Pushpendra Rana

Natural Climate Solutions, complex systems, machine learning, causality. Cofounder: @Indian Forest Service @UIUCGeogGIS Alum.

[Pushpendra Rana @_PushpendraRana]

Carbon offset credits claim to represent precise climate benefits.

But many forest projects operate in complex social-ecological systems where causal attribution is deeply uncertain.

This tension sits at the heart of the carbon market.

My interview:
https://t.co/37T2uSaLVt

[Pushpendra Rana @_PushpendraRana]

Our analysis shows why forest carbon offsets struggle at the project level: uncertainty overwhelms effects, impacts reverse under modest bias, and multiple interacting pathways violate key causal assumptions.

This is a structural limitation, not a data or methods problem.

[Pushpendra Rana @_PushpendraRana]

1/ 🧵 Carbon offsets don’t fail because data or methods are weak.
They fail because project-level causal effects are not identifiable in complex forest systems—yet credits demand that precision.

Evidence from Brazil & India (50-day free access):
🔗 https://t.co/tO7ZEkSsQ9 👇

[Pushpendra Rana @_PushpendraRana]

Estimated average effects overlap zero, while uncertainty is large and exceeds the effects themselves.
That alone should trigger serious caution about credit issuance based on point estimates.

Sorting CATEs shows why: uncertainty explodes due to non-linear dynamics, not noise.

[Pushpendra Rana @_PushpendraRana]

Forests matter enormously for climate.

But using them to offset fossil emissions requires project-level causal precision that complex systems cannot deliver.

Finance forests.
Reduce emissions.
Treat them as complementary—not interchangeable.

[Pushpendra Rana @_PushpendraRana]

A deeper problem: non-interference fails.

Offset methods assume outcomes respond only to the project itself (SUTVA).

We find multiple causal pathways operate simultaneously—farms, grazing, fire,—invalidating project-level isolation. Therefore, project attribution collapses.

[Lav Varshney @lrvarshney]

I'm up next on the big stage at Supercomputing (SC25) in St. Louis, speaking about "High-Performance Creativity". Wish me luck!

There are lots of intriguing high-performance / distributed computing problems in creativity and scientific discovery.

[Forrest Fleischman @ForrestFleisch1]

Several colleagues and I are organizing a panel on "From global restoration goals to people's visions for the future: Capturing diverse imaginaries of ecosystem restoration" at the upcoming Pollen conference in Barcelona (June 29-July 3, 2026).
https://t.co/1AVeLcHML9

[Pushpendra Rana @_PushpendraRana]

@nancanvass Agreed. We’re seeing similar enthusiasm from many others. Here’s another example — a Mahila Mandal-led planting in Palampur, where local communities are embracing tree planting as a sacred cultural and religious act, rather than merely a natural climate solution..

[Pushpendra Rana @_PushpendraRana]

One on-going plantation by a Mahila Mandal in Kullu, Himachal under RG Van Samvardhan Yojana. 600+ Women groups and youth collectives are planting trees in 1400+ hectares this monsoon.Also using https://t.co/tOXKCAVP74 to select sites for planting..

[Forrest Fleischman @ForrestFleisch1]

Drawing on data from over 400 forest plantations in Himachal Pradesh, India, we show that improved forest outcomes forest cover, forest-based livelihoods) are strongly associated with long-term participation by local people in decision-making
https://t.co/zTY37jvymK

[Pushpendra Rana @_PushpendraRana]

@ForrestFleisch1 That’s a challenge. We did try ePlantation Site Assistant (ePSA) mobile app - https://t.co/qZWTtPN8W0 - but it did not scale as it involved cumbersome steps. We built on that experience and hope this ChatGPT like - easy to use bot will help improve restoration in the field.

[Forrest Fleischman @ForrestFleisch1]

This seems to be the really useful kind of "AI" application - providing data-based decision support to people facing real challenges. Whether it improves govt. tree planting in Himachal remains to be seen, but its a start.

[Pushpendra Rana @_PushpendraRana]

@bijay_eco Not yet.

[Pushpendra Rana @_PushpendraRana]

I’ve supervised millions of trees planted across the Himalayas.
But here’s the hard truth: most don’t survive.
Why? -> Wrong trees. Wrong places.
This monsoon, forest rangers in Himachal are using AI + Machine Learning to plant smarter.
Here’s how it works 🧵

[Pushpendra Rana @_PushpendraRana]

Any questions? Ideas? Want to collaborate?
DM the team or try the bot: https://t.co/9asmazDxmt
#AI #ClimateTech #Restoration #India #Sustainability #Reforestation #MachineLearning #Forestry

[Pushpendra Rana @_PushpendraRana]

8/ Forests need more than good intentions.
They need good data.
->WhereToPlant is bringing AI/ML to the roots of forest restoration.
Start planting smart:
https://t.co/tOXKCAWmWC