Amber Cowans (she/her)

An Atlantic Puffin stands on a narrow, rocky cliff ledge streaked with lichen and tufts of grass. Its black back and white chest contrast with the pale stone, and its bright orange bill catches the light. The blurred backdrop of steep coastal rock faces hints at a rugged seabird colony, one of the special coastal habitats that provide vital nesting sites for wildlife.

Which is why the habitats they fight to return to must be protected.

Weakening protections for key sites for nature would push our already struggling wildlife closer to collapse.

Yet the UK Government is considering doing just that.

Tell your MP to stop this: action.rspb.org.uk/page/186397/... ‼️

The first issue of 2026 is now out! 🚨

This issue contains our new special feature 'Conservation, ecology and artificial intelligence: Advances and symbiotic solutions'. 🌍 🧪

Read the whole issue here 👇

buff.ly/AmWPK0C

Improving the integration of artificial intelligence into existing ecological inference workflows Artificial intelligence (AI) has revolutionised the process of identifying species and individuals in audio recordings and camera trap images. However, despite developments in sensor technology, m...

Great paper on AI and statistical ecology - well worth a read. Improving the integration of artificial intelligence into existing ecological inference workflows 🌎🌐 besjournals.onlinelibrary.wiley.com/doi/full/10....

🌍🌱🧪🔬🦠🧫📈🧬🖥
Scary how many #MicrobiomeSky studies ignore this
"Ecological data are inherently noisy and sparse [...] As such, it is not unexpected that robust inference of complex ecological processes like species interactions necessitates large sample sizes."
doi.org/10.1002/ecy....

In the simplest scenario, there is high bias in the interaction parameter (used for co-occurrence inference) with less than 100 sites at high detection, and 400–1000 sites at low detection, depending on interaction strength. Strong co-occurrence is detected consistently above 200 sites with high detection probabilities, but weak co-occurrence is never consistently detected even with 2980 sites. We demonstrate that the mean predictive ability of the co-occurrence model is less affected by sample size

Sample size considerations for species co-occurrence models #ecopubs @esajournals.bsky.social

'While occupancy patterns are often robust to limited sample size, reliable inference about co-occurrence demands substantially larger datasets than many studies currently achieve'

doi.org/10.1002/ecy....

Hi Alex, I'm working with Strava metro to look at recreational effects on wildlife in the Cairngorms! Happy to chat more

A new #OpenAccess Statistical Report in "Ecology"!👀

🚨 So "How many sites do I need"? The answer = it depends on inference objective! We provide recommended minimum sample sizes for different contexts based on our simulation outputs here ⬇️ P.s. all our code is freely available on OSF, so have a play around!

Take home 4: In all scenarios, far fewer sites were needed to estimate the conditional and marginal occupancy probabilities (i.e. prediction), compared to the exact interaction term (i.e. inference)

Take home 3: Adding covariates to the model also upped the sample sizes needed to estimate interaction terms without bias 💻🌲

Take home 2: Adding more species to the models reduced the power to detect interactions of a similar strength. For example, a null model with 5 species needed 400 sites compared to 250 sites with 3 species

Take home 1: Under the simplest model structure, we needed around 200 sites to detect and estimate strong interactions without bias. These requirements increased when (1) detection probability was lower and (2) species interactions were weaker

We did an extensive simulation study to show how well the Rota et al multispecies occupancy model estimates interaction terms under different detection probabilities, interaction strengths, interaction directions (positive/negative) and model structures, from 20 to ~3000 sites

Sample size considerations for species co‐occurrence models Multispecies occupancy models are widely applied to infer interactions in the occurrence of different species, but convergence and estimation issues under realistic sample sizes are common. We conduc...

📝 Are you using multispecies occupancy models to investigate interactions in species occupancy (i.e. co-occurrence)? 🦁🦓

Check out our new paper for advice on the number of sites you need to reliably detect interactions under different scenarios ⬇️

Improving the integration of artificial intelligence into existing ecological inference workflows Artificial intelligence (AI) has revolutionised the process of identifying species and individuals in audio recordings and camera trap images. However, despite developments in sensor technology, m...

New paper alert ⚠️ Using #AI tools like #megadetector and #birdNET to process camera trap images or audio recordings?

Read our perspective piece for some considerations and guidance on
📊 working with 0-1 confidence scores
🤔 making thresholding decisions
🧑‍💻 and navigating AI-labelling errors

I'll be talking at the @pintofscience.uk festival in St Andrews on May 19! Come along to learn about how AI works, how it's revolutionising ecology and what it's costing the earth 🌍 pintofscience.co.uk/event/ale-go...

Now accepting applicants for 25-26 intake of our #StatisticalEcology MSc: bit.ly/3ooHNyc. A unique opportunity to develop skills at the interface of #statistics and #ecology (some partial scholarships available too). Please help me share!

The Los Angeles wildfires are climate disasters compounded Conditions for a January firestorm in Los Angeles like this have not existed before now, writes a meteorologist and climate journalist

The Los Angeles wildfires are climate disasters compounded, writes @ericholthaus.com

- the conditions for a January LA firestorm have never existed until now

#climatecrisis #LAfires
www.theguardian.com/world/2025/j...

We just posted our first round of fully funded interdisciplinary challege-led projects!

netgain.wp.st-andrews.ac.uk/phd-projects/

PhDs available at: @aberdeenunilib.bsky.social, @durhamuniversity.bsky.social, @glasgowunisrc.bsky.social, @uniofstandrews.bsky.social, @ukri.org

Bridging Philosophy and Ecology: Ethical Decision-Making in Wild Salmonid Conservation at University of St Andrews on FindAPhD.com PhD Project - Bridging Philosophy and Ecology: Ethical Decision-Making in Wild Salmonid Conservation at University of St Andrews, listed on FindAPhD.com

I am advertising a *fully funded* and genuinely interdisciplinary PhD looking at integrating ecology, statistics and environmental ethics. The project is part of the amazing @netgain.bsky.social. Please get in touch and share with anyone who might be interested! www.findaphd.com/phds/project...

Chris Sutherland describes simulations to assess the power of species co-occurrence models under a variety of conditions.

The value of simulation before analysis: @chrissuthy.bsky.social talks through work led by @ambercowans.bsky.social, yielding some cautionary lessons for users of species co-occurrence models.
www.biorxiv.org/content/10.1...
#BES2024

Wow! Amazing opportunity here for South Africa based 🇿🇦 researchers to develop their quantitative skills! 🤩💻

Two workshops:

• An intro to stats in R
• Applied hierarchical modelling

The course is FREE! These opportunities rarely come about in 🇿🇦, so share widely & sign up!

#conservationscience 🌍

a group of people standing next to each other with the words right perspective on the bottom right ALT: a group of people standing next to each other with the words right perspective on the bottom right

Huge congrats to @ambercowans.bsky.social on the acceptance of her first PhD chapter in @methodsinecoevol.bsky.social:

'Improving the integration of AI into existing ecological inference workflows'

It's a perspective piece, but if you cant wait, the preprint is here:

biorxiv.org/content/10.1...