Development of an AI image classifier for detecting vulnerability of African ecosystems under changing climates using ancient data
PI: Sarah Ivory
The level of effort appropriate for the proposed project: 2 semesters at 50% RA or 1 semester plus summer 2026
Plan for funding tuition: existing NSF grant or I would seek support from the department or EESI
A list of specific areas of computational and/or data science expertise or skills: image classification algorithms, app development
Any other requirements or expectations: attendance at regular group meetings and weekly one-on-one meetings
A list of specific objectives for work:
• Evaluate existing training libraries for image classification of pollen
• Compile and expand training library with pollen reference collection
• Build image classification algorithm using TensorFlow or similar ML model focused on small subset of pollen taxa in Africa
• Validate image classification on fossil pollen samples from Lake Mahoma, Uganda
• Build a user friendly app for image classification of microscope images
At least one medium to long-term goal: this is a publication, but also serves as a proof of concept for work that is being developed for a European Research Council grant proposal
A short statement explaining the connection of the project to ICDS’s mission: Climate is changing at historically unprecedented rates with important implications for ecosystem stability. Information about the response of vegetation to climate in the past from fossil pollen plays an important role in understanding natural ranges of variability and ecosystem vulnerability under changing conditions not observed in the historical record, but much of this work is done manually and is extremely time intensive. In this project, we seek to develop a proof of concept for AI image classification of 4 African pollen taxa common in the fossil record. This project connects to the mission of ICDS to “address research questions of scientific and social importance” that can only be approached through “multi and interdisciplinary teams”.
A paragraph summarizing team member’s recent and/or planned engagement with ICDS: Sarah Ivory gave a short lightning talk at ICDS Day this academic year to find collaborators in applied and fundamental research in global environmental change and paleoecology in Africa. She would like to establish a more lasting collaboration with ICDS affiliates taking the form of grant proposals and jointly advised students.
Project Summary:
Significance: Climate is changing at historically unprecedented rates with important implications for ecosystem stability. In Africa, in particular, people depend daily on services provided by natural ecosystems, like medicine, food, and fresh water. Thus, understanding ecosystem stability and predicting changes in ecosystem services is critical, as most economies aren’t prepared to adapt once resources are gone. Information about the response of vegetation to climate in the past from fossil pollen plays an important role in understanding natural ranges of variability and ecosystem vulnerability under changing conditions not observed in the historical record.
Problem: Pollen analysis today suffers from one critical problem for being useful data that can be implemented in conservation and management planning: pollen data is generated by a slow, manual, bespoke process that involves a person doing morphological classification at a microscope. However, machine learning and AI tools, particularly image classification, stand to revolutionize the way that fossil data is generated, the volume of data available, and techniques for generating predictions from sparse, uncertain data.
The proposition: New methods for automating data collection and quantifying spatial and temporal uncertainty could change the way fossil workers do business and their impact; however, they often do not have the data sciences background to implement these techniques. Proposals that are currently under development could support some of these parts, but many pieces of bringing together paleontology and machine learning are feasible now. For example, automated scanning microscopes are available in the Department of Geosciences and may be useful for rapidly generating large volumes of pollen images from samples. Additionally, a reference library of modern African pollen types is currently in the Ivory Paleoecology Lab for building a training library. For this project, a student might develop a small pilot to develop an image classifier of four common and morphologically distinct pollen taxa (such as Poaceae, Celtis, Podocarpus, Ericaceae). Then, this classifier could be applied to fossil samples from Lake Mahoma, Uganda, to evaluate its performance on real samples. Samples from this lake have already been analyzed manually (Ivory et al., 2024) and include important changes in abundance of many common, morphologically distinct taxa. Additionally, the development of a simple web app could make this tool accessible to paleoecologists working in similar settings.
This project would potentially leverage all four pillars of ICDS, especially data sciences and AI. Further, the research themes of this project with a focus on impacts of climate change, ecosystems, and risk aligns with many research interests of ICDS students and faculty, but who are not currently leveraging fossil data.