This is the system we are building at Get Zero Private Limited, a climate tech company from India. Here is the architecture at a high level.

Planting records vs survival checks

A planting record says what was planted, where and when. Geo-tagging apps and most government portals do this. A survival check goes back to the same site again and again and measures how much of it is alive. A useful plantation tracking system in India has to do both. The CAG found no new tree cover at 70% of the Green India Mission sites it checked by satellite, which is the gap a survival check is meant to close.

Constraints

• Coverage. Sites can be anywhere in India. A site in the Thar desert should cost the same to check as one next to a city.
• Cadence. Saplings die within weeks in a dry spell, so a yearly survey is too slow. We wanted a weekly check.
• Cloud. Most planting in India happens in the monsoon, exactly when optical imagery is blocked.
• No hardware. No sensors or tags. A pin, a boundary, a KML file, a shapefile or a list of GPS points should be enough.

Data sources

Sentinel-2 (optical, multispectral). Revisits about every five days. Red and near infrared bands give a vegetation index (NDVI). A moisture index measures plant water content. The roughly 10 metre resolution sets our grid size.

NDVI is the standard normalised difference of near infrared and red reflectance:

NDVI = (NIR - Red) / (NIR + Red)

Sentinel-1 (C-band SAR). Radar passes through cloud, so it keeps the time series running through the monsoon.

Drones. At 10 metres, one pixel holds a group of saplings. For tree level counts we fly drones, twice a month on sites that need it.

The 10 metre grid

Each site boundary is split into 10 metre cells, and every cell is scored on its own. A site level mean can hide a dying corner behind a healthy centre, and the dying corner is what the field team needs to find.

Three signals per cell

1. Greenness from the vegetation index
2. Water stress from the moisture index
3. Growth trend, whether growth is speeding up or slowing down

Change over time, not single readings

A single reading is noisy. Each week, every cell is compared with its own history, and a sustained decline raises a WhatsApp alert. Cells that stay green and keep growing count as surviving. Cells that fade and stay bare count as lost. Drone counts confirm the number at tree level.

From cells to a tree survival rate

The headline number is still the classic tree survival rate formula:

Survival rate (%) = live saplings counted / saplings planted x 100

For example, 770 live saplings out of 1,100 planted is a 70% survival rate. The satellite tells us where survival is falling. Drones and field plots supply the tree counts that go into the formula.

Processing

Everything runs on Google Earth Engine, which is built for planetary scale raster processing. Adding a hundredth site costs compute time, not a field trip.

Output

• A live dashboard with a health map per site
• WhatsApp alerts that name the failing patch in plain words
• Survival rate per site and project, a survival curve since planting, dated maps and survival certificates

FAQ

What is a plantation tracking system? A plantation tracking system keeps a record of every planted site, with its location, boundary, species and planting date, and then checks the site again over time to show how many trees are alive, healthy and growing.

Can satellites measure tree survival? Sentinel-2 sees the ground in 10 metre squares, so it measures vegetation health across groups of trees rather than single saplings. It spots where a site is declining. Drones or field plots count individual trees.

Is a geo-tagging app enough to track a plantation? No. A geo-tagged photo proves a sapling was planted at a place on a date. It does not show whether it is alive a year later unless someone goes back.

Read choosing a plantation tracking system in India, how to measure tree survival rate and how satellite tree monitoring works. Get Zero is built by Ritwik Joshi (CTO) and Gagan Sharma (CEO).