This document describes a wireless sensor network for monitoring environmental carbon dioxide (CO2) levels using laser-based sensor nodes. The network aims to provide high spatial resolution real-time CO2 concentration data with unprecedented sensitivity and selectivity. Laboratory and field tests of a two-node and three-node network demonstrated the long-term stability and cross-correlation performance of the sensor nodes, which achieved temperature drift correction. Future work involves deploying a larger sensor network for long-term trace gas monitoring in environmental field studies.

1. Wireless Sensor Network for
Environmental CO2 Monitoring
Clinton J. Smith1*, Wen Wang1, Stephen So1,2, and Gerard Wysocki1**
1) Dept. of Electrical Engineering, Princeton University, Princeton, NJ 08544
2) Sentinel Photonics Inc., Monmouth Junction, NJ 08852, USA
* clintons@princeton.edu; ** gwysocki@princeton.edu
Motivation Sensor Node Long-Term Performance Two-Node Sensor Network Monitoring
 The CO2 impact on the greenhouse gas effect requires global and In-Lab Activity
local monitoring capability which would greatly benefit from
availability of sensors that are lightweight, portable, robust, highly
0.5m Base
sensitive, and selective. Station 3.3m
 For study of the Carbon Cycle, these sensors should also be low-
FL Sensor
power/battery operated and capable of being wirelessly networked
and autonomous. GL Sensor
 Large area wireless networks of laser-based trace-gas sensors will
provide high spatial resolution of real time concentration data with
unprecedented sensitivity and selectivity to the target molecular
species.
 These sensors are expected to maintain a high degree of long-term
stability in the field, despite changing environmental conditions.
Background CO2 concentration time series show the degree to which temperature
 We have built a laser spectroscopic sensor for CO2 detection and correction improves sensor-node performance. Allan deviation
demonstrated its performance in preliminary laboratory and field calculation of long term concentration measurements allows quantifying  Laboratory activity can be interpreted from this data
tests [1]. the sensor long-term stability. Adaptive linear regression to remove set.
sensor temperature dependence enables potentially indefinite assurance  The high concentration events near 0 hours as shown by the
 The sensor achieves higher long-term sensitivity by locking to
of short term (<10 sec.) sensitivity. FL Sensor are caused by the operator working at the base
the targeted absorption line. station to configure the WSN.
 We identified environmental temperature as a main source of drift Multi-Node Long-Term Cross-Correlation Performance  The baseline reflects the overall activity in the room while
and applied a linear regression technique to correct for it. the high short-time concentration spikes refer to
 To demonstrate wireless sensor network (WSN) capability, both a individuals working in the vicinity of the sensor.
two-node and a three-node network similar to [2] for long-term  Both at the beginning (hour 0) and at the end (hour 8) the
real-time monitoring of CO2 have been investigated. CO2 concentration exhibits a low baseline level
Base corresponding to low human activity in the lab.
 The multi-node sensor cross-correlation was calculated. Station
Wireless Sensor Network & Deployment Summary and Future Directions
Node 1 Node 2 Node 3
E-QUAD at Princeton University  We have demonstrated a two and a three sensor-node
350 m range directional
antennas are used network for long-term real-time CO2 concentration
monitoring.
 Sensor node cross-correlation is comparable to that of
commercial products
 A temperature correction technique was demonstrated
Test Sight: Crop field in Princeton, NJ
to remove the temperature impact on sensor drift.
Future directions:
Three calibrated sensor nodes are placed in one box and sample  A large area WSN is being deployed in the field for long-
Three locations selected to monitor term trace-gas monitoring.
approximately the same air sample. Continuous CO2 measurement was
coupled local environments:
performed over a period of 10 hours. The cross-node correlation R2 values
1. Adjacent to the local road: car traffic
CO2 sensor-node. The
2. In the inner courtyard: local between the three nodes while line-locking (120 min. – 550 min.) are: References:
total size is less than [1] C. J. Smith, S. So, and G. Wysocki, "Low-Power Portable Laser Spectroscopic Sensor
vegetation
that of a shoebox.  Node1, Node2 = 0.87 Standard deviation among the for Atmospheric CO2Monitoring," in Conference on Laser Electro-Optics:
3. On the roof of the building Applications, OSA Technical Digest (CD) (Optical Society of America, 2010), paper
Acknowledgements: This material is based upon work supported by the National Science Foundation
 Node1, Node3 = 0.94 nodes ranges from 2 to 6 ppmv JThB4.
under Grant No. EEC-0540832, an NSF MRI award #0723190 for the openPHOTONS systems and
National Science Foundation Grant No. 0903661 “Nanotechnology for Clean Energy IGERT.”  Node2, Node3 = 0.93 [2] S. So, A. A. Sani, Z. Lin, F. Tittel, and G. Wysocki, "Demo abstract: Laser-based trace-
gas chemical sensors for distributed wireless sensor networks," in Information Processing