Comparing soil moisture sensors for smart irrigation systems

In this project, different sensors for the quantification of water availability in plans are compared for smart irrigation applications.

Things used in this project

Hardware components
	5 probe NPK sensor	×	1
	Industrial Soil Moisture & Temperature & EC Sensor	×	1
	Soil conductivity sensor	×	1
	DFRobot Gravity: Analog Capacitive Soil Moisture Sensor- Corrosion Resistant	×	1
	Seeed Grove – Round Force Sensor (FSR402)	×	1
	Texas Instruments General Purpose Dual Op-Amp	×	1
	Arduino Mega 2560	×	1
	SparkFun I2C DAC Breakout – MCP4725	×	1
	PH Value Detection detect Sensor	×	1
	Peristaltic Liquid Pump with Silicone Tubing – 12V DC Power	×	1

Story

1. Synopsis

The development of irrigation systems is essential in agriculture for the right maintenance of plants and to improve the crop production. To date, multiple sensors are available commercially for the measurement of the moisture in soil by different means (i.e. conductivity, capacitance). However, the performance of these sensors can vary depending on the system used for the detection of water. In this project, we compare different systems for the measurement of water needs in plants. We divided this study into 2 parts; Measurement of the soil moisture and measurement of water inside plants. Both approaches could be used to compare the accuracy of invasive and non-invasive methods in the determination of the water needs of plants. In all cases, the tested plant was watered twice with 250 mL using a peristaltic pump. We first used dry soil where the plant being tested was placed (Calathea). This soil was watered with 250 mL using a peristaltic pump and measured the response of the sensors over 1 h. We then provided an additional 250 mL while measuring the changes detected by the water sensors. The response was then recorded and we compared the devices in terms of sensitivity, noise levels and response time.

2. Soil-based methods for sensing water content

2.1. 5 probe NPK sensor

The 5-probe NPK sensor is a versatile device, able to measure up to 7 different parameters from the soil (Temperature, Nitrogen, Phosphorous, Potassium, pH, Conductivity and Moisture). It consists of 5 metallic probes that can be inserted on the soil, and it can be used interfaced with an arduino microcontroller using an LTT device, since it operates using RS485 communications.

5-probe NPK sensor for soil analysis

As such, it provides a good characterisation of the soil, enabling an estimation of the amount of employed fertiliser, quality and moisture. However, when we tested a dry soil sample before adding any water, the Nitrogen, Phosphorous, Potassium, electrical conductivity and moisture sensors provided a response signal of 0, while the pH was set to 7. After watering the soil with 250 mL, the sensors started sending readings of the different soil parameters. An initial peak of signal was obtained, and the values stabilised on a specific value, which increased slightly when the soil was watered for the second time with a further 250 mL. Below are the obtained values obtained for these components before and after watering the soil sample, and the signal noise (σ) after the signal stabilisation.

Graphs obtained from the 5-probe sensor after recording the signal from Nitrogen, Phosphorous, Potassium, pH, Conductivity and Humidity values respectively.

A low signal noise of 0.38 mg/kg was observed in these sensors in the case of N, P, K, Conductivity and Humidity measurements. After adding the water, these sensors took in the range of 20 mins for the signals to stabilise. This value was used as the response time of the devices, and was calculated for all the sensors tested in this project. However, the pH signal showed a high noise and greatly changed between 7 to 9. The obtained values from phosphorous and nitrogen concentrations were identical, and the humidity and conductivity values were low compared to the values obtained by the other sensors we tested. Thus, although this system could detect the presence of water within the soil samples, we foresee that the sensitivity of the devices is low, and more suitable for environments where larger amounts of water are being supplied.

2.2. Soil Moisture & Temperature & EC Sensor

The configuration of this industrial-grade soil sensor is similar to the previously described 5-probe soil sensor. In this case, the system contains 3 corrosion resistant probes that can be inserted in the soil. This device can measure both the electrical conductivity and the soil moisture using a series of metallic probes, and can be interfaced with arduino microcontroller using an LTT device.

Soil Moisture & Temperature & EC Sensor for soil characterisation.

Contrary to the previous case with the 5-probe sensor, this device was able to measure the conductivity of dry soil even before the water was added, indicating a higher sensitivity, and lower detection limit. However, the signal noise obtained was relatively higher than in the previous case, with 0.41 uS/cm. In addition, the measured values for the conductivity were high compared to the ones obtained using the 5-probe sensors. The results of the changes in the conductivity over time before and after adding the water can be seen below.

Plot obtained after the monitoring of electrical conductivity in soil samples.

Upon watering the soil samples with 250 mL, the signal increased to approximately 3250 uS/cm. The response time of the sensors was significantly shorter, with 11 mins, and the response in dry soil was in the range of 154 uS/cm. This result indicates a higher sensitivity than the 5-probe sensor. However, the range of parameters that this sensor can measure (electrical conductivity and humidity) is limited compared to the previous case.

2.3. Conductivity sensor

Another testing of the measurement of moisture in soil was conducted using an analog conductivity sensor (see below). In this case, a voltage is applied between 2 electrodes, and the changes in current are measured. In principle, such changes in current can be attributed to the presence of water in soil, which decreases its electrical resistance. However, this conductivity can also be influenced by the concentrations of salts in the soil, which reduces the specificity of the detection.

Electrical conductivity Sensor for soil characterisation.

Contrary to the 2 approaches previously described, this device can be powered using an arduino microcontroller only, and does not require additional hardware for the signal collection, such as the LTT. Thus, it can be incorporated easier on our project. After watering the dry soil, the response time of the device was similar to the one obtained in the previous case, in the range of 9 mins.

Plot obtained after the monitoring of electrical resistance in soil samples using an arduino microcontroller.

Despite the relatively short response time of the devices, the noise levels were higher than in the previous case, leading to standard variation on the voltage of 0.14 V/min. In addition, the changes in the received voltage upon watering the plant were small, varying from 2.11 to 1.84 V. These results indicate a lower sensitivity and detection limit of the device compared to the previous approaches.

2.4. Capacitive sensor

The capacitance-based sensors for soil moisture measurements are one of the most widespread sensors for smart gardening applications. This device consists of a single probe, and its capacitance changes in the presence of water due to the changes in the dielectric constant of the medium. As such, the signal is less influenced by the presence of ions in the soil medium, and more specific towards the changes in the water content in soil.

Capacitive Sensor for soil characterisation.

Similar to the previous case, this device could be interfaced with an arduino microcontroller without the need of additional hardware or external power. In addition, the measured noise levels of this device were significantly lower than in the previous case with the conductivity sensor, with 0.01 V/min.

Plot obtained after the monitoring of capacitance in soil samples using an arduino microcontroller.

Upon adding water to the soil, the obtained signal changed from 3.05 to 2.66V, indicating a higher sensitivity than the one obtained in the case of the conductivity sensors. In addition, the response time was the lowest among the devices tested for soil measurements, being about 1 min.

3. Plant-based methods for sensing water content

3.1. Force sensors attached to leaves

A promising non-invasive way to obtain information about the water status of a plant is the analysis of turgor. The structure of stem and leaves of a plant can be maintained due to the fluid pressure inside them. However, when the amount of available water is low, the water pressure decreases, and the pressure inside the plant structures tend to be lower. These differences in the water pressure can be indirectly assessed by the bending of leaves, using a force sensing resistor as shown below.

Force sensor used for leaf characterisation.

This force sensing resistor was placed at the surface of the leaves, and the resistance was assessed using a reference resistor as shown below. This setup allowed us to determine the water pressure inside the plant through the exerted force on the sensor from the leaves. Such force was a consequence of the leaf bending due to the moisture conditions. Although this approach allowed us to detect when the plants were watered in a non-invasive way, the sensitivity was low, and the signal changed from 3.35 to 3.33 V after adding 250 mL of water. However, we foresee that these changes will be significantly higher if the plant is kept under dry conditions during long periods of time, with a marked bending of the leaves. The monitoring of the signal forces from the leaves is shown on the graph below.

Device setup, and plot obtained after the monitoring of the force using an arduino microcontroller.

The voltage obtained from the sensor decreased from 3.35 V to 3.33V upon watering the plant. Despite the low sensitivity of the sensor when applied to measuring the movement of leaves, the noise levels were significantly lower than the previous approaches, with 40 mV/min. In addition, the response time was the shortest among the tested sensors, with 40s. Thus, despite the low sensitivity of this system, the measurement of turgor through the forces exerted by the leaves is still a promising non-invasive approach to measure the water needs of plants.

3.2. Potentiostat

The changes in the ion concentrations inside the xylem sap could also be indicative of the water content in plants. Such changes are typically reflected in a change of the electrical resistance inside the stems. When the concentration of salts is high, which is a consequence of a low quantity of water inside the stems, the electrical resistance decreases given the higher conductivity of the tissues. On the contrary, when the volume of water is higher, which can happen when the plants are being watered, this electrical resistance decreases. As such, this measurement can provide a good indication of the water needs of our plants. For this measurement, we employed a custom-made potentiostat following the indications of our previously reported work here. The final setup was similar to the one shown below.

Figure taken from our previous project, “A low-cost potentiostat for sensing applications”.

The 3 electrode cell was inserted inside the stem of the tested plants, and a stripping voltammetry assay was conducted, by measuring the received signal after applying multiple voltages to the stem. The slope of the obtained plot was then calculated, and the resistance of the stems could then be extrapolated. As expected, this resistance increased when the plant was watered, going from an initial value of 370.4 kOhm up to 500.0 kOhm and increased steadily over time after the the first amount of water was added. No effects on this increasing trend were observed after the plants for a second time.

Results from the resistance characterisation of the potentiostat, and plot obtained after monitoring the stem of the plant before and after watering.

In this case, the response time of the device to the changes in the water status could not be estimated since the measurements were taken every 15 mins. As observed, when the plant was watered for the second time, the resistance of the stem kept increasing at the same rate. This behaviour could indicate a constant increase in the uptake of water by the plants, even when the water abundance in the soil changed.

3.3. Open Circuit Potential measurement of stems

A final method for measuring the water uptake of plants was conducted by using a open circuit potential measurement (OCP). This OCP is related with the potential of an electrode compared to a reference when no current is flowing between both. Since ions in solution are charged, they can produce an electrochemical potential when they are adsorbed onto the electrode that can be measured and correlated with the concentration. For conducting these measurements, we adapted a standard potential meter from an arduino pH sensor, and connected the reference and working electrodes to a screen printed electrode similar to the one used in the previous method using a low-cost potentiostat.

OCP sensor used for plant-based water measurements.

The changes on the voltage of the electrode were then recorded over time. In this case, the noise levels of the sensors were in the range of 0.29 V/min, one of the highest noise levels found within the studied sensors. Upon watering the plants, the OCP of the device decreased, from 2.44 up to 2.03 V, indicating a decrease in the concentration of ions in the stem.

As such, this method could also be used for the determination of the watering needs of plants with a good sensitivity. By measuring the changes in ions inside the plant stem, a more accurate assessment of the water needs can be made. However, the noise levels were high, which could translate into a poor sensitivity of the device.

4. Summary

A summary of the performances of the different tested sensors is provided below.

Whilst the 5-probe sensor provided the largest amount of information among the tested, the sensitivity was low, and they needed a relatively long time to reach a stable reading. On the contrary, the capacitive sensor for soil water measurements combined a fast response, with a relatively low noise and good sensitivity. However, the measurement was limited to only 1 type of soil parameter (humidity).

In the case of the plant-based sensors, the received signal was more indicative of the physiological status of the plant. This method allowed us to measure more accurately the water needs of the plants since, as we observed, the signal were not affected by the conditions of the soil, but the availability of water inside the plants themselves. However, these methods were in general noisy. In the case of the potentiostat, only 1 measurement every 15 mins were taken, due to the high volume of data points collected during the voltammetric method. Although this challenge of continuous monitoring of electrolytes inside plants could be overcome by using the OCP measurements, this approach lead to a high electrochemical noise. The advantages and limitations of the sensors we tested are summarised below.

Credits