Costanza Zanghi
banner
coszan.bsky.social
Costanza Zanghi
@coszan.bsky.social
Interested in all things fish and behavioural ecology. Love nature, gelato & my dogs. She/them
📮I've been matched with a Pre-Scientist pen pal!🧪✍️✨
This year, I’ll be exchanging handwritten letters with an 8th grader in the US, sharing what it’s like to work in science & answering their Qs about pursuing a career in STEM.
💌Check out LPS website to donate or volunteer:
prescientist.org
September 23, 2025 at 11:59 AM
Huge thanks to
@mollyaclark.bsky.social & Ella Waples for their help in those early fish-lab pandemic days,
‪@iplanass.bsky.social for python support‬‬,
Jo Attwell who introduced me to the Optical Flow() in Matlab, saving this project when it looked doomed!
And ofc ~800 🐟 who made it all possible!
July 14, 2025 at 12:01 PM
Using this new tool gave us objective movement data we hope to, one day, align with real predators vision models 👀
8/9
July 14, 2025 at 12:01 PM
It was great fun to apply optical flow analysis to this experiment. An interesting challenge for me to try this new approach, after years of tracking fish movement from above or manually scoring behaviours.
see here👇 royalsocietypublishing.org/doi/full/10.... & www.nature.com/articles/s42...
7/9
July 14, 2025 at 12:01 PM
There was no ⚡ between 🌡️ & 🌫️: temp dominated the magnitude response, while turb dominated the duration/No. of events. But together? A cumulative antagonistic effect emerges:
Turbidity makes guppies harder to find…
…but if spotted in warm water, they’re more conspicuous = acute predation risk!
5/9
July 14, 2025 at 12:01 PM
🌡️Higher temperature = higher mean magnitude = more conspicuous guppies.
They swam faster & were more active, so easier to spot when in view!
🌫️Turbidity reduced the frequency of detection, and the duration of the “interaction” (ie time 🐟 were visible).
4/9
July 14, 2025 at 12:01 PM
We used optical flow analysis ↗️ to quantify how visible guppies were from one video frame to the next.
Our behav metrics:
🌈conspicuousness =mean magnitude of pixel displacement
📸how many times guppies were within view =detection events #
⏲️For how long they were on camera =encounter duration
3/9
July 14, 2025 at 12:01 PM
With a fully factorial design, we filmed guppies from the perspective of a hypothetical 🐠ambush predator (a 📷)
The water in our treatments:
💧clear & ambient (control),
🌡️clear & warmer (warm),
🌫️turbid & ambient (turbid),
⚡turbid & warmer (interaction).

2/9
July 14, 2025 at 12:01 PM