UC Davis

For a Chinook salmon restoration project in the lower Yuba River in California, we designed and deployed a wireless sensor network to monitor river temperatures at micro-habitat scales. The study required that temperatures be measured along a 3 km study reach, across the channel, and into off-channel areas. To capture diel and seasonal fluctuations, sensors were sampled quarter-hourly for the full duration of the six-month juvenile salmon winter residency. This sampling duration required that nodes minimize power-use. We adopted event-based software on MSP430 micro-controllers with 433 MHz radio and minimized the networking duty-cycle. To address link failures, we included network storage. As the network lacked real-time clocks, data were timestamped at the destination. This, coupled with the storage, yielded timestamp inaccuracies, which we re-aligned using a novel algorithm. We collected over six months of temperature data from 35 sensors across seven nodes. Of the packets collected, we identified 21% as being incorrectly timestamped and were able to re-align 41% of these incorrectly timestamped packets. We collected temperature data through major floods, and the network uploaded data until the flood destroyed the sensors. The network met an important need in ecological sampling with ultra-low power (multi-year battery life) and low-throughput.