AI for Agriculture and Environment

Using AI to Solve Agricultural and Environmental Problems

For millennia, new tools have enabled major advancements in agricultural efficiency and environmental management. Purdue Agriculture researchers are using artificial intelligence (AI) and machine learning (ML) to enhance their ability to address agricultural and environmental issues. AI enables computers to mimic human intelligence, while ML, a subset of AI, allows computers to learn from data and identify patterns.

Brady Hardiman, associate professor of forestry and natural resources and environmental and ecological engineering, uses AI and ML in his research. He co-edited a special issue of Frontiers in Ecology and the Environment that highlighted the importance of applying AI and ML to process large 3D datasets. Hardiman studies urban ecosystems, noting that 80% of the US population lives in urban areas. He says that cities' different histories, cultures, policies and infrastructure make them complicated to study. His group analyzes remote sensing data and high-resolution imagery of urban landscapes, using AI and ML to detect patterns invisible to the naked eye. For example, they identify tree species to improve urban forest management.

Tracking Invasive Species

In Chicago, Hardiman’s team uses aerial light detection and ranging (LiDAR) data and machine learning to track invasive buckthorn shrubs in city forest preserves. This mapping helps forest managers conserve forests.

In rural areas, ML helps researchers identify tree species for timber production and other applications.

AI-Enhanced Medical Robot for Cattle

Upinder Kaur, assistant professor of agricultural and biological engineering, is creating an AI-enhanced medical robot to collect accurate data on cattle nutrition, sustainability practices, and methane output. “This robot is the first such medical robot for animals. It can swim inside the stomach of the cow. It can monitor methane, temperature, pH and other biomarkers to give you a much richer detail of how the rumen is working,” Kaur says. AI reduces the computational resources needed to operate the robot, improving its consistency and lifespan. The robot provides detailed methane emission data, which is more comprehensive than current methods that use masks to monitor individual cows for short periods.

Kaur’s team aims to make the robot small enough for cows to swallow. Currently, it is inserted through a cannula. The robot also enhances animal welfare by providing frequent biomarker data, enabling quick responses to changes in a cow’s condition, such as during birth. Kaur says their goal is to improve animal lives and reduce burdens on farmers.

Kaur, along with Catherine Hill and Maria Murgia, is also developing a robot “dog” that uses sensors to locate ticks and identify tick activity hotspots. This helps control tickborne diseases.

Simulating Crop Yields with Machine Learning

Sajad Jamshidi, a postdoctoral scientist in agronomy working with Diane Wang, simulates crop yields under various climate scenarios. He co-authored a paper that combined 10 machine learning models to assess the effect of public breeding trials on rice yields under future climate conditions. The researchers initially used one model but then decided to use an ensemble approach to combine the strengths of multiple models for more accurate results.

Wang’s team analyzed data from the Southern U.S. rice growing region from 1970 to 2015 and found that public breeding programs developed more resilient rice varieties. These breeders unintentionally developed varieties resistant to a warming climate while focusing on higher yields. The Purdue team’s framework allows breeders to quickly test rice variety responses to climate, temperature, and rainfall through simulations.

AI and Decision Support Tools for Crops

Ankita Raturi, assistant professor of agricultural and biological engineering, develops technologies to help agricultural stakeholders access useful data. She created a web-based recommender system called “Netflix for crops,” which allows users to filter data by location, soil, weather, and goals to identify the right crop for the right time and place. Her Agricultural Informatics Lab focuses on human-centered approaches to building innovative technologies for food system stakeholders.

Raturi’s projects include using agent-based models to represent foodsheds for data-driven policymaking and emphasizing regenerative or small-scale, diversified agriculture. These models help policymakers and food coordinators balance food security outcomes with their needs.

Detecting Crop Diseases with Machine Learning

Somali Chaterji, associate professor of agricultural and biological engineering and electrical and computer engineering, uses ML to build algorithms that help answer questions in computing, health, and agriculture. ML can assist farmers in scouting for crop diseases more efficiently. Chaterji’s work on semi-supervised semantic segmentation is ideal for detecting rare disease outbreaks. Her models start with expert-annotated leaf images and then automatically tag other photos, expanding the training set without extra human effort.

This technology can be used in networks of ground sensors and drone fleets for continuous agricultural surveillance, enabling farmers to target treatments precisely. Chaterji’s Innovatory for Cells and Neural Machines (ICAN) lab developed Agile3D, a LiDAR-based perception algorithm for resource-constrained platforms, allowing devices to run sophisticated models without draining batteries or needing constant connectivity. Her projects focus on resource-aware training and inference, designed to use minimal computing resources.

ICAN has also incorporated training to make algorithms resilient to noise. Chaterji’s research is funded by a National Science Foundation CAREER award, and she presents her work at top conferences. Her work focuses on processing data close to where it is generated, enabling real-time decisions, cutting energy use, and broadening access to AI.