Operations Update 2023


Kath Quayle


March 13, 2023

Reading time: 2 minutes

The Galiwatch project was initiated in February 2021 to monitor pollinators and weather patterns at a mid-Island site on Galiano Island.

The initial goal was to monitor pollinator activity on Oregon grape (Mahonia nervosa) and salal (Gaultheria shallon), two native flowering plants that grow abundantly in the forest understory.

Our approach was to collect timelapse pictures and video, and to collect weather and air quality data using raspberry-pi technology and an Accurite weather station system, as described on the website.

A neural network that identified bees in our images was investigated to analyse the data.

Year 1: 2021

Data collection

  • Oregon grape - we successfully collected pictures at 1 min intervals throughout the Oregon grape blooming period (mid April to early May), as well as video footage (4 of 5 mins) during the peak blooming period.

  • Salal - data collection for salal was less successful due to a camera failure.

  • Weather - data collection with enviro-pi and Accurite weather station was successful.

Lessons learned

The volume of data generated in monitoring two different plant systems is unmanageable at this point.

  • Decision taken to focus on Oregon grape as the single focus for long-term monitoring of pollinator activity at the study site.

Video footage showed most bees spend between 10s and 35s on a flower before moving on.1 min time-lapse photographs captured approximately 20-25% of bees seen in the videos. One minute time-lapse photography is insufficient to capture changes in pollinator numbers due to the low number of positive hits. Videoing continually on the other hand is unmanageable in terms of data handling capacity.

  • Decision taken to take 15s time-lapse pictures for 2022 season with two cameras.

Occasional gaps in weather and air quality data collection with the enviro-pi occurred during weather-related internet drops.

  • Decision taken to continue as is

Year 2: 2022

Data collection

  • We successfully collected data at 15 sec intervals throughout the Oregon grape blooming period with two cameras in operation, each focused on a cluster of flower spikes.

  • The enviro-pi weather data collection system was not reliable and started failing in late spring with only patchy PM2.5 data collection attainable. This was replaced by a Purple Air monitor in August 2022. We filled the data gap by accessing data from a neighbouring system on Salt Spring Island approximately 3 km away.

Lessons learned

15s time-lapse photography produces a manageable data set, but we need some video coverage to determine proportion of bee visits captured by 15s time-lapse pics and inform estimates of total visits.

  • Decision taken to proceed with 15s time-lapse (two cameras) and video 4 of 5 mins for 5 days mid-way through blooming in 2023.

The weather station and Purple Air system for monitoring air quality are very reliable.

  • Weather data collection to continue as is.

Training of a neural network for identifying bees is challenging!

  • A new strategy for training the neural work to identify bees in time-lapse pics is underway