Bayesian Modeling for Agricultural Data

Bayesian models are now used for applied data analysis almost as regularly as classic methods such as t-tests, regression, and ANOVA. Data generated from agricultural systems, whether from a designed experiment, on-farm trials, or opportunistic observations, can benefit from using Bayesian models. Bayesian statistics enables researchers to build bespoke statistical models tailored to the specific research question or application. Furthermore, Bayesian models enable fully probabilistic and statistically valid inference not only on model components (e.g., slope parameters) but also on other indirect quantities of interest (e.g., probability yield is below a certain threshold). In this workshop, we aim to enable practitioners to understand the basics of Bayesian models, demystify standard Bayesian techniques such as Markov chain Monte Carlo, and provide real-world, hands-on agricultural data examples where Bayesian models enable new and essential insights.

Workshop Session Time and Location: 

This workshop will take place from 2:00pm - 4:00pm on Sunday, 21 July 2024.

Manhattan, KS 66506

 

---

Instructors:

Dr. Trevor Hefley, Dr. Josefina Lacasa, Francisco Palmero

---

 

Dr. Trevor Hefley

Trevor J. Hefley is an associate professor in the Department of Statistics at Kansas State University. He earned a joint PhD in Statistics and Natural Resource Science at the University of Nebraska-Lincoln and focuses on developing and applying spatio-temporal statistical methods to inform environmental decisions.

 

Dr. Josefina Lacasa

Josefina Lacasa is an Assistant Professor in the Department of Statistics at Kansas State University. Josefina holds a MSc in Statistics and a PhD in Agronomy. Her work aims to enhance quantitative methods in agricultural sciences through teaching, consulting and research. Overall, her research focuses on agricultural statistics, where she develops statistical methods for agricultural researchers and adapts existing techniques to agricultural applications.

 

Francisco Palmero

Francisco Palmero is a PhD candidate in the Department of Agronomy at Kansas State University. He earned an MSc in Soil Fertility and Crop Nutrition from the Universidad de Buenos Aires. His current research focuses on crop physiology and nitrogen management in corn hybrids.

GIS-based Spatial Interpolation Methods

21 July 2024

Modern GIS software allows users to apply a range of spatial analysis models across a spectrum of analytical sophistication from simple (but informative) descriptive statistics to powerful explanatory models. In this workshop, we’ll examine spatial interpolation methods as one approach to predictive modeling that helps practitioners determine the value of an important agricultural or environmental variable where it hasn’t been measured using a control point dataset of known values recorded at specific locations. Spatial interpolation methods can be categorized broadly into deterministic and geostatistical approaches. We will explore techniques within each category to better understand their purpose, appreciate their strengths and weaknesses, and how to evaluate which technique is best for a given modeling scenario. Learning will take place for two one-hour studio sessions featuring both lecture discussion and practical hands-on work. A final practical exercise will be used to reinforce concepts and allow participants to work independently with a new dataset to produce predictions with cross-validated model performance metrics.

Workshop Session Time and Location:

This workshop will take place from 9:00am - 12:00pm on Sunday, 21 July 2024.

Manhattan, KS 66502

---

Instructors:

Dr. Shawn Hutchinson, Carlos Hernandez, Dr. Trevor Hefley

---

 

Carlos Hernandez

Carlos Hernandez was born in the city of Rio Cuarto, Argentina. During his undergraduate career he obtained his degree in agronomist engineering from the National University of Río Cuarto, Argentina. During his professional career he has been part of various work teams as an advisor on applied technologies and precision agriculture, providing support to farmers in both the academic and private sectors in various companies. Over the past few years he has reinforced his knowledge of geospatial technologies and topics such as predictive agriculture, data science and artificial intelligence. Currently its focus is on the research and development of algorithms and data products that can help farmers in the decision-making process such as the estimation of quality maps in soybeans or the analysis of nitrogen dynamics in the cropping system.

 

Dr. Shawn Hutchinson

Shawn Hutchinson is a Professor of Geography and Geospatial Sciences, Director of Kansas State University’s Geographic Information Systems Spatial Analysis Laboratory, and founding Co-Director of the Institute for Digital Agriculture and Advanced Analytics (ID3A) at Kansas State University. He is a geographer and environmental scientist who seeks to maximize the power of GIS, remote sensing, and computational methods with an emphasis on real- and near-real time environmental monitoring. His research program focuses on developing, evaluating, and forecasting metrics that characterize the sustainability of working lands, the status and quality of surface water resources, and in assessments of agricultural biosecurity hazards at regional and national scales.

 

Dr. Trevor Hefley

Trevor J. Hefley is an associate professor in the Department of Statistics at Kansas State University. He earned a joint PhD in Statistics and Natural Resource Science at the University of Nebraska-Lincoln and focuses on developing and applying spatio-temporal statistical methods to inform environmental decisions.

Object Detection 101: A Data-to-Deployment Workshop

21 July 2024

Machine learning, specifically deep learning approaches, can be useful to detect, classify, and segment different objects using imagery collected from different devices. Currently, this technology is rapidly growing, and the need to address different agricultural tasks without intense labor can be solved with automation using deep learning. For example, different detection models including in the YOLOv5 family can exceed in detecting small and big objects. This beginner-friendly crash course on developing deep learning models objects in agricultural spaces will equip you with essential skills to harness the power of object detection in the context of precision agriculture. In this workshop you will understand the basics of collecting high-quality data using sensors on mobile devices but with pathways to adopt similar strategies to data collected from drones, satellites, and ground-based sensors. Participants will also explore the capabilities of cutting-edge, open-source software like Roboflow to build your initial object detection models through hands-on activities and will gain practical experience in developing and fine-tuning object detection models. Participants will also learn how these tools are currently applied in addressing real-world challenges in precision agriculture and integrated into robotic systems, such as pest and disease detection, sense-and-spray technologies, yield estimation, and crop health monitoring.

Workshop Session Time and Location:

This workshop will take place from 10:00am - 12:00pm on Sunday, 21 July 2024.

Manhattan, KS 66506

---

Instructors:

Dr. Ivan Grijalva, Dr. Brian Spiesman, Dr. Brian McCornack

---

 

Ivan Grijalva

Ivan Grijalva is a Postdoctoral Researcher in Entomology at Kansas State University. He obtained his Ph.D. in Entomology from Kansas State University, specializing in Integrated Pest Management and Digital Agriculture. Additionally, he holds a certification in Geographic Information Systems. His research employs machine learning and digital agricultural tools to automate pest management strategies. He is particularly interested in using computer vision models to detect insects by remote sensors and robotics. Presently, Grijalva is engaged in various digital agriculture projects, such as developing models for aphid detection in sorghum and using drone technology for Japanese beetle detection in soybeans. He has authored several research articles in various journals, focusing on implementing computer vision models to automate insect detection. These contributions are significant for advancing automation within Entomology and related disciplines.

 

Dr. Brian McCornack

Brian McCornack is a Professor and Department Head for Entomology at Kansas State University (K-State) with research, extension and teaching responsibilities. He is also a Co-Director of Engagement for the Institute for Digital Agriculture and Advanced Analytics (ID3A) at K-state. He has an integrated research program that facilitates the discovery and application of tangible solutions to emerging pest issues, including endemic and invasive species impacting soybean, corn, sorghum, wheat, and other major crops across the Southern Great Plains. McCornack has developed a wide range of Integrated Pest Management (IPM) tools for use in commercial agriculture, including widely adopted economic injury levels and thresholds for invasive species, sampling plans for key agronomic pests, and sampling strategies for managing economically important species, natural enemies, and pollinators using computer vision.

 

Dr. Brian Spiesman

I study relationships between insects, plants, and the environment. My research focuses on how species are distributed in space and time, how networks of interactions help structure communities, and the consequences of those interactions for community function and stability. Pollinators, many of which are in decline, are essential for the function of many natural and managed habitats. It is therefore important to understand how pollinators respond to disturbances such as habitat loss and climate change, but also how they respond to conservation efforts aimed at preserving their biodiversity. I use a combination of large- and small-scale experiments, observational studies, and mathematical modeling to explore questions in basic and applied ecology.