Google DeepMind exports Indian-trained agricultural models to six African countries

The ALU and AMED models, designed to read Indian plots from the satellite, are now available in Kenya, Uganda, Ghana, Rwanda, Nigeria and Zambia.

Google DeepMind's AnthroKrishi team extends two agricultural analysis models developed on Indian soil to Africa. The first, Agricultural Landscape Understanding (ALU), mapping and characterizing cultivated plots; The second, Agricultural Monitoring & Event Detection (AMED), tracks crop activity and detects anomalies. Six countries are involved: Kenya, Ghana, Rwanda, Nigeria, Uganda and Zambia.

The models are displayed via APIs and integrated into Google Earth, where the ALU layer is among the most popular worldwide. Several actors have already put them at work: CarbonFarm combines ALU and Gemini to automatically produce analyses of rice plots in 12 countries, with a target of twenty; the Terrastack platform covers more than 140 million hectares of Indian agricultural land to bring cadastre together with estimates of agricultural income; The State of Telangana's Agriculture Data Exchange uses it to detect crop stress among more than five million farmers.

In Karnataka, ALU and AMED are deployed on 2.6 million hectares for water and watershed management. A geoAI4stats initiative with FAO, supported by $2.5 million by Google.org, also provides these data sets in CROPGRIDS to monitor the climate resilience of agricultural systems.