Rapid urbanization in Florida is increasing the demand for domestic

Performance of a New Capacitance Soil Moisture Probe in a Sandy Soil Lawrence R. Parsons Wije M. Bandaranayake* Univ. of Florida IFAS Citrus Research and Education Center 700 Experiment Station Rd. Lake Alfred, FL 33850 Rapid population growth and increasing urban demand reduce the availability of water for agriculture in Florida. The water-holding capacity of sandy soils in the Central Florida Ridge area is very poor (<0.10 m 3 m 3 ). Improved soil water monitoring probes can help growers manage irrigation more efficiently and conserve water. This study evaluated a new soil water probe (ECH 2 O EC-5 sensor, Decagon Devices, Pullman, WA) in terms of probe-to-probe signal variability, response to fertilizer-induced salinity, and changes in soil temperature, soil volume sampled, sensitivity to pockets of air or dry soil, and performance in the field. Results were compared with an earlier version of a Decagon probe, the EC-20. Results indicated that the new probe has several advantages. The EC-5 was not sensitive to salinity or temperature fluctuations. Probe output change was almost zero when the salinity of the soil was increased by adding fertilizer to a soluble solids concentration of 14 g kg 1. When temperature was changed gradually from 3 to ~38 C, probe output increased by only about 1%. Soil volume sampled by the probe was about 15 cm 3. The change in probe response was negligible when soil cores up to 0.95 cm in diameter near the probe surface were removed. Probes responded well to changes in soil water content in the field. The EC-5 probe output increased noticeably when bulk density was increased from 1.1 to 1.6 Mg m 3. Probe-to-probe output signal and response to bulk density variations can affect the estimation of field water content unless necessary correction factors are utilized. These probes can be useful for monitoring soil water movement, estimating soil water content, and scheduling irrigation. Abbreviations: EC, electrical conductivity. SOIL & WATER MANAGEMENT & CONSERVATION Rapid urbanization in Florida is increasing the demand for domestic water. Consequently, water permitted for agricultural irrigation has been reduced by some water management districts. More than 75,000 ha of Florida citrus are grown on well-drained sandy soils (>95% fine sand) that have available water contents of <0.08 m 3 m 3. Maintaining high irrigation efficiency while keeping the cost of irrigation affordable is a challenge in these soils. Soil water probes can play an important role in tracking the soil water status between irrigations so that excessive irrigation can be avoided. Leaching of irrigation water and nutrients below the root zone can be reduced in these sandy soils if a reliable soil water probe is used to monitor water movement. Some companies have improved soil moisture probes based on feedback from field performance. Decagon Devices (Pullman, WA) has produced several capacitance-type soil moisture sensors called ECH 2 O probes. The ECH 2 O EC-20 probe has a blade about 20 cm long, while the newer ECH 2 O EC-5 and TE-5 probes have blades about 5 cm Soil Sci. Soc. Am. J. 73:1378-1385 doi:10.2136/sssaj2008.0264 Received 14 Aug. 2008. *Corresponding author (wijeb@ufl.edu). Soil Science Society of America 677 S. Segoe Rd. Madison WI 53711 USA All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Permission for printing and for reprinting the material contained herein has been obtained by the publisher. long. The TE-5 is similar to the EC-5 but has the additional capability of measuring soil electrical conductivity (EC) and temperature. These probes are modified to minimize sensitivity to fertilizer-induced salinity, soil temperature variations, and electrical interference in the field. Operational frequency is the main difference between the 20- and 5-cm probes. The EC-20 operates at?10 MHz measurement frequency while the EC-5 and TE-5 operate at 70 MHz (www.decagon.com/ag_research/soil/ soil.php and personnel communication). At higher frequencies, probe sensitivity to salinity and temperature fluctuations can be reduced, but production costs limit the extent to which the operational frequency can be modified. Decagon has produced the new EC-5 probes that operate at a higher frequency at a moderate price (close to the price of EC-20 probe). All three probe types are easy to install, relatively low cost, and require little or no maintenance. Bandaranayake et al. (2007) evaluated EC-20 probes and found that probe-to-probe output variation reduced the precision of volumetric soil water content (θ v ) measurements. Fertilizerinduced salinity and temperature changes also influenced probe output considerably. If dry soil lenses developed due to hydrophobicity (Buczko and Bens, 2006) or uneven wetting next to the sensor surface, then these dry lenses could interfere with the probe response and prevent appropriate characterization of the normal wetting and drying of the bulk soil. The new EC-5 probe is an improved version that was designed to minimize many of these problems. Therefore, the objectives of this study were to evaluate this new probe in terms of (i) probe-to-probe signal variability and its effect on θ v estimation, (ii) probe response to fertilizer-induced salinity, (iii) the soil volume sampled, (iv) 1378 SSSAJ: Volume 73: Number 4 July August 2009

probe sensitivity to pockets of air or dry soil, (v) probe response to compaction (bulk density changes), and (vi) probe response to changes in soil temperature, and (vii) where possible, compare the results with the already tested EC-20 probe (Bandaranayake et al., 2007). MATERIALS AND METHODS This study was conducted in Candler fine sand, a hyperthermic, uncoated Lamellic Quartzipsamment commonly found on the Central Florida Ridge. This soil has >95% sand, <3% clay, and <1% organic matter in the top 2 m. The bulk density of this soil is?1.5 Mg m 3 and water content at field capacity (θ fc ) is 0.08 m 3 m 3. We collected soil from a citrus grove (located at the University of Florida Citrus Research and Education Center, Lake Alfred, FL; 28.101864 N, 81.713366 W) that was selected for probe installation. Soil was air dried and sieved to remove roots and debris. Laboratory evaluations of the ECH 2 O EC-5 soil water probe were similar to those done with the ECH 2 O EC-20 soil water probe (Bandaranayake et al., 2007). Several tests were conducted to evaluate the following: (i) probe-to-probe output variation and its implications for θ v estimation, (ii) probe response to fertilizer-induced salinity, (iii) soil volume sampled, (iv) probe response to a dry soil lens close to the sensor surface, (v) probe response to macropores (Luxmoore, 1981) close to the sensor surface, (vi) probe response to bulk density variations (microporosity), and (vi) probe response to soil temperature changes. Test 1: Probe Response in Air, Water, and Soil We connected the EC-5 probes to Decagon EM50 dataloggers and measured the raw count (when connected to Campbell CR10X dataloggers [Campbell Scientific, Logan, UT], the output is in millivolts) from 12 individual EC-5 probes in air and tap water separately. The probes were held in the air or tap water and the raw count output was continuously recorded for 8 or 16 h (8 h if it was during daytime and 16 h if it was at night) at 15-min intervals using EM50 dataloggers. We continued to record air and water values alternately for 4 d as indicated above. We used fresh tap water each day. Standard procedures used by Paltineanu and Starr (1997) were considered during calibration of the probes in the laboratory. The amount of water required to bring the soil to 2, 4, 6, 8, 10, and 12 m 3 m 3 θ v was calculated by using the weight of the air-dry soil and assuming that the bulk density of the soil reached about 1.5 Mg m 3 after packing. Airdry soil was poured into a rotating concrete mixer and water was added slowly with a spray bottle. The soil was mixed thoroughly for about 15 min, and then the mixed soil of known θ v was poured into a 75- by 45- by 45-cm (length, width, and height) rectangular container. About one-fifth of the total soil was transferred at a time while pounding on the soil surface with a wooden tamper to compact the soil. Each time after the soil was mixed to a different θ v, it was brought to approximately the same bulk density by compacting it to the same height in the rectangular container. Probes were installed vertically about 15 cm apart to avoid any probe-to-probe interference. Sensor output for each θ v was collected with an EM50 datalogger that was programmed to acquire data every 1 min. Mean output values during a 6-min period for each increment of moisture content was used to develop a calibration equation. After each measurement, six soil cores were extracted from different parts of the container to measure the gravimetric water content (Gardner, 1986) and dry bulk density (Blake and Hartge, 1986). The gravimetric water content was then multiplied by the dry bulk density of each core sample to obtain θ v. The raw count was then regressed against the measured θ v using SAS (SAS Institute, Cary, NC). Test 2: Probe Response to Salinity This test was conducted using a soil column (25.7-cm diameter and 60-cm height). The procedure was similar to that described by Bandaranayake et al. (2007). Three EC-5 probes and one TE-5 probe were placed in sets at depths of 12 and 27 cm below the soil surface (the second set of probes was 15 cm below the first set). After applying fertilizer solution to the soil, water was applied to the surface to move the fertilizer close to the sensors. The water input was set to the lowest flow rate possible (measured at 10.8 L h 1 ) until all the fertilizer was drained from the bottom of the soil column. The raw count was compared with that of a control treatment that was run without fertilizer. The salinity level influenced by the added fertilizer solution in the column was measured using the TE-5 probe. Test 3: Soil Volume Sampled Tests 3 through 7, which measured the sensitivity of sensors to identified soil variables, were conducted using a rectangular plastic box that measured 28.5 by 10.5 by 5.0 cm. To test the volume of soil sensitive to θ v changes, wet soil close to θ fc was packed in the plastic box to 1.5 Mg m 3 bulk density and separated from the sensor surface area with?3.0-cm space from either side of the sensor surface using plastic dividers during packing. Two EC-5 probes connected to an EM50 datalogger were placed in the middle of the box. Dry soil (θ v =?0.025 m 3 m 3 ) was packed on both sides of the probe to the same bulk density (1.5 Mg m 3 ). Data were recorded at 3-min intervals for?30 min. If there was no distinct change in output, the wet soil was moved by 0.5 cm toward the sensor blade from either side. When the probe output appeared to change, the distance from the sensor surface was noted and used to estimate the volume of soil sensitive to θ v changes. Test 4: Probe Response to Air-Dry Soil Lens Trapped Close to Sensor Surface To evaluate the probe response to a very dry soil layer (lens) next to the sensor surface, the plastic box was packed with airdried soil (θ v =?0.005 m 3 m 3 ) and three probes were installed at the center line. Wet soil (θ v =?0.07 m 3 m 3 ) was introduced at decreasing distances along the length (Fig. 1), and sensor output was measured for different thicknesses of dry soil associated with the sensor surface. Test 5: Probe Response to Air Pockets Close to Sensor Surface To create macropores, soil cores 4.6 cm long were removed from the soil (0.055 m 3 m 3 θ v ) packed to 1.5 Mg m 3 bulk density in the plastic box (Fig. 1). Cores were removed along the length of the probe 0 to 0.2 cm away from the sensor surface using 0.71- and 0.95-cm-diameter samplers. The volumes of these single cores were?1.82 and 3.23 cm 3, respectively. Sensor response was mea- SSSAJ: Volume 73: Number 4 July August 2009 1379

Fig. 1. The EC-5 performance laboratory tests: (a) the plastic box used in the tests and (b) the plastic box filled with dry or wet soil. sured by changing the size and number of cores along the probe length to evaluate the macropore effect on probe output. Test 6: Probe Response to Bulk Density Changes A known weight and θ v of a soil was compacted to known volumes using calibrated marks on the sliding panels (Fig. 1). Sensor output from three probes was measured in response to changes in compaction. Total porosity (m 3 m 3 ) and air-filled pore space (m 3 m 3 ) were estimated as explained by Danielson and Sutherland (1986) and Bandaranayake et al. (2007). Test 7: Temperature Effect Two kilograms of soil brought to 0.07 m 3 m 3 θ v was packed in the plastic box (Fig. 1) to 1.5 Mg m 3 bulk density. After partial packing, two EC-5 probes and one TE-5 probe (for monitoring temperature) were installed along the center line of the box and were connected to an EM50 datalogger. After the soil was packed, the plastic box with the wet soil and sensors was wrapped with polyethylene film to minimize soil water loss by surface evaporation. The datalogger was programmed to measure the output from the EC-5 sensors and the soil temperature from the TE-5 probe. The plastic box with the sensor datalogger setup was placed in a temperature-controlled chamber; the air temperature in the chamber was gradually increased from?1.7 to?37.8 C and back to 3 C. The sensor output and the soil temperature were recorded at 5-min intervals. Probe Response to Irrigation and Rainfall after Installation in the Field Probes were installed at two field locations with the same soil type but managed differently as part of a larger study designed to observe field performance (Fig. 2). RESULTS Probe-to-Probe and Within-Probe Output Signal Variations The CR10X dataloggers recorded the output signal from the EC-20 probes in millivolts (Bandaranayake et al., 2007), while the EM50 recorded the output signal from the EC-5 probes in raw count (no units). When the sensor response was measured in water and air during a 4-d period, the output signal from individual sensors showed both probe-to-probe variation and within-sensor variation. The mean, minimum, and maximum sensor output from each of 10 EC-5 probes are given in Table 1. The mean output in water from different EC-5 sensors ranged from 1289 to 1400 and the mean output in air ranged from 328 to 362 (Table 1). The water minus air value for each probe ranged from 929 to 1071. When variation within the same probe was considered, the probe in water with the most variation had an output that ranged from 1242 to 1349 during the 4-d time period, and the probe with the least variation had output that ranged from 1313 to 1320 for the same time period. The maximum and minimum variation in air was not always associated with the same two sensors that produced the maximum and minimum variation in water (Table 1). Fig. 2. The EC-5 performance field tests: diagrams of the EC-5 and TE-5 probes and how they were installed in the field (connected to the EM50 datalogger). Individual Probe Performance during Calibration The calibration curve using the measured θ v against the sensor output under laboratory conditions showed a strong linear relationship, with an r 2 1380 SSSAJ: Volume 73: Number 4 July August 2009

of 0.948 (Fig. 3a) and RMSE of 0.008. The individual sensor variations could not be minimized further by normalizing (Bandaranayake et al., 2007) the sensor output values (Fig. 3b) due to random output variations when measured in water (Table 1) and in soil (discussed below) after packing the columns with soil of different water contents. The degree of output variability among probes and within the same probe when measured in a soil medium were identified during laboratory calibration. Figure 3a shows the spread of output from different probes for each measured θ v when plotted against the raw sensor Table 1. The EC-5 probe performance laboratory tests: mean, minimum, maximum, and range (maximum minus minimum) of the sensor output signal (raw count) when measured alternately in water and in air during four consecutive days, and the difference between the air and water values for each probe. ECH 2 O Sensor output in water Sensor output in air Difference (water air) probe Mean Min. Max. Range Mean Min. Max. Range Mean Min. Max. Range EC 5 1 1379 1375 1382 7 328 321 340 19 1050 1038 1058 20 EC 5 2 1343 1334 1348 14 343 336 356 20 1000 989 1008 19 EC 5 3 1400 1394 1404 10 328 323 332 9 1071 1065 1077 12 EC 5 4 1391 1383 1396 13 328 316 360 44 1062 1032 1077 45 EC 5 5 1317 1313 1320 7 333 330 337 7 985 979 989 10 EC 5 6 1338 1327 1341 14 342 339 347 8 995 988 999 11 EC 5 7 1297 1280 1302 22 362 360 366 6 931 920 937 17 EC 5 8 1298 1242 1349 107 345 332 364 32 929 906 971 65 EC 5 9 1342 1338 1345 7 353 351 356 5 988 986 992 6 EC 5 10 1289 1278 1301 23 330 324 341 17 965 948 977 29 Mean 1339 1326 1349 22 339 333 350 17 998 985 1009 23 Min. 1289 1242 1301 59 328 316 332 16 929 906 937 6 Max. 1400 1394 1404 10 362 360 366 6 1071 1065 1077 65 SD 40 50 36 12 14 12 50 51 47 Mean of readings for 4 d. Mean of 10 probes. output. Figure 3b shows the spread when each measured θ v was plotted against the normalized sensor output. The maximum or minimum response for each measured θ v was not consistently associated with a specific probe across the range of θ v values tested and did not follow the same response pattern when they were tested in water or air. This result may be the reason why normalizing did not minimize the data spread at each measured θ v during calibration. Individual sensor variations can cause difficulties when the sensors operate within a narrow soil water range, as in the case of this sandy soil with an available water range of 0.02 to 0.08 m 3 m 3. Figure 4 shows the difference between the measured θ v and the estimated value for each individual probe using the calibration equation from Fig. 3a. The maximum θ v variation from different probes in the range of different θ v values tested was ±0.016 m 3 m 3 (Fig. 4). This variation accounted for about 26% of the available water in the soil. Furthermore, Fig. 4 indicates that the variation displayed by each probe did not indicate a pattern (trend) but varied randomly across different θ v points. Therefore, probe-to-probe variations due to manufacturing can be overshadowed by other external factors that cause output variations. One external factor that was identified in this study was the degree of soil packing. During calibration, probes were randomly inserted in the packed soil each time the soil was brought to a new θ v. Soil packing at each θ v level was done manually; thus, the soil was not always packed to a consistent bulk density within the container. The bulk density varied between 1.55 and 1.64 Mg m 3 spatially within the container and at different packings. Probe response was very sensitive to bulk density (details discussed below and in Bandaranayake et al., 2007), and appeared to override probe-to-probe output variations during manufacturing. Because of random variation in soil bulk density, it can be erroneous to determine the true field water content using a laboratory calibration by relying on output from an individual probe. A sensor raw count output difference of 13 for an EC-5 probe was equivalent to a 0.01 m 3 m 3 change in water content according to the calibration. With the EC-20, the output difference for a 0.01 m 3 m 3 change in available water content was 17 mv (Bandaranayake et al., 2007). Figure 5 and Table 2 show how the probes performed when installed in the field at different depths. In general, these probes were more stable with time in the field than the EC-20 probes (Bandaranayake et al., 2007). The graph and the table indicate the response pattern of these probes to three consecutive irrigation events (early mornings of 19, 23, and 27 June) and a light Fig. 3. Results of the EC-5 performance laboratory tests: measured soil water content plotted (a) against probe output as a raw count and regression equation, and (b) against normalized probe output (raw count) and the regression equation. SSSAJ: Volume 73: Number 4 July August 2009 1381

Fig. 4. Results of the EC-5 performance laboratory tests: the difference between the actual (measured in the field) and estimated soil water contents (θ v, m 3 m 3 ), using the laboratory calibration equation, for Probes no. 1 through 12. rainfall (evening of 19 June). The TE-5 probe output was similar to the EC-5, with the exception that the TE-5 produced a higher output to θ v changes while simultaneously measuring soil temperature and salinity. At the 90-cm depth, the change in θ v was not as great as at the shallower depths. This suggests that little irrigation water reached the 90-cm depth and that water use was also minimal there. The EC-5 probes at 30- and 45-cm depths fluctuated similarly. The estimated average θ v before the three irrigations ranged between 0.086 and 0.037 m 3 m 3 among probes (Table 2). When the TE-5 (because it was a different probe type) and the EC-5 at the 90-cm depth (assuming this depth was below the main root zone) probes were excluded, the depthaveraged θ v before irrigation was 0.046 (indicating that θ fc was depleted by 43% before irrigation). If we account for the maximum probeto-probe θ v variation of ±0.016 (Fig. 4), the field was irrigated when θ fc was depleted between 23 and 63%. If an irrigation was delayed until θ fc was depleted to the extreme end, which was 63% during the spring (flowering and fruit set stages), then this level of water stress could potentially reduce fruit yield. Fig. 5. Results of the EC-5 performance field tests: output from four EC-5 probes placed at 15-, 30-, 45-, and 90-cm depths and one TE-5 probe placed at 15-cm depth in response to irrigation and rainfall from 17 to 29 June 2007. Table 2. The EC-5 probe performance in the field: probe type, installation depth, maximum output and estimated volumetric water content (θ v ) after an irrigation, and minimum output and estimated θ v before the next irrigation. Probe Depth Max. θ v Min. θ v Difference θ v drop cm raw count m 3 m 3 raw count m 3 m 3 raw count % TE-5 15 768 0.167 668 0.086 101 7.7 EC-5 15 741 0.131 632 0.058 109 8.4 EC-5 30 694 0.100 606 0.037 88 6.7 EC-5 45 704 0.107 612 0.042 91 7.0 EC-5 90 668 0.083 648 0.071 20 1.5 Average 715 0.118 633 0.046 82 6.3 Average probe output as raw count from three consecutive irrigations. Estimated θ v using calibration equation, Drop from saturation until next irrigation. Average excluding TE-5 probe and electrical conductivity at 90-cm depth. Response to Fertilizer- Induced Salinity The output from the EC-20 probes changed dramatically after a fertigation and remained unstable until the next irrigation or significant rainfall (Bandaranayake et al., 2007). In laboratory tests, the EC-5 probes responded differently and were not affected by fertilizer. From soil column tests, we observed that added fertilizer during irrigation and drainage did not alter probe-estimated θ v (Fig. 6). The maximum sensor output when irrigated with nonsaline tap water was 860. After injecting fertilizer, the maximum output from the sensors was 861. Figure 6 indicates that the maximum salinity induced by the added fertilizer was equivalent to an EC of 21 ds m 1. This EC is equivalent to?14,000 mg L 1 of dissolved salts (1 ds m 1 is equivalent to approximately 700 mg L 1 ; Hanlon et al., 1993). Test results showed that the EC-5 sensors were not affected by salt in the range of salinity tested. The lack of smoothness of the curve was due to the output variation in the water supply. Sensor response to fertilizer-induced salinity could be a problem if 1382 SSSAJ: Volume 73: Number 4 July August 2009

irrigation is automated and sensor output determines the irrigation timing. Since these probes are not sensitive to salinity, they would be more suitable than the EC-20 probes for use in automated irrigation systems. Sampled Soil Volume and the Sensitivity to Pockets of Air or Dry Soil When the sampling volume is greater, the θ v of the surrounding soil is better represented. The EC-5 probes started to detect water when the wet soil was?1.0 cm away from the probe surface (Fig. 7). Therefore, the sampling volume amounted to about 15 cm 3 (length of sensor width sensing distance from sensor surface 2) (Fig. 2), which was smaller than the volume sampled by the EC-20 probes (<128 cm 3 ; Bandaranayake et al., 2007). Because of the smaller sampling volume, it is important to make sure that the sensors are placed close enough to the sprinklers or drippers so that they can adequately measure the average wetting. During irrigation or rainfall, soil wetting may not be even at all times due to various external factors. If the soil closely associated with the sensor surface does not get wet, the sensors may indicate a dry soil when the soil surrounding the probe is actually wet. The tests conducted indicated that the sensors did not respond to wet soil when an air-dry thin soil layer (0.3 cm) was trapped between the sensor surface and wet soil (Table 3). With this thin, dry soil layer, there was very little change in raw count or estimated θ v during the 45- min measuring time. This uneven wetting is more common with sandy or structured soil types, and sensor unresponsiveness due to a dry soil lens can be minimized by wetting the soil close to the sensor manually if a weak response from a sensor is noticed and there are no other obvious reasons for the problem. The negative influence on probe output due to larger pores in the soil was minimal (Table 4). Macropore volumes representing 17 to 60% of the sampling soil volume did not affect the sensor reading. The effect of larger pores exceeding 1 cm in diameter was not tested because such large pores are not common in this sandy soil. Response to Bulk Density The laboratory tests indicated that the probes were very sensitive to changes in soil bulk density. Probe output increased sharply within the range across which the bulk density was increased (compaction) (Fig. 8, Table 5). Although the number of data points was not sufficient, it appears that the response decreases at higher bulk densities (high compaction) and can approach a plateau (Fig. 8). Therefore, it is very important to pack the soil to the natural bulk density during probe installation. Some soils can settle faster with time, and therefore a small difference in bulk density during probe installation may not have a great impact. In some finetextured soils, however, the settling rate may be very slow and soil bulk density may not reach the average field value during the study period. If the bulk density during calibration is different from that in the field, sensor readings may have to be adjusted to represent the true field water status. When gravimetrically measured field θ v was compared with θ v estimated using the equation derived from the laboratory calibration (Fig. 3a), it appeared that there was a positive shift, i.e., the estimated θ v was higher than the actual field θ v (Fig. 9). The average bulk density in the Fig. 6. Results of the EC-5 performance laboratory tests: probe output with changing electrical conductivity (EC) vs. time, indicating the probe response to fertilizer-induced salinity. Fig. 7. Results of the EC-5 performance laboratory tests: probe output with time when a wet soil (0.07 m 3 m 3 ) layer is advanced toward a sensor surface that is surrounded by drier (<0. 025 m 3 m 3 ) soil. Arrows indicate the distance from the sensor surface to the wet soil. Table 3. The EC-5 probe performance laboratory tests: sensor response (raw count) when a 0.3-cm dry soil layer separates the probe surface and soil with a volumetric water content (θ v ) of 0.06 m 3 m 3. Time Wet soil (θ v = 0.06 m 3 m 3 ) Estimated θ v 0.3 cm air-dry soil between sensor surface and wet soil Estimated θ v Sensor no. 9 Sensor no. 12 Sensor no. 9 Sensor no. 12 min m 3 m 3 m 3 m 3 10 626 651 0.063 553 568 0.012 20 628 649 0.063 553 568 0.012 35 628 649 0.063 554 569 0.013 45 627 650 0.063 554 570 0.013 Table 4. The EC-5 probe performance laboratory tests: the effect of size and number of large pores (macropores created by extraction of cores) on sensor response to volumetric soil water content (θ v ). Core diameter Cores Core volume Core volume Mean sensor ratio output Estimated θ v θ v difference (before after) cm no. cm 3 % 0.71 2 3.64 17 625 0.055 0.000 0.95 2 6.46 30 640 0.065 0.010 0.71 4 7.29 34 622 0.053 0.002 0.95 4 12.91 60 631 0.059 0.004 Percentage of sampling volume. Difference in θ v before and after cores were extracted. SSSAJ: Volume 73: Number 4 July August 2009 1383

Fig. 8. Results of the EC-5 performance laboratory tests: probe output plotted against bulk density. field was measured at 1.65 (SE = 0.007) and the average bulk density during laboratory calibration was 1.58 (SE = 0.004). As seen in Fig. 9, probe output values could be higher in the field for the same water content due to soil factors including bulk density. When these higher output values were applied to the laboratory calibration, the estimated water content appeared to be higher than the actual (an overestimation). A bulk density difference of 0.07 Mg m 3 can influence the probe noticeably (Fig. 8). When corrected using the output vs. bulk density curve, the θ v correction appeared to be 0.013 m 3 m 3. Figure 9 indicates that the field θ v was overestimated by?0.02 m 3 m 3. Signal output from probes in the field may also be affected by roots and pockets of soil with high organic matter (not measured in this study). Response to Changes in Soil Temperature The response of EC-5 probes to soil temperature was less than that of the EC-20 probes (Bandaranayake et al., 2007). When the temperature was changed gradually from 3 to?38 C, the probe output increased from 600 to about 606 (Fig. 10). When the sensors and the soil in the same setup were allowed to cool, the output curve appeared to decrease at a greater rate, lowering the probe output more than that seen in the warming curve. One explanation for why the cooling curve lies below the warming curve is that during warming, sufficient water was evaporated from the soil and condensed on top between the polyethylene cover and the soil surface (see above). During cooling, this condensed water did not move back and mix evenly with the soil. Initially, the θ v of the soil was 0.075 m 3 m 3. When the θ v was lower than the initial θ v, the probe output was also lower. Fig. 9. Results of the EC-5 performance in the laboratory and field tests: field-measured volumetric water content (θ v ) plotted against estimated θ v using the regression equation derived from laboratory calibration. According to Fig. 10, the soil appeared to lose about 1% of its soil water by volume. CONCLUSIONS In this sandy soil, the ECH 2 O EC-5 soil water probe demonstrated many advantages over the EC-20 probe. Overall, the EC-5 was more stable in the field and continued to perform well during a 1-yr period of testing. None of the probes required replacement due to malfunctioning. The output from this probe did not fluctuate noticeably in the field due to electrical interference, as was the case with the EC-20 (Bandaranayake et al., 2007). Despite a smaller sampling volume than the EC-20, the response to soil water changes due to irrigation or rainfall was clear. Signal output change is determined by the saturation level at each depth, and is more or less similar for a particular depth. Probe-to-probe output variance can be a problem when the soil water range monitored is very narrow. This variance is not easy to explain due to the randomness of influences on the sensor from the microenvironment in which the probe is operating. Sensor Table 5. The ECH 2 O performance laboratory tests: sensor response (raw count) to changing bulk density (compaction), porosity, air-filled porosity, total porosity sensor output, and estimated and real volumetric water content (θ v ). Bulk Air-filled Porosity density porosity Total porosity Mean sensor output Estimated θ v Real θ v Mg m 3 m 3 m 3 % m 3 m 3 1.08 0.59 0.51 7.7 557 0.01 0.08 1.25 0.53 0.45 6.8 584 0.03 0.08 1.3 0.51 0.43 6.5 594 0.03 0.08 1.37 0.48 0.40 6.0 632 0.06 0.08 1.47 0.45 0.37 5.6 659 0.08 0.08 1.6 0.4 0.32 4.8 672 0.09 0.08 1.69 0.36 0.28 4.2 673 0.09 0.08 Percentage of sampling volume. Fig. 10. Results of the EC-5 performance laboratory tests: sensor output plotted against soil temperature. 1384 SSSAJ: Volume 73: Number 4 July August 2009

unresponsiveness related to pockets of dry soil near the sensor surface was not observed from sensors that were in the field. They also did not respond to the removal of large soil cores ( 1.0-cm diameter) that were equivalent to the size of dead root channels or soil fauna burrowings. The EC-5 probes did not indicate sensitivity to changing salinity or temperature; however, the probes were very sensitive to bulk density changes due to compaction differences. Although spatial bulk density variations are minimal in this sandy soil, such variation could cause problems in other soils. The laboratory calibration also indicated that the bulk density to which the soil is packed can greatly influence the calibration results. The EC-5 probes have the advantages of low to moderate cost, ease of installation, low maintenance, and reliable performance in the field. When used in conjunction with EM50 dataloggers, they can be a useful tool for monitoring water movement in the soil, estimating θ v, and scheduling irrigation. ACKNOWLEDGMENTS We acknowledge J.D. Holeton for his help in data acquisition and processing. REFERENCES Bandaranayake, W.M., L.R. Parsons, M.S. Borhan, and J.D. Holeton. 2007. Performance of a capacitance-type soil water probe in a well-drained sandy soil. Soil Sci. Soc. Am. J. 71:993 1002. Blake G.R. and K.H. Hartge. 1986. Bulk density. p. 363 366. In A. Klute (ed.) Methods of soil analysis. Part 1. Physical and mineralogical methods. 2nd ed. Agron. Monogr. 9. ASA and SSSA, Madison, WI. Buczko, U. and O. Bens. 2006. Assessing soil hydrophobicity and its variability through the soil profile using two different methods. Soil Sci. Soc. Am. J. 70:718 727. Danielson, R.E., and P.L. Sutherland. 1986. Porosity. p. 443 461. In A. Klute (ed.) Methods of soil analysis. Part 1. Physical and mineralogical methods. 2nd ed. Agron. Monogr. 9. ASA and SSSA, Madison, WI. Gardner, W.H. 1986. Gravimetry with oven drying. p. 493 507. In A. Klute (ed.) Methods of soil analysis. Part 1. Physical and mineralogical methods. 2nd ed. Agron. Monogr. 9. ASA and SSSA, Madison, WI. Hanlon, E.A., B.L. McNeal, and G. Kidder. 1993. Soil and Container media electrical conductivity interpretation. Florida Coop. Ext. Serv., IFAS, Univ. of Florida, Gainesville. Circular 1092. Luxmoore, R.J. 1981. Micro-, meso-, and macroporocity of soils. Soil Sci. Soc. Am. J. 45:671 672. Paltineanu, I.C., and J.L. Starr. 1997. Real-time soil water dynamics using multisensor capacitance probes: Laboratory calibration. Soil Sci. Soc. Am. J. 61:1576 1585. SSSAJ: Volume 73: Number 4 July August 2009 1385