Integrated Sensor for Estimating in situ Soil Water Content in Vertical Profile

Many agricultural and hydrological processes require the detailed knowledge of soil water content (SWC) in the vertical profile. Quantifying real-time and in situ SWC is difficult due to time, cost, toil, and technical issues. This paper describes the development of a multi-depth SWC monitoring sensor which can estimate the SWC from 4 vertical depths simultaneously. The probe is a type of electromagnetic (EM) sensor that indirectly measures the SWC on the basis of dielectric theory. The sensor was calibrated with soil samples of three distinct topographical locations. The calibration models were established by fitting linear order equations. The performance of the sensor was evaluated in situ field conditions. A multi-depth SWC curve was investigated to examine the impact of continuous estimations of SWC at specified depths on the sensor performance. The sensor was integrated with vertical interpolation technique to improve the measurement accuracy. The results indicated the optimal range of the SWC measurements, and the estimation error was less than 5%, except irrigation cycles. The linear fit coefficient of determination (R) ranged from 0.957 to 0.993 and root mean square error (RMSE) was ranging from 1.565 to 4.456. The results showed that the sensor performed consistently better for at least 4 months within acceptable soil conditions. The sensor will be advantageous for continuous estimations of SWC, and managing the irrigation practices.

In recent years, a large number of automatic techniques for scaling the SWC on point and large scale have been developed.The performance of SWC monitoring devices has been tested for a wide variety of soil types in laboratory and field conditions (Evett, Heng, Moutonnet, & Nguyen, 2008;Luhr & Kleisinger, 1998;Romano, 2014;Walker, Willgoose, & Kalma, 2004).The estimation of the dielectric constant of a soil has emerged as an elegant method for measuring the SWC.EM methods, in general, either time domain reflectometry (TDR) for local estimations, ground penetrating radar (GPR), capacitance probes, and Frequency domain reflectometry (FDR) all measure the SWC by estimating the dielectric constant (Dean, Bell, & Baty, 1987;Topp, 2003;Topp, Davis, & Annan, 1980).The other devices such as impedance probes, tensiometers (Chow et al., 2009;Ling, 2004), resistance blocks (Hignett & Evett, 2008;Ling, 2004), heat dissipation type sensor (Hignett & Evett, 2008;Ling, 2004), and neutron sensors could be used for measuring the SWC (Evett et al., 2008).However, the above mention techniques have some problems such as the GPR active microwave and passive microwave operate on scattered EM waves penetrating into the soil or being reflected from it.The vegetation and surface coarseness limit the sensitivity and only can scan the upper depths (~10 cm) (Entekhabi et al., 2010;Li, Toll, Zhan, & Cosgrove, 2012).TDR is a widely used technique to monitor SWC at point scale however it destroys the soil structure and also impacted by material heterogeneity and electrical conductivity at different soil depths.In addition, it is an expensive approach (Baldwin, Manfreda, Keller, & Smithwick, 2017;Greco & Guida, 2008;Stangl, Buchan, & Loiskandl, 2009).The other concerns related to EM based technology include; over-under estimations of SWC with factory specifications, impacted by soil salinity and electric conductivity, and small air gaps could impact the measurements significantly.The poor performance of the probes at 0.35-0.50m soil depths has been reported, and at deeper soil depths (0.60-0.80 m) monitoring ability of the dielectric sensors have been questioned (Chow et al., 2009;Kumar et al., 2009;Leib, Jabro, & Matthews, 2003;Mittelbach, Lehner, & Seneviratne, 2012;Parsons & Bandaranayake, 2009;Romano, 2014;Topp, 2003).In recent decades, various statistical approaches were applied to predict the temporal distribution of SWC.The linear and cubic interpolation techniques were found to be the best for infilling randomly missing moisture values (Fernández-Gálvez, Simmonds, & Barahona, 2006;İmamoğlu & Sertel, 2016;Kornelsen & Coulibaly, 2012).
This study describes the development of a high-resolution integrated sensor for open and controlled environments, which can continuously measure the SWC from 4 vertical depths (0.20 to 0.80 m) simultaneously in different environmental and terrain conditions.At the same time, the sensor can transfer the measured data to the user terminal and data bank.The main objectives of the study were to: (i) design and develop a multi-depth SWC monitoring sensor with a minimum damage to the original soil structure; (ii) integrate with the automatic irrigation systems; (iii) calibrate the sensor with different soil types; (iv) analyze the effect of continuous estimations of SWC at specified depths within the soil horizon on the sensor performance; and, (v) evaluate the performance of the sensor in situ.

Sensor Development and Measurement Principle
A multi-depth (4 depths) SWC monitoring probe was developed in this study.The developed probe is a type of EM sensor that indirectly estimates the SWC on the basis of the dielectric property of soil (Topp et al., 1980).The capacitance of the probe can be measured by two methods.The traditional method uses the frequency measurement technique.The frequency changes with the capacitance, which is influenced by the dielectric constant of the medium (Dean et al., 1987).The dielectric constant of various media also varies such as water is 80, soil (2.4-3.5), and soil minerals dielectric constant ranged from 2.7 to 5.0.Therefore, the change in SWC will impact the soil dielectric constant, subsequently.The second method measures the electrical impedance of the soil at a definite excitation frequency.The sensor functioned at a 100 MHz frequency (Dean, 1994;Kargas & Kerkides, 2009;Kelleners et al., 2004;Stacheder, Koeniger, & Schuhmann, 2009).Figure 1 describes the monitoring principle circuit for SWC monitoring.The sensor outputs a DC voltage that is converted into SWC by the calibration equations already embedded into the sensor.The probe impedance in the medium is determined by the following Equation (1): Where, Z C the impedance of the probe in the air, L represents the probe length, λ 0 wavelength of the sine wave signal in the air; ε represents the dielectric constant of the soil around the probe.In terms of angular frequency of electromagnetic waves ω, then Equation (1) can be simplified as Equation (2): The resistance of the sensing probe is related to the SWC when the probe is placed in air.It can be expressed as follows: jas.ccsenet.

U
Where U 1 , perception electromag dry soil, at The U 1 = 0    Agricultural Sci e. C x is the equ c waves, u i re ce will minimu l frequency 10 developed by a linear fit equation to investigate the structural imbalance impact on the continuous estimations of SWC at a specified number of depths on the sensor performance, when converting sensor output voltage to SWC.

Greenhouse Experiment
The sensor was installed in a commercial greenhouse from June 2017 to September 2017 and SWC was monitored from 4 different 0.05, 0.15, 0.25, and 0.35 m depths all the day long at a 1 h interval; the irrigation system was installed at 0.15 m depth.Three irrigation treatments were applied; irrigation time, irrigated water and the gaps between irrigations were also recorded.

Kunming Peach Field Experiment
The sensor was installed in a Kunming peach field for an open field test.The SWC was measured at 0.20, 0.40, 0.60, and 0.80 m depths all the day long at the 24 h interval from March 2017 to June 2017 which includes the spring and early summer seasons.

Vertical Interpolation Technique
Many agricultural processes require the detailed knowledge of the vertical distribution of SWC.The sensor was integrated with an interpolation technique.The SWC time series were measured at 0.20, 0.40, 0.60 and 0.80 m depths and at 0.02 m increments depth intervals in Kunming peach field on occasional events for the interpolation.The interpolation method uses the detailed time series of SWC at a given depth from one location (used as reference) and rescales it so as to adjust the occasional time series measurements for other location.If at the given depth and at a time "t" the reference time series SWC value is θ t (where "t" represents the 1 h interval) and θ t is the interpolated value for the neighboring location.In order to measure occasional variations in SWC among locations, the measurements between two succeeding time intervals were linearly interpolated.If any pair of values is taken at two consecutive times t = t 1 , t = t 2 then, Where C t1 and C t2 are the ratios of SWC at a given depths for each time.The coefficients were supposed to differ linearly, which produces Equation 9 for the interpolation of SWC at any time × between t = t 1 , t = t 2 .
The sensor measured SWC data from 4 depths were compared with the linear interpolated SWC values at same depths within the same time interval.The corresponding depths measured SWC data on different time interval showed similar patterns, then for each 0.02 m depth interval "i", the ratio (γ i ) between the actual (θ t ) and the interpolated (θ t ) SWC was remarkably close.Therefore, the mean value of each depth (γ i ) used as a scale to the interpolated data by:

Sensor Calibration
Three calibration experiments showed significant results and determined that the sensor could perform equally better in different moist conditions.The measured voltage and the relevant SWC were fitted by a linear equation.
Figure 4 shows the linear calibration curves.The corresponding equations are shown in Table 1.Where x w 0.993, and Fig

Perfor
The perfor environme accuracy a

Mult
(1) CAU F The measu significant correspond Where x is  There wer and depth

System
The system values of o the irrigati rate was le

SWC xcept error
Our research reported that the most of the results were in close agreement with reference findings, and revealed the potential of the developed sensor for the SWC measurements.The sensor responded markedly well for all measured soil depths, especially at the 0.60~0.80m depths.Whereas at the few points the variations in sensor measurements were subjected to irrigation events, clay and sandy contents in in situ soil structure.The temporal variability of SWC (dielectric constant steeps with water contents) could be influenced by the soil hydraulic and matric potential, infiltration process, subsequently will influence the sensor measurements (Evett et al., 2012;Evett, Tolk, & Howell, 2006;Rudnick, Djaman, & Irmak, 2015).
The results determined that the sensor captured the major magnitude of SWC in the vertical profile in different environmental and terrain conditions.It can be used to develop the smart irrigation strategies in the research area following the site-specific calibrations for various soil horizons.The developed sensor will be more advantageous if similar soil moisture management strategies prevail across the year then it becomes possible to plot a field calibration curve to resolve the aforementioned issues associated with sensing technology.The un-calibrated sensor can be used for scheduling the irrigation management but it will limit the research work (Irmak & Irmak, 2005).

Conclusions
This study presents a development of a multi-depth wireless SWC sensor based on the dielectric theory.A particular estimation method, as well as the hardware, has been introduced.The performance of the developed sensor was thoroughly tested under various in situ environmental conditions.The sensor was calibrated by using soil samples collected from different sites.The linear order SWC prediction models were established.A multi-depth SWC curve was established to examine the impact of continuous estimations of SWC at a specified number of depths on the sensor performance.An interpolation technique was used to estimate the SWC in the vertical profile.The interpolated SWC patterns were also correlated with the sensor measured SWC values, but at some points, measurements showed slight variations in individual results.In situ experiments indicated that estimation error was less than 5%, except irrigation events.It is recommended that the sensor should be used after the infiltration of irrigated water for better results.The sensor can work and transfer consistently for at least 4 months with 2100mAh/3.6Vbattery.Hence, the newly developed multi-depth sensor is advantageous for estimating and managing SWC under actual environmental conditions.

Figure
Figure 1.Th and A ted data and pr from the sens modules.The in ugh a voltage p ience Figure 5 p SWC data projected s and terrai variations Brakensiek to derive p installed in also help a (2) Greenh The green monitored

Table 1 .
Linear fit equations Note.Null hypothesis has been tested at a 5% significance level.jas.ccsenet.