Guess who’s back? Dr. Seamus Anderson! Seamus completed his PhD at Curtin in 2023 and afterwards moved to the U.S. in January to work at NASA Goddard Space Flight Centre. He managed to escape back to sunny Perth and join us for a meteorite hunt. Destination? The Nullarbor Plain! Objective? Test out some new software and hopefully find a meteorite. Let’s go!
If this is your first GFO blog, welcome! Read about some previous meteorite searching trips if you’d like, such as the first ever meteorite found with with a drone, or the inaugral meteorite hunt using remote image sorting.
Day 1
After a lovely first day of driving, our first casualties occurred at 5:01pm between Norseman and Balladonia. A unit of 36 beers fell from the great height of 1.25 meters when their safety net was cast open (we raised the ute’s tray cover). The box suffered 2 fatalities. A gruesome site was beheld as their remnants drained into the dirt. Be assured their corpses were disposed of responsibly before we left the area.
Day 2
We drove through some rain while passing through Madura and Mundrabilla. “A clean solar panel on the roof!”, we thought. After a stop at lunch for our final kitchen-cooked, cheese-filled meal in Eucla, we turned off the sealed road and headed north. We soon realised that the clean solar panel was short lived. The recent rain also filled our 4wd track with puddles, and Rhonda gathered a healthy coating of mud.
Day 3
Seamus began the low-resolution overview drone survey, and we tested the image upload via Starlink. Another new feature we were using for the first time was an RTK (Real-Time Kinematic positioning) unit. This unit increases the location accuracy of the images in the drone survey. Aside from the expected, minor teething issues, the surveying and image upload went very smoothly. Seamus switched from the overview survey to the main survey around mid-morning, even though the overview survey wasn‘t completed yet. This was to prioritise uploading the images of the main survey which needed to be fed to a machine learning model. After the model receives the survey images, it can be trained to detect small, round objects in the images. All of the images from the main survey that our model thought had a black rock in them were then sent to the team back at Curtin to begin stage 1 candidate selection. At camp, we busied ourselves with drone flights, battery swapping, image uploads, and setting up camp. We finished up the day sitting around doing stage 1 selection once the sun was too low for surveying. Long shadows make for over eager rock detection! We were visited by a beautiful Wedge-tailed eagle flying around our camp in the evening.
Day 4 - All hands-on deck!
Stage 1 selection was well under way. Between the team at Curtin and us three in the field, we sorted through 26,610 images by the afternoon of day 4. The sorting happened simultaneously with the survey image upload, so it was started and completed as early as possible.
In the afternoon, Seamus and Martin tested out some stage 3 candidate follow up with the drone. It was becoming apparent that our fall site had a LOT of black rocks, so speedy drone visits to eliminate as many as possible was an appealing idea as the sun climbed higher in the sky.
Meanwhile, Sophie tested the wifi extender and did some stage 4 follow-up for candidates close to camp. Ideally, we would be able to visit stage-4 candidates on the ground within an internet signal. This would allow us to load the drone images on a tablet and record our answer ‘meteorite’ or ‘not a meteorite’ straight away to the survey database. Headphones on and juice box in hand, it was an enjoyable afternoon! Interjected with moments of frustration at walking in circles for particularly tricky to locate candidates…
Day 5
We began the day finishing stage 2 candidate selection. Our body clocks were syncing with the sun, which made for great sunrise views at 4:30am AWST! Unfortunately, this shifted our team out of sync with people back in Perth, so when we finished breakfast at 6am and wanted to start the day, we didn’t have Hadrien ready to debug our upload when it stopped, or all the stage 2 candidates finished. This gave us a quiet morning around camp in cool weather conditions to complete stage 2 candidate selection.
Now that the entire field area had been parsed through the detection model and sorted in stages 1 and 2 of candidate selection, we had our final set of images which were the most likely to contain a meteorite! There was one, fairly large problem though… There were ~800 candidates! It didn’t look like we were going to be strolling out into the field and immediately picking up this meteorite… Pursuing our goal of finding a meteorite, we spent the afternoon preparing our tablets with data and software to aid us in the stage 4 candidate follow up. We spent a little bit of time visiting stage 4 candidates close to camp. In the evening we made the decision to spend one full day tomorrow visiting as many candidates as possible.
Day 6 - Treasure Hunt
From sunup to sundown, we searched the field. Dividing up the search area, plotting our search routes, and following targets on a map, it felt a bit like a treasure hunt! By the end of the day, we had visited 245 candidates! Unfortunately, we did not find the recently fallen meteorite. Martin scored an older meteorites, and Seamus found both an old meteorite and a very large tektite!
Day 7 & 8
We had a very drama-free pack up and drive home. We appreciated our first showers when we stopped in Ballandonia, and welcomed the rain washing Rhonda as we through some rain. The joy was short lived, as the rain brought out the insects! And once again, Rhonda was once again in need of a bath.
