Temperature compensation in a low cost frequency domain (capacitance based) soil moisture sensor

G Model

ARTICLE IN PRESS

AGWAT-4662; No. of Pages 8

Agricultural Water Management xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Agricultural Water Management journal homepage: www.elsevier.com/locate/agwat

Temperature compensation in a low cost frequency domain (capacitance based) soil moisture sensor M.J. Oates ∗ , A. Fernández-López, M. Ferrández-Villena, A. Ruiz-Canales Engineering Department, Miguel Hernández University of Elche, Ctra. Orihuela-Beniel, km 3.2, Orihuela, 03312, Spain

a r t i c l e

i n f o

Article history: Received 2 July 2016 Received in revised form 30 October 2016 Accepted 1 November 2016 Available online xxx Keywords: TDR Irrigation water management Calibration Sustainability

a b s t r a c t Frequency Domain Analysis is a well established technique in soil moisture determination, using the change in electrical capacitance of probes inserted into the soil caused by the presence of water. However it is known that temperature affects the determination of this capacitance. Here two different techniques are used, the first passing a fixed frequency through the soil via insulated probes, then measuring the amplitude of the resultant signal. The second uses the soil capacitance as the controlling component in a variable frequency oscillator, measuring the resultant times to charge and discharge. The measured capacitance is seen to be affected both by the temperature of the soil and, due to the sensitive nature of the monitoring electronics, also the temperature of critical components in the measurement circuits. Results from these experiments show that these two effects are complementary, soil temperature adding to the measured capacitance, whilst electronics temperature effectively decreases the measured capacitance. The daily profiles of the soil and electronics temperatures, whilst both rising during the day, and falling at night, show significant phase difference and therefore do not simply cancel out. Further, the strength of temperature compensation required is shown to vary with technique and moisture level. This paper explores these phenomena using results from a recently developed, four probe Frequency Domain capacitance based sensor costing around 12 Euros. These measurements are compared to those achieved by a commercial soil moisture system costing over 250 times this price. Preliminary results are presented from temperature compensation algorithms intended to minimize these effects. © 2016 Elsevier B.V. All rights reserved.

1. Introduction In irrigation management, the determination of water requirements of a crop over time is based on soil water balance measurements, plant measurements and meteorological data. The entire parameters are used for estimating the irrigation water scheduling in a specific crop. The irrigation water scheduling of a crop determines the quantity of water and the time when this quantity is applied. One part for determining the soil water balance is the measuring of soil moisture. The use of soil-based water measurements have been adopted as an adequate strategy for water balance estimation and many methodologies to measure water fluxes from crops have been traditionally developed (Ojha et al., 2015; Jaguey et al., 2015; Navarro-Hellin et al., 2015; Tarange et al., 2015).The ultimate objective of these techniques is to provide farmers with information about the most appropriate volumes of irrigation to apply in each phenological period of the crop, depending on the

∗ Corresponding author. E-mail address: [email protected] (M.J. Oates).

desired yield levels and other parameters. For determining soil moisture, a great range of sensors are used (Vienken et al., 2013). There is a wide range of electrically based soil moisture measurement techniques well established in the fields of geophysical surveying (Linck and Fassbinder, 2014; Lehmann et al., 2014) and agronomy (Fatas et al., 2014; Baghdadi et al., 2014) These including resistivity based methods such as the Wenner (Jiao-Jun et al.,2014) and Schlumberger Arrays (Mosuro et al., 2012), and capacitive based methods such as Frequency Domain Reflectometry (FDR) (AlAsadi and Mouazen, 2014; Jaria and Madramootoo, 2013) and Time Domain Reflectometry (TDR) (Janik et al.,2014) as well as Radiation based techniques such as the Neutron Probe (Kodikara et al., 2014). Whilst low cost implementations of resistive based sensors have been suggested in the past (Igboama and Ugwu, 2011), commercial implementations of these units (for example the Landviser Landmapper) are expensive (typically $500–$1600), lack integrated data-logging capabilities, or are simply unavailable. Previous experiments (Oates et al., 2014) have established the baseline potential of a low cost resistivity based Wenner Array sensor design, giving highly correlated results (>95%) with a commercial Hydra II probe. Further, a simple compensation formula has

http://dx.doi.org/10.1016/j.agwat.2016.11.002 0378-3774/© 2016 Elsevier B.V. All rights reserved.

Please cite this article in press as: Oates, M.J., et al., Temperature compensation in a low cost frequency domain (capacitance based) soil moisture sensor. Agric. Water Manage. (2016), http://dx.doi.org/10.1016/j.agwat.2016.11.002

G Model AGWAT-4662; No. of Pages 8 2

ARTICLE IN PRESS M.J. Oates et al. / Agricultural Water Management xxx (2016) xxx–xxx

been derived and published by the authors (Oates et al., 2015) to compensate for changes due to temperature. However this resistivity based technique is well known to be susceptible to a variety of differing soil conditions such as composition (Hanson and Peters, 2000; Kibria and Hossain 2014), texture (Hadzick et al., 2011; Nadler, 1991), varying pH (Ishada and Makino, 1999; Islami et al., 2012), salinity (Austin and Rhoades, 1979; Velstra et al., 2011; Read and Cameron, 1979) and temperature (Afa and Anaele, 2010; Everwand et al., 2014; Newill et al., 2014). As the FDR technique depends on the dielectric constant of the soil (i.e. its electrical capacitance) rather than its conductivity (reciprocal of resistivity), it is theoretically less susceptible to the salinity of the soil. Thus low cost Frequency Domain sensors were constructed to explore this technique.

The principle of operation of the Frequency Domain capacitance probe relies on the fact that the dielectric constant between water and air differs by a factor of 80. Thus the presence of water in the soil between the probe plates produces a highly significant change in its capacitance, the higher the water concentration, the higher the capacitance. This capacitance can then be measured by electrical means. As the probe is electrically insulated, there is no direct current flow within the soil, and thus the conductive effect of ion based salts in the soil is minimized. 2. Materials and methods Two electrical methods are used to determine the effective capacitance of the probe (see Fig. 1). The first involves using the

Fig 1. Circuit diagram of the two Frequency Domain capacitive probe methods.

Fig. 2. Uncompensated Charge and Discharge times in Mulch based soil.

Please cite this article in press as: Oates, M.J., et al., Temperature compensation in a low cost frequency domain (capacitance based) soil moisture sensor. Agric. Water Manage. (2016), http://dx.doi.org/10.1016/j.agwat.2016.11.002

G Model AGWAT-4662; No. of Pages 8

ARTICLE IN PRESS M.J. Oates et al. / Agricultural Water Management xxx (2016) xxx–xxx

3

Fig. 3. Uncompensated Signal Strength values in Mulch based soil.

Fig. 4. Temperature compensated values in Mulch based soil.

probe as the capacitive component of a low pass filter. The microcontroller outputs a fixed frequency square-wave of 250 KHz on pin D3 (and later 125 KHz, 62.5 KHz and 31 KHz). These frequencies were selected to be as high as practical, given the limited Gain Bandwidth Product of the chosen low cost Operational Amplifiers, and also to be aligned with the corner frequency of the resulting low pass filter. The signal is then passed into a 47 K  resistor which then connects to one terminal of the Frequency Domain capacitive probe. The other end of the probe is connected to the electrical ground. The junction of the probe and the 47 K resistor is also connected to the +ve input of one of the operational amplifiers. This amplifier is configured as a near unity gain buffer with a 330 R feedback resistor from the output to the −ve input. The output is then connected via a 1N4001 diode to a ‘peak detector’ circuit consisting of a 1.8 M resistor and 100 nF capacitor connected in parallel and both connected to the electrical ground. The output is also connected to an analog to digital converter (ADC) input of the

microcontroller. This voltage is sampled after a stabilization period of 20 ms after the square-wave signal is applied. Once the signal strength has been sampled, the square-wave signal is stopped and the microcontroller pin returned to a high impedance state. Thus when the Frequency Domain capacitive probe is in air or very dry soil, its capacitance is low (of the order of 30pF). Therefore the ‘corner frequency’ of the low pass filter is 113 KHz and so we would expect just over 3 dB signal attenuation of, for example, the 125 KHz signal. This attenuation will be higher for the 250 KHz signal and lower for the 62.5 and 31 KHz signals. When the Frequency Domain capacitive probe is in very wet soil, its capacitance rises to more typically 200 to 400 pF giving a lower ‘corner frequency’ of between 17 and 8.5 KHz. At these corner frequencies, the attenuation of all four test signals is far higher and so a lower reading is seen at the ADC. Although the theoretical change in perfect capacitance could be as high as 80, this is not seen in practice due to the contribution of the dielectric constant of the insulating varnish on

Please cite this article in press as: Oates, M.J., et al., Temperature compensation in a low cost frequency domain (capacitance based) soil moisture sensor. Agric. Water Manage. (2016), http://dx.doi.org/10.1016/j.agwat.2016.11.002

G Model AGWAT-4662; No. of Pages 8 4

ARTICLE IN PRESS M.J. Oates et al. / Agricultural Water Management xxx (2016) xxx–xxx

Fig. 5. Uncompensated Signal Strength values in Clay based soil.

Fig. 6. Temperature compensated reciprocal Signal Strength values in Clay based soil.

the sensor, and the longitudinal geometry of the sensor. Further, the corner frequencies and performance of the low pass filter are further influenced by the 270 K feedback resistor used in the second circuit amplifier. During the first method readings, this second amplifier output is forced to 0 v by the microcontroller D4 control pin. Strictly speaking, the theoretical 3 dB ‘corner frequency’ attenuation only applies to sine wave signals whereas here we are using square waves. This means there are significant, but decreasing amplitude, component harmonics of the base frequency at odd multiples. As we are only sampling peak value signal strength, and these harmonics are usually above the corner frequency of the low pass filter, this is not a significant issue in the performance of the sensor. The second method uses the probe as the capacitive element in an oscillator circuit, repeatedly charging and discharging the capac-

itor as the voltage passes between two controlled thresholds. The time taken for the voltage to rise and fall between these thresholds is measured on pin D5 over four, half wave oscillation cycles to provide an indication of the capacitance value. For this method the 47 K resistor and the first amplifier play no significant part as both are followed by high impedance pathways. The Frequency Domain capacitive probe is connected between electrical ground and the −ve input of the second amplifier. This is in turn connected via a 270 K feedback resistor to the output of the second amplifier. This output is also fed back to the +ve input of the second amplifier to change the target threshold at which the amplifier switches output state. This threshold is also determined by the midpoint of the potential divider from the microcontroller control pin output D4, and 0 v, consisting of two further 470 K resistors. The oscillator is effectively switched on by setting control pin output D4 high. The +ve input to the second amplifier is now raised

Please cite this article in press as: Oates, M.J., et al., Temperature compensation in a low cost frequency domain (capacitance based) soil moisture sensor. Agric. Water Manage. (2016), http://dx.doi.org/10.1016/j.agwat.2016.11.002

G Model AGWAT-4662; No. of Pages 8

ARTICLE IN PRESS M.J. Oates et al. / Agricultural Water Management xxx (2016) xxx–xxx

to around 2.5 v, plus or minus a contribution from the feedback resistor connected to the amplifier output. If this output is also high, the +ve threshold can be expected to be of the order of 3.2 v and the Frequency Domain capacitive probe will now begin to charge via the 270 K feedback resistor on the −ve input side. When this charging reaches just over 3.2 v, the output of the second amplifier will swing low. This will cause the Frequency Domain capacitive probe to begin to discharge via the 270 K resistor. The 470 K feedback resistor on the +ve side now pulls the +ve input threshold down to around 1.8 v (as the amplifier output is now low). Thus the circuit waits until the charge on the Frequency Domain capacitive probe has been brought down to just below 1.8 v. At this point the amplifier output switches back to high and the cycle repeats. The output of the second amplifier is monitored by a microcontroller input pin. The time taken to charge and discharge the Frequency Domain capacitive probe between these thresholds is determined by its capacitance value, which in turn is influenced by the amount of water in the immediate area around the probe. When the control pin is low, if the output of the second amplifier is high, the +ve threshold of the second amplifier is forced to less than 1.6 v, immediately switching the second amplifier output low, which in turn lowers the threshold down to close to 0 v. In this state, the Frequency Domain capacitive probe can no longer charge and the oscillation is stopped. Theoretically, when the first method square-wave oscillator is operating, it is possible for the voltage at the −ve input of the second amplifier to fall below the threshold on the +ve input. This depends on characteristics of the operational amplifier, in particular how close to the 0 v rail it is capable of forcing its output to become. In practice, with the operational amplifiers used here, this has not been seen to present a problem. Five different operational amplifiers were tested, a twin Op Amp LM358 and an OPA2134, and quad Op Amps TL074, MC33079 and TLC274. Prices per amplifier range from 4.5c (TL074) to 68c (OPA2134), current consumption from 0.25 mA (LM358) to 8 mA (MC33079) and Gain Bandwidth Product from 1 MHz (LM358) to 16 MHz (MC33079). It should be noted that all operational amplifiers used in the lab experiments were in 0.1 inch, DIL packages inserted into standard breadboard. This was for ease of prototyping. For a production model, surface mount variants of these devices are likely to be used together with closer PCB tracking. This will inevitably change the capacitance of the wiring from the Frequency Domain capacitive probe to the input of the amplifiers. This can be expected to have an impact on the overall performance of the sensor; however careful design and layout should not cause this to be detrimental. Four different Frequency Domain capacitive probe heads were investigated. The first consists of a 54 mm by 39 mm double sided rectangular PCB with twin ‘interlocking finger’ tracks. The PCB is constructed from standard 1.6 mm FR4 board and is insulated by dipping into marine varnish, which is allowed 24 h to dry before a second dipping. The probe is then allowed at least 7 days to cure before use. This is connected to the electronic interface by twin cables of length 40 cm. The area of tracking is approximately 40% of the surface of the boards therefore giving an approximate size of 840 mm2 per plate. This is herein referred to as the ı´Flat Marine´ı probe. The second probe consisted of the same physical PCB, this time hand-painted with two coats of cosmetic acrylic varnish. This produced a thinner layer of insulation, but less uniformly coated. This is referred to as the ı´Flat Cosmetic´ı probe. The third probe consisted of a 20 mm * 60 mm PCB with two double sided, parallel 7 mm wide tapered prongs of insulated plates, with the PCB, but not the plates joined together at the top. The two prongs are separated by an air gap of 6 mm. The PCB is insulated using the same marine varnish technique as described before. The effective plate area is 500 mm2.

5

This is referred to as the ı´Prong Marine´ı probe. The fourth probe consisted of a hand-painted, two coat cosmetic acrylic varnish version of this second PCB design and is referred to as the ı´Prong Cosmetic´ı probe. The microcontroller used in these experiments was an Arduino Pro Mini PCB, running at 5 v and 16 MHz. Temperature measurements were made using a DS18B20 waterproof digital temperature probe, inserted 3 cm into the soil. This device communicates with the microcontroller via the ‘One Wire’ protocol. The entire sensor system can cost less than 7 Euros including IP56 rated box, Frequency Domain capacitive probe, DS18B20 temperature sensor, LM358 operational amplifier, electronic interfaces and microcontroller. For the lab experiments the microcontroller sent data via a CH340G USB cable based serial port connection to be logged on a PC from which the sensor was powered, however in the field experiments, for an additional 3 Euros the system was adapted to use 4 AA batteries and an NRF24L01 2.4 GHz radio transceiver. Alternatively the data could be logged directly onto an SD card. Using a low quiescent current regulator, direct pin control of the LM358 power supply, removing the Arduino ‘Power On’ LED and placing the radio and microcontroller in sleep mode between readings, this can provide power for in excess of 4 months, providing a stand alone, RF based sensor system for less than 10 Euros. Where multi-depth readings are required, a CD4052 CMOS dual 4 way switch can be inserted between the amplifier inputs/ground pairs and the physical probes, allowing up to 4 probes to be operated by the amplifier and microcontroller within the same enclosure. For each amplifier, tests were performed to gather baseline readings with the Flat Cosmetic and Flat Marine sensors in air and in water together with a range of capacitance values. Tests were also performed with all 4 probe designs under air, water, and salt water conditions. These tests identified the most promising candidate amplifier/probe head combinations. Finally 3 and 4 day trials of various combinations of probe and amplifier were conducted in two different soil types, recording every 5 min and base-lined against readings from the commercial soil moisture recorder. 3. Results and discussion 3.1. Initial results in the laboratory The ı´Time´ı row of Table 1 shows the number of microseconds to perform 4 half cycle oscillations for the TLC274 amplifier with the Flat Cosmetic probe under Air, Water and Saline conditions and with a range of Capacitor values used in place of the probe. This is the ı´variable frequency´ı technique. The following rows give the raw ADC readings (averaged over 4 readings) for the probe and capacitors using the 4, fixed frequencies. It is clear that the time to oscillate increases with moisture and capacitance; also that signal strength at all frequencies decreases with moisture and capacitance; and that signal strength is lower when excited by higher frequencies. The probe in air can be seen to have an effective capacitance of around 30 pF and in water of around 300 pF. Table 2 shows the difference in each pair of readings between the Flat Marine probe being in Air and Water using each of the

Table 1 Absolute readings for TLC274 Op Amp with Flat Cosmetic Probe and Capacitors. Air

Water

Saline

22pF

47pF

100pF

220pF

470pF

74 528 646 742 777

298 398 429 488 587

313 393 422 480 579

68 555 675 756 780

89 491 599 716 772

134 437 510 627 738

226 402 443 521 642

437 383 400 439 514

T2cyc (us) SS@250KHz SS@125KHz SS@62KHz SS@31KHz

Please cite this article in press as: Oates, M.J., et al., Temperature compensation in a low cost frequency domain (capacitance based) soil moisture sensor. Agric. Water Manage. (2016), http://dx.doi.org/10.1016/j.agwat.2016.11.002

G Model

ARTICLE IN PRESS

AGWAT-4662; No. of Pages 8

M.J. Oates et al. / Agricultural Water Management xxx (2016) xxx–xxx

6

Table 2 Difference between Air and Water readings for the Flat Marine probe. LM358

OPA2134

TL074

MC33079

TLC274

118 102 176 116 13

87 77 56 12 5

124 34 86 129 19

90 84 149 177 97

111 93 156 173 100

Delta time(us) Delta 250 K Delta 125 K Delta 62 K Delta 31 K

chosen amplifiers. For maximum resolution, these values need to be as high as possible. For the variable frequency technique, the TL074 gives the most encouraging results, whilst at 125 KHz it is the LM358 and at 62 KHz the MC33079 appears to perform best. Tables 3, 4 and 5 show the difference in each pair of readings between 22 pF and 47pF; 47 pF and 100pF; and 100 pF and 220 pF capacitors using each of the chosen amplifiers. These are deemed representative of dry, moist and wet soil conditions respectively. For maximum resolution, the highest value is beneficial. In Table 4 the TL074 grossly underperforms, with the LM358 and TLC274 giving best performance at 125KHz. Table 6 shows the difference in readings for all 4 probes between Air and Water conditions using the TLC274 Op Amp. Clearly the Cosmetic varnish insulation gives the highest values, with the Prong probe performing better at higher frequencies. Table 7 shows the difference in readings for all 4 probes between Water and Saline conditions (1 g per 100 ml) using the TLC274 Op Amp. Clearly the Prong Cosmetic varnish insulation gives the highest susceptibility to salt in the water, whilst the Prong Marine probe has moved outside the range of useful values. Table 3 Difference between 22 pF and 47 pF – equivalent to dry soil conditions. LM358

OPA2134

TL074

MC33079

TLC274

22 41 70 4 1

18 33 7 2 1

21 0 −2 −3 −1

19 49 69 46 10

21 64 76 40 8

Delta time Delta 250 K Delta 125 K Delta 62 K Delta 31 K

Table 4 Difference between 47 pF and 100 pF − equivalent to moist soil conditions. LM358

OPA2134

TL074

MC33079

TLC274

46 52 88 46 3

35 44 27 5 2

47 19 45 52 3

40 46 81 96 38

45 54 89 89 34

Delta time Delta 250 K Delta 125 K Delta 62 K Delta 31 K

Table 5 Difference between 100 pF and 220 pF equivalent to wet soil conditions. LM358

OPA2134

TL074

MC33079

TLC274

94 32 64 100 31

69 29 56 21 4

92 31 65 107 42

80 29 60 102 107

98 35 67 106 96

Delta time Delta 250 K Delta 125 K Delta 62 K Delta 31 K

Table 6 Air to Water deltas for TLC274 with all 4 probes. Flat Marine

Flat Cosmetic

Prong Marine

Prong Cosmetic

111 93 156 173 100

224 130 217 254 190

56 131 157 98 30

206 188 262 253 183

Delta Time Delta 250 K Delta 125 K Delta 62 K Delta 31 K

Table 7 Water to Saline deltas for TLC274 with all 4 probes. Flat Marine

Flat Cosmetic

Prong Marine

Prong Cosmetic

5 4 5 5 3

15 5 7 8 8

2 18 8 −5 −5

23 41 39 23 9

Delta Time Delta 250 K Delta 125 K Delta 62 K Delta 31 K

Tests were then performed with various Probe, Amplifier and Soil Type combinations over 3 day periods. It was found that the Marine varnish had a tendency to expand under wet soil conditions. This lead to ripples forming in the insulation and gave slightly erratic performance. The Flat probes also demonstrated a susceptibility to a film of moisture clinging to the surface of the probe, which due to the interwoven nature of the plates also gave poor results. Results with the two prong cosmetic varnish probes proved to be most stable, however all results demonstrated significant variation with both the temperature of the soil and the temperature of the electronics. 3.2. Further results under field conditions To validate the usefulness of the cosmetic varnish probe in actual soils, two further experiments were carried out. The first using the probes in a mulch based soil over 3 days with 1 irrigation event, and the second in a clay based soil over 4 days with 2 irrigation events. To exaggerate soil temperature effects, the first of these experiments was conducted in a 25 cm tall pot of upper diameter 25 cm and lower diameter 22 cm. From the results of this experiment, temperature compensation formulae were derived. For the final experiment, the probes were buried in an experimental plot at UMH, Orihuela, where results from a co-located commercial device were available for comparison. For these experiments all values were rescaled within the Arduino to fit into the byte range 0–255, by subtracting constants and dividing the results. Cycle times were broken down into separate ı´Time to Charge´ı and ı´Time to Discharge´ı times. These are referred to as ı´Th´ı and ı´Tl´ı respectively, and although under normal circumstances these 2 values maintain a fixed ratio, under fault conditions this ratio is seen to change. This effect is explored elsewhere (Oates et al., 2016). Whilst ı´Th´ı and ı´Tl´ı values are seen to increase with higher water content (i.e. greater capacitance), signal strength (SS) values are seen to fall with higher water content. Therefore to normalize these results, in Fig. 6 the reciprocal of the Signal strengths is displayed, multiplied by 8000, to give results of similar value to ı´Th´ı and ı´Tl´ı. Preliminary experiments reported in (Oates et al., 2016) had indicated that the measured capacitance was affected both by the temperature of the soil and, due to the sensitive nature of the monitoring electronics, also the temperature of critical components in the measurement circuits. Results from these experiments showed that these two effects were complementary, soil temperature adding to the measured capacitance, whilst the temperature of the electronics effectively decreased the measured capacitance. The daily profiles of the soil and electronics temperatures, whilst both rising during the day, and falling at night, showed significant phase difference and therefore did not simply cancel out. The operational amplifiers used in these experiments were chosen for low cost, low voltage and low current considerations, but are known to have an input offset voltage variation typically in the range of 6–20 uV per degree C. It is possible that high precision amplifiers (such as the OP200 with input offset values typically 0.1–0.5 uV per degree C) or an instrumentation amplifier configuration could reduce this effect, however such designs would significantly increase cost, current consumption or require a higher operating voltage.

Please cite this article in press as: Oates, M.J., et al., Temperature compensation in a low cost frequency domain (capacitance based) soil moisture sensor. Agric. Water Manage. (2016), http://dx.doi.org/10.1016/j.agwat.2016.11.002

G Model AGWAT-4662; No. of Pages 8

ARTICLE IN PRESS M.J. Oates et al. / Agricultural Water Management xxx (2016) xxx–xxx

Fig. 2 shows the uncompensated values for Time to Charge and Time to Discharge both before and after an irrigation event. Clearly the temperature of the soil (DsT´ıC) has a stronger effect under wetter conditions than dry and thus the degree of compensation must be influenced by both the temperature and the approximate soil moisture level. Fig. 3 shows the uncompensated values for the Signal Strength at each of the four frequencies. Although the SS250 signal appears to be affected to a greater extent than the other frequencies, this is artificial as the scaling factor used to reduce this data to the useful 0–255 range divides the SS250 value by half the amount used for the other frequencies. This is shown here in this form to emphasize the temperature effect, particularly on the day following irrigation where a more complex ı´double peaked´ı temperature effect is plainly visible. The sharp drop in the temperature of the electronics (Tint ı´C) seen each afternoon is caused by the sensor falling into the shadow of a nearby building. In an attempt to remove the temperature induced variations, the following compensation formulae were derived: • • • • • •

Thcomp = Th + (Tint´ıC − 12)/4.4–(DsT´ıC − 20) * (Th − 40)/50 Tlcomp = Tl + (Tint´ıC − 12)/1.1–(DsT´ıC − 20) * (Th − 50)/40 SS250comp = SS250 - (Tint´ıC − 12)/1.33–(DsT´ıC − 20)/2.4 SS125comp = SS125 - (Tint´ıC − 12)/2.67–(DsT´ıC − 20)/1.6 SS83comp = SS83 - (Tint´ıC − 12)/2.67–(DsT´ıC − 20)/1.8 SS62comp = SS62 - (Tint´ıC − 12)/2.67–(DsT´ıC − 20)/1.5

When this compensation is applied to the Mulch based data, the results shown in Fig. 4 are obtained. The Th and Tl values are seen to initially rise by a very small amount. This is typical of initial sensor stabilization. They then exhibit a step change increase when irrigation is applied, followed by a flat response for the rest of that day. During the following day, the charge and discharge times are seen to fall slightly as solar induced evaporation reduces the soil moisture content. In contrast, the Signal Strength readings are seen to initially fall slightly during stabilization, fall dramatically on irrigation and then flatten out. The following day solar induced evaporation causes soil moisture levels to fall and hence Signal Strengths are seen to rise. For the final experiment the sensor was moved to the research plot at EPSO Orihuela with the probe buried approximately 15 cm into high silica based clay typical of the Vega Bahia region of South

7

Eastern Spain. The plot is irrigated for 15 min every 2 days in the morning and is monitored by a network of commercial soil moisture recorders. Fig. 5 shows the uncompensated Signal Strengths recorded by the low cost sensor over a 4 day period from the 12th to the 15th of May 2016. Soil temperature (DsT´ıC) is seen to have much smaller variation than in the mulch and pot based experiment as the large volume of ground provides a dampening heat reservoir, however electronics temperatures are seen to vary widely from day to night. Again the 250 KHz signal is recorded and displayed with twice the resolution of the other 3 signals to make best use of the 0–255 data range and emphasize the temperature effects. Fig. 6 shows this Clay soil data turned into their reciprocals and compensated with exactly the same temperature compensation formulae derived from the Mulch soil. This clearly shows the two irrigation events and their subsequent drying phases, followed by solar induced evaporation phases as the temperature rises on the intermediate days with no irrigation. Finally Fig. 7 shows the results from the co-located commercial device with 2 probes inserted at different depths into the soil (10 cm and 20 cm from the surface). The profiles of Figs. 6 and 7 show remarkable visual similarities and when X-Y plotted for linear correlation give an R squared result in excess of 0.6, the most significant differences being the shape of the trailing edges following irrigation events. The prototype sensor shows a rapid decline in soil moisture level as water drains past the sensor, whilst the commercial device shows a more gradual decline. This may be due to the geometry of the commercial probe which consists of a PVC tube of 150 cm in length and 56.5 mm diameter inserted vertically into the soil. Such sensors are prone to attracting available moisture to the surface of the PVC tube, which can take an extended period of time to drain. 4. Conclusions There is clearly a range of performances achievable by different combinations of probes and amplifiers. An amplifier with a higher Gain Bandwidth Product does not always give best performance at higher frequencies in this application. There is a balance to be struck between maximizing resolution under drier soil conditions and minimizing sensitivity to salinity. These results suggest that a Prong probe with at least 2 coats of Cosmetic varnish can provide good performance at 125 KHz with the LM358 Op Amp. For dual

Fig. 7. Results from commercial device in Clay based soil.

Please cite this article in press as: Oates, M.J., et al., Temperature compensation in a low cost frequency domain (capacitance based) soil moisture sensor. Agric. Water Manage. (2016), http://dx.doi.org/10.1016/j.agwat.2016.11.002

G Model AGWAT-4662; No. of Pages 8

ARTICLE IN PRESS M.J. Oates et al. / Agricultural Water Management xxx (2016) xxx–xxx

8

probes, the TLC274 gives good performance over a wider range of frequencies but at higher cost and current consumption. The physical characteristics of the capacitive probe are clearly as important as its electrical characteristics when in a field application. The Frequency Domain capacitive plates require a significant physical separation to prevent ‘bridging’ by thin films of water which can accumulate around the probe surface. Further, the coating material must demonstrate physical and chemical stability when exposed to wet soil conditions. The readings of Time to Charge and Discharge for the variable frequency technique and the readings of resultant Signal Strength for the fixed frequency technique are all seen to be affected by temperature. This is true for both the temperature of the soil and the temperature of the monitoring electronics, and is further complicated by the actual soil moisture content. These results suggest that it is possible to derive temperature compensation formulae to significantly reduce these effects in at least two soil types with very different characteristics. The temperature compensated moisture level determined by this low cost Frequency Domain capacitive probe can be used to control highly localized (if not individual plant) irrigation schedules, leading to cost efficient management of scarce water resources. Once perfected, the low cost of this sensor will make it highly suitable for large scale agronomic deployment where large numbers of plants need to be individually managed.. Acknowledgments The authors wish to acknowledge, with gratitude, the technical assistance provided by Telenatura EBT, S.L. References Afa, J.T., Anaele, C.M., 2010. Seasonal variation of soil resistivity and soil temperature in Bayelsa State. Am. J. Eng. Appl. Sci. 3 (4), 704–709. Al-Asadi, R.A., Mouazen, A.M., 2014. Combining frequency domain reflectometry and visible and near infrared spectroscopy for assessment of soil bulk density. Soil Tillage Res. 135, 60–70. Austin, R.S., Rhoades, J.D., 1979. A compact low cost circuit for reading four-electrode salinity sensors. J. Soil Sci. Soc. Am. 43, 808–809. Baghdadi, N., Dubois-Fernandez, P., Dupuis, X., Zribi, M., 2014. Sensitivity of main polarimetric parameters of multifrequency polarimetric SAR data to soil moisture and surface roughness over bare agricultural soils. IEEE Geosci. Remote Sens. Lett. 10 (4), 731–735. Everwand, G., Fry, E.L., Eggers, T., Manning, P., 2014. Seasonal variation in the capacity for plant trait measures to predict grassland carbon and water fluxes. Ecosystems 17 (6), 1095–1108. Fatas, E., Vicente, J., Latorre, B., Lera, F., Vinals, V., Lopez, M.V., Blanco, N., Pena, C., Gonzalez-Cebollada, C., Moret-Fernandez, D., 2014. TDR-LAB 2.0 Improved TDR Software for soil water content and electrical conductivity measurements. Procedia Environ. Sci. 19, 474–483 (four decades of progress in monitoring and modeling of processes in the soli-plant-atmosphere system: applications and challenges). Hadzick, Z.Z., Guber, A.K., Pachepsky, Y.A., Hill, R.L., 2011. Pedotransfer functions in soil electrical resistivity estimation. Geoderma 164 (3–4), 195–202. Hanson, B.R., Peters, D.W., 2000. Soil types affects accuracy of dielectric moisture sensors. Calif. Agric. 54 (3), 43–47. Igboama, W.N., Ugwu, N.U., 2011. Fabrication of resistivity meter and its evaluation. Am. J. Sci. Ind. Res. 2 (5), 713–717.

Ishada, T., Makino, T., 1999. Effects of pH on dielectric relaxation of montmorillonite, allophane, and imogolite suspensions. J. Colloid Interface Sci. 212, 152–161. Islami, N., Taib, S.H., Yusoff, I., Ghani, A.A., 2012. Integrated geoelectrical resistivity, hydrochemical and soil property analysis methods to study shallow ground-water in the agriculture area, Machang, Malaysia. Environ. Earth Sci. 65 (3), 699–712. Jaguey, J.G., Villa-Medina, J.F., Lopez-Guzman, A., Porta-Gandara, M.A., 2015. Smartphone irrigation sensor. IEEE Sens. J. 15 (9), 5122–5127. Janik, G., Skierucha, W., Blas, M., Sobik, M., Albert, M., Dubicki, M., Zawada, A., 2014. TDR technique for estimating the intensity of effective non rainfall. Int. Agrophys. 28 (1), 23–37. Jaria, F., Madramootoo, C.A., 2013. Thresholds for irrigation management of processing tomatoes using soil moisture sensors in Southwestern Ontario. Trans. ASABE 56 (1), 155–166. Jiao-Jun, Z., Hong-Zhang, K., Gonda, Y., 2014. Application of Wenner configuration to estimate soil water content in pine plantations on sandy land. Pedosphere 17 (6), 801–812. Kibria, G., Hossain, M.S., 2014. Investigation of geotechnical parameters affecting electrical resistivity of compacted clays. J. Geotech. Geoenviron. Eng. 138 (12), 1520–1529. Kodikara, J., Rajeev, P., Chan, D., Gallage, C., 2014. Soil moisture monitoring at the field scale using neutron probe. Can. Geotech. J. 51 (3), 332–345. Lehmann, P., Gambazzi, F., Suski, B., Baron, L., Askarinejad, A., Springman, S.M., Holliger, K., Or, D., 2014. Evolution of soil wetting patterns preceding a hydrologically induced landslide inferred from electrical resistivity survey and point measurements of volumetric water content and pore water pressure. Water Resour. Res. 49 (12), 7992–8004. Linck, R., Fassbinder, J.W.E., 2014. Determination of the influence of soil parameters and sample density on ground-penetrating radar: a case study of a Roman picket in Lower Bavaria. Archaeol. Anthropol. Sci. 6 (1), 93–106. Mosuro, G.O., Bayewu, O.O., Oloruntola, M.O., 2012. Application of vertical electric soundings for foundation investigation in a basement complex terrain: a case study of Ijebu Igbo, Southwestern Nigeria. In: Near-Surface Geophysics and Environment Protection., pp. 29–34. Nadler, A., 1991. Effect of soil structure on bulk soil electrical conductivity (Eca) using the TDR and 4P techniques. Soil Sci. 152, 199–203. Navarro-Hellin, H., Torres-Sanchez, R., Soto-Valles, F., Albaladejo-Perez, C., Lopez-Riquelme, J.A., Domingo-Miguel, R., 2015. A wireless sensors architecture for efficient irrigation water management. Agricultural Water Manage. 151, 64–74. Newill, P., Karadaglic, D., Podd, F., Grieve, B.D., York, T.A., 2014. Electrical impedance imaging of water distribution in the root zone. Meas. Sci. Technol. 25 (5) (Number of article:055110) http://dx.doi.org/10.1088/0957-0233/25/5/ 055110. Oates, M.J., Vazquez de Leon, A.L., Edwards, N.M., 2014. A minimal cost soil moisture measurement system. In: Procs. of Sensornets 2014, Jan 2014, Lisbon, Portugal, pp. 373–380. Oates, M.J., Vazquez de Leon, A.L., Intrigliolo, D.S., Molina Martinez, J.M., 2015. Evaluation of an experimental system of soil moisture registration for irrigation management in potted vineyard (Vitis vinifera L, CV Bobal) of multi-depth temperature compensation based in resistivity measurements. Agric. Water Manage. 151, 126–135. Oates, M., Fernandez Lopez, A., Ruiz Canales, A., Vazquez de Leon, A.L., 2016. Temperature compensation in a low cost FDR based Soil Moisture Sensor. Procs of SNIH16, II Simposio Nacional De Ingenieria Horticola (in press). Ojha, T., Misra, S., Raghuwanshi, N.S., 2015. Wireless sensor networks for agriculture: the state-of-the-art in practice and future challenges. Comput. Electron. Agric. 118, 66–84. Read, D.W.L., Cameron, D.R., 1979. Relationship between salinity and Wenner resistivity for some dryland soils. Can. J. Soil Sci. 59, 381–385. Tarange, P.H., Mevekari, R.G., Shinde, P.A., 2015. Web based Automatic Irrigation System using wireless sensor network and Embedded Linux board. In: International Conferenced On Circuits, Power And Computing Technologies (iccpct-2015), Noorul Islam Univ, Noorul Islam Ctr Higher Educ, Dept Elect & Elect Engn; IEEE. Velstra, J., Groen, J., De Jong, K., 2011. Observations of salinity patterns in shallow groundwater and drainage water from agricultural land in the northern part of the Netherlands. Irrig. Drain. 60, 51–58. Vienken, T., Reboulet, E., Leven, C., Kreck, M., Zschornack, L., Dietrich, P., 2013. Field comparison of selected methods for vertical soil water content profiling. J. Hydrol. 501, 205–2012.

Please cite this article in press as: Oates, M.J., et al., Temperature compensation in a low cost frequency domain (capacitance based) soil moisture sensor. Agric. Water Manage. (2016), http://dx.doi.org/10.1016/j.agwat.2016.11.002