The ‘solar canals’ making smart use of India’s space Solar energy is clean, but it usually takes up huge tracts of land. In India, an alternative is turning the country’s canals into glittering trails of solar panels.

Great to see California catching up to India!

www.bbc.com/future/artic...

Where can I learn more about the PDF crawl problem? It feels relevant to a working group I’m leading on applied AI for evidence synthesis.

H/t to my collaborators at Conservation Science Partners & @diogoverissimo.bsky.social, and thanks to the organizations who funded this work (On The Edge Conservation).

Map of spp discourse

Ultimately, we hope that our pipeline can make it easier for conservation advocates and researchers to track how the public is relating to our fellow members of kingdom #Animalia. Link to the open-access paper: doi.org/10.1111/cobi... 🧵 5/5

Chord diagram of different topics discussed regarding different species

We rapidly classified how people discuss wildlife w/COVID outbreak as a case study. We used open-source transformer language models (#BART) along with more traditional #NLP approaches. We saw marked differences across spp, highlighting the benefits versus risks of notoriety for wildlife. 🦠🧵4/5

Our common name solution method

Using common names for wildlife required resolving which search terms actually relate to distinct families/genera (e.g. lion is a substring of sea lion & mountain lion). We used the longest common substring problem, and show how you can implement this with iterative human expert feedback. 🧵3/5

Sankey diagram of the number of mentions for different species that are relevant or not, and processed at different stages of our pipeline

One hurdle was discerning species from sports teams, companies, or celebrities. Also, the public rarely uses formal Latin binomial species names. So, we aimed to filter out irrelevant entries while broadening our search to include the real ways that the public talks about #wildlife. 🧵2/5

Machine learning pipeline for extracting data on public perceptions of biodiversity from different data sources

New(ish) research alert: How does the public actually talk about wildlife online? Turns out, ‘Tiger’ is ~everywhere~…but people aren’t usually talking about that big Panthera cat 🐅. We created an open-source #ML and transformer #LLM pipeline to track #publicperceptions of #biodiversity. 🧵1/5

Is the public actually talking about biodiversity? 🐾

Our new #OA paper in @ConBiology introduces an automated NLP pipeline to measure wildlife saliency and attitudes from digital media

Essential for tracking GBF targets! 📈

conbio.onlinelibrary.wiley.com/doi/epdf/10....

#ConservationScience #AI

Looking for collaborators, testers, and skeptics to help shape this work and our products. H/t to my collaborators, Adam Pearson, Jingyi Li, DataKind, Caitlin Augustin, and others -- really excited to see where this work takes us all!

We're also training the next generation and running practitioner workshops through our @cu-esiil.bsky.social working group. If you work in conservation evidence synthesis, environmental policy, or just have strong opinions about AI and NLP in ecology, we’re keen to hear from you.

We aim to have every output trace back to source documents so users can evaluate credibility themselves. No black boxes that falsely claim they can replace human judgment.

To that end, we're tackling this problem with practitioners, ethicists, and data sovereignty experts to figure out how we can best use AI to help augment (not replace!) expert opinions.

I know that many of us in ecology, climate, & the environment are skeptical of "AI", and for good reason (water and energy footprints, plagiarism machines,...the list goes on).

We plan to build open-source AI and NLP tools that can link evidence across scientific papers, gray literature, and project reports (e.g., connecting specific agroforestry practices to carbon outcomes, or restoration approaches to species recovery).

Happy to share that we have received funding from the Seaver Institute to tackle an increasingly urgent problem in #conservation and #climateaction: how do you synthesize mountains of research when decisions need to happen rapidly?

Bluesky has genuinely revived my love for science communication online. After watching the same posts go nowhere on Twitter, seeing them resonate here has been such a good reminder of why I do this. Thanks for being here 🖤

Introduce yourself with 5 animals you’ve seen in the wild:

Giant Ibis
Dhole
White-throated babbler
Diademed sifaka
Spoon-billed sandpiper

Funny to me that our paper on the mismatch between venture capital and climate impact is getting the most engagement on LinkedIn.

Maybe because everything else on there is AI slop?

iopscience.iop.org/article/10.1...

This is a fascinating study showing how misaligned venture capital can be when it comes to climate. Instead of chasing high-impact carbon solutions, it tends to chase fads.

We need to do a better job aligning our money with a more substantial climate impact.

iopscience.iop.org/article/10.1...

Thank you @globalecoguy.bsky.social for the important work you've done to advance environmental stewardship! I routinely teach your papers on food security & climate in my classes. Makes for a great interactive coding session for the students too, using @ourworldindata.org and R/Python.

#climate #paperskygest #conservation #naturebasedsolutions #naturalclimatesolutions

Home Project Drawdown drives meaningful climate action by advancing science-based solutions and strategies.

This research builds on exceptional datasets from Project Drawdown (drawdown.org) and Vibrant Data Labs. H/t to @pomonacollege.bsky.social for funding & Paul Hawken, @globalecoguy.bsky.social, & Eric Berlow for making these data available. 🧵 9/9

📄 Read the full paper here (no paywall!): doi.org/10.1088/2753... 🧵 8/9

Public finance, strategic investors, and conservation organizations can lead where traditional VC has failed. There's substantial potential to improve climate finance outcomes, and much more to explore. 🧵 7/9

As Rohan notes: VC is optimizing for market familiarity and short-term returns, **not climate outcomes**. But this misalignment reveals an opportunity: nature-based solutions and other high-mitigation technologies are 𝘂𝗻𝗱𝗲𝗿𝘃𝗮𝗹𝘂𝗲𝗱 and 𝘂𝗻𝗱𝗲𝗿𝗰𝗮𝗽𝗶𝘁𝗮𝗹𝗶𝘇𝗲𝗱. 🧵 6/9

Swarm plot of investment by company stage (pre-seed/seed through more mature stages such as scaling).

𝗧𝗵𝗲 𝗩𝗮𝗹𝗹𝗲𝘆 𝗼𝗳 𝗗𝗲𝗮𝘁𝗵 𝗽𝗲𝗿𝘀𝗶𝘀𝘁𝘀: Only 4.6% of companies reached middle-stage funding, with median capital far below what's needed for scaling and commercialization. 🧵 5/9

𝗙𝘂𝗻𝗱𝗶𝗻𝗴 𝗶𝗴𝗻𝗼𝗿𝗲𝘀 𝗰𝗹𝗶𝗺𝗮𝘁𝗲 𝗶𝗺𝗽𝗮𝗰𝘁: No statistical relationship between investment dollars and carbon mitigation potential or technological maturity—the metrics that should drive deployment. 🧵 4/9

Table 1 from the manuscript, showing ROI results in percentages for different mitigation sectors.

𝗡𝗮𝘁𝘂𝗿𝗲-𝗯𝗮𝘀𝗲𝗱 𝘀𝗼𝗹𝘂𝘁𝗶𝗼𝗻𝘀 𝘀𝘁𝗮𝗿𝘃𝗲𝗱 𝗼𝗳 𝗰𝗮𝗽𝗶𝘁𝗮𝗹: Food, agriculture, land use, and ecosystem restoration received minimal investment despite high sequestration potential. Early ROI data suggests promise, though the sample remains small and companies are still maturing. 🧵 3/9

Bar chart of funding distribution by Project Drawdown climate solution category

𝗪𝗲'𝗿𝗲 𝗯𝗲𝘁𝘁𝗶𝗻𝗴 𝗼𝗻 𝗳𝗮𝗺𝗶𝗹𝗶𝗮𝗿 𝗺𝗮𝗿𝗸𝗲𝘁𝘀, 𝗻𝗼𝘁 𝗰𝗹𝗶𝗺𝗮𝘁𝗲 𝗼𝘂𝘁𝗰𝗼𝗺𝗲𝘀: 71% of funding went to three sectors with existing private market incumbents, with the bulk of funding directed toward EVs (over 40%), even though such solutions represent just 3.5% of mitigation potential. 🧵 2/9