Three different instruments were used to simultaneously measure soil EC_a (mS/m) within the same area. These included a GCR sensor Veris Quad EC 1000 (Veris Technologies, Inc., Salina, Kansas, USA; VERIS-EC) shown in Figure 1 and two EMI instruments: DUALEM-21S (Dualem, Inc., Milton, Ontario, Canada) and EM-38 (Geonics Limited, Mississauga, Ontario, Canada) shown in Figure 2 . Table 1 summarizes the main parameters of these instruments.

The VERIS-EC used in this study consisted of four rolling coulters and provided output related to shallow (0-30 cm) soil EC_a. The DUALEM-21S consisted of a 2.41 m long tube and had one transmitter coil and four receiving coils. Two of these four coils form a horizontal coplanar (HCP) array at 1 m (DUALEM_HCP-1) and 2 m (DUALEM_HCP-2) distances whereas the other two form a perpendicular coplanar (PRP) array at 1.1 m (DUALEM_PRP-1.1) and 2.1 m (DUALEM_PRP-2.1) distances. The sensing depths for all configurations can be found in Table 2 . Finally, the EM-38 had only one pair of coplanar coils 1 m apart. The unit can be positioned in a horizontal dipole or a vertical dipole mode producing EC_a measurements related to 0.75 and 1.55 m deep soil profiles, respectively. This unit was calibrated before each use according to the manufacturer’s recommendations. Since the vertical dipole is the same as HCP, EM–38_HCP-1 and DUALEM_HCP-1 measurements are comparable (21), the EM-38 instrument was tested only in the vertical dipole configuration. All instruments went through the warming up period for about 5 minutes before each test event.

A LabView (National Instruments, Corp., Austin, Texas, USA) application has been developed to automatically log data from the three sensors at individual data rates. A Watch Dog 2700 weather station (Spectrum Technologies, Inc., Aurora, Illinois, USA) was used to record ambient conditions that might affect instrument performances. Monitored ambient parameters were logged with a 5-min interval and included: air temperature and humidity, wind speed and direction, and rainfall. The same station was used to monitor soil temperature and water content 30 cm below the surface using an installed SMEC 300 (Spectrum Technologies, Inc., Aurora, Illinois, USA) stationary probe.

The instruments were placed in stationary positions approximately 8 m apart and about 6 m away from the data logging station, as shown in Figure 3 . The setup distance was used to prevent signal distortion and interference with sensor readings. The test area at Macdonald Farm of McGill University, Quebec, Canada, was a regularly cut lawn approximately 2 m from the edge of a corn field. The soil type at the test location was identified as Chicot series, sandy loam with moderate water holding capacity, and moderate to poor drainage (37) and had relatively low soil EC_a.

A series of five 4.5-h data recordings were conducted from August to October. Each time, the instruments were placed in the same marked locations. The GCR coulter disks were pushed down gently (about 5 – 10 cm deep) to ensure good contact with the soil. At the same time, the EMI instruments were placed on the flat ground with the roll and pitch of the instruments as close to 0° (normal position) as possible. Another set of 5-min data recordings was conducted over several days from September to November with artificially introduced operational noise. Evaluated factors included: a) 10 cm height above the ground simulating an inconsistent distance between the instrument and soil surface, b) +10° and -10° pitch simulating potential raising of one end of the instrument, and c) +10° and -10° roll simulating deviation of the instrument from its vertical orientation ( Figure 4 ) during a typical mapping exercise. The 0 cm height with 0° roll or pitch represents the typical normal position of the EMI sensors when placed on the ground during the mapping operation. The Table 3 summarizes all data acquisition events that allowed five replicates of temporal and three replicates of operational tests for every instrument.

Data analysis was based on a comparison of 1-s data obtained at the highest possible rate without any filtering. While the temporal tests quantify the potential data drift from the beginning to the end of a single mapping exercise, the operational tests reveal the influence of typical uncertainties of the position of the instrument with respect to the ground. In addition, the test replicates show the influence of ambient conditions along with the possible uncertainties of sensor repositioning and other feasible inconstancies between test replicates.

For both temporal and operational tests, descriptive statistics, such as mean and standard deviation (STD) of each test replicate, were calculated. Root mean square errors (RMSE) for the temporal tests were estimated using the following equation:

where n is the number of 1-s measurement averaged within any specific data log e.g. for 4.5 h; m is the number of different logging events of the test replicates e.g. 5 replications.

The Levene’s test of equal variances (e.g. 38) was conducted to compare mean square error (MSE) values corresponding to different instruments using the raw dataset. Due to a very large number of data records, high degrees of freedom made relatively similar variance estimates significantly different from each other. Therefore, a subjective grouping of similar RSME estimates was performed to facilitate the discussion. Thus, RMSE values less than 0.01 mS/m will be considered temporarily the most stable measurements, 0.01-0.50 mS/m as temporarily relatively stable measurements, and 0.5 mS/m or more as temporarily relatively unstable measurements. A simple linear regression was applied to the relationships between EC_a measurements and ambient conditions, including soil and air temperature, soil water content, air humidity, and internal temperature of the DUALEM-21S instrument. In terms of the operational test, a t–test using α = 0.05 was used to compare the means of three operational test replicates to the mean of nine replicates representing normal operation of the instrument (i.e., zero height, roll and pitch).

Figure 5 indicates the range of air and soil temperatures, relative humidity, soil water content, and recorded internal instrument temperature of the DUALEM–21S during each 4.5-h temporal test. These tests generally cover all reasonable operational conditions when soil EC_a data are normally collected. The weather data captured from the weather station showed ambient and soil temperatures varying from 23.3 °C to freezing (-0.1 °C) and 29.5 to 7.6 °C, respectively. The latter measurements slightly vary within the same measurement date; however, they change greatly from one test event to another. The internal temperature of the DUALEM-21S ranged from 6 to 40 °C across the test dates. Assuming the ambient temperature effect is similar on all instruments, this temperature data then was used as a basis for the comparison between all these three instruments. The increase in soil moisture on 9-Oct was due to rainfall events during the two days prior to the test (6 mm of total precipitation).

Figure 6 illustrates data logs for four different measurements obtained during the 9-Oct test. The ranges (minimum and maximum) for unprocessed soil EC_a measurements for the entire temporal test are presented in Figure 7 . Table 4 summarizes the average, STD, and RMSE (Equation 1) values. The most stable soil EC_a measurements were from the GCR instrument. Earlier, Serrano et al. (33) observed a similar level of consistency of CGR measurements. Both DUALEM PRP measurements produced RMSE values 5-10 times smaller than those from the EM-38 or DUALEM HCP measurements. In addition to the 4.5-h drift of EC_a measurements, there were noticeable changes from day to day. For an unknown reason, the most apparent reduction in EC_a measurements was on 18-Sep for both DUALEM HCP measurements, but not for PRP. That day, the initial internal and ambient temperatures were similar (10.4 and 11.6°C), but a steady increase of the ambient temperature with relatively low wind speed (around 2 km/h) may have resulted in rapid solar warming of the instrument. This typically reduces soil EC_a measurements. However, it is not certain what caused this sensor behaviour.

Table 5 summarizes the correlation coefficients for a linear regression between ambient conditions and recorded measurements. Figure 8 demonstrates the relationships between air, soil and internal DUALEM instrument temperatures with several EC_a measurements. It is obvious that an anomaly, such as the 18-Sep drop in DUALEM HCP measurements, affected the observed relationships. This anomaly cannot be explained by ambient conditions and may be affiliated with a number of unaccounted for factors, such as instrument positioning and the conditions of surrounding vegetation. When disregarded, it appears that the EM-38 measurements are negatively correlated with ambient and internal temperatures. According to Allred et al. (19), low soil water content and high temperature normally reduces soil EC_a. Sudduth et al. (18) reported that the drift over 10% of EC_a observed during field mapping using the EM-38 might be due to a change in internal temperature rather than ambient temperature variations. Corwin and Lesch (11) recommend converting EC_a measurements at a specific temperature to measurements at a reference temperature (e.g., 25°C). Naturally, this would mean that temperature-compensated VERIS-EC and DUALEM-21S measurements would not be affected by ambient conditions to the extent of non-compensated EM-38 measurements. However, the presented data have not revealed temperature-induced changes in EM-38 measurements greater than other effects, such as instrument repositioning. The effects of soil temperature and water content are less quantifiable since they did not change significantly during individual tests.

Figure 9 provides the results of the operational tests (minimum and maximum values) for both EMI sensors for height (0 and 10 cm), roll (−10°, 0°, +10°) and pitch (−10°, 0°, +10°) tests. Each 5-min data log represented a particular test configuration that was repeated in random order on three different occasions during at least two different days. Since normal operation (zero height, pitch and roll) was part of each operational test, this configuration has been replicated nine times. Table 6 shows the individual soil EC_a test average, STD and t-test p-values. In this case, the average of three operational test replicate means were compared with the average of nine normal operation means.

It appears that raising the instrument did not contribute to greater EC_a measurement changes than the differences between replicates. In most cases, HCP measurements decrease when the instrument is raised in the air, but this may not be the case for some sensor configuraions if high EC_a soil overlays less conductive subsoil. A marginal significance of the drop in average EC_a caused by the raised instrument was found for DUALEM_PRP-2.1 which may have been due to the relatively low EC_a difference between replicates rather than the magnitude of this change. In terms of the pitch and roll tests, it appears that the 10° deviations from the normal operation did not have a significant effect on the recorded measurements. The exceptions were EM-38_HCP-1 and DUALEM_PRP-2.1 when the end of the instrument containing the transmitting coil was raised above the ground.

From a practical standpoint, the results of this study indicate that GCR sensing of EC_a may be less sensitive to temporal effects than EMI measurements in many situations and may have be appealing for many environments test. However, the non-contact nature of EMI measurements provides versatility with respect to the measurement environment and, when designing the deployment platform, such as sled (32, 39), these instruments should stay close to the ground with zero pitch and roll. It was determined to be very important to keep the transmitting coil close to the ground. Minor deviations from these conditions do not affect measurements to a greater degree than temporal test replications.