1. Experiences and Challenges in Campaign Style Deployments using Wireless Sensor Networks Jayant Gupchup †, Scott Pitz *, Douglas Carlson †, Chih-Han Chang *, Michael Bernard *, Andreas Terzis †, Alex Szalay ±, Katalin Szlavecz * Department of Computer Science, Johns Hopkins University † Department of Physics and Astronomy, Johns Hopkins University ± Department of Earth and Planetary Sciences, Johns Hopkins University*

2. Campaign Style Deployment 2 2/22/2016 Jayant Gupchup Quito

3. Requirements / Constraints No access to line power or internet Collect data at a high rate (e.g. 30s) –Deployment order of days to weeks Researchers require access to data in field 3 2/22/2016 Jayant Gupchup

4. A Typical Sensor Network Gateway/ Basestation Stable Storage ……. 7 Ah

5. Differences / Challenges On-site decisions Ad hoc hardware reconfigurations Dealing with high data rates in the field (using a netbook) Use of high power sensors –Vaisala CO 2 sensors –Driven by a car battery 5 2/22/2016 Jayant Gupchup Power ~ 4W

6. Deployment Details - I Location : Quito, Ecuador Goal : Understand tropical soil respiration Duration : 16 days 6 2/22/2016 Jayant Gupchup

7. Deployment Details -II 30s Sampling Interval Data retrieved over the air using a netbook 20 Sampling locations –12 Soil CO 2 –8 Soil Temperature & Moisture Each CO 2 Location –3 depths (12 locations, 3 depths = 36 sensors) –Vaisala GMT 220 Series –CO 2 Powered by 12 V / 45 Ah Car Battery 7 2/22/2016 Jayant Gupchup

8. A CO 2 Set-Up 8 2/22/2016 Jayant Gupchup 3 m

9. Under The Hood 9 2/22/2016 Jayant Gupchup Antenna TelosB Mote CO 2 Sensors Mote Battery CO 2 Sensor Power Connector

10. Power Consumption Lead acid car batteries 12V / 45Ah Each battery serviced 9 CO 2 sensors Total current draw :1A Lasted 36 hours after recharge 10 2/22/2016 Jayant Gupchup

11. Power Cycling Motivation 11 2/22/2016 Jayant Gupchup Batteries needed replacement/recharged every other day Carried every other day for ~ 3 Km 12V/40Ah car battery weighs 14Kg Power Cycling! –Warm up time: 15 min

12. Some Data 12 2/22/2016 Jayant Gupchup Power Loss Sensors lack range

13. Ad Hoc Replacements At 11 locations, sensors lacked range to sense the phenomenon Distribution of CO 2 hardware –10000 ppm : 26 –20000 ppm : 12 –30000 ppm : 3 –100000 ppm : 3 Researchers placed sensors initially –Reconfigured sensors if range was not good enough Final data calibration requires accurate metadata: – sensor type and date of reconfiguration Motivates need for self-Identifying sensors (Dallas 1-wire protocol) 13 2/22/2016 Jayant Gupchup

14. High Data Rates Decisions in field are driven by the data Researchers used netbook to download and view data 10 days of 30s sampling: ~ 576000 rows –Spreadsheet and word processing software unable to handle this volume –Researchers stopped looking Downsample: Provide low-resolution “view” in the field 14 2/22/2016 Jayant Gupchup

15. System Performance 15 2/22/2016 Jayant Gupchup Automated parts from existing system worked well.

16. Conclusions WSN technology is mature enough to be driven by scientists Challenges / Lessons Learned –Power cycling for high power sensors –Self-Identifying sensor design –Researchers require low-resolution data in field 16 2/22/2016 Jayant Gupchup

17. Credit NSF- MIRTHE, NSF- IDBR Microsoft Research Betty and Gordon Moore Foundation 17 2/22/2016 Jayant Gupchup

18. Questions 18 2/22/2016 Jayant Gupchup