Weeds and invasive plants reduce crop yields, displace native species, and can negatively affect human health. Yet weed management remains heavily dependent on herbicides. In the United States, corn and soybean production alone receives more than 110 million kg of herbicides annually across an area roughly twice the size of Germany, contributing to widespread herbicide resistance and environmental concerns.
Recent advances in sensing, robotics, and artificial intelligence are creating new opportunities for more precise and sustainable weed management. In this UNU-INWEH Science Talk, Dr. Mohsen Mesgaran will explore a suite of digital tools for weed identification, prediction, detection, and knowledge access.
These include WeedChat, a source-grounded AI assistant that provides weed-science answers with citations and generates on-demand research syntheses; image- and key-based identification tools covering approximately 3,500 weed species; and hydrothermal-time models that use soil temperature and moisture to predict weed emergence. The talk will also present a computer-vision tool that uses Google Street View imagery to detect and map invasive plants along roadsides, as well as a comprehensive database of regulated plants covering 104 jurisdictions worldwide. Together, these innovations support more informed, precise, and timely management of weeds and invasive plants.
You may watch the talk on the UNU-INWEH YouTube channel (no registration required).
Speaker
Dr. Mohsen Mesgaran
Lead, Ecological Modeling and Food Security
