Geospatial Artificial Intelligence (GeoAI) & Advanced Simulation

The RCC’s GIS and scientific computing teams partner with University of Chicago researchers to apply advanced computational methods to complex geospatial questions. We specialize in combining geospatial data with artificial intelligence, deep learning, and sophisticated environmental modeling to create new methodologies and accelerate research.

Our work emphasizes a collaborative model, integrating domain expertise from fields like archaeology and environmental science with advanced computing expertise to analyze vast datasets and test hypotheses in ways not possible through traditional methods.

Our Capabilities

Our projects demonstrate expertise in a range of advanced computational and geospatial techniques:

• Deep Learning & Computer Vision: We implement deep learning models to detect and classify features in high-resolution satellite and aerial imagery.
• Hydrological & Environmental Simulation: We employ advanced models (like GeoWEPP, HEC-HMS, HEC-RAS) to simulate environmental processes such as water flow and soil erosion on digitally reconstructed landscapes.
• High-Resolution 3D Landscape Reconstruction: We use LiDAR and photogrammetry data to build and digitally modify high-resolution Digital Elevation Models (DEMs) for “what-if” scenario analysis.
• Large-Scale Data Management: We develop workflows for researchers to both produce and consume data at scale. This includes command-line and Python-based tools for uploading datasets, which can be automatically converted to cloud-optimized formats.
• Collaborative Research Infrastructure: We architect platforms designed to foster collaboration by closing the gap between data production and consumption. This enables teams to share and version large-scale products like satellite imagery or derived maps.

Selected Projects

Afghan Heritage Mapping Partnership (AHMP)

This project demonstrates a successful collaboration between archaeologists (AHMP) and computer scientists (RCC) to dramatically accelerate the identification of archaeological sites across Afghanistan. By using deep learning, the partnership developed a new methodology for cultural heritage preservation.

Challenge: Manually identifying archaeological sites over Afghanistan’s 652,860 km² is a monumental task. The traditional method required 5 years for the initial survey effort.

Computational Approach: A deep learning model was trained to detect features like mounds, caravanserais, forts, and qanats using satellite and LiDAR imagery. The model processed over 70 million image tiles, with AI-predicted sites then validated by archaeologists in a collaborative, iterative workflow.

Impact: The AI-assisted methodology increased the speed of site identification by up to 10 times. The team was able to survey the remaining 288,423 km² in just 1.5 years. The resulting methodology is accessible, reproducible, and adaptable to other regions.

Reconstructing and Modeling Petra’s Ancient Water Management

This project integrates archaeological field knowledge with advanced computational science to quantitatively analyze the sophisticated water management systems of ancient Petra. The research tests the effectiveness of ancient infrastructure by digitally reconstructing the landscape and simulating environmental processes.

Challenge: To understand how Petra’s inhabitants supported agriculture and protected their city from flash floods in a hyper-arid region, without the benefit of long-term historical climate data.

Computational Approach: The team creates high-resolution 3D maps from LiDAR data and uses GIS to digitally add ancient terraces, dams, and walls back into the landscape. To test these “what-if” scenarios, we developed a 100-year stochastic climate model specifically for Petra and use it to power advanced hydrological and erosion simulations.

Impact: This research provides quantitative measurements of the effectiveness of ancient engineering, showing how small-scale infrastructure significantly altered runoff and reduced soil erosion. It is one of the first applications of this integrated suite of simulation tools in the Petra region, creating a replicable framework for studying human-environment interactions worldwide.

The A3RD Initiative: A Global Qanat Detection Project

This project, in collaboration with the University of Chicago’s CAMEL Lab, exemplifies our approach to building cutting-edge research tools for the humanities. A3RD is a global initiative focused on the automated detection of ancient water systems known as qanats.

Challenge: Manually locating specific, widespread archaeological features like qanats is a slow, laborious process. Furthermore, advanced computational tools often remain inaccessible to humanities scholars who are the domain experts.

Collaborative Approach: The project brings together international collaborators with the CAMEL and RCC teams, as highlighted in a recent A3RD workshop in Paris. The focus is on improving AI models for qanat detection through methods like transfer learning and refining model architecture to handle regional variations.

Collaboration

To learn more about our GeoAI and simulation capabilities or to discuss a potential project, please contact the Research Computing Center.