Juan Nathaniel

phd @Columbia | prev @IBMResearch, @NUSingapore

[Juan Nathaniel @juannat7]

We present our #CVPR2024 work on #EarthVision: Deep Generative Data Assimilation in Multimodal Setting that calibrates Earth system model state🌎 with diverse observations🛰️📡 using diffusion

Paper: https://t.co/whx4xScmLn
Code: https://t.co/KCyF77ZO6S

🧵1/6

[Pierre Gentine @PierreGentine]

Excited to share our new paper using transfer learning to better predict pCO2 in data sparse conditions (in the real world), by Shaun Kim, @juannat7, Zhewen Hou and @tz33cu https://t.co/NIOymrdaez

[Kara Lamb @karadlamb]

Glad to see our paper on using data-driven reduced order modeling to develop cloud microphysics schemes published in JAMES this week! https://t.co/77DGjN6COB
@LeapStc @CUSEAS @columbiaclimate

[Juan Nathaniel @juannat7]

🏆🥇 Happy to share that our work on SLAMS won the Best Student Paper Award at @EarthVisionWS #CVPR2024. We will present it in ✈️Seattle next week.

Also, happy to have ☕️chats with folks interested in #AI4Science broadly!

[Juan Nathaniel @juannat7]

We present our #CVPR2024 work on #EarthVision: Deep Generative Data Assimilation in Multimodal Setting that calibrates Earth system model state🌎 with diverse observations🛰️📡 using diffusion

Paper: https://t.co/whx4xScmLn
Code: https://t.co/KCyF77ZO6S

🧵1/6

[Juan Nathaniel @juannat7]

@tungnd_13 @PierreGentine @YongquanQu2340 @LeapStc Thank you Tung!!

[Pierre Gentine @PierreGentine]

1/2 We hear a lot about "Earth twins" these days. There are still major challenges on the way before we can get to Earth's twins whether using high-resolution Earth system models or AI-based models but progress is happening fast, especially on the AI front. Some comparisons:

[Juan Nathaniel @juannat7]

🎇Credits to amazing co-authors, @YongquanQu2340 (co-first), Shuolin Li, @PierreGentine for the opportunity to work on this fun project! @LeapStc

6/6

[Juan Nathaniel @juannat7]

We present our #CVPR2024 work on #EarthVision: Deep Generative Data Assimilation in Multimodal Setting that calibrates Earth system model state🌎 with diverse observations🛰️📡 using diffusion

Paper: https://t.co/whx4xScmLn
Code: https://t.co/KCyF77ZO6S

🧵1/6

[Juan Nathaniel @juannat7]

🎟️Bonus: we perform feature importance study to identify the relative contribution of each observation modality. For example, assimilating ex-situ OLR🌞, rather than in-situ precipitation🌧️, better constrain top-of-the-atmosphere temperature 🌡️

5/6

[Juan Nathaniel @juannat7]

Excited to share our new work on applying meta-learning to upscale global carbon fluxes! https://t.co/iWv2UTMO8F

#carbon #metalearning #machinelearning @ScientificData

[Pierre Gentine @PierreGentine]

Excited to share our new global terrestrial carbon flux product based on metalearning led by Juan Nathaniel, a great student in our group: https://t.co/MXfNLgu5KV, called MetaFlux. This product has improved capacity in data poor regions such as the tropics and dry regions.