This paper assesses trending AI foundation models, especially emerging computer vision foundation models and their performance in natural landscape feature segmentation. While the term foundation model has quickly garnered interest from the geospatial domain, its definition remains vague. Hence, this paper will first introduce AI foundation models and their defining characteristics. Built upon the tremendous success achieved by Large Language Models (LLMs) as the foundation models for language tasks, this paper discusses the challenges of building foundation models for geospatial artificial intelligence (GeoAI) vision tasks. To evaluate the performance of large AI vision models, especially Meta’s Segment Anything Model (SAM), we implemented different instance segmentation pipelines that minimize the changes to SAM to leverage its power as a foundation model. A series of prompt strategies were developed to test SAM’s performance regarding its theoretical upper bound of predictive accuracy, zero-shot performance, and domain adaptability through fine-tuning. The analysis used two permafrost feature datasets, ice-wedge polygons and retrogressive thaw slumps because (1) these landform features are more challenging to segment than man-made features due to their complicated formation mechanisms, diverse forms, and vague boundaries; (2) their presence and changes are important indicators for Arctic warming and climate change. The results show that although promising, SAM still has room for improvement to support AI-augmented terrain mapping. The spatial and domain generalizability of this finding is further validated using a more general dataset EuroCrops for agricultural field mapping. Finally, we discuss future research directions that strengthen SAM’s applicability in challenging geospatial domains.

We introduce critical reflection mapping as a novel and hybrid research methodology for examining the sociospatial perceptions of researchers in new research settings, particularly international ones. The methodology, theoretically situated within the critical geographic information systems literature, combines two existing research methods (qualitative sketch mapping and critical reflection) to elicit original ways in which researchers can critically reflect on an area new to them while spatially linking these qualitative place-based reflections to sketch maps. The methodology allows for synergistic data sets to inform each other and to be analyzed together rather than separately. Through critical reflection mapping, we demonstrate how multiple data sets and methods are combined so that critical reflection and word clouds add significant intellectual value by making another layer of textual information immediately accessible to qualitative sketch mapping data analysis. We present two case studies in Belize and Panama from our current community geography research agendas to demonstrate the viability as well as the caveats of this novel methodology for understanding and representing the immediate sociospatial perceptions of researchers. In the context of international research experiences discussed in this article, the methodology captures individual responses to features of the built environment including walkability and sustainability; documents the changing emotions a newly immersed researcher has in a largely unfamiliar geographic setting; and connects new experiences in a foreign research setting to an individual's everyday lived experiences, positionality, and multiple identities. It also makes these experiences more visible to fellow researchers in a large research team and thus lends itself as a potential forum for shared reflection.

COVID-19 dashboards with geospatial data visualization have become ubiquitous. There is a growing sense of responsibility to report public health data pushing governments and community organizations to develop and share web-based dashboards. While a substantial body of literature exists on how these GIS technologies and urban analytics approaches support COVID-19 monitoring, their level of social embeddedness, quality and accessibility of user interface, and overall decision-making capabilities has not been rigorously assessed. In this paper, we survey 68 public web-based COVID-19 dashboards using a nominal group technique to find that most dashboards report a wealth of epidemiologic data at the state and county levels. However, these dashboards have limited emphasis on providing granular data (city and neighborhood level) broken down by population sub-groups. We found severe inadequacy in reporting social, behavioral, and economic indicators that shape the trajectory of the pandemic and vice versa. Our survey reveals that most COVID-19 dashboards ignore the provision of metadata, data download options, and narratives around visualizations explaining the data’s background, source, and purpose. Based on these lessons, we illustrate an empirical experiment of building a dashboard prototype—the COVID-19 Economic Resilience Dashboard in Arizona. Our dashboard project demonstrates a model that can inform decision-making (beyond plain information sharing) while being accessible by design. To achieve this, we provide localized data, drill-down options by geography and sub-population, visualization narratives, open access to the data source, and accessible features on the interface. We exhibited the value of linking pandemic-related information with socioeconomic data. Our findings suggest a pathway forward for researchers and governments to incorporate more action-oriented data and easy-to-use interfaces as they refine existing and develop new information systems and data analytics dashboards.

This study aims to examine the spatially varying relationships between social vulnerability factors and COVID-19 cases and deaths in the contiguous United States. County-level COVID-19 data and the Centers for Disease Control and Prevention social vulnerability index (SVI) dataset were analyzed using local Spearman’s rank correlation coefficient. Results suggested that SVI and four social vulnerability themes have spatially varying relationships with COVID-19 cases and deaths, which means spatial heterogeneity is an essential factor that influences the relationship, and the strength of association varies significantly across counties. County hot spots that were subject to all four social vulnerability themes during the pandemic were also identified. Local communities and health authorities should pay immediate attention to the most influential social vulnerability factors that are dominant in their region and incorporate measures tailored to the specific groups of people who are under the greatest risk of being affected during the COVID-19 pandemic.

Patterns of racial, ethnic, and gender participation in higher education are diverse and vary spatially, yet they fail to fully reflect the changing demographic landscape of the United States. This article explores two key questions: (1) How does diversity within geography compare to diversity within U.S. higher education more broadly? (2) How does the participation of underrepresented minorities and women in geography programs differ across regions and settings? We conclude that departmental strategies and goals for enhancing diversity should be sensitive to institutional and geographic contexts that shape individual programs’ unique opportunities and constraints.