Grand challenges for the spatial information community

Leye Wang, Ouri WolfsonJ. Spatial Inf. Sci 2020 · DOI

Abstract

The spatial information (SI) community has an opportunity to address major societal and scientific problems including public health, climate change, air pollution, transportation, and others. Beyond the significant contributions made by the SI community, more can be done by focusing the efforts of the community and generalizing them. Focus can be achieved by an IMAGENET-like spatial information database and competition. Generalization can be achieved by solving spatio-temporal information problems in disciplines such as neuroscience, chemistry, biology, astronomy, and engineering.

Introduction

The spatial information (SI) community currently faces a paradox: while data-driven spatial studies are booming, the field remains fragmented by incomparable datasets and a narrow focus on traditional geography. To address global crises such as climate change and pandemics, the SI community must evolve. This paper proposes a dual-track strategy: adopting a standardized benchmarking culture similar to Computer Vision’s ImageNet and generalizing spatio-temporal tools to "non-geographic" domains like neuroscience and molecular biology.

The Problem: Incomparability and Domain Silos

Despite the proliferation of spatial data, the SI community faces two major hurdles:

  1. Lack of Standardized Benchmarking: Unlike Computer Vision (CV), where the ImageNet dataset provides a "gold standard" for testing algorithms, SI researchers often utilize disparate datasets with varying objectives. This lack of standardization makes it nearly impossible to objectively determine if a new method offers a genuine improvement over existing ones without exhaustive re-implementation.
  2. Narrow Disciplinary Focus: SI techniques, such as trajectory mining or spatial indexing, are traditionally confined to maps and GPS data. However, many complex systems—from the human brain to the expanding universe—are fundamentally spatio-temporal. By remaining in traditional silos, the SI community misses opportunities to lead in these adjacent scientific fields.

Strategic Roadmap: Focus and Generalization

The authors propose a strategic roadmap centered on two pillars:

  • The "Focus" Strategy: Rather than a single "one-size-fits-all" dataset, the authors suggest a distributed, theme-based competition framework. Annual competitions, potentially seeded by organizations like ACMGIS, could focus on specific societal challenges like contact tracing for public health or congestion modeling for autonomous vehicles.
  • The "Generalization" Strategy: The authors advocate for "SI+X" interdisciplinary research. This involves applying spatial coordinate systems (e.g., MNI or Talairach in brain mapping) and temporal signal processing to fields where the "map" is not the Earth's surface, but a molecule or a neural network.

The Data Landscape

The foundation for this vision rests on existing and future data infrastructure:

  • Existing Repositories: OpenStreetMap, NYC Taxi & Limousine Commission (TLC) trip records, and the NYC Open Data Portal provide valuable starting points.
  • Niche Benchmarks: Projects like Movebank (animal tracking) and GeoLife (human trajectories) have already demonstrated the value of specialized datasets.
  • The Proposed Goal: A "spatial information warehouse" that integrates diverse data types—including satellite imagery, trajectories, and geo-tagged social media—into a unified model for cross-disciplinary research.

Key Insights: Beyond Geography

The paper highlights transformative applications of SI tools in other fields:

  • Neuroscience as Spatial Data: The human brain consists of 100 billion neurons, each with specific 3D coordinates. Treating brain activity as a spatio-temporal data problem could unlock mysteries of consciousness and improve Artificial Neural Networks.
  • Molecular Dynamics as Trajectory Mining: Protein folding is essentially a spatio-temporal process. The same tools used to track delivery fleets can be adapted to model the "trajectories" of atoms within a folding protein.
  • The Complexity of "ImageNet for SI": SI data is inherently more heterogeneous than the pixel-based data of Computer Vision, involving points, polygons, rasters, and graphs. Thus, a spatial benchmark must be a more complex, multi-modal warehouse.

Current Progress and Societal Impact

While primarily a vision paper, it acknowledges the community's successful foundations:

  • GeoAI: Geospatial Artificial Intelligence has already revolutionized place-name extraction and satellite image processing.
  • Infrastructure: Systems like SpatialHadoop and JUST have demonstrated the ability to handle large-scale spatial data.
  • Societal Value: The urgency is underscored by the fact that air pollution exposure caused approximately 7 million deaths in 2012. Improved SI research is not just about academic metrics; it is about saving lives.

Challenges and Future Directions

Significant barriers remain for the community to overcome:

  • Data Complexity: Designing a general-purpose database that serves researchers across diverse sub-fields (e.g., geo-tagged tweets vs. satellite hydrology) is a major technical challenge.
  • Management: A distributed competition model requires sustained, multi-year effort and coordination from international organizations.
  • Interdisciplinary Barriers: Generalizing SI to fields like astronomy or chemistry requires researchers to bridge significant gaps in vocabulary and coordinate systems.

Conclusion

By focusing on standardized benchmarking and generalizing its methodologies to new domains, the spatial information community can transition from a niche discipline to a central pillar of modern scientific inquiry, addressing the most pressing challenges of our time.

Summary updated

BibTeX
@article{wang2020grand,
  author = {Leye Wang and Ouri Wolfson},
  title = {{Grand challenges for the spatial information community}},
  journal = {J. Spatial Inf. Sci.},
  year = {2020},
  doi = {10.5311/josis.2019.20.667},
}
Figure