The Trials of Drone Field Work

Last post I introduced our cross-country road trip and some of the awesome field work we are doing. We spent yesterday flying a site in Northern New Brunswick. We learned this morning that these sites cross New Brunswick. I find it exciting to contribute to such a large research project, but am still learning site details. In lieu of science, this post describes a typical day of drone field work. Throughout this post I wil point out places were things can AND will (or rather did) go wrong. Let’s dive in!

A day of Drone Field Work:

The Night Before

The night before fieldwork starts with an equipment list. On the scantiques road show we are driving a Ford Ranger which is stacked to the top with cases for our sensitive instruments and safety equipment. We even have a case with our personal gear strapped on top! So what is packed into this truck?

  • DJI M300 drone and controller
  • 1-2 battery cases storing batteries, controller, and GPS base-station
  • 1-2 DJI mini-drones
  • A large-long case storing the GPS base-station
  • A Trimble GPS base-staton (yes we need two)
  • Sensors including:
    • DJI Zenmuse L1 (Lidar)
    • DJI P1 (a very fancy camera)
    • A multiband camera (this is in-lie of the hyperspectral camera I described yesterday)
  • field laptop
  • Starlink internet
  • generator (to re-charge drones throughout the day)
  • gas can (to fill the generator)
  • rangefinder
  • hardhats
  • cruising safety vests
  • Radios (and chargers)
  • Cooler
  • Pop-up tent for rain and sun
  • Camping chairs for the crew on long field days
  • All of our personal gear for 3 weeks away

To protect this mountain of gear - we bring all of it into our hotel rooms. In the hotel room, we dedicate all outlets for drone gear charging creating a constant whirr of charging electronics. Before bed, the pilot in the command plans the next day’s flight on drones remote controller by down loading a shapefile for the site and looking at satellite imagery.

The Morning Of

Many of the field sites we visit are 1-2 hours from the forest sites where we are working. We often wake up early, visit the local Tim Hortons, and head out to the field. Here is where we ran into issue (1) of the day - the need for an oil change. On Sunday, when we turned the truck key, we got a flashing light for OIL CHANGE REQUIRED, but Monday was a federal holiday. Open garages were hard to find. Canadian Tire seemed like it might be open, so we booked an appointment for 9 am yesterday. While it first seemed as though the Canadian Tire was also closed, they eventually agreed to change the oil. It took two hours, and my field team can provide you with a full review of Canadian Tire patio furniture - but we made it on the road before the afternoon.

At the site

We made it to our first of two sites just before noon. My field crew operates much like a race car pit-crew. The race car is the drone. To run, a drone needs a pilot with a flight plan (me - Sarah), connection to satellites and the base-station, a fresh set of batteries, and the sensor installed. Generally, the pilot runs all things data heavy. I get to the site and verify the site plan. The crew sets up the GPS base-station, takes the drone out of the case and installs new batteries. Last, I install the sensor, turn on the drone, and pray to the heavens. No, literally I look to the sky in hope and anguish. Any number of errors can come up at this point, the most frequent for closed forests of Canada is poor satellite connection. The drone will not fly with less than 10 satellites and an active connection with the GPS base station getting signal from those same satellites. Luckily at the first site the odds were in our favor 30+ satellites quickly connected. We were ready to fly!

Mid-flight

Once the drone takes off I monitor it as it goes to its flight start point. During the flight the drone executes the flight plan with little direction from the pilot. My role as the pilot is to make sure the drone has enough battery power, is not in danger from obstacles or birds (it happens), and is collecting the correct data. While in flight, I keep line of sight and connection by climbing on top of the field truck or running down the road. On the controller, I check to make sure that the drone is not close to trees. Every 20-30 minutes the drone has to land for a battery change. Here, the pit-crew returns to rapidly replace the battery so we can get the drone back up in the air. They take care of managing the number of batteries we have left, when we run low on batteries the pit crew will start the generator to make sure that we don’t run out.

End of flight

After the drone finishes the flight path it automatically returns home. I collect the data from the sensor and we take down the small drone-focused camp that surrounds the truck. We pile it all into the truck and continue onto the next site. At the end of the day, I back up the data to a computer and a harddrive and make sure that all of the data is there.

Summmary

This is what happens when all goes as planned. This is exactly what happened for the first site - not the case for the second site. Here, what should have been a 1.5-hour flight turned into a 5-hour adventure. The first thing to go wrong was access to the site. Anyone who has walked a trail in the early spring knows just how many trees can fall in the winter. While humans have the ability to clamber over trees and branches, clearing the path for a truck requires saws, tools, and a couple hours. When trying to get to the second site we quickly found that no-one had been there in a while - countless trees blocked what we thought would be perfect site access.

Luckily, there was a road on the other side of a small stream from the site. We thought we could fly from there. We drove over, set up, and turned on the drone. This time, the satellites were not in our favor. There was not enough signal for the drone to fly. We were deep in a forested river valley, so when the drone system told us to fix the error by ‘moving to a more open area or power-cycling the drone,’ our only option was to power cycle and wait. Which we did. Once. Twice. Three times. A bathroom break and a soda-water later, just one more time. And again. After an hour had passed, we had to try something new. I walked down the road to try and find another spot - higher ground, fewer trees, anything. I found a site that was about the same quality as where we were and radioed the team to pack up the truck - we were moving. We moved and set up for another attempt. We set up and turned on the drone again - no luck. Another power cycle. Some contemplation on what to do when I look down and see - we had satellite! Moving quickly, I pulled up the flight plan, did my pre-flight check, and launched the drone. We were off!

We were flying! For a while… The first problem was the site was long. To keep line-of-sight on the drone, I had to fast-walk alongside the flight route, which paralleled the road. I bid adieu to the pit-crew in pursuit of the drone. All was well until the first battery switch. After landing and hot-swapping the batteries, the drone flew to the pause-point in the flight plan and immediately lost satellite connection. To maintain accurate data collection, the drone and the GPS station need constant satellite connection - the two work in tandem to collect data at sub-centimeter level precision and accuracy. After trying to reconnect, I was forced to land and see if it might connect again. After 15 minutes of ground time, the drone remarkably connected again. Take off part 2! We successfully flew another 50% of the 1.5-hour flight before we lost satellite connection. RATS! Another landing, more downtime, and another remarkable reconnection! Take off part 3! At this point, the clock read 4:45, and thunderclouds loomed in the distance. Time was of the essence. The flight neared completion (97%) - almost there! And lo, another lost satellite connection. Disaster! But with thunder threatening, I finished the flight, praying that the redundancy I built in the flight plans (overlap and buffer areas) will salvage the overall accuracy of the lidar final product.

So there we have it - one day, one perfect drone flight execution, and a second drone flight that felt like it would never end. Tune in tomorrow for an update on our last field site in New Brunswick - a drought exclusion collaboration with University of New Brunswick and JD Irving!