A Computerized System for Remote Monitoring of ... - BioOne

HOUSEHOLD AND STRUCTURAL INSECTS

A Computerized System for Remote Monitoring of Subterranean Termites near Structures NAN-YAO SU Ft. Lauderdale Research and Education Center, University of Florida, Institute of Food and Agricultural Sciences, Ft. Lauderdale, FL 33314

J. Econ. Entomol. 94(6): 1518Ð1525 (2001)

ABSTRACT Sensors comprising wooden stakes painted with conductive circuits of silver particle emulsion were inserted in Sentricon stations in soil near structures. Sensors were wired to a datalogger that was programmed to test for circuit breakage every 2 h and store the data in its memory. A host computer was programmed to access the datalogger through telephone communication lines for data download every 4 d. The computerized monitoring system was tested in three remote sites, and site visits were conducted monthly for 6 mo to examine system accuracy in detecting termite activity. The mean monthly accuracy for the system to correctly report the presence (true positive) or absence of termites (true negative) in the stations was 85%, but the accuracy at 6 mo after system installation ranged from 41 to 79%. Mean sensor longevity, deÞned as the time for a sensor circuit to break in the absence of termites, was ⬇4.4 mo. KEY WORDS Isoptera, datalogger, automated termite monitoring system, colony elimination

UNLIKE THE TRADITIONAL barrier treatments using liquid insecticides, a monitoring-baiting program such as the Sentricon system (Dow AgroSciences, Indianapolis, IN) relies on routine inspection by pest control professionals to protect structures from subterranean termites (Su 1994). Because the insecticidal baits are used only when termites are detected, the adoption of monitoring-baiting programs in the past few years has reduced pesticide use by the termite control industry. Such reduction was predicted by La Fage (1986). For a ⬇200-m2 house, 5Ð10 kg termiticides may be applied to soil to create a chemical barrier for exclusion of soil-borne termites, whereas only ⬇1 g of hexaßumuron may be used in the Sentricon system to eliminate a colony of subterranean termites (Su 1994). A house is protected by a monitoring-baiting program that relies on a routine monitoring for early detection of termite activity. Manual monitoring, however, is labor-consuming and costly because a technician has to be on site to open each station for visual inspection. For some termite species, a frequent inspection may disrupt termite feeding in the stations. Moreover, some homeowners often question if the monthly or quarterly on-site inspection is frequent enough to prevent termite damage before detection and subsequent baiting. This may be one reason why the Sentricon system, for example, has been used mostly for remedial control rather than preventative control (Grace and Su 2001). These factors highlight a need for an automated monitoring system that can frequently monitor termite activity near a house. In a previous report (Su et al. 2000), we described a sensor consisting of a wooden stake painted with a conductive circuit of silver particle emulsion. When

placed in an in-ground Sentricon station, circuit breakage was detected by a circuit tester when the silver traces were damaged by termite feeding. In the current study, similar sensors were wired to a datalogger that was programmed to record the sensor circuit continuity at predetermined intervals. A host computer was then programmed to download circuit continuity data from the datalogger through telephone communication lines. This paper reports the Þeld results of this computerized remote monitoring system. Materials and Methods Computerized Remote Monitoring System. As shown in Fig. 1A, a sensor consisted of a monitoring device (spruce stake, 21.5 by 2 by 1 cm) on which a thin line ((2 mm wide) of conductive silver particle (⬍10 ␮m) emulsion was painted in a serpentine pattern with a conductive pen (Chemtronics, Kennesaw, GA) (Su et al. 2000). An extractor, sandwiched between a monitoring device and a sensor, was inserted into each Sentricon station housing installed at soil level. Several sensors were wired together using 20AWG (0.41 mm diameter) electric cable to form a continuous loop of an electric circuit that was then connected to the I/O (input/output) outlet of a datalogger (CR10X, Campbell ScientiÞc, Logan, UT) (Fig. 1B). Cables from each sensor exited through predrilled holes on each side of the station housing at ⬇2 cm below soil level (Fig. 1A). Because there were Þve I/O outlets for a datalogger, stations placed at 2to 10-m intervals at each site were grouped into Þve zones, and the continuous loop of sensor circuit for

0022-0493/01/1518Ð1525$02.00/0 䉷 2001 Entomological Society of America

December 2001

SU: REMOTE MONITORING OF SUBTERRANEAN TERMITES

1519

Fig. 1. A sensor consisting of a monitoring device with a conductive circuit was inserted in a Sentricon station in soil near a structure (A). Several sensors in a predesignated zone of a site were connected together to form a continuous loop of an electric circuit that was then connected to the datalogger (B). There were Þve zones per site. For site SHL, a voice/data router was installed so that the existing telephone line can be shared between the datalogger and existing telephone equipment.

each zone was connected to one of the Þve I/O outlets of the datalogger. Stations were grouped according to logistic consideration when connecting them with the datalogger using underground cables. Some stations could not be connected together with cables due to obstacles such as concrete foundation, pavement or tree roots. The number of stations in each zone varied from 3 to 10 (Fig. 2). CR10X, a programmable datalogger equipped with nonvolatile memory and a modem, was programmed to apply 2,500-mV current through the circuit loop of each sensor zone and record the return voltage every 2 h. When the circuit was intact and the electric resistance was negligible, the return voltage remained ⬇2,500 mV. A decline in return voltage (⬍2,500 mV) signiÞed circuit breakage within the zone. Records of circuit breakage that lasted only brießy (⬍2 h) were excluded from the breakage count for calculating accuracy because such nonlasting, single-event breakages were typically caused by wood expansion as reported by Su et al. (2000), and they were distinguishable from circuit breakages caused by termite feeding. The bi-hourly return voltage readings were stored in the datalogger memory. A host computer located at the Ft. Lauderdale Research and Education Center, University of Florida, was programmed to dial the datalogger for data download every 4 d for 6 mo.

Testing Sites and Termite Survey. The monitoring system was installed at three remote sites including one in Ft. Lauderdale (SHL) and two in Vero Beach, FL (VER and VR0). Wooden (Picea sp.) stakes (2 by 3 by 30 cm) were driven into the ground at each site to detect termite activity. When subterranean termites were found in survey stakes, underground monitoring stations (plastic collars 17 cm in diameter and 15 cm in height) containing feeding blocks (6 spruce boards [7 by 13 by 2 cm] nailed together) (Su and Scheffrahn 1986) were installed to monitor foraging activities of the eastern subterranean termite, Reticulitermes flavipes (Kollar). When more than three monitoring stations were infested with termites, worker termites collected from a station with high activity (⬎5,000 termites) were fed Þlter disks (Whatman No. 1, 5.5 cm) stained with 0.05% (wt:wt) Nile Blue A (Su et al. 1991) for 3 d before being released back into the same station. One month after the release, stations in each site were examined for the presence of marked termites. The foraging territory of each colony, deÞned as the area encompassed by interconnected stations, was determined by the presence of marked termites. The site SHL was a residential duplex with signs of previous infestation by subterranean termites, Reticulitermes spp., but no live termites were found during the testing period. Eighteen Sentricon stations in-

1520

JOURNAL OF ECONOMIC ENTOMOLOGY

Vol. 94, no. 6

Fig. 2. Wooden sensors placed in Sentricon stations (asterisks) near a structure were grouped in Þve zones and wired to a datalogger. Tests were conducted at three sites, SHL, VER, and VR0. No termites were found at site SHL, but one colony each of R. flavipes was found in sites VER and VR0. Colony foraging ranges (shaded areas) were determined by marking and releasing termites through underground monitoring stations (solid circles).

stalled in planters along a partial perimeter of the structure were divided into Þve groups (Fig. 2), and sensors placed in the stations of these zones were connected to a datalogger mounted on the exterior wall. A voice/data router (SR3 Selective Ring Router, Multi-Link, Lexington, KY) was installed at this site so that the existing telephone line could be shared with the datalogger. The site VER was a parking structure at the Florida Medical Entomology Laboratory, University of Flor-

ida, in Vero Beach, FL. Numerous foraging tubes of the eastern subterranean termite were found on the interior walls of a storage room connected to this structure and termite activity was so extensive that cellulosic laboratory supplies in the room were severely damaged. The mark-recapture study identiÞed a colony of R. flavipes foraging beneath the parking structure (Fig. 2). Twenty-seven Sentricon stations containing sensors placed in soil surrounding the structure were grouped into Þve zones and connected

December 2001


1521

Fig. 3. Bi-hourly voltage records from Þve zones of sensors for site SHL for 6 mo. A signiÞcant decline of voltage from 2,500 mV indicates circuit breakage of sensor(s) in a zone. Because no live termites were found at SHL, all circuit breaks occurred in the absence of termites and were false positives (FP).

to a datalogger mounted on the interior wall of the storage room. Wooden sensors infested with termites were removed and replaced with cellulose baits containing 0.5% hexaßumuron (Recruit II, Dow AgroSciences, Indianapolis, IN). Cables that were attached to the infested sensors were removed from the sensors and reconnected to the datalogger so that the circuit continuity of the zone was restored. The baited station was not monitored by the datalogger during the baiting period. Substantially consumed baits (⬎75% visual estimate) were replaced with fresh baits. When termite activity was no longer observed in any Sentricon stations or underground monitoring stations, baits were replaced with fresh sensors that were reconnected to the circuit loop of each zone for continuous monitoring of termite activity by the datalogger. At site VR0, live termites were found beneath many railroad ties on the ground southeast of the administrative building of the Florida Medical Entomology Laboratory (Fig. 2). The mark-recapture study identiÞed a colony of R. flavipes extending from the railroad ties to planters east and north of the building. Despite such extensive foraging range for this colony, no termite activity was found inside the structure.

Twenty-nine Sentricon stations containing sensors placed in the soil around the building and near the railroad ties were grouped in Þve zones and connected to a datalogger mounted on the western exterior wall of the building beneath a wooden deck. During the monthly site visit, sensors with broken circuits were replaced with fresh sensors that were reconnected to the datalogger for continuous monitoring. No bait was applied to this colony at site VR0 so that the sensorsÕ ability to detect termites could be tested throughout the entire 6-mo period. Data Collection and Analysis. Monthly foraging activity was determined by the numbers of active stations (with live termites or signs of termite activity such as feeding damage), including both the monitoring stations and Sentricon stations with monitoring devices. Sentricon stations with Recruit II baits were excluded from the count. During the monthly site visit, Sentricon stations were manually opened to examine the presence of termites. Site inspection results were compared with datalogger records for assessment of sensor accuracy. Results were assigned to one of four categories: true negative (TN), circuit intact in the absence of termites; true positive (TP), circuit break-

1522 Table 1. Site

SHL

Site total T/F VER

Site total T/F VRO

Site total T/F Overall T/F


Vol. 94, no. 6

Sensor accuracy of a computerized remote termite monitoring system tested in three sites Month

Truea

Falseb

TN

TP

FN

FP

1 2 3 4 5 6

18 17 15 16 17 11

0 0 0 0 0 0

0 0 0 0 0 0

0 1 3 2 1 7

1 2 3 4 5 6

23 25 21 25 23 11

2 2 4 0 0 0

2 0 0 0 0 0

1 2 3 4 5 6

26 23 21 22 18 17

1 3 3 7 2 6

2 0 0 0 3 1

94

136

149 379

14

26

25 65

0 0 2 2 4 16 0 3 5 0 6 5

% accuracyc (mean) 100 94 83 88 94 61 (87) 93 100 93 93 85 41 (84) 93 90 83 100 69 79 (86) (85)

Wilcox ranked testd Z

P

⫺2.201

0.0277

⫺1.992

0.0464

⫺2.201 ⫺3.642

0.0277 0.0003

a

TN, true negative, circuit intact in the absence of termites; TP, true positive, circuit breakage in the presence of termites. FN, false negative, circuit intact in the presence of termites; FP, false positive, circuit breakage in the absence of termites. % accuracy ⫽ 100 ⴱ (TP ⫹ TN)/(TP ⫹ TN ⫹ FP ⫹ FN). d Standard normal distribution (Z) and signiÞcance level (P) between true (T ⫽ TN ⫹ TP) and false (F ⫽ FN ⫹ FP) responses. b c

age in the presence of termites; false negative (FN), circuit intact in the presence of termites; and false positive (FP), circuit breakage in the absence of termites. Accuracy rate (A) was calculated as the percentage of both true categories within all categories, or A ⫽ 100 ⴱ (TN ⫹ TP)/(TN ⫹ TP ⫹ FN ⫹ FP). The results were analyzed monthly for signiÞcant differences between true (T ⫽ TN ⫹ TP) and false (F ⫽ FN ⫹ FP) responses for all sites and for the entire 6-mo test period with the WilcoxonÕs signed-ranks test (SAS Institute 1998). Sensor longevity was calculated as the mean time before a sensor produced a false

Fig. 4. Foraging activity of R. flavipes colonies at sites VER (solid circles and line) and VR0 (open circles and dash line) as represented by the number of active loci (number of monitoring stations and Sentricon stations with termites, excluding Sentricon stations containing bait). Baits containing 0.5% hexaßumuron (Recruit II) were placed in Sentricon stations at VER from March to May (arrows).

positive response, namely sensor circuit breakage caused by factors other than termite feeding. Only sensors installed at the beginning of the test were included in calculating sensor longevity. This excludes sensors placed later and thus were tested at shorter period than the original sensors.

Results and Discussion Site SHL. Two weeks after system installation, small (⬍0.3 V) and/or brief (⬍2 h) voltage declines were recorded from zone one (Fig. 3). But the on-site inspection did not reveal obvious circuit breakage caused by termite feeding. Instead, small circuit fracturing as reported by Su et al. (2000) was observed on some sensors in zone 1. Because the brief breakages reconnected by themselves, these miniscule intermittent voltage declines were probably caused by minute circuit fracturing when wooden sensors expanded and contracted in response to moisture ßuctuations. With the exception of the small breakage in zone 1, sensors in all zones remained intact for the Þrst month. At 2 mo, the small circuit separation in a sensor of zone 1 began to break substantially. Because there was no termite activity at SHL, all circuit breakage, as indicated by the voltage declines, were categorized as false positive responses (Fig. 3). By the third month, one sensor each of the zones 3, 4, and 5 broke, yielding an accuracy rate of 83% (Table 1). By 6 mo, however, more sensors broke in the absence of termites, resulting in a 61% accuracy rate. The mean accuracy of sensors for the entire 6-mo period at site SHL was 87%.

December 2001


1523

Fig. 5. Bi-hourly voltage records from Þve zones of sensors for site VER for 6 mo. TN, true negative, circuit intact in the absence of termites; TP, true positive, circuit breakage in the presence of termites, FN, false negative, circuit intact in the presence of termites; FP, false positive, circuit breakage in the absence of termites.

Site VER. Termites were found in four monitoring stations in November 1996, but activity ceased during the winter months (Fig. 4). Within a month of installing the monitoring system, two substantial voltage declines (1,200 Ð1,500 mV) were recorded in zones 3 and 5 in March 1997 (Fig. 5). A site visit conÞrmed circuit breakage in one sensor each of these two zones by termites resulting in true positives. Termites were also found in one sensor each in zones 1 and 2, but termites apparently had just entered these stations and had not yet broken the circuit, resulting in false negative responses (Fig. 5; Table 1). In addition to these four Sentricon stations, termites were found in two monitoring stations; totaling six active loci for this R. flavipes colony in March (Fig. 4). Sensors in all four active Sentricon stations with termite activity were replaced with Recruit II baits in March. From April through May (second to third month), six additional stations infested by termites were correctly recorded by the datalogger (Table 1). Because there were multiple stations within each zone, circuit breakage was sometimes caused by a combination of true positive (termite damage) and false positive (no termites)

readings as shown in the datalogger output in May for zones 2 and 5 (Fig. 5). Recruit II was placed in all six stations with termite activity in April and May, and after a 3-mo baiting period, termite activity ceased from all Sentricon and monitoring stations in June (Fig. 4). In contrast, a R. flavipes colony at nearby site VR0 that did not receive hexaßumuron baits remained active through 1998. Even in the absence of termites at site VER, some sensor breakage were recorded sporadically 4 Ð5 mo after system installation, and by 6 mo, circuits of ⬇60% of the sensors in site VER broke in the absence of termites (Table 1) as indicated by the voltage declines in zones 2, 3, 4, and 5 (Fig. 5). The accuracy rate declined to 41% at 6 mo, but the overall accuracy rate for VER was 84% with signiÞcantly more true than false (positive or negative) results (Table 1). Site VR0. A stake survey initiated in the spring of 1997 yielded two active stations in May, and the numbers of active loci (monitoring stations and Sentricon stations with termite activity in monitoring devices) increased steadily throughout the summer (Fig. 4). One month after system installation, one sensor in zone 3 was broken by termite feeding (Fig. 6), and a

1524


Vol. 94, no. 6

Fig. 6. Bi-hourly voltage records from Þve zones of sensors for site VR0 for 6 mo. TN, true negative, circuit intact in the absence of termites; TP, true positive, circuit breakage in the presence of termites, FN, false negative, circuit intact in the presence of termites; FP, false positive, circuit breakage in the absence of termites.

site visit revealed that one station each in zones 1 and 2 was also infested by termites, but the sensors were still intact (false negative) (Table 1). Sensors broken by termites were replaced by fresh sensors that were reconnected to the circuit loop of each zone. Because baits were not used, termites continued foraging in several stations and their feeding activity was recorded by the datalogger (true positive) (Fig. 6). As time progressed, many sensors also broke in the absence of termites. Such false positive circuit breakage was in most cases caused by moisture-induced wood expansion (Table 1) as visually conÞrmed during the on-site visit. On many occasions, the declines in voltage readings included both false positive and true positive results. Five to 6 mo after system installation, false positive response increased and the accuracy declined to 69 Ð79% (Table 1). The mean monthly accuracy for the system to correctly report the presence (TP) or absence (TN) of termites in the stations at the three sites was 84 Ð 87%, with an overall mean accuracy of 85% (Table 1). Although there were signiÞcantly more true results (TP⫹TN) than false results (FN⫹FP), the accuracy

substantially declined 5Ð 6 mo after system installation. In most cases moisture-induced wood expansion appeared to be responsible for circuit breakage. Such circuit breakages were visually conÞrmed by the minute fracturing on the silver traces as reported by Su et al. (2000). The mean longevity of the sensors installed at the start of the test was 4.4 mo. In a previous study (Su et al. 2000), similar wooden sensors were used in the sandy soil of Little Cayman Island, and the accuracy of 89% at 7 mo was better than that in current study. Soil in site VER and VR0 was mucky and saturated with water, which may have contributed to the faster circuit breakage than the well-drained sandy soil present in Little Cayman Island. With the exception of the wooden sensors, other components of the system including computer hardware, customized software for automated download, telephone communication elements, datalogger and associated software, performed ßawlessly. All of these components are commercially available. The current system was used to monitor in-ground activity of termites, but can be adopted to monitor above-ground activity with appropriate sensors. There are other ter-

December 2001


mite detection devices that may be incorporated into the automated systems, including an acoustic emission detector (Scheffrahn et al. 1993) and electronic odor devices (Lewis et al. 1997). For an automated monitoring system to be practical, however, a simple and inexpensive detector such as the wooden sensors described in this study may be most feasible. Wooden sensors were used in this study because termites readily feed on wood and because the wooden monitoring device is currently used in the commercial Sentricon system. The use of wood as the sensor medium, especially in wet environments, yielded a short sensor longevity of 4.4 mo and accuracy of 41Ð79% at 6 mo, which is probably unacceptable for commercial use. Aside from the cost factor (current monitoring system exceeds $10,000 per unit), future studies need to address improving the sensor longevity if the system is to become a useful tool for the termite control industry.

Acknowledgments We thank R. Baker (Florida Medical Entomology Laboratory, University of Florida) for logistical assistance, P. Ban and R. Pepin (University of Florida) for technical assistance, H. Puche (University of Florida) for assistance in statistical analysis, J. Perrier (University of Florida) for Þgure illustrations, and R. H. Scheffrahn and B. Cabrera (University of Florida) for review of the manuscript. This research was supported by the Florida Agricultural Experiment Station and a grant from Dow AgroSciences, and approved for publication as Journal Series No. R-07945.

1525

References Cited Grace, J. K., and N.-Y. Su. 2001. Evidence supporting the use of termite baiting systems for long-term structural protection. Sociobiology 37: 301Ð310. La Fage, J. P. 1986. Subterranean termites: a personal perspective. Proceedings, National Conference on Urban Entomology, College Park, MD, 24 Ð27 February 1986. Lewis, V. R., C. F. Fouche, and R. L. Lemaster. 1997. Evaluation of dog-assisted searches and electronic odor devices for detecting the western subterranean termite. For. Prod. J. 47: 79 Ð 84. SAS Institute. 1998. StatView reference. SAS Institute, Cary, NC. Scheffrahn, R. H., W. P. Robbins, P. Busey, N.-Y. Su, and R. K. Mueller. 1993. Evaluation of a novel, hand-held, acoustic emissions detector to monitor termites (Isoptera: Kalotermitidae, Rhinotermitidae) in wood. J. Econ. Entomol. 86: 1720 Ð1729. Su, N.-Y. 1994. Field evaluation of hexaßumuron bait for population suppression of subterranean termites (Isoptera: Rhinotermitidae). J. Econ. Entomol. 87: 389Ð397. Su, N.-Y., and R. H. Scheffrahn. 1986. A method to access, trap, and monitor Þeld populations of the Formosan subterranean termite (Isoptera: Rhinotermitidae) in the urban environment. Sociobiology 12: 299 Ð304. Su, N.-Y., P. M. Ban, and R. H. Scheffrahn. 1991. Evaluation of twelve dye markers for population studies of the eastern and Formosan subterranean termites (Isoptera: Rhinotermitidae). Sociobiology 19: 349 Ð362. Su, N.-Y., P. M. Ban, and R. H. Scheffrahn. 2000. Control of Coptotermes havilandi (Isoptera: Rhinotermitidae) with hexaßumuron baits and a sensor incorporated into a monitoring-baiting program. J. Econ. Entomol. 93: 415Ð 421. Received for publication 27 February 2001; accepted 17 July 2001.