Field calibration of the ECH2O soil moisture probes at the Brookhaven National Laboratory / NCAR meteorological tower site
Robert Tardif National Center for Atmospheric Research / Research Applications Program
September 2003
-2-
1) Introduction The ECH2O EC-20 probes, manufactured by Decagon Devices Inc., are used to provide measurements of soil moisture content at various depths at the Brookhaven National Laboratory in Upton NY. The sensors are part of a site instrumented for ceiling and visibility studies. The ECH2O EC-20 probes are designed to measure the dielectric constant of the soil medium in contact with the probe and thus estimate its volumetric water content. Since the dielectric constant of water (80) is higher than the one of air (1) or soil minerals (4), the dielectric constant of the soil is a sensitive measure of its water content. The probes are essentially composed of copper traces sealed between two pieces of fiberglass having dimensions of 20 cm x 2 cm. The electromagnetic field generated by the traces travels through the fiberglass and into the soil surrounding the probe. The probe is typically supplied with a 10 ms excitation of 2500 mV and the subsequent rate of change of voltage on the sensor is monitored to estimate the dielectric constant of the surrounding soil medium. The probe produces an output voltage that depends on the dielectric constant of the soil in contact over the entire length and width of the probe.
2) Calibration procedure A relationship that relates the probe output voltage to the actual volumetric water content of the soil surrounding the probe is required. A standard calibration was determined by the manufacturer and is said to be valid for most types of soil with low to moderate sand content. Tests have shown that this calibration relationship leads to significant errors in soils with somewhat high sand content. Field calibration of the probes is suggested if the soil of interest is composed of high proportions of sand or when greater accuracy is needed. Indeed, prior experience with field deployments of such probes, using a calibration relationship derived in the laboratory using glass beads (sandlike), suggests that a field calibration using the actual soil of interest is preferable if a higher degree of accuracy of the soil moisture estimates is desired (R. Cuenca, personal communication).
-3-
Thus, a probe-by-probe, soil type dependent calibration has been performed in the field at the Brookhaven National Laboratory meteorological tower site. Inspection of the soil in the layer from the surface to a 1m depth showed that three different soil types are present over this layer. A preliminary visual inspection of the composition of the topsoil layer (0-25cm) showed the presence of some organic matter (roots), silt particles and some clay particles. A number of sand particles were also apparent in the core samples taken from this layer. The middle layer (25 to 80 cm) is mostly composed of clay, while the deeper layer (depths larger than 80 cm) is composed mostly of sand with a fair amount of small rocks having diameters ranging from 1 to 10 cm. The bulk densities of the natural and dried core samples taken at various depths are presented in table I. There is a tendency for the bulk density to increase with depth, as deeper layers experience greater compaction.
Table I. Bulk densities of core soil samples taken in a soil column from the surface to a depth of 1 m at the Brookhaven meteorological tower site. Data taken on September 8th 2003. Depth (cm) 3 to 6 7 to 10 10 to 13 14 to 17 24 to 27 44 to 47 59 to 61 93 to 96
Wet bulk density (kg m-3) 1631.7 1634.6 1853.0 1762.7 1765.7 1738.0 1804.9 1797.7
Dry bulk density (kg m-3) 1333.3 1406.1 1569.1 1611.4 1480.3 1518.2 1572.1 1694.3
The calibration procedure of the ECH2O probes consists of taking readings of output voltages while the probes are inserted in the particular soil type they will be installed later on. This is done for various water contents to obtain a number of data points, on which a least-squares fit is performed. The resulting linear regression is the calibration relationship that will be used to relate probe output (mV) to volumetric water content (Θ).
-4-
Soil samples were collected from each layer with different soil types. The soil was spread out on a large wooden platform and left to dry as much as possible under the sun for a few hours. The soil was then poured into a plastic “cement mixing” pan. The soil was then prepared by breaking up the larger clumps into finer elements to prevent the presence of air pockets within the soil into which the probes were inserted. The soil-toprobe contact is critical for good measurements with this type of probes. The soil was then compacted into the pan and the particular probes identified for the corresponding layer of soil type were then inserted in the soil, so the entire length of the probe was completely buried. The probes were oriented with their surface perpendicular to the soil surface. The soil was then compacted by hand around either side of each probe. Fingers were used to carefully pack the soil along the entire length of the probes on each side. More soil was then added on top so the probes were buried at least 3 cm below the soil surface. The soil surface was packed again by applying pressure with clenched fists. Once inserted, the probes were left to stabilize for a few minutes. Once this was achieved, 3-min averages of 5 sec. Voltage readings were collected and subsequently averaged over about a 30-minute period. A CR10X datalogger was used to collect the probe measurements. Once the ECH2O probe readings were recorded, the probes were taken out and two soil samples were taken where the probes were located by fully inserting a brass ring having a volume of 68.7 cm3. Soil cores were removed and promptly placed in glass jars. The jars were quickly sealed to prevent evaporation. The glass jars containing the soil cores are then weighted on a precision balance (0.1g precision). The cores are then dried by placing them in a microwave oven for 10 minutes at full power. Jar lids were replaced immediately after drying, and the samples were left to cool back to near ambient temperature. The soil samples were weighted again to determine their dry weight. The jars themselves were also weighted for a tare weight. The weight difference between the wet and dry samples represents the gravimetric mass of water (mw) contained in the soil core samples. Volumetric water content (Θ) is then determined with
md mw ρ d mw V = mw 1 , Θ= = md ρ w md ρ w V ρw
-5-
where ρ d is the bulk density of the sample ( ρ d =
md , with md being the mass of the V
dried sample and V the volume of the soil sample). ρ w is the density of water and is taken to be 1000 kg m-3. The results from the two samples are averaged to obtain a representative value of volumetric soil water content. To obtain multiple data points in order to build a calibration relationship, these steps were repeated 4 or 5 times after adding a certain amount of water to the soil sample and thoroughly mixing it to obtain uniform wetting. The target depths of the six measurement levels are 3, 8, 15, 30, 45 and 90 cm. Thus the top three probes (probes labeled P04-1, P04-2 and P04-3) were calibrated in the topsoil taken in the 0-25 cm soil layer near the surface. Probes P04-4 and P04-5 were calibrated in the clay soil dug up from the 25-80 cm layer, while probe P04-6 was calibrated in the sand taken about 1 m deep. Probe P04-5 was also used while probe P046 was calibrated to obtain a basis for comparison.
3) Calibration results The results from the calibration procedure outlined above are shown in this section. Figure 1 shows the calibration results (symbols) for the probes assigned to the topsoil layer (0-15 cm), while figures 2 and 3 correspond to results obtained for the clay layer (probes P04-4 and P04-5) and the sand layer (probes P04-6) respectively. The P04-5 probe was also used in the sand to obtain a basis for comparison with probe P04-6. The data points suggest a linear relationship between probe output and volumetric soil water content. A linear least-square fit was calculated for each soil type and results are presented on each figure (solid lines). Linear fits for each probe taken individually are summarized in Table II. Probes P04-1, P04-2 and P04-3 generally provided similar values in the topsoil. Although great care was taken during the calibration procedure, random variations in the data are expected due to the difficulty of obtaining accurate volumetric water content measurements, good probe-to-soil contact and uniform wetting of the soil. Nevertheless, the good agreement between the three probes suggests that a single linear fit to the data would provide estimates of volumetric water content with -6-
sufficient accuracy. The linear fit obtained by taking all data points into account is shown in Figure 1 (solid line), along with the standard calibration proposed by the manufacturer (dashed line). It is observed that results from the field calibration show that the probes are somewhat less sensitive to the volumetric water content than what is suggested by the standard calibration. The slope of the linear fit is larger than in the standard calibration (0.001092 vs 0.000695). Also, there is a significant discrepancy between the intercepts of the linear fit. The regression on the experimental data shows that the probes would output about 330mV with a zero volumetric water content, while the standard calibration yields a value of 417mV. According to these results, the use of the standard calibration with probes P04-1, P04-2 and P04-3 would induce an underestimation in the observed volumetric water content by about 50 to 60%. The results obtained for probes P04-4 and P04-5 inserted in clay are shown in Figure 2. Here, we observe a systematic difference between the results from the two probes. Probe P04-4 consistently provides a higher voltage compared to probe P04-5 for the same water content. The difference is about 25 mV. By calculating a linear fit on the data from individual probes, the difference between the estimated volumetric water content from both probes is about 0.03 m3/m3, which represent a 10% error for moderate to high water contents, and about 30% for low water contents. Thus, it is argued that given the expected experimental errors, a single fit can still be used as the calibration relationship. The linear regression obtained by taking all data points is shown in Figure 2 as the solid line. Again, the standard calibration relationship is shown for comparison. Once again, a gross underestimation of the water content would be obtained by using this relationship. Figure 3 shows the calibration results for probes P04-5 and P04-6, obtained using the sand from the soil layer located below 80cm in depth. Similar values are obtained from both probes, with a slight tendency for probe P04-5 to output a lower voltage. Nevertheless, differences between the two probes do not warrant the use of individual calibration relationships. Thus a single linear fit was calculated using all available data points (solid line in Figure 3). Again, significant differences are observed compared to the standard calibration.
-7-
The results from the field calibration indicate that the use of distinct calibration relationships should be used for each layer characterized by a particular soil type.
0.45
P04-1
0.40
P04-3
3
3
Volumetric water content (m /m )
0.50
0.35 0.30
P04-2 y=0.001092x-0.3620 Stand. cal.
0.25 0.20 0.15 0.10 0.05 0.00 350
400
450
500
550
600
650
700
750
800
Voltage (mV)
FIG. 1. ECH2O calibration in the topsoil from the soil layer from the surface to 25cm depth at the Brookhaven meteorological tower site. A linear fit to the data is shown (solid line), along with the standard calibration (dashed line).
P04-4
0.45
P04-5
3
3
Volumetric water content (m /m )
0.50
0.40
y=0.001004x-0.3675 Lab. cal.
0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 350
400
450
500
550
600
650
700
750
800
Voltage (mV)
FIG. 2. Same as Fig. 1, but for the P04-4 and P04-5 ECH2O probes inserted in the clay from the 25-80cm soil layer. -8-
0.45
3
3
Volumetric water content (m /m )
0.50
0.40 0.35
P04-5 P04-6 y=0.000905x-0.3413 Lab. cal.
0.30 0.25 0.20 0.15 0.10 0.05 0.00 350
400
450
500
550
600
650
700
750
800
Voltage (mV)
FIG. 3. Same as Fig. 1 but for the P04-5 and P04-6 ECH2O probes inserted in the sand from the soil layer below 80cm in depth.
Table II. Calibration relationships relating the probe output (mV) and volumetric soil water content (Θ) for the six ECH2O probes, derived from data from a field calibration experiment at the Brookhaven meteorological tower site. Results from individual probes and overall results for each soil type are shown, along with the standard calibration from the manufacturer. Topsoil: P04-1: P04-2: P04-3: All: Clay: P04-4: P04-5: All: Sand: P04-5: P04-6: All: Standard calibration:
Θ = 0.001143 ⋅ mV − 0.3812 Θ = 0.001115 ⋅ mV − 0.3745 Θ = 0.001011 ⋅ mV − 0.3233 Θ = 0.001092 ⋅ mV − 0.3620
(1) (2) (3) (4)
Θ = 0.001025 ⋅ mV − 0.3929 Θ = 0.001013 ⋅ mV − 0.3605 Θ = 0.001004 ⋅ mV − 0.3675
(5) (6) (7)
Θ = 0.000899 ⋅ mV − 0.3313 Θ = 0.000922 ⋅ mV − 0.3965 Θ = 0.000905 ⋅ mV − 0.3413 Θ = 0.000695 ⋅ mV − 0.29
(8) (9) (10) (11)
-9-
4) First field measurements Once all probes were installed in the ground, some data was collected to make sure the probes were functioning properly. Figure 4 shows the temporal evolution of the measured volumetric soil water content over a 24-hour period on September 13th 2003 (day following the installation of the probes). Equation (4) was used relate probe output to volumetric water content for probes at depths of 3, 8 and 15 cm, equation (7) was used for probes at 30 and 45 cm while equation (10) was used for the 90 cm level. Measurements show constant water contents throughout the day, except for a slight increase for the two topmost levels starting at 1830 UTC. Rain showers occurred at around that time. Figure 5 shows a comparison of volumetric water content profiles derived from the initial core soil samples and from the ECH2O probes once they were installed. Given the fact that a few warm and sunny days had elapsed between the time the core samples were taken and the installation of the probes, some evaporation from the top of the soil layer is to be expected. Thus the fact that ECH2O measurements show smaller values than the core samples is not inconsistent. Values for the middle layer (~50 cm) show good agreement, while values for the deepest measurement level suggest that the ECH2O probe tends to provide an estimate that is too low. This could be due to poor soil-to-probe contact, as it was difficult to insert the probe into the soil due to the presence of numerous rocks. Nevertheless, given the quite imperfect comparison, it can be concluded that general good agreement is obtained, indicating that the ECH2O probes seem to provide good measurements. On September 23rd 2003, precipitation was observed as a front passed through the region. Figure 6a presents data from the GEONOR rain gauge. It shows the occurrence of two precipitation events, one starting at 1500 UTC and the other a little after 1600 UTC. Figure 6b shows the corresponding temporal evolution of soil moisture. An immediate increase in the near-surface soil moisture is observed as precipitation occurs, showing the expected sensitivity of the probe. The probe located at a depth of 8cm first detects an increase in soil moisture after a delay of about 30min. This suggests an infiltration rate within the soil of about 10cm per hour. The 0.6 inch of rain that fell on that day lead to an increase in the superficial soil moisture from 0.25 m3m-3 to 0.30 m3m-3. - 10 -
BNL - Sept. 13 2003
0.25
3
3
Volumetric water content (m /m )
0.30
0.20
0.15
3cm 8cm 15cm
0.10
30cm 45cm 90cm
0.05
0.00 0
1
2
3
4
5
6
7
8
9
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time (UTC)
FIG. 4. Time series of soil volumetric moisture content obtained from ECH2O probes on September 13th 2003, from 0000 to 1200 UTC. 3
3
Volumetric water content (m /m ) 0.00 0
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
-10 -20 depth (cm)
-30 -40 -50 -60 -70 -80 -90
from core samples (Sept. 8) ECH2O data (Sept. 13)
-100
FIG. 5. Comparison of profiles of volumetric water contents derived from the core soil samples taken on September 8th, and the average of measurements from the ECH2O probes taken on September 13th.
- 11 -
a)
b)
FIG. 6. Temporal evolution of liquid water equivalent from the GEONOR rain gauge (a) and volumetric water content from the ECH2O probes (b) on September 23rd 2003. - 12 -