Testing the data transfer capabilities of WildSense: A GPS-based wildlife tracking network

Background/Question/Methods

The construction of low-cost, advanced GPS telemetry systems for wildlife tracking is growing in popularity, especially systems that can communicate with each other to track contacts and, more recently, transfer data. This novel function represents a step forward from current technology because it allows researchers to retrieve data from collars that have been damaged or lost. It also elucidates broad networks of interactions between individuals to monitor disease spread and social preference. We tested the communication and data transfer capabilities of a low-cost, custom-built GPS telemetry system with an on-board wireless sensor network (i.e. WildSense) using people and captive bighorn sheep as an experimental model. Our goal was to determine the data transfer capabilities of our collars at several transmission strengths and with several types of obstruction. We then placed our collars on four bighorn sheep in a captive setting in order to compare collar communications data to ground truth observations using a digital camcorder.

Results/Conclusions

When we modeled data transfer success by distance, we found that data loss occurs in a largely sigmoid pattern. Under conditions of high transmission strength and low obstruction (optimum), a log logistic decay curve was the best model for data transfer success, but under conditions of low transmission strength and high obstruction (sub-optimum), data transfer was less reliable. Animal position adversely affected the baseline data transfer success rate, with bighorn sheep lowering baseline data transfer success to less than 50% at low transmission strengths. Obstruction also affected the optimum distance of data transfer for each transmission strength, but bighorn sheep had less of an impact than did humans. Obstruction adversely affected maximum distance of data transfer for each transmission strength, but bighorn sheep and humans did not differ in their effects. When we used captive sheep to compare contact data to video ground truths, we found that contact distances were accurate to ±5m, with variation in contact distance largely being attributed to the baseline accuracy of the GPS system. Based on the modeled distance of data transfer, there were no false contacts, although redundant contact logs were prevalent at higher rates of communication attempts. This preliminary testing on captive bighorn sheep demonstrates that our system is on track for use in a field setting in the near future.