Google has launched India-First AI models for agriculture through its Google DeepMind AnthroKrishi team, introducing two models designed to generate field-level agricultural intelligence from satellite imagery. Agricultural Landscape Understanding (ALU) maps agricultural fields and land patterns, while Agricultural Monitoring and Event Detection (AMED) tracks changes and agricultural events over time. The technology is already being applied across more than 140 million hectares of farmland by TerraStack, while Karnataka is using the data across 2.6 million hectares of irrigated land for water management.

The models are designed to identify agricultural fields, trees and water bodies across regions and then monitor how those areas change. Their outputs are available through APIs for developers and organisations, as well as through a visual data layer on Google Earth.

ALU and AMED Build a Field-Level Agricultural Picture

Agricultural Landscape Understanding, or ALU, focuses on understanding the physical agricultural landscape. It can identify field boundaries and land patterns, creating granular information that can be used for mapping, land analysis and digital agriculture services.

Agricultural Monitoring and Event Detection, or AMED, adds the time-based monitoring component. Once agricultural fields are identified, the model can help track crop conditions, land-use changes and events such as sowing and harvesting. Together, the models provide a framework for understanding not just where agricultural land exists, but what is happening across it.

Also Read: China’s Solar Capacity Surges To 1.28 TW, Reaching 31.5% Of Total Power Capacity

Agricultural Intelligence Reaches Google Earth and APIs

Google has integrated the data produced by ALU and AMED into Google Earth as a visual layer. Partners can also access the underlying information through APIs, allowing them to incorporate agricultural intelligence into their own applications and services.

The India-First AI models for agriculture are already being used for several practical applications.

• CarbonFarm has combined the ALU API with Gemini for low-carbon rice cultivation, with an ambition to cover 2 million hectares by 2030.
• TerraStack, an IIT Bombay-incubated startup, has used the APIs to map more than 140 million hectares of farmland.
• Karnataka’s Water Resources Department is using the data for water management across 2.6 million hectares of irrigated land.
• The Agricultural Data Exchange (ADeX) in Telangana has incorporated the models into digital agriculture services.

Also Read: Declining Natural Carbon Sinks Could Complicate Net Zero: Wood Mackenzie

Google Expands the Models Beyond India

Although the initiative was developed with India’s agricultural ecosystem in mind, Google has begun extending the technology internationally. The models have been shared with trusted testers across 11 countries in Asia-Pacific and Africa, including Indonesia, Japan, Malaysia, Vietnam, Kenya, Uganda, Ghana, Rwanda, Nigeria and Zambia.

The expansion reflects Google’s broader objective through AnthroKrishi, which is focused on organising agricultural information and making it more accessible and useful. The team is applying Google’s geospatial modelling capabilities to areas where better agricultural data can support credit access, crop advisory and farm-level decision-making.

According to Google DeepMind’s Alok Talekar, the models provide wall-to-wall coverage that can identify individual fields and then monitor them over time. As the technology expands to more countries, Google sees potential applications extending from food security to agricultural resilience.

Also Read: UN Warns Rising Seas Could Displace 14 Million In South Asia