of
(see Section 2.4 for more details). The comparison with changes about 0.7 mm (6%) and an underestimation of about 2.5 mm
in lysimeter estimates shows that the TRIME-IT/-EZ performs best (14%) for the first and second event, respectively.
for the first two events (Fig. 6a and b) with an overestimation of
about 10 mm (11%) and underestimation of about 0.8 mm (4%),
respectively. By contrast, for the third event (Fig. 6c) the change 4. Discussion
in soil moisture is overestimated by about 15.5 mm (35%) with this
sensor type. Nevertheless, this overestimation is likely related to Two years of field measurements with a TDR sensor and three
the wetter initial conditions compared to events 1 and 2: The ante- non-TDR low-cost sensor types operated side-by-side are used to
cedent rainfall for the third event is 118 mm (10 days of rainfall) evaluate the performance of the low-cost sensors using the respec-
compared to 33 mm (5 days of rainfall) and 15 mm (7 days of rain- tive calibration functions provided by the manufacturer. This study
fall) for the first and second event, respectively. Furthermore, the clearly shows that none of the investigated low-cost sensors can
outflow from the lysimeter during the third event is higher and still satisfactorily capture the actual VWC (VWCTRIME) under the given
continuously decreasing during the whole period. The initial condi- conditions. With a RMSD of up to 0.3 m3/m3 in particular in
tions of the third event and the resulting non-steady outflow leads near-surface layers, none of the sensors justifies the performance
to the conclusion that the soil water balance approach to estimate specified by the user manual. A seasonal dependency of the RMSD
evapotranspiration is not appropriate for this event. Hence, further is found with smallest error in the summer season. Nonetheless,
comparisons focus on the first two events. even in the summer season, this error can still be as large as
The 10HS sensor underestimates the evapotranspiration for the 0.2 m3/m3. The large RMSD of the measurements is far larger than
first and second event by 5.5 mm (48%) and 10.2 mm (58%), respec- the requirements for the calibration and validation of e.g. passive
tively. This can be mainly explained by the missing drying in the remote sensing algorithms (Jackson et al., 2010). Previous studies
upper depths due to the small sensitivity of the 10HS under the gi- concerning the evaluation of soil moisture sensors identified the
ven moisture conditions. The CS616 overestimates the first event measurement frequency to affect the determination of the permit-
by 2.6 mm (22%) but strongly underestimates the change in soil tivity and thus the performance of VWC measurements to a large
moisture by 11 mm (63%) for the second event. The strong under- part (see e.g. Blonquist et al., 2005; Escorihuela et al. 2007; Kizito
estimation of the second event is likely due to the temperature et al., 2008). We find that each low-cost sensor has specific issues
dependency of the CS616, which affects the measured changes in in accurately measuring VWC and that none clearly outperforms
soil moisture for the rather steady VWC at depths below 25 cm. the others applying the calibration functions provided by the
Furthermore, for the second event the daily change in soil temper- respective manufacturers.
ature over the whole soil column (change in mean soil temperature The 10HS sensor has two main limitations: On the one hand, it
over 5–110 cm relative to the duration of the event) is nearly three fails to measure VWC above 0.4 m3/m3 and on the other hand
times higher and the event itself is twice as long as the first event. it presents a decreasing sensitivity in measuring VWC with
H. Mittelbach et al. / Journal of Hydrology 430–431 (2012) 39–49 47

a 0 b0 c0
−20 −20 −20

storage (mm)
−40 −40 −40

−60 −60 −60

−80 ΔT5 -110 cm= +0.05 K/d −80 ΔT5 -110 cm = +0.14 K/d −80 ΔT5 -110 cm = -0.34 K/d
(mm)

1.5 1.5 1.5

LY S I

1 1 1
0.5 0.5 0.5
0 0 0
2 4 2 4 6 8 10 2 4 6 8
day of event day of event day of event
TRIME 10HS CS616wT CS616woT SISOMOP lysimeter outflow

Fig. 6. Changes in absolute soil moisture and changes in temperature per day for the three precipitation-free periods (a) 8 September–13 September 2009, (b) 24 September–
5 October 2009, and (c) 25 June–4 July 2010 for the TRIME-IT/-EZ, 10HS, CS616wT, CS616woT and SISOMOP sensors. The blue line indicates the change by the lysimeter for each
period. The bottom figures show the daily outflow of the lysimeter. (For interpretation of the references to colour in this figure legend, the reader is referred to the web
version of this article.)

increasing VWC (see also Mittelbach et al., 2011). Both issues result suggest a soil specific calibration with temperature compensation
in a poor ability of the sensor in representing the variability in in particular for fine-grained soils. Rüdiger et al. (2010) established
VWC for moist conditions. This confirms the findings of Mittelbach a general equation for the CS616 measurements including the soil
et al. (2011), where this dependency was shown to be induced by type, based on the Australian soil classification. However, although
the lack of sensitivity of the sensor reading under these conditions. the soil texture of the investigated site is known, it is based on an-
These limitations and the given moist conditions at the Rietholz- other soil classification scheme (here USDA) and the soil type-spe-
bach site lead to an underestimation of changes in soil water stor- cific parameters of the formula can thus only be estimated.
age for dry-down periods of around 50%. Furthermore, under moist Consequently the application of the equation results in inadequate
conditions, it is barely possible to distinguish between the temper- values for the present measurements (not shown).
ature dependency of the 10HS and its problems in measuring the The performance of the SISOMOP sensor regarding the mea-
VWC. Nevertheless, a temperature dependency of the 10HS is ex- sured VWC and temperature dependency is of similar magnitude
pected as it operates at a low measurement frequency (70 MHz) as for the other two (more commonly used) low-cost soil moisture
which affects the sensitivity of sensors measurements to tempera- sensors tested in this study. It thus seems to be an equivalent alter-
ture (Kelleners et al., 2005; Kizito et al., 2008). Furthermore, its native within this type of sensors taking into account the overall
forerunner model (EC-5) was shown to be sensitive to temperature limitations documented here.
variations (Bogena et al., 2007). The above discussion concerns the performance of the tested
The CS616wT estimates, i.e. the CS616 measurements with the sensors in capturing the absolute VWC. In climate research, the
applied temperature correction, are found to overcorrect (exces- VWC variability as well as extreme VWC conditions are generally
sive increase) the original VWC estimates (CS616woT) under the gi- more relevant than the absolute VWC (Seneviratne et al., 2010).
ven field conditions. In addition, the manufacturers correction With respect to the VWC anomalies, all low-cost sensors have in
superimpose an unrealistic strong seasonal soil temperature signal, common strongly lower RMSD and more similar cdfs compared
which has a major influence on the CS616wT estimates in depths to those for the absolute values. Hence, they are found to perform
with little VWC variation but significant soil temperature varia- better for long-term anomalies, i.e. to capture the dynamics of
tions, and leads to an overestimation of VWC above 10 °C and an VWC. The maximum RMSD in summer is about 0.07 m3/m3 at
underestimation of VWC below 10 °C. Based on our analysis we depths of 5 and 15 cm. Our results confirm reported better perfor-
identified this behavior as spurious temperature dependency, mance of low-cost electromagnetic soil moisture sensors for dry
which is not directly related to effects on the permittivity, e.g. conditions than for saturated conditions (e.g. Evett et al., 2006;
based on measurement frequency, soil texture, and electrical con- Mittelbach et al., 2011). It has been suggested that the error in both
ductivity (e.g. Persson and Berndtsson, 1995; Kizito et al., 2008), the absolute VWC and its anomalies can be reduced for this sensor
but in this case is induced by the temperature correction of the type with a site-specific calibration (e.g. Ventura et al., 2010).
manufacturer. By not considering this temperature correction, Nonetheless, also the actual sensitivity of the sensor needs to be
the CS616woT is not able to represent the variability of the VWC considered and may impair such calibrations (Mittelbach et al.,
and in addition, it shows the lowest correlation to the TDR mea- 2011).
surements. Compared to the TDR-based sensor the CS616woT The spurious temperature dependency of the daily data was
showed a low temperature dependency, which confirms findings particularly strong for the CS616wT estimates, and especially af-
from several studies. Blonquist et al. (2005) identified a similar fects the evapotranspiration estimates derived with the water-bal-
behavior but under temperature-controlled test conditions. In their ance approach in Section 3.3 (Fig. 5), particularly for longer dry-
study, the over- to underestimation change point of the CS616 was down periods.
found at 25 °C for a media with higher permittivity than that of the One should note that beside the accuracy of soil moisture sen-
Rietholzbach site. The identified temperature dependency for the sors, also their design, and in particular their geometrical shape,
clayey soil confirms the study by Benson and Wang (2006), which is relevant for field applications (see Section 1). A first difficulty
48 H. Mittelbach et al. / Journal of Hydrology 430–431 (2012) 39–49

exists for sensors with long rods (e.g. CS616), which are in principle Basara, J.B., Crawford, T.M., 2000. Improved installation procedures for deep-layer
soil moisture measurements. J. Atmos. Ocean. Technol. 17 (6), 879–884.
advantageous because of their larger measurement volume, but
Benson, C.H., Wang, X., 2006. Temperature-compensating calibration procedure for
their installation in stony and clayey soil can be challenging. The Water Content Reflectometers. In: Proc. TDR 2006, Purdue University, West
second difficulty is found for sensors with relatively large and hea- Lafayette, USA, September 2006, Paper ID 50.
vy sensor bodies (e.g. TRIME-IT/-EZ) whose centroid is not close to Blonquist Jr., J.M., Jones, S.B., Robinson, D.A., 2005. Standardizing characterization of
electromagnetic water content sensors: Part 2. Evaluation of seven sensing
the sensor rods. The resulting pressure on the sensor rods can lead systems. Vadose Zone J. 4 (4), 1059–1069.
to a decreasing contact of the installed rods with the surrounding Bogena, H.R., Huisman, J.A., Oberdörster, C., Vereecken, H., 2007. Evaluation of a
soil. We do not assess the impact of these effects on the measure- low-cost soil water content sensor for wireless network applications. J. Hydrol.
344, 32–42.
ments, but these could also explain some of the identified Brocca, L., Melone, F., Moramarco, T., Morbidelli, R., 2010. Spatio-temporal
discrepancies. variability of soil moisture and its estimation across scale. Water Resour. Res.
46, W02516.
Campbell Scientific, 2006. CS616 and CS625 Water Content Reflectometers,
5. Conclusions Instruction Manual, Revison: 8/06.
Czarnomski, N., 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
This study is one of the first studies comparing the 10HS sensor two forest soils of the Pacific Northwest. Can. J. Forest Res. 35 (8), 1867–1876.
with other frequently used soil moisture sensors over a more than De Jeu, R., Wagner, W., Homes, T.R.H., Dolman, A.J., van de Giesen, N.C., Friesen, J.,
1 year measurement period. The results confirm the need for a site- 2008. Global soil moisture patterns observed by space borne microwave
radiometers and scatterometers. Surv. Geophys. 29 (4–5), 399–420.
specific calibration of low-cost sensors, including temperature cor-
Decagon Devices, 2009. 10HS Soil Moisture Sensor Operator’s Manual, Version 2.0.
rections. This is in agreement with results from previous studies, Diffenbaugh, N.S., Pal, J.S., Giorgi, F., Gao, X., 2007. Heat stress intensification in the
which also used 10HS and CS616 sensors as well as their forerun- Mediterranean climate change hotspot. Geophys. Res. Lett. 34 (11), 1–6.
Dirmeyer, P.A., Koster, R.D., Guo, Z., 2006. Do Global Models Properly Represent the
ner (EC-5 and CS615, respectively), but that were partly conducted
Feedback between Land and Atmosphere? J. Hydrometeorol. 7 (6), 1177–1198.
under different soil and meteorological conditions (e.g. Benson and Dorigo, W.A., Wagner, W., Hohensinn, R., Hahn, S., Paulik, C., Drusch, M.,
Wang, 2006; Bogena et al., 2007; Logsdon, 2009; Rüdiger et al., Mecklenburg, S., van Oevelen, P., Robock, A., Jackson, T., 2011. The
2010; Mittelbach et al., 2011). For the temperature correction, International Soil Moisture Network: a data hosting facility for global in situ
soil moisture measurements. Hydrol. Earth Syst. Sci. Discuss. 8, 1609–1663.
one should note that parallel installed temperature sensors would Escorihuela, M.J., De Rosnay, P., Kerr, Y.H., Calvet, J.C., 2007. Influence of bound-
be of advantage. If a site-specific calibration is established, low- water relaxation frequency on soil moisture measurements. IEEE T. Geosci.
cost sensors may be a viable alternative to TDR sensors for certain Remote 45 (12), 4067–4076.
Evett, S.R., Tolk, J.A., Howell, T.A., 2006. Soil profile water content determination:
environmental applications (Seyfried and Murdock, 2004). None- Sensor accuracy, axial response, calibration, temperature dependence, and
theless, our results highlight significant weaknesses of these sen- precision. Vadose Zone J. 5 (3), 894–907.
sors such as (dependent on the sensor) a lack of sensitivity in Famiglietti, J.S., Devereaux, J.A., Laymon, C.A., Tsegaye, T., Houser, P.R., Jackson, T.J.,
Graham, S.T., Rodell, M., van Oevelen, P.J., 1999. Ground-based investigation of
moist soil moisture regimes or a spurious dependency on soil tem- soil moisture variability within remote sensing footprints during the Southern
perature. These imply intrinsic issues with the measurements de- Great Plains 1997 (SGP97) Hydrology Experiment. Water Resour. Res. 35 (6),
rived with this type of instruments. 1839–1851.
Findell, K.L., Gentine, P., Lintner, B.R., Kerr, C., 2011. Probability of afternoon
The results claim for a combination of high-accuracy and low-
precipitation in eastern United States and Mexico enhanced by high
cost sensors in the design of soil moisture measurement networks evaporation. Nat. Geosci. 4 (7), 434–439.
and highlight the importance to evaluate and compare soil mois- Fischer, E.M., Seneviratne, S.I., Vidale, P.L., Lüthi, D., Schär, C., 2007. Soil moisture–
atmosphere interactions during the summer 2003 heatwave. J. Clim. 20, 5081–
ture sensors under different soil characteristics (texture, tempera-
5099.
ture, bulk density, and salinity) and under different moisture Hirschi, M., Seneviratne, S.I., Alexandrov, V., Boberg, F., Boroneant, C., Christensen,
regimes. This would allow a better quantification of the accuracy O.B., Formayer, H., Orlowsky, B., Stepanek, P., 2011. Observational evidence for
of in situ measurements. This consideration is particularly critical soil-moisture impact on hot extremes in southeastern Europe. Nat. Geosci. 4 (1),
17–21.
for a number of environmental, climate, and hydrological applica- Hohenegger, C., Brockhaus, P., Bretherton, C., Schär, C., 2009. The soil moisture–
tions, including the assessment of remote sensing measurements precipitation feedback in simulations with explicit and parameterized
and the evaluation of land surface, hydrological and climate mod- convection. J. Climate 22 (19), 5003–5020.
Hupet, F., Bogaert, P., Vanclooster, M., 2004. Quantifying the local-scale uncertainty
els. In particular, the error ranges of the respective sensors should of estimated actual evapotranspiration. Hydrol. Process. 18 (17), 3415–3434.
be assessed in such applications. IMKO, 2006. TRIME-EZ /-EZC /-IT /-ITC. User manual, IMKO Micromodultechnik
GmbH.
Jackson, T.J., Cosh, M.H., Bindlish, R., Starks, P.J., Bosch, D.D., Seyfried, M., Goodrich,
Acknowledgments D.C., Moran, M.S., Du, J.Y., 2010. Validation of Advanced Microwave Scanning
Radiometer Soil Moisture Products. IEEE T. Geosci. Remote 48 (12), 4256–4272.
The SwissSMEX project is supported by the Swiss National Jaeger, E.B., Seneviratne, S.I., 2011. Impact of soil moisture–atmosphere coupling on
European climate extremes and trends in a regional climate model. Clim.
Foundation (project 200021#120289). We gratefully acknowledge Dynam. 36 (9–10), 1919–1939.
the support from Liane Krauss in providing the sensor specific cal- Jung, M., Reichstein, M., Ciais, P., Seneviratne, S.I., Sheffield, J., Goulden, M.L., Bonan,
ibration for the SISOMOP instruments. We are grateful to two G., Cescatti, A., Chen, J., de Jeu, R., Dolman, A.J., Eugster, W., Gerten, D., Gianelle,
D., Gobron, N., Heinke, J., Kimball, J., Law, B.E., Montagnani, L., Mu, Q., Mueller,
anonymous reviewers for helpful comments on the manuscript. B., Oleson, K., Papale, D., Richardson, A.D., Roupsard, O., Running, S., Tomelleri,
E., Viovy, N., Weber, U., Williams, C., Wood, E., Zaehle, S., Zhang, K., 2010. Recent
decline in the global land evapotranspiration trend due to limited moisture
Appendix A. Supplementary material supply. Nature 467, 951–954.
Kelleners, T.J., Robinson, D.A., Shouse, P.J., Ayars, J.E., Skaggs, T.H., 2005. Frequency
Supplementary data associated with this article can be found, in dependence of the complex permittivity and its impact on dielectric sensor
calibration in soils. Soil. Sci. Soc. Am. J. 69 (1), 67–76.
the online version, at doi:10.1016/j.jhydrol.2012.01.041.
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
References low-cost capacitance soil moisture sensor. J. Hydrol. 352 (3–4), 367–378.
Koster, R.D., Dirmeyer, P., Guo, Z., Bonan, G., Chan, E., Cox, P., Gordon, C., Kanae, S.,
Kowalczyk, E., Lawrence, D., Liu, P., Lu, C., Malyshev, S., McAvaney, B., Mitchell,
Albergel, C., Rüdiger, C., Pellarin, T., Calvet, J.-C., Fritz, N., Froissard, F., Suquia, D.,
K., Mocko, D., Oki, T., Oleson, K., Pitman, A., Sud, Y., Taylor, C., Verseghy, D.,
Petitpa, A., Piguet, B., Martin, E., 2008. From near-surface to root-zone soil
Vasic, R., Xue, Y., Yamada, T., Team, G.L.A.C.E., 2004. Regions of strong coupling
moisture using an exponential filter: an assessment of the method based on
between soil moisture and precipitation. Science 305, 1138.
in situ observations and model simulations. Hydrol. Earth Syst. Sci. 12, 1323–
Koster, R.D., Mahanama, S., Yamada, T.J., Balsamo, G., Boisserie, M., Dirmeyer, P.,
1337.
Doblas-Reyes, F., Gordon, C.T., Guo, Z., Jeong, J.H., Lawrence, D., Li, Z., Luo, L.,
Albertson, J.D., Montaldo, N., 2003. Temporal dynamics of soil moisture variability:
Malyshev, S., Merryfield, W., Seneviratne, S.I., Stanelle, T., van den Hurk, B.,
1. Theoretical basis. Water Resour. Res. 39 (10), 1274.
H. Mittelbach et al. / Journal of Hydrology 430–431 (2012) 39–49 49

Vitart, Wood, E.F., . The contribution of land initialization to subseasonal Rietholzbach research catchment and lysimeter: Analysis of 32-year
forecast skill: First results from the GLACE-2 Project. Geophys. Res. Lett 37, hydroclimatological time series and 2003 drought.Water Resour. Res., 2012.
L02402. Seyfried, M.S., Murdock, M.D., 2004. Measurement of Soil Water Content with a 50-
Koster, R.D., Mahanama, S.P.P., Livneh, B., Lettenmaier, D.P., Reichle, R.H., 2010b. MHz Soil Dielectric Sensor. Soil Sci. Soc. Am. J. 68, 394–403.
Skill in streamflow forecasts derived from large-scale estimates of soil moisture Taylor, C.M., Gounou, A., Guichard, F., Harris, P.P., Ellis, R.J., Couvreux, F., De Kauwe,
and snow. Nat. Geosci. 3 (9), 613–616. M., 2011. Frequency of Sahelian storm initiation enhanced over mesoscale soil-
Krauss, L., Hauck, C., Kottmeier, C., 2010. Spatio-temporal soil moisture variability in moisture patterns. Nature Geosci. 4 (7), 430–433.
Southwest Germany observed with a new monitoring network within the COPS Teuling, A.J., Troch, P.A., 2005. Improved understanding of soil moisture variability
domain. Meteorol. Z.,19(6), 523–537. dynamics. Geophys. Res. Lett. 32 (5), L05404.
Lehner, I., Teuling, A.J., Gurtz, J., Seneviratne, S.I., 2010. Long-term water balance in Teuling, A.J., Seneviratne, S.I., Stöckli, R., Reichstein, M., Moors, E., Ciais, P., Luyssaert,
the prealpine Rietholzbach catchment: First comparison of evapotranspiration S., van den Hurk, B., Ammann, C., Bernhofer, C., Dellwik, E., Gianelle, D., Gielen,
estimates. in Status and Perspectives of Hydrology in Small Basins, 30 March–2 B., Grünwald, T., Klumpp, K., Montagnani, L., Moureaux, C., Sottocornola, M.,
April 2009, IAHS Publ. 336, Goslar-Hahnenklee, Germany. Wohlfahrt, G., 2010. Contrasting response of European forest and grassland
Logsdon, S.D., 2009. CS616 Calibration: Field versus Laboratory. Soil. Sci. Soc. Am. J. energy exchange to heatwaves. Nature Geosci. 3, 722–727.
73 (1), 1–6. Topp, G.C., 2003. State of the art of measuring soil water content. Hydrol. Process.
Lorenz, R., Jaeger, E.B., Seneviratne, S.I., 2010. Persistence of heat waves and its link 17 (14), 2993–2996.
to soil moisture memory. Geophys. Res. Lett. 37, L09703. Van den Hurk, B.J.J.M., van Meijgaard, E., 2010. Diagnosing land–atmosphere
Margesin, R., Schinner, F. (Eds.), 2005. Manual for Soil Analysis-Monitoring and interaction from a Regional Climate Model simulation over West Africa. J.
Assessing Soil Bioremediation. Springer Verlag, Berlin, p. 366. Hydrometeorol. 11 (2), 467–481.
Martínez-Fernández, J., Ceballos, A., 2005. Mean soil moisture estimation using Vautard, R., Yiou, P., D’Andrea, F., de Noblet, N., Viovy, N., Cassou, C., Polcher, J., Ciais,
temporal stability analysis. J. Hydrol. 312, 28–38. P., Kageyama, M., Fan, Y., 2007. Summertime European heat and drought waves
Mittelbach, H., Casini, F., Lehner, I., Teuling, A.J., Seneviratne, S.I., 2011. Soil moisture induced by wintertime Mediterranean rainfall deficit. Geophys. Res. Lett. 34 (7),
monitoring for climate research: Evaluation of a low-cost sensor in the 5.
framework of the Swiss Soil Moisture Experiment (SwissSMEX) campaign. J. Veldkamp, E., O’Brien, J.J., 2000. Calibration of a frequency domain reflectometry
Geophys. Res.-Atmos., 116: 11. sensor for humid tropical soils of volcanic origin. Soil Sci. Soc. Am. J. 64 (5),
Persson, M., Berndtsson, R., 1995. Texture and electrial conductivity effects on 1549–1553.
temperature dependency in time domain reflectometry. Soil Sci. Soc. Am. J. 62, Ventura, F., Facini, O., Piana, S., Rossi Pisa, P., 2010. Soil moisture measurements:
887–893. comparison of instrumentation performances. J. Irrig. Drain. E-Asce 136 (2), 81–
Plauborg, F., Iversen, B.V., Laerke, P.E., 2005. In situ comparison of three dielectric 89.
soil moisture sensors in drip irrigated sandy soils. Vadose Zone J. 4 (4), 1037– Vereecken, H., Kollet, S., Simmer, C., 2010. Patterns in soil-vegetation-atmosphere
1047. systems: monitoring, modeling, and data assimilation. Vadose Zone J. 9 (4),
Robinson, D.A., Campbell, C.S., Hopmans, J.W., Hornbuckle, B.K., Jones, S.B., Knight, 821–827.
R., Ogden, F., Selker, J., Wendroth, O., 2008. Soil moisture measurements for Verhoef, A., Fernandez-Galvez, J., Diaz-Espejo, A., Main, B.E., El-Bishti, M., 2006. The
ecological and hydrological watershed scale observatories: A review. Vadose diurnal course of soil moisture as measured by various dielectric sensors:
Zone J. 7, 358–389. effects of soil temperature and the implications for evaporation estimates. J.
Roth, K., Schulin, R., Fluhler, H., Attinger, W., 1990. Calibration of time domain Hydrol. 321, 147–162.
reflectometry for water-content measurements using a composite dielectric Vinnikov, K.Y., Robock, A., Speranskaya, N.A., Schlosser, A., 1996. Scales of temporal
approach. Water Resour. Res. 26 (10), 2267–2273. and spatial variability of midlatitude soil moisture. J. Geophys. Res.-Atmos. 101
Rüdiger, C., Hancock, G., Hemakumara, H.M., Jacobs, B., Kalma, J.D., Martinez, C., (D3), 7163–7174.
Thyer, M., Walker, J.P., Wells, T., Willgoose, G.R., 2007. Goulburn river Wagner, W., Naeimi, V., Scipal, K., de Jeu, R., Martinez-Fernandez, J., 2007. Soil
experimental catchment data set. Water Resour. Res. 43 (10), 1–10. moisture from operational meteorological satellites. Hydrogeol. J. 15 (1), 121–
Rüdiger, C., Western, A.W., Walker, J.P., Smith, A.B., Kalma, J.D., Willgoose, G.R., 131.
2010. Towards a general equation for frequency domain reflectometers. J. Walker, J.P., Willgoose, G.R., Kalma, J.D., 2004. In situ measurement of soil moisture:
Hydrol. 383, 319–329. a comparison of techniques. J. Hydrol. 293 (1–4), 85–99.
Schlaeger, 2007a. SISOMOP –Control unit, the operation and analysis device for Weisheimer, A., Doblas-Reyes, F.J., Jung, T., Palmer, T.N., 2011. On the
SISOMOP. Schlaeger measurement techniques & science. predictability of the extreme summer 2003 over Europe. Geophys. Res. Lett.
Schlaeger, 2007b. SISOMOP – Simple Soil Moisture Probe. Schlaeger measurement 38, L05704.
techniques & science. Western, A.W., Grayson, R.B., Blöschl, G., Willgoose, G.R., McMahon, T.A., 1999.
Scott, H.D., 2000. Soil physics: Agricultural and environmental applications. Iowa Observed spatial organization of soil moisture and its relation to terrain indices.
State University Press, Ames, 421 pp. Water Resour. Res. 35 (3), 797–810.
Seneviratne, S.I., Lüthi, D., Litschi, M., Schär, C., 2006. Land–atmosphere coupling Western, A.W., Zhou, S.L., Grayson, R.B., McMahon, T.A., Blöschl, G., Wilson, D.J.,
and climate change in Europe. Nature 443, 205–209. 2004. Spatial correlation of soil moisture in small catchments and its
Seneviratne, S.I., Corti, T., Davin, E., Hirschi, M., Jaeger, E.B., Lehner, I., Orlowsky, B., relationship to dominant spatial hydrological processes. J. Hydrol. 286 (1–4),
Teuling, A.J., 2010. Investigating soil moisture-climate interactions in a 113–134.
changing climate. Earth-Sci. Rev. 99 (3–4), 125–161. Wraith, J.M., Or, D., 1999. Temperature effects on soil bulk dielectric permittivity
Seneviratne, S.I., Lehner, I., Gurtz, J., Teuling, A.J., Lang, H., Moser, U., Grebner, D., measured by time domain reflectometry: experimental evidence and
Menzel, L., Schroff, K., Vitvar, T., Zappa, M., 2012. The Swiss pre-alpine hypothesis development. Water Resour. Res. 35 (2), 361–369.