More than one-quarter of world’s Greenhouse Gas (GHG) emissions come from agriculture, forestry, and land use change. The agriculture's role in GHG is widely known but not well understood. In order to understand farming factors to GHG, we present a framework for automated identification of spatio-temporal carbon hot-spots and associate factors by discovering cohorts of farms based on different parameters such as farming practices, weather conditions, farm characteristics, yield and GHG emissions. The framework also generate farming practices recommendations to reduce the GHG. To demonstrate the framework, we will present a case study on palm oil farming in Indonesia.