Environmental effects on explosive detection threshold of domestic dogs

Abstract

Detection canines are deployed to detect explosives in a wide range of environmental conditions. These environmental conditions may have negative impacts on canine capabilities as a sensor. This study leveraged an air dilution olfactometer to present controlled odor concentrations of four different energetic materials (double base smokeless powder, Composition C4, ammonium nitrate, and flake Trinitrotoluene) to dogs working in a range of high temperature, standard, and low temperature conditions with high and low humidity conditions. The air dilution olfactometer controlled concentrations independent of environmental condition. Dogs' detection threshold limits were measured using a descending staircase procedure. We measured dogs' threshold twice for each energetic under each environmental condition. Results indicated heterogeneity in effects based on energetic, but all odors were detected at their lowest concentrations under standard conditions. Smokeless powder detection was reduced under all environmental conditions compared to standard and was least detectable under high temperature and humidity conditions. AN detection was poorest under high temperature high and low humidity conditions. C4 in contrast, was least detectable at low temperatures with high humidity. TNT detection was difficult under all conditions, so decrements due to environmental conditions were not statistically detectable. Additional measures were also found to be associated with detection limits. Under high temperature conditions, correlations were observed between canine mean subcutaneous temperature and detection limits, such that dogs experiencing greater temperature increases showed poorer detection limits. In addition, dog's latency to sample the odor port from the onset of a trial was longest in the high temperature conditions. Longer latencies were also predictive of poorer detection performance. Overall, dogs showed deficits in detection sensitivity limits under all environmental conditions for at least one energetic material when the concentration of that energetic material was not directly impacted by the environmental conditions. These results suggest that behavioral factors related to environmental exposure can have important impacts on canine detection sensitivity and should be considered in operational environments.

Fig 1. Schematic diagram of environmental chamber.

Testing environment set up, containing a built-in heater and AC unit. An environmental sensor was attached to the back wall to monitor the environmental conditions while testing to maintain proper conditions. A go/ no go olfactometer testing apparatus was placed in the front of the room to preform training and testing.

Fig 2. Diagram of mass air flow dilution olfactometer.

The path of air flow is shown from left to right. Clean air travels through a warm water bath heated to replicate the same temperature of the odorant water bath. The clean air then travels to the manifold and is dispersed to the MFC A, MFC B and MFC E. Air from MFC A is pushed into the vial containing an odorant located in a separate water bath to collect the headspace from the vial and then pushed out of the vial to a three-way junction connected to MFC B. The air from the vial is mixed with clean air from the manifold the travels to the remaining mass air flow controllers. The odorant is systematically diluted by the mass air flow controllers and then pushed down the odor line to be delivered to the port.

Fig 3. The dog Jack following the 3-down 1-up adaptive threshold procedure for C-4 in standard conditions.

After three correct responses, the concentration was decreased by a half-log. Once Jack responded incorrectly when presented with 0.01 dilution of C4, the concentration was increased by a half-log. Jack correctly alerted 0.01 dilution of C4 but was not able to detect lower concentrations. Eight reversals were reached during testing and threshold was calculated as the geometric mean of the last six reversals.

Fig 4. Average threshold of each odor.

Displaying average dilution of vapor saturation threshold across all dogs for both testing sessions in standard conditions of each odor. SP, C4, AN, and TNT. Error bars show 95% bootstrap estimated confidence intervals.

Fig 5. Threshold (log proportion of vapor saturation) for each environmental condition and odor.

Error bars show bootstrap estimated 95% confidence intervals.

Fig 6. Individual threshold difference for each environment condition compared to standard condition for dogs detecting SP.

Average threshold between both testing sessions for each condition.

Fig 7. Implanted skin temperature during testing.

Error bars show the 95% confidence interval.

Fig 8. Threshold relation to mean subcutaneous temperature.

Line shows the best fit regression.

Fig 9. Respiratory effort ratings during the threshold assessments.

Error bars show the boot strap estimated 95% confidence intervals.

Fig 10. Relationship between mean RE and log threshold.

Fig 11. Latency to sample odor port by environmental condition.

Error bars show the 95% bootstrap estimated confidence intervals.

Fig 12. Relationship between mean latency to initiate a trial in a session with overall threshold.