Soil Sensor Prototype Initial Results

Overview

In mid-November 2007 we installed a prototype soil sensor mote on the Wildflower Trail. This report presents initial results on soil moisture, conductivity, and temperature sensed by the prototype, including data from the first significant rains of the season.

The mote is located just inside the trail entrance, on the left in Bed 2. The mote is situated near the crest of a mound, and is shaded to the south and east. The soil in Bed 2 is a sandy loam, and has been worked to a depth of around twelve inches to include additional organic matter.

Next we review the equipment and methods used to collect the readings and convert them into useful information. Then we report on the moisture, conductivity, and temperature values generated over the first couple of weeks of reliable data collection.

Instrumentation and Methods

The sensor mote consists of two main elements as shown in the photos above. The sensor probe is installed horizontally at around nine or ten inches depth. The enclosure houses a Tmote Sky control/wireless module, custom-built probe interface circuitry, and batteries. The enclosure is elevated around twenty inches above ground.

The probe is an ECH2O-TE from Decagon Devices [1]. We collect dielectric permittivity (moisture), electrical conductivity, and temperature readings every 15 minutes. In addition we collect temperature and humidity readings for the ambient air within the mote enclosure, which is vented. For this prototype we store readings in the control module and periodically download them manually to a laptop. However once operational the control module will tranmit readings wirelessly to the base station at the Development building.

Data collection failed initially during the cold nighttime period. We isolated the problem to faulty synchronization in the communication between the probe and control module. Since Nov. 28 data collection has worked flawlessly. In addition power usage thus far indicates we can expect several months of use from the four AA batteries.

Next we describe how the raw readings from the sensor are converted to useful information, and the checks we performed to verify the accuracy of the readings.

Moisture

Moisture is described as volumetric water content (VWC). We calculate VWC using a linear equation for generic mineral soils:
θ = 1.087 * 10-3 * P - 0.629
where P is the dielectric permittivity reading from the sensor.

We confirmed that the raw counts in air and water match the manufacturer's specs, which indicates accuracy of approximately ±3% for volumetric water content (VWC). We can perform more precise calibration for this soil in the lab by wetting a known amount with increasing quantities of water.

Electrical Conductivity

Electrical conductivity (EC) is described as pore water EC, which measures the conductivity of the water in the pore space of the soil. This value is not the same as solution EC, which commonly is returned by a soil test. Solution EC is calculated by wetting the soil to saturation with distilled water, and then measuring the conductivity of that solution. According to Decagon, it is difficult to compare pore water EC with solution EC.

Pore EC does not measure salinity per se. However, we are most interested in changes to pore EC. In this case the main agent of change is the water applied to the soil, and we expect the saline irrigation water to increase pore EC and distilled rain water to decrease pore EC.

Pore EC is calculated using all three of the probe's readings:

σp = εp σb / (εb - εb=0)

σp is Pore EC measured in dS/m
εp = 80.3 - 0.37 * (Tsoil - 20), where Tsoil is the soil temperature in °C
σb is the sensor bulk electrical conductivity
εb = 7.64 * 10-8 * P3 - 8.85 * 10-5 * P2 + 4.85*10-2 * P - 10, where P is the sensor dielectric permittivity
εb=0 is assumed to be a generic value of 6
For more background, see the ECH2O-TE manual [2] and Hilhorst (2000) [3].

We performed some rough verification of the sensor's electrical conductivity readings. We found the conductivity for distilled water, tap water, and potting soil within expected values. Decagon claims accuracy of ±20% for Pore EC for moist soils.

Temperature

The probe measures temperature using a thermistor and reports the reading directly. We confirmed the soil sensor readings at room temperature are with ±1°C. For production we plan to verify over a greater range of temperatures.

We also verified the accuracy of the control module's temperature sensor. We found it to be ±1°C at room temperature, and no more than ±2°C at 0°C. The temperature sensor's specs specify ±1°C at 0°C; however we are limited by the accuracy of our reference thermometer at this temperature.

[1] http://www.decagon.com/ag_research/soil/5te.php
[2] http://www.decagon.com/literature/manuals/ManualECH2O-TE_EC-TM.pdf
[3] M.A. Hilhorst. A Pore Water Conductivity Sensor. Soil Sci. Soc. Am. J. 64:1922-1925 (2000).

Soil Moisture

This graph of VWC over time clearly shows the effect of the two big rain events in late November and early December.

The sandy loam soil appears to drain quickly from the initial soaking and then slowly loses moisture content over the next several days. One would expect good drainage in such soil especially considering the mote's position near the crest of a mound. A review from a soil expert at Decagon confirms the readings are as expected under these conditions. Also we notice the second peak is not as sharp as the first one, as that rain occurred over a longer time period.

Soil Electrical Conductivity

This graph of pore water EC also shows the impact of the two rain events.

The main result is the reduction in conductivity with each rain. Before the first rain, pore water EC is roughly 4 dS/m. After the first rain it reduces to around 3 dS/m, and then after the second rain it reduces to 2.1 dS/m.

The fluctuations in value within these three regions come from two sources. Before the first rain, small changes in the probe moisture reading generate the variation. The soil moisture is down around 12% VWC, which causes the denominator in our calculation to become small and thus magnify these small changes.

After the rains we see more consistent readings but with discontinuities, for example early on Dec. 4 and early on Dec. 10. These discontinuities result from a single digit change in the bulk EC readings. For example on Dec. 10, the reading changes from 0.10 dS/m to 0.09 dS/m. This level of granularity is reflected in the probe manufacturer's claim of ±20% accuracy for pore water EC. However given this degree of accuracy, the results still show a clear decrease in conductivity with each rain.

Soil and Air Temperature

This graph includes three measures of temperature -- in the soil, in the above-ground enclosure, and at the Garden's WeatherBug weather station.

First we focus on the soil readings. The period from Dec 3-7 shows how a daily variation of 15°C in air temperature is dampened to 1° in the soil. This period also shows the soil temperature max/min values lag the enclosure max/min by a little less than half a day. For the absolute soil temperature values, it should be significant that the mote location is fairly shaded by trees. The long term trend in soil temperature roughly follows the average enclosure temperature.

We also notice an unexpected relationship between the enclosure temperature and the Garden's weather station. Although the enclosure maximum temperatures approximate the weather station's, the minimum temperatures often are significantly lower. This trend is especially clear for Dec. 2-6. Perhaps cold air pools where the mote is located, toward the bottom of a slope. In addition the enclosure is only twenty inches above ground although it is located on a mound. Some of the variation also may be in the measurement itself since our verification of the accuracy of the mote's temperature sensor is limited to 2°C at 0°C.

Conclusions and Future Work

The sensor probe passed our initial calibration tests, and we are collecting probe readings reliably. Already we have learned valuable information about soil conditions, especially related to significant rain events.

Now that we have seen results for this particular installation, we can prioritize our interest in observing variations under other conditions like topography, sensor depth, soil type, and sun exposure. For soil moisture and electrical conductivity it also will be interesting to compare the effect at this depth of regular watering. Given the sensitivity of some desert plants to low temperatures, it may be valuable to follow-up on our observation of localized reduction of minimum air temperature.

Version History
6/2008Update for links to Decagon Devices.
12/2007Initial version.