About

Fangzheng Lyu is associated with the Center for Geospatial Information Technology (CGIT) at Virginia Tech, which collaborates across research, education, and outreach with a transdisciplinary approach, addressing complex problems with geospatial science. The center focuses on applying geospatial science to improve quality of life, environment, and community through smart decision making. CGIT utilizes extensive knowledge in Geographic Information Systems to provide powerful geospatial tools with an easy-to-use interface, transforming spatial data into secure, intuitive decision-making tools that empower agencies, researchers, and communities across the Commonwealth. The research at CGIT, where Fangzheng Lyu is involved, fuses geospatial science, software engineering, and user experience design to develop applications that translate complex datasets into practical insights. These tools support decision-makers in mapping risk, tracking infrastructure, forecasting change, and enhancing safety, efficiency, and strategic planning. Key projects include redesigning the DMV Geocoding Tool for address accuracy and safety analytics, developing the Virginia State Police Crash Analysis Platform to visualize crash data, and creating dashboards that connect data across sectors for smarter policy and community development. The center's work emphasizes intelligent mapping, data visualization, secure GIS and automation tools, AI and analytics for predictive modeling, and decision-support systems for transportation, environment, and security, contributing to data-driven decision-making for a more informed, resilient, and secure future.

Research topics

• Computer Science
• Artificial Intelligence
• Geography
• Cartography
• Machine Learning
• Remote sensing
• Environmental health
• Geology
• Business
• Meteorology
• Medicine
• Economic growth
• Environmental science

Selected publications

• Converging high-performance computing and machine learning for geospatial discovery and innovation

  Annals of GIS · 2026-03-03

  articleOpen access1st author

  The synthesis of high-performance computing (HPC) and machine learning has been critical for addressing complex geospatial problems and enabling geospatial knowledge discovery. This paper conducts a systematic review of 289 selected literature indexed in the Web of Science Core Collection from 1996 to 2024 that integrates HPC and machine learning (ML) for geospatial discovery and innovation. Starting in 2015, there has been a significant increase in studies combining HPC including supercomputing, parallel computing, cloud computing and fog computing with machine learning models for geospatial research across domains. This paper categorizes prior work based on the purposes of leveraging machine learning and HPC for geospatial knowledge discovery including speedup, accuracy improvement, spatial and temporal resolution improvement, scaling up, real-time analysis and novel model development. In addition, we propose a future research agenda including five key research questions focusing on scaling geospatial foundation models on HPC systems, integrating ML with physics-based models while preserving fidelity, quantifying uncertainty and ethical risks in ML predictions, balancing computational intensity with energy efficiency and ensuring HPC-ML pipelines are FAIR and cross-sector usable as well as four interconnected research thrusts: scalable geospatial data fabrics, geospatial foundation models, domain knowledge and ML integration, and responsible and transparent geospatial AI, along with their future implementation strategies and anticipated impacts.

  PublisherDOI

• Customized decision-making support per location's characteristics of spatial accessibility: A case study of emergency healthcare in New York City

  Computers Environment and Urban Systems · 2026-05-23

  articleOpen access

  • Locations can be classified based on the different characteristics (the combination of temporal changes and different levels) of their accessibility. • These clusters can identify diverse needs across locations to effectively reduce access inequality. • Our work can facilitate the optimal resource allocation and support customized and flexible policy responses.

  PublisherDOI

• Evaluating the Evaluation Matrices: Integrating Spatial Assessment in Geospatial AI Model Training and Evaluation

  2025-07-21

  articleOpen access1st authorCorresponding

  This paper examines the limitations of current evaluation metrics in GeoAI. Through two case studies on deep learning models—a building detection classification problem and a remote sensing image fusion regression problem—this paper demonstrates how traditional statistical evaluation matrices alone can be misleading in geospatial problems. The findings indicate that traditional metrics (e.g., RMSE, MAE) used in current GeoAI models can have difficulty capturing the spatial dimensions inherent to geospatial problems. This paper suggests that the model evaluation process in GeoAI should move beyond traditional evaluation matrices by integrating spatial thinking throughout the modeling pipeline—not only incorporating spatial accuracy in model evaluation but also embedding it within optimization functions in model structure and model training.

  PublisherOA PDFDOI

• Building Machine Learning Challenges for Anomaly Detection in Science

  ArXiv.org · 2025-03-03

  preprintOpen access

  Scientific discoveries are often made by finding a pattern or object that was not predicted by the known rules of science. Oftentimes, these anomalous events or objects that do not conform to the norms are an indication that the rules of science governing the data are incomplete, and something new needs to be present to explain these unexpected outliers. The challenge of finding anomalies can be confounding since it requires codifying a complete knowledge of the known scientific behaviors and then projecting these known behaviors on the data to look for deviations. When utilizing machine learning, this presents a particular challenge since we require that the model not only understands scientific data perfectly but also recognizes when the data is inconsistent and out of the scope of its trained behavior. In this paper, we present three datasets aimed at developing machine learning-based anomaly detection for disparate scientific domains covering astrophysics, genomics, and polar science. We present the different datasets along with a scheme to make machine learning challenges around the three datasets findable, accessible, interoperable, and reusable (FAIR). Furthermore, we present an approach that generalizes to future machine learning challenges, enabling the possibility of large, more compute-intensive challenges that can ultimately lead to scientific discovery.

  PublisherOA PDFDOI

• Examine the environmental inequity impact of urban heat mitigation on redlining legacy: case study of Charlotte’s retrofitting, 2001–2020

  UNC Libraries · 2025-02-05

  articleOpen access

  Climate adaptation policies have received attention in major due to the dual challenges of external factors like global warming, and internal factors related to the transition from rapid urbanization to sustainable development. However, previous research on heat or climate mitigation has often focused on external factors, neglecting the internal factors throughout the process of urban development and planning history. Research has revealed that city center where urban heat island phenomena is prominent, are subjected to external factors of intense heat exposure, as well as deeply influenced by the internal factors “urban development legacy.” An increasing body of research note that the inequitable legacy from urban development could impact environmental equity outcomes of cities. Based on this, we argue that urban heat mitigation research should adopt the perspective of the urban development process. We then utilize the Heat Mitigation Framework to examine the tangible outcomes of environmental equity over an extended period of urban development. This study focuses on the Charlotte city center that have undergone multiple processes of redlining policies and rapid urbanization, using a research framework for environmental equity-oriented urban heat management to examine whether a series of heat mitigation policies have effectively reduced heat exposure and whether they have truly benefited heat-vulnerable groups. Based on 20 years of multi-source heat exposure and urban spatial data, this paper provides evidence of ongoing enhancements to the heat exposure environment in the Charlotte city center. However, despite these improvements, heat vulnerable group that are particularly susceptible to the negative effects of heat exposure did not experience commensurate benefits. The conclusion of this article validates the ongoing trends of global sustainable studies in nature-based solutions and social-ecological systems, highlighting the issue of environmental equity evaluation.

  PublisherDOI

• Analysis of the Multi-Scale Spatial Heterogeneity of Factors Influencing the Electric Bike-Sharing Travel Demand in Small and Medium-Sized Cities

  Sustainability · 2025-11-21 · 1 citations

  articleOpen access

  The spatial heterogeneity of the electric bike-sharing (EBS) travel demand in small and medium-sized cities is influenced by a combination of the built environment, socio-economic gradients, transportation accessibility, and residents’ travel behavior patterns, and is significantly different from the shared travel characteristics of developed cities. In order to explore the influencing mechanisms of the EBS travel demand under different travel distance scales in small and medium-sized cities, this paper utilizes multi-source data from Tongxiang, Zhejiang Province, including operational data of EBS and built environment data. This paper analyzes the impact of the built environment on the EBS travel demand and its spatial heterogeneity across various distance scales from a local perspective. The results demonstrate that the fit of the multiscale geographically weighted regression (MGWR) model is superior to that of the geographically weighted regression (GWR) and the ordinary least squares (OLS) model. The explanatory variables exhibit significant spatial heterogeneity in their influence on the demand for EBS trips across different distance scenarios. The density of primary roads demonstrates a positive correlation with EBS travel demand in the western urban core area, but it is negatively correlated with travel demand in the eastern urban core area. Accommodation services show a negative correlation with long-distance EBS travel demand in the urban core area and the northern city, but they are positively correlated with short-distance EBS travel demand in the urban core area. There is competition between long-distance EBS and public transportation in city centers. However, short-distance EBS and public transportation exhibit a complementary relationship in the urban periphery. The research findings are beneficial for gaining a deeper understanding of the patterns of change in the EBS travel demand and promoting the refined and sustainable development of shared transportation.

  PublisherOA PDFDOI

• Converging GeoAI and CyberGIS for Human Geography

  Springer geography · 2025-01-01

  book-chapter1st authorCorresponding
  PublisherDOI

• Revitalizing Cities: The 5R Framework Approach to Urban Retrofitting and Big Data Insights

  Growth and Change · 2025-01-03 · 2 citations

  articleOpen access

  ABSTRACT Urban retrofitting is a fundamental approach for achieving sustainable and resilient urban development in the face of contemporary challenges. The increasing prevalence of urban big data presents an opportunity to establish a robust analytical framework for urban retrofitting, enabling more effective comparative studies and informed decision‐making. This paper introduces a comprehensive 5R framework—Re‐inhabitation, Re‐building, Re‐transportation, Re‐capitalization, and Re‐greening—to provide a multidimensional perspective on urban retrofitting. The 5R framework facilitates a holistic understanding of urban transformation processes and establishes standardized metrics for analyzing urban retrofitting initiatives using diverse urban big data sources. To demonstrate the adaptability and effectiveness of the 5R framework in a real‐world context, we conduct a case study of Charlotte, North Carolina. By applying innovative methods for the integration and analysis of extensive datasets, our study offers new insights into the evaluation of urban retrofitting efforts, such as transportation accessibility, micro‐scale building improvements, investment patterns, green space enhancements, and overall livability. This approach addresses existing research gaps by providing a structured set of indicators that assess each dimension of urban transformation comprehensively. Beyond academic advancements, the 5R framework offers practical tools for policymakers and urban planners to evaluate retrofitting interventions, quantify their outcomes, and understand the dynamics of evolving urban spaces. The insights gained through our research highlight the importance of using big data to enhance the scope and impact of urban development strategies, ultimately bridging the gap between theoretical concepts and real‐world urban retrofitting applications. Our findings demonstrate the potential of the 5R framework to serve as a guiding model for more sustainable, data‐driven urban growth and revitalization.

  PublisherOA PDFDOI

• A Video Machine Learning Framework for Spatiotemporal Analysis of Complex Urban Dynamics

  Transactions in GIS · 2025-08-01

  articleOpen access1st author

  ABSTRACT Urban dynamics is complex and interconnected across various social and environmental systems. To better understand such dynamics, this study proposes a scalable and flexible video machine learning framework for spatiotemporal analysis of urban dynamics. The framework is based on a space–time cube representation and decomposes the cube structure along the temporal dimension into a sequence of time‐series spatial aggregation, similar to a video. State‐of‐the‐art video machine learning models including ConvLSTM, predRNN, predRNN‐V2, and E3D‐LSTM are utilized for spatiotemporal modeling and prediction. The scalability of this cyberGIS‐enabled framework is shown by its applicability to diverse geographic regions, its ability to address various urban problems, and its capacity to integrate heterogeneous geospatial data. Moreover, the framework's flexibility is further enhanced by adjustable spatial and temporal granularity. The framework's effectiveness is validated through two case studies: (1) a real‐world urban heat analysis in Cook County, Illinois, USA in 2018, which achieved an RMSE of 0.60535°C, representing a 46% improvement over established benchmarks; and (2) a simulated dataset analysis demonstrating the framework's adaptability for spatial heterogeneity and temporal changes. A series of evaluations demonstrate the effectiveness of the proposed framework in spatiotemporal analysis of complex urban dynamics.

  PublisherOA PDFDOI

• From Individual Motivation to Geospatial Epidemiology: A Novel Approach Using Fuzzy Cognitive Maps and Agent-Based Modeling for Large-Scale Disease Spread

  Sustainability · 2024-06-13 · 5 citations

  articleOpen access

  In the past few years, there have been many studies addressing the simulation of COVID-19’s spatial transmission model of infectious disease in time. However, very few studies have focused on the effect of the epidemic environment variables in which an individual lives on the individual’s behavioral logic leading to changes in the overall epidemic transmission trend at larger scales. In this study, we applied Fuzzy Cognitive Maps (FCMs) to modeling individual behavioral logistics, combined with Agent-Based Modeling (ABM) to perform “Susceptible—Exposed—Infectious—Removed” (SEIR) simulation of the independent individual behavior affecting the overall trend change. Our objective was to simulate the spatiotemporal spread of diseases using the Bengaluru Urban District, India as a case study. The results show that the simulation results are highly consistent with the observed reality, in terms of trends, with a Root Mean Square Error (RMSE) value of 0.39. Notably, our approach reveals a subtle link between individual motivation and infection-recovery dynamics, highlighting how individual behavior can significantly impact broader patterns of transmission. These insights have potential implications for epidemiologic strategies and public health interventions, providing data-driven insights into behavioral impacts on epidemic spread. By integrating behavioral modeling with epidemic simulation, our study underscores the importance of considering individual and collective behavior in designing sustainable public health policies and interventions.

  PublisherOA PDFDOI

Frequent coauthors

• Shaowen Wang

  University of Illinois Urbana-Champaign

  18 shared

• Shaohua Wang

  14 shared

• Jeon‐Young Kang

  Kyung Hee University

  6 shared

• Jin-Woo Park

  Kyung Hee University

  5 shared

• Su Yeon Han

  5 shared

• Anand Padmanabhan

  5 shared

• Alexander Michels

  University of Illinois Urbana-Champaign

  5 shared

• Vincent L. Freeman

  University of Illinois Chicago

  3 shared

Labs

• Center for Geospatial Information TechnologyPI

  Not provided

Education

• PhD, Geography and GIS

  University of Illinois Urbana-Champaign

  2024

• MS, Geography and GIS

  University of Illinois Urbana-Champaign

  2021

• BE, Computer Engineering

  The University of Hong Kong

  2018