Trends in Soil Conductivity

Overview

We installed a Decagon Devices ECH2O-TE soil probe in fall 2007 at the Desert Botanical Garden in Phoenix. The probe measures soil temperature, volumetric water content (VWC), and pore water electrical conductivity (EC). The ongoing graphs of soil temperature and water content against time provide immediately useful information. In contrast, the meaning of the conductivity measurements is not immediately clear. Conductivity depends heavily on water content, which changes relatively quickly. Yet we expect conductivity to change slowly -- on the order of weeks or months rather than hours. So to understand the conductivity readings, we need more analysis to find its longer-term trends.

This report analyzes the EC readings collected thus far. We find a trend of increasing conductivity values over the summer of 2008, and speculate the trend may be part of a longer-period cycle in soil conductivity over the course of a year.

What Is Pore Water Conductivity? [1]
The sensor is designed to measure the conductivity of water in the pore space of the soil. This value is not the same as solution EC, which often is measured by a soil test. Solution EC is calculated by wetting the soil to saturation with distilled water in the lab, and then measuring the conductivity of that solution. Pore water EC and solution EC are related by the bulk density of the soil; however according to Decagon, it is difficult to compare these values.

The theoretical calculations to relate pore water EC and solution EC:

Solution EC = (σpθ + σd(φ - θ)) / φ
φ = 1 - (ρb / ρs)
σp is pore water EC, θ is volumetric water content, σd is distilled water EC (σd = 0), φ is porosity, ρb bulk density, and ρs is mineral density.

[1] This material is largely drawn from the ECH2O-TE manual, available at http://www.decagon.com/literature/manuals/ManualECH2O-TE_EC-TM.pdf

High-level Review of Readings

Tthe ongoing graphs of EC readings show a generally inverse relationship with water content. This relationship is seen in the graph below, which includes all of the readings taken to date. As water content (in blue) decreases after a rain or irrigation peak, conductivity (in brown) increases. This increase is expected as ionic concentration increases in the solution.

To understand the relationship better, the next plot shows EC directly against VWC, and a data point represents these two values from the same moment in time. The readings are organized by month and smoothed.

EC varies widely at low water content levels near 10%, and Decagon describes these EC values as unreliable. However above 15% VWC the smoothed EC values diverge, suggesting some month-to-month variation.

The graph above also shows an exponential function fitted to match the overall relationship of EC to VWC. In other words water content is the main determiner of conductivity at a given moment. Therefore elimination of water content as a variable should allow us to see longer-term changes in conductivity. So let's look at the magnified view below in vertical slices of constant VWC.

At 20% VWC and greater, the EC readings for April, August and September are noticeably higher. Below those months is December EC, and beginning at around 26% VWC, July EC also rises above the others. Interestingly the EC readings at the 30% VWC mark show EC increasing nearly uniformly with time. The lowest line at that water content level is for March, and above that is May, June, etc. Only April is out of order, with the highest EC values. There appear to be some trends here; so next we take a more statistical look at the data.

Analytical Method

The analysis below focuses on VWC levels from 20% - 30%. Water content from rainfall last winter did not rise above 30%, so this range allows us to include the full ten-month period. As mentioned above we wish to compare EC at fixed VWC levels; so we slice the VWC range of interest into ten 1% pieces from 21% to 30%. For example for 23% VWC we include EC readings taken when VWC ranged from 22.5% to 23.5%.

We also want finer time granularity than the monthly data above. As the first graph at the top of in this report shows, in cooler months the soil maintains the 20% - 30% moisture range for a longer period of time than in summer. Therefore we choose the time unit to be an individual rain/irrigation event, and average all of the EC readings at a given VWC level over a two day period. Readings are taken every 15 minues, so for example we might create a data point by averaging together 7 readings from 24.5% to 25.5% from January 2 and 3, which resulted from rainfall on January 2.

Increase Since July

Before we look at the full ten-month period as a whole, let's focus on a period that showed a significant change in conductivity. Beginning in July a clear trend emerged of increasing conductivity, as shown in the table below, which only includes data from July 1 to the last readings on September 12. We performed linear regression on the data points at each VWC level to find the best fitting line for the conductivity increase. For example at the 28% level, the slope of 0.014 dS/m/day translates into a monthly increase of 0.42 dS/m. The mean correlation coefficient of 0.82 indicates a reasonably strong correlation between VWC and EC.

VWC
(%)
Slope
(dS/m/day)
Std Dev Corr Coeff
210.0150.250.78
220.0160.270.77
230.0160.260.79
240.0150.150.91
250.0140.170.87
260.0130.160.87
270.0130.170.87
280.0140.240.80
290.0130.270.70
300.0180.240.83
Mean0.0150.220.82

To illustrate this trend, below is the graph of EC readings from individual rain/irrigation events at the 28% VWC level. The regression line shows the approximate 0.4 dS/m increase per month described above.

An Annual Cycle?

Any trend in pore water EC readings from December to July is not as clear as from July to the present.The graph below shows EC readings at the 23% VWC level, and a sine regression curve with a standard deviation of 0.30.

Readings induced by the first rain of the season at the beginning of December showed high conductivity, but subsequent rains showed much reduced EC levels. This trend continued until irrigation began in March. In April and May two waterings produced unusually high conductivity readings, only one of which logged a reading in the VWC range of the graph above. Finally this graph again shows the significant increase in EC from July to present.

If we expect the winter rain to lower pore water EC, then it seems reasonable to expect the September values of around 2.6-2.8 dS/m to trend back down to 2.0. The high reading from the first rain last December supports this idea, but at the same time it's just a single event. We need several more months of data to confirm an annual cycle in EC readings.

Version History
Oct 2008  Initial version.
20 Oct 2008  Below is an update based on data from the period mid-September to mid-October. The graph below shows EC when VWC is at the 23% level, against time. In general the EC readings over the past month have remained elevated although the most recent reading is relatively low.

4 Jul 2009  Below is an update based on data through April. The graph below shows EC as a function of time for VWC at the 21% level. We show the 21% level because the soil moisture level reached 23%, the level shown in the October 2008 graph above, only twice since mid-November, while the 21% level was achieved four times.

The graph shows an overall trend of increasing EC levels, at a rate of approximately 0.07 dS/m/month in the linear fit. Since the last report in October, a group of EC readings below the fit were collected in October and November, and a group of EC readings above the fit were collected in January through March. These two groupings also are apparent at all other VWC levels from 17% through 23%.