Code in the Canopy

How Madhya Pradesh’s AI experiment in forest monitoring could become a model for the world

Madhya Pradesh

In the heart of India, a forest officer has achieved a coup of sorts that ministries and tech firms across the world are still mulling. It is marrying artificial intelligence with satellite imagery and field-level accountability to fight deforestation in real time.

Madhya Pradesh, the state with India’s largest forest cover, is now the first in the country to pilot an artificial intelligence (AI)-driven, cloud-based forest alert system. Developed not in some private lab but by a public servant - a young Indian Forest Service officer named Akshay Rathore - it is an experiment worth watching. If it works, the system could become a template not only for India’s fragile forests but also for endangered ecosystems across the developing world.

The system, already rolled out across five forest divisions notorious for illegal tree-felling and encroachment in Guna, Shivpuri, Khandwa, Burhanpur and Vidisha, uses satellite images from Google Earth Engine, compares them across three dates, and applies a custom-built AI model to detect changes as minute as a 10-by-10 metre patch of tree cover. Alerts are pushed directly to beat guards via a mobile app. They are then expected to physically verify the site, upload geo-tagged photos and audio comments, and close the feedback loop.

This is a system designed for a country where manpower is stretched and terrain is often inaccessible and given that traditional monitoring methods, usually paper-based or relying on bureaucratic relays, are ill-suited to respond to dynamic threats.

While Madhya Pradesh may have 85,724 sq km of forest and tree cover (according to the Forest Survey of India’s 2023 report), it also leads the country in forest loss, with 612 sq km lost that year alone. Rathore’s system does more than just flag these changes. It classifies them, analyses their vegetation index (NDVI, SAVI, and EVI, for those who like acronyms) and sets up the ground force for real-time response.

Rathore, an alumnus of IIT Roorkee, built the initial Python scripts himself by using ChatGPT to streamline some of the scripting while leaning on lessons learned from an earlier encroachment flare-up in Guna.

The best innovations in governance are often not those with the largest budgets or biggest private partners, but those born out of institutional urgency and local knowledge. Consider Kenya’s use of blockchain to verify land titles or Indonesia’s ‘One Map’ policy to integrate spatial data for forest governance. India, with its complex land politics and mounting ecological pressures, needs more such bottom-up, tech-enabled models.

Still, the Madhya Pradesh model is far from perfect. Human verification, though necessary for now, slows down the system and leaves room for neglect. But Phase 2 of the project, which proposes to use drones and historical seasonal data to train predictive models, could address this.

There are reasons for caution. AI-based governance tools often raise concerns about surveillance, data misuse and overreliance on algorithms. But the potential here is vast. The approach blends precision with scalability. A 10-by-10 metre resolution is good enough to catch most illegal activities without overwhelming field staff. The alert-to-action loop means the system is not just diagnostic but operational. Over time, as the model learns from on-ground feedback, it promises to be self-improving. Think of it as the Waze of forest governance except instead of navigating traffic, it is routing patrols to illegal loggers and encroachers.

The Indian state is often accused of being sluggish, reactive and under-resourced. But this experiment shows what is possible when the state leverages both its local intelligence and cutting-edge tech.

Whether Madhya Pradesh’s system will scale to other parts of India remains to be seen. Bureaucratic rivalries, budget constraints and technical hurdles are real. But in a country where forests are both sacred groves and political battlegrounds, and where climate change is no longer a future threat but a lived reality, Rathore’s AI system offers a fighting chance to protect forests.