Tana
Data-Driven Forest Conservation Impact System
Created on 24th November 2025
•
Tana
Data-Driven Forest Conservation Impact System
The problem Tana solves
The Problem It Solves
Across many green initiatives—including GBM, WMF, and county-led tree-growing programs—there is a persistent gap between planting trees and ensuring they survive, thrive, and get monitored. Today, most teams face challenges such as:
1. Inaccurate or incomplete forest data
Tree-planting events are well documented, but long-term tracking of survival, growth, and ecosystem change is limited. Many organizations still rely on:
Manual spreadsheets
Scattered field reports
Infrequent site visits
No satellite-based verification
This leads to poor visibility into what is actually happening on the ground.
** 2. No integrated system for seedling → forest lifecycle**
Existing tools focus on either nursery management, or planting records, or satellite analytics.
There is no unified platform that tracks:
Nursery species & capacity
Tree distribution
Planting locations
Forest recovery over time
Survival and risk prediction
As a result, programs struggle to answer critical questions like:
“Which school or community needs more seedlings?”
“Where are trees dying?”
“Which forests will be at risk next season?”
3.** Limited ability to use AI or satellite intelligenc**e
Most organizations cannot easily access or use platforms like Earth Engine, AlphaEarth, or Sentinel data.
This creates major barriers:
No rapid deforestation alerts
No carbon or biomass estimation
No automated monitoring
No predictive insights
Decision-making becomes reactive rather than proactive.
4. Community engagement is difficult to sustain
Schools and communities lack tools to:
Report tree status
Upload images and GPS points
Receive feedback or training
Participate in digital restoration tracking
This weakens long-term survival and reduces ownership of planted trees.
Our Solution (What We Built)
We solve these challenges by creating an integrated system that combines AlphaEarth AI, community data collection, and forest analytics into a single platform that can:
Track seedlings from nursery → planting → forest
Map deforestation, regrowth, and survival clusters
Predict future forest conditions (risk, health, survival)
Support multi-user teams: schools, counties, NGOs
Visualize impact and generate automated insights
Provide AI assistance for educators, forest officers, or community champions
This creates the first AI-augmented forest intelligence platform rooted in Kenya's real environmental context.
Challenges we ran into
1. Access & limitations of AlphaEarth layers
Not all datasets are public or available in our region.
Some required switching to:
Sentinel-2
Sentinel-1
Forest canopy models
NDVI/NDWI layers
Land cover datasets
This required redesigning parts of the workflow.
2. Training clustering & prediction without a custom ML model
Creating survival clusters required:
Careful feature stacking
Cloud compute optimization
Fixing Earth Engine sampling errors
Ensuring AOI boundaries were correct
Some collections caused access errors or exceeded memory limits.
**
3. Visualization issues inside notebooks**
Map layers sometimes rendered blank due to:
Missing geometries
Incorrect CRS
Mis-ordered bands
Earth Engine throttling
We had to include fallbacks and debugging tools.
4. Multi-user data flow design
Designing for schools + foresters + admins required:
Role-based access
Shared data but private contributions
Synchronizing field data with satellite analytics
Deciding what gets stored locally vs cloud
This shaped the architecture significantly.
Tracks Applied (1)