ABSTRACT This study proposes a set of pedagogical principles for implementing thick mapping in spatial design studios, based on insights from three workshops involving 166 students. Grounded in collective intelligence and framed within an action research methodology, it examines the impact of thick mapping strategies, supported by geospatial and bio-physiological tools, on co-creation processes and spatial understanding in complex urban environments. The research also evaluates the usability of the Field Maps mobile application, comparing open and geofenced modes for spatial data collection, effectiveness, and user engagement. Findings indicate that thick mapping fosters situated, multimodal, and narrative-driven engagement with complex urban environments, while improving students’ contextual insights and soft skills. Ultimately, the study positions thick mapping as an adaptive, creative, and critical methodology in spatial design education, with the integration of bio-physiological data emerging as a particularly promising approach to enriching site analysis.

ABSTRACT Automated map generation, especially generating maps from natural language descriptions, not only democratizes access to geographic data but also strengthens decision-making, improves communication and allows for customization. However, map generation faces some challenges, such as lowering the professional threshold, improving generation quality, and ensuring geographic consistency. In recent years, large generative models (e.g. text-to-image models) have excelled in the field of image generation. However, since these models are primarily trained on natural image data, they exhibit significant gaps when generating map data with unique layouts and symbols designed. To address this, we propose a text-to-map generation framework based on diffusion model, called MapGenerator. By using a strategy that combines self-instruct and expert refinement, we construct the training dataset MGTrain and the evaluation dataset MGEval, containing 1000 and 100 pairs of maps and their corresponding detailed descriptions, respectively. Based on the training data, we employ a Parameter-Efficient Fine-Tuning (PEFT) strategy to fine-tune a pre-trained general text-to-image model, enhancing its performance in map generation tasks. Experimental results show that MapGenerator achieves the best FID and CLIP Score among all models, and expert evaluations confirm its superior ability to accurately capture geographic objects and spatial relationships described in the text. The study confirms the feasibility and effectiveness of the diffusion model-based text-to-map generation approach, offering new solutions and technical support for text-driven geographic information generation.

ABSTRACT The labeling of point features on a map is a well-studied topic. In a static setting, the goal is to find a non-overlapping label placement for (a subset of) point features. In a dynamic setting, the set of point features and their corresponding labels change, and the labeling has to adapt to such changes. To aid the user in tracking these changes, we can use morphs, here called transitions, to indicate how a labeling changes. Such transitions have not gained much attention yet, and we investigate different types of transitions for labelings of points, most notably consecutive transitions and simultaneous transitions. We give (tight) upper bounds on the number of overlaps that can occur during these transitions. When each label has a non-negative weight associated to it, and each overlap imposes a penalty proportional to the weight of the overlapping labels, we show that it is NP -complete to decide whether the penalty during a simultaneous transition has weight at most k . Finally, we consider geotagged data on a map, by labeling points with rectangular or square labels. We developed a prototype implementation to evaluate different transition styles in practice, measuring both number of overlaps and transition duration.

ABSTRACT Artificial intelligence (AI) is widely used in geographic information science (GIS) and has spawned a new research direction, Geospatial Artificial Intelligence (GeoAI). While there is a growing awareness of potential negative impacts of AI and increased research on AI ethics, systematic ethics research on GeoAI is still lacking and needs the integration of quantitative analytical capabilities. However, existing research on GeoAI ethics lacks a methodological approach for deriving quantitative analytical results, which in turn limits the ability to obtain conclusions with numerical significance. To address this challenge, we collect news data to obtain raw cases of GeoAI ethics. Subsequently, an exploratory framing serves as the guideline for data coding, resulting in a recoded and quantifiable dataset, which is intended for conducting an exploratory analysis of the themes and content of GeoAI ethical issues as they appear in news data. By integrating the designed icon system, we conduct a visual analysis of the dataset. Based on the findings, an infographic framework for GeoAI ethics is proposed to strengthen the structured development of ethical GeoAI. The exploratory framing, together with the infographic framework, contributes to the systematic identification of ethical issues and the formulation of policy guidelines within the GeoAI community.

ABSTRACT The study focuses on the pedagogical applications of literary maps and the didactic orientation within the broader field of literary geography. The research question is, “What functionality should a digital map offer in educational settings to facilitate the exploration of place names from literary texts?” To answer the question, we developed an experimental tool to investigate place names mentioned in Norwegian literature 1814–1905, and arranged a session with literature students and a session with Norwegian teachers to explore user needs and the tool’s pedagogical potentials in literary geographies. Our results show that both the literature students and the teachers considered the tool inspiring and triggering their curiosity. Our study confirms suitability of design choices, with the main map panel accompanied with frequency panel, and equipped with a moderate level of interactive functions that support building corpora, plotting extracted place names on an interactive map and listing them in a supplementary overview panel. Moreover, while the criticism received primarily concerned failed geolocations and slow system responses, suggestions for further development mainly focused on the search mechanism for place names in specific locations and on providing tutoring prior to the use of the map in upper secondary education.

ABSTRACT Open geographic information science (GIScience) aspires to make new forms of research and discovery possible by facilitating the reproduction, replication, reanalysis, and extension of prior studies. We used open science practices to conduct a series of open GIScience studies stemming from a single study of spatial accessibility to COVID-19 healthcare in Illinois. We conducted the studies with students while establishing a reusable model for open GIScience research practices, including an executable research compendium for use with Git version control and a public CyberGIS system. Contradicting the perceived burdens and low value of reproduction and replication studies, we used open GIScience to improve research quality and support discovery. We tested the ability to repeat the study methods replicating it with different data in Connecticut and reproducing it with the same data in Illinois and Chicago. We successfully repeated the study with modifications to improve reproducibility, manage large file sizes, and accommodate changes in computational environments and volunteered geographic data. We then reanalyzed the study to improve computational efficiency and improve validity in terms of data errors, missing data, edge effects, and boundary effects. Finally, we extended the study to research spatiotemporal accessibility of pharmacies in Vermont, finding that the lack of after-hours access exacerbates and exceeds urban-rural access disparities. The cascading improvements observable across our studies underscore the importance of reproducibility as a catalyst for research quality and innovation in GIScience, particularly in the context of unique phenomena and research challenges across geographic contexts.

ABSTRACT Generative artificial intelligence (AI) holds significant promise for cartography, but its application is hindered by a fundamental tension between creative freedom and rule-based design. This study addresses this challenge by proposing and evaluating a model-agnostic “template-render” framework that decouples conceptual design (“template”) from visual synthesis (“render”), creating a more controllable workflow. We implement this with a two-step approach: a Large Language Model (LLM) generates a structured symbol description, which a Text-to-Image (T2I) model then renders. Our baseline evaluation demonstrates that while the unguided approach is technically feasible, its outputs are often cartographically unsuitable. We then show that by introducing a knowledge-guided prompt to the template stage, the quality, clarity, and fitness of the symbols are significantly improved. We further present the Map Symbol Agent (MSA), a prototype that automates this pipeline. Our work validates the effectiveness of this framework, while also systematically identifying critical future challenges, such as ensuring stylistic consistency and mitigating model biases. This study serves as a crucial exploratory step, charting a promising path and defining a research agenda for developing more controllable generative systems in specialized domains.

ABSTRACT Historical map georeferencing, especially when dealing with maps that exhibit high levels of local distortion, remains a time-consuming process. This paper introduces a pipeline that automates much of this process by transferring georeferencing information from georeferenced maps (Anchor) to maps lacking georeferencing (Target). At its core, the method employs deep-learning algorithms for image registration (SuperPoint and SuperGlue) alongside tailored modules to exclude outliers and enhance match density. Specifically, RANSAC is combined with a Delaunay-based procedure to discard erroneous matches and preserve consistent spatial relationships. To address the reduction in keypoints following outlier emoval, we incorporate a patch-based local image registration, enabling multiscale matching. After a final outlier-removal step, the resulting high-quality matches are used to assign real-world coordinates to the Target map. We evaluated the pipeline on 86 georeferenced historical maps of Jerusalem and obtained a root mean square error (RMSE) below 1% of the map diagonal for 71 of them. Moreover, the final georeferencing accuracy was closely tied to the number of matching keypoints, with a threshold of 100 serving as a strong indicator of reliable results. Extending the pipeline to an additional 113 non-georeferenced maps, we found that 86 were successfully georeferenced based on this keypoint threshold.

ABSTRACT Understanding the distinct mobility patterns of expectant parents is critical for public health and urban planning. However, research has traditionally overlooked fathers, and studies using digital trace data often fail to account for residential relocations and multi-destination trips. This study addresses these gaps by analyzing 434,180 geotagged tweets from a U.S. cohort of 416 mothers and 511 fathers across pre-pregnancy, pregnancy, and postpartum phases. Using an enhanced methodology that captures residential moves and complex travel, we analyzed three facets of mobility: migration, long-distance travel, and daily activities. The results show that the majority of migrations occurred during the second and third trimesters, followed by the postpartum period. Mothers significantly reduced long-distance and international travel during pregnancy and further in the postpartum period, whereas fathers did so only after birth. However, fathers demonstrated more substantial adjustments to their daily activities during the pregnancy. These parental mobility and activity patterns were also significantly associated with socioeconomic status. The findings offer a more holistic understanding of the journey to parenthood and underscore the importance of engaging both parents in perinatal health and planning initiatives. The enhanced mobility measures developed for this study also present a valuable new tool for broader human mobility research.

ABSTRACT Interactivity and narration are valuable strategies for enhancing cognitive load management in map animations. When integrated into a web mapping application, these elements provide a mixed approach, allowing for user and author control. However, users may need guidance when exploring large temporal datasets and animations. Current web mapping applications utilizing animations and interactivity often fail to provide users with a guided exploration experience, resulting in a lack of linearity. To fill this gap, in this paper we present an application that includes animations that depend on the user, which is guided to interact with a generalized time-aware layer representing active fires of 2023. The application is structured into successive animations, developed with the ArcGIS Maps SDK for JavaScript to ensure a structure based on successive animations and a well-defined resolution. Guidance is presented through text blocks positioned outside a 3D globe, while the narrative is conveyed through text blocks integrated on the globe’s surface. Thus, we constructed an application combining fragmentation, guidance, and interactivity to reduce the disadvantages of map animations created for storytelling. The suggested approach offers an alternative solution to non-interactive animations that overstimulate users with excessive information or for applications that prioritize only user control and exploration.