С����Ƶ

Housed at the Geography Department’s research ecosystem is the Geospatial Human-Centered Artificial Intelligence (GeoHAI) Lab, where innovation in spatial data science meets urgent societal needs. Directed by Geography professor Dr. Morteza Karimzadeh, the GeoHAI Lab is a vibrant community of interdisciplinary researchers dedicated to advancing both the theory and practice of geospatial AI.

Rooted in the disciplines of Geography, Computer Science, and Information Science, the lab’s work spans a broad array of environmental and social science application domains. True to its mission of use-inspired research, the lab develops methods that not only push the boundaries of scientific knowledge but are also grounded in real-world applications. GeoHAI’s projects are frequently co-designed with stakeholders, ensuring that the lab’s innovations are usable, relevant, and impactful.

A Lab Built on Collaboration and Innovation

The GeoHAI Lab thrives on collaboration—both within its team and across disciplinary lines. Dr. Karimzadeh's leadership draws on his expertise in spatiotemporal machine learning, remote sensing, spatial statistics, and human-centered visual analytics, integrating these fields to address critical challenges facing society today.

From sea ice mapping in the Arctic to forecasting the geographic spread of diseases, the lab’s projects harness non-linear spatial and temporal relationships using cutting-edge AI techniques. But the work doesn’t stop at model development. Whether through designing interactive visualizations or collaborating on epidemiological algorithm design, the lab's products are meant to be adopted and used by practitioners, not just published in academic journals.

Projects at the Intersection of Science and Society

GeoHAI’s projects are as diverse as they are impactful:

• Air Pollution Estimation for Public Health: Leveraging satellite remote sensing and ground-based monitoring, the lab develops deep learning models to estimate daily PM2.5 and ozone concentrations at high spatial resolution. These models support epidemiological research and policy efforts to mitigate health risks from air pollution.
• Sea Ice Mapping and Uncertainty Quantification: The lab’s research in the cryosphere domain combines SAR, passive microwave sensing, and ICESat-2 laser altimetry to improve sea ice charting. Importantly, they integrate uncertainty quantification to help decision-makers understand model confidence—critical for navigational safety and climate monitoring.
• Forecasting Post-Fire Vegetation Recovery: In collaboration with land managers and ecologists, GeoHAI students develop state-of-the-art deep learning models to predict how landscapes recover after wildfires. These forecasts support adaptive land management and conservation planning in fire-prone regions. This work impacts how and if land managers conduct reseeding, reforestation or mudslide prevention activities after wildfire events.
• Multi-Source Foundation Models for Environmental Monitoring: The lab has recently started developing “foundation models” that fuse data from multiple satellite sources to enable training deep learning tasks with fewer ground-truth labeled samples, suitable for a variety of environmental monitoring tasks such as flood mapping and land use classification, sea ice mapping, change detection, etc.
• COVID-19 Forecasting: During the pandemic, GeoHAI played a pivotal role in contributing spatiotemporal machine learning models and weekly forecasts to the US COVID-19 Forecast Hub, forecasting hospitalizations and case counts at state and county levels. These were in turn used by the Centers for Disease Control and Prevention for communication and reporting, to aid states in planning. By incorporating social media connectivity to capture spatial interactions in deep learning, the models outperformed national ensemble forecasts during critical surges.
• Visual Analytics for Disease Epidemiology: GeoHAI’s GeoDEN tool enables spatiotemporal exploration of dengue serotype dynamics, offering epidemiologists an interactive platform to analyze disease spread patterns for dengue, a disease that has potential to cause a worldwide pandemic. The lab’s focus on human-centered design and collaboration with epidemiologists ensures that such tools directly address user needs, and the developed tools have the potential to apply to other diseases.
• Network Visualization for Power Grids: Collaborating with the National Renewable Energy Laboratory (NREL), the lab researchers developed novel network-weighted contour maps to visualize electrical grid voltages, helping planners and engineers interpret complex grid behaviors at scale. This work is intended to address the perpetual challenge of difference of visualizations in a geographic map and the topology of electrical grid lines.
• Spatiotemporal Graph Neural Networks for Human Action Segmentation: In collaboration with Intel Labs, GeoHAI develops graph-based deep learning methods to recognize human activities from video, depth sensors, and wearable accelerometers. This work models spatiotemporal relationships across modalities to create efficient, scalable algorithms for health monitoring, manufacturing, and workplace safety.

Sea ice charting example. Top row: high-resolution Sentinel-1 SAR image. 2nd row: Low- resolution passive Microwave imagery. 3rd row: manually generated ice charts. 4th row: Deep-learning-based generated ice charts, along with uncertainty map to the right.

��

GeoDen (available at geoden.net): A tool developed at GeoHAI for Exploratory analysis of spatiotemporal spread and interaction of dengue serotypes.

��

GeoHAI lab members enjoying group lunch on the Hill. Left to right: Kevin Lane, Sepideh Jalayer, Jen MacDonald, Morteza Karimzadeh, Juliar Romero, Zhongying Wang, Daehyeon Han, Isaiah Lyons-Galante

Students and Postdocs: Driving the Lab’s Success

GeoHAI’s strength lies not only in its projects but primarily in its people. The lab is home to an exceptional team of Ph.D. students, master’s students, undergraduate students, and postdoctoral researchers with backgrounds ranging from geography, engineering, computer science and environmental modeling. These emerging scholars bring a wealth of experience—from working at Nokia Bell Labs and Johns Hopkins APL to startups and software engineering roles—enriching the lab’s research with fresh perspectives.

Students in the lab work on sea ice classification, multi-modal machine learning, infectious disease forecasting, air pollution mapping and more. They play a central role in every stage of research, from algorithm design and data processing to stakeholder engagement and scientific publication.

��

Building a Sustainable and More Equitable Future

GeoHAI researchers are committed to open science. The lab regularly publishes datasets, open-source code, and visualization tools to foster reproducibility and broaden the impact of its research. Whether advancing environmental sustainability or supporting public health, the lab’s work exemplifies how AI and geospatial science can serve as catalysts for equity and societal resilience.

To learn more or explore collaboration opportunities, visit the GeoHAI Lab at CU С����Ƶ: