🌎 AGU25 Call for Abstracts! 🌿 I’m an early career convener for our session on: Atmosphere-Hydrology-Ecology Interactions During Extreme Events. Join us to explore how Earth systems connect under climate extremes! #AGU25 #Climate #Hydrology #Ecology

🔥 AGU25 Call for Abstracts! 🌱🌍 I’m lucky to co-convene a session with friends from SNU: Deciphering Land Carbon Sinks: Integrating Data, Models, and Observations Let’s bridge top-down & bottom-up estimates and tackle uncertainty together! #AGU25 #CarbonCycle #ClimateChange #LandCarbonSinks

Wow this is cool!

Good luck to LEMONTREE's @jianingfang.bsky.social who has just arrived at Seoul National University for his 1-month secondment with @youngryel.bsky.social! Looking forward to hearing more about their work on implementing Differentiable BESS @pierregentine.bsky.social @columbiauniversity.bsky.social

Excited to share our first paper on the optimal complexity of process representation in hybrid-ML models, with soil moisture limitation as a case study! 🚀We present a flexible, differentiable framework to test new process ideas & hypotheses. Huge thanks to my advisor Pierre Gentine for his guidance!

Pierre Gentine@pierregentine.bsky.social · 2y ago

Excited to share the first paper on our new differentiable land model called DifferLand which allows learning entire functional form only by assimilating indirect data dx.doi.org/10.1029/2024.... Led by a great student: Jianing Fang. Some new science is possible with this model.

I am thrilled to share Jianing Fang's new differentiable hybrid land model called DifferLand, which can learn both parameters but also new structural relationships directly from data and can fuse local and global data arxiv.org/abs/2411.09654. Many new questions can be addressed with this model.

Differentiable Land Model Reveals Global Environmental Controls on Ecological Parameters

Accurate modeling of terrestrial carbon and water exchange requires robust ecological parameters that capture vegetation responses and adaptations to the local environment. The current generation of l...

arxiv.org