While eXplainable Artificial Intelligence (XAI) has significant potential to glassbox Deep Learning, there are challenges in applying it in the domain of Geospatial Artificial Intelligence (GeoAI). A land use case study highlights these challenges, which include the difficulty of selecting reference data/models, the shortcomings of gradients to serve as explanation, the limited semantics and knowledge scope in the explanation process of GeoAI, and underlying GeoAI processes that are not amenable to XAI. We conclude with possibilities to achieve Geographical XAI (GeoXAI).
Geodemographic classifications are an important tool to support public-service decision making. While people are the focal point of geodemographics, classifications are often built on variables that describe populations rather than individuals. Synthetic populations, model-based approximations of the individual makeup of small census areas, remain largely unused for geodemographic classification, yet they can provide a more direct and holistic understanding of localized resource needs than existing approaches. This paper develops a new method for performing individual-centered geodemographic classifications using synthetic populations. The building blocks of this approach are abstractions of the synthetic population attributed to each small census area via affinity matrices computed from similarities in both the size and attributes among individuals. Using a rank-1 spectral decomposition of an area’s affinity matrix enables rapid computation of a dissimilarity metric which is compatible with cluster analysis techniques used in traditional geodemographic classifications. Using data from the American Community Survey (ACS), an example classification is developed for the Knoxville, TN, USA Public-Use Microdata Area (PUMA) to illustrate how distinctions can be drawn among small census areas in terms of specific types of representative individuals, providing a more tailored view of the groups that serve to benefit from spatial policy interventions. Beyond improving traditional public-domain geodemographic classifications, this approach provides a novel open-source alternative to commercial neighborhood segmentation products with added flexibility for custom research applications.
Public transportation in cities is less popular than the private car due to lower personal flexibility, perceived comfort or the unavailability of infrastructure. The latter one is an issue in Augsburg with regard to outer districts since the existing star-shaped network layout requires a route through the inner city. A recent proposal called "Verkehr4.0" aims to extend the layout of the existing infrastructure by adding new express bus lines to connect outer city districts. This research paper investigates the direct traffic flow between the outer districts Stadtbergen and Göggingen in contrast to the existing flow via the central hub "Königsplatz". We implement an agent-based simulation comparing waiting times, travel times and total times spent on trips in the two scenarios. Furthermore, we model a measure dubbed "happiness" of the people as well as their willingness to change their mode of transport. The preliminary results of our simulation show that waiting time for public transport users decreases, while total time, travel time and happiness reveal no statistical difference through the introduction of an express bus line.
Virtual reality technology has the potential to be a revolutionary addition to the field of Geographic Information Science. The application of virtual reality to GIScience has been discussed for decades, however adoption has been limited until recently. Virtual reality GIScience represents an interdiscip- linary approach, incorporating fields such as video game development. In this paper, we introduce Locative Reality, a virtual reality software that presents users with immersive 360° video experiences of forest environments. It incorporates spatial information into the virtual environment so that data generated by virtual research can be directly linked to real-world locations. The implications for the field of GIScience include virtual research tools and educational experiences, accessible to anyone anywhere in virtual reality.
Data related to households or addresses needs be published in an aggregated form to obfuscate sensitive information about individuals. Usually, the data is aggregated to the level of existing administrative zones, but these often do not correspond to formal models of privacy or a desired level of anonymity. Therefore, automatic privacy-preserving spatial clustering methods are needed. To address this need, we present algorithms to partition a given set of locations into k-anonymous clusters, meaning that each cluster contains at least k locations. We assume that the locations are given as a set T ⊆ V of terminals in a weighted graph G = (V, E) representing a road network. Our approach is to compute a forest in G, i.e., a set of trees, each of which corresponds to a cluster. We ensure the k-anonymity of the clusters by constraining the trees to span at leastterminals each (plus an arbitrary number of non-terminal nodes called Steiner nodes). By minimizing the total edge weight of the forest, we ensure that the clusters reflect the proximity among the locations. Although the problem is NP-hard, we were able to solve instances of several hundreds of terminals using integer linear programming. Moreover, we present an efficient approximation algorithm and show that it can be used to process large and fine-grained data sets.
Landmarks serve to structure the environment we experience, and therefore they are also critically important for our everyday movement through and knowledge acquisition about space. How to effectively visualize landmarks to support spatial learning during map-assisted pedestrian navigation is still an open question. We thus set out to assess how landmark visualization styles (i.e., abstract 2D vs. realistic 3D) influence map-assisted spatial learning of expert wayfinders in an outdoor navigation study. Below we report on how the visualization of landmarks on mobile maps might influence wayfinder’s gaze behavior while trying to find a set of landmarks along a given route in an unfamiliar environment. We find that navigators assisted with mobile maps showing realistic-looking 3D landmarks more equally share their visual attention on task-relevant information, while those assisted with maps containing abstract 2D landmarks frequently switch their visual attention between the visualized landmarks and the mobile map to complete the navigation task. The presented analysis approach for the assessment of wayfinder’s gaze patterns has the potential to contribute ecologically valid insights for the understanding of human visual attention allocation during outdoor navigation, and to further understand how landmark depiction styles on mobile maps might guide wayfinders’ visual attention back to the environment to support spatial learning during map-assisted navigation.
Polycentricity is a critical characteristic of the spatial organization of cities. Many indices have been proposed to measure the degree of morphological polycentricity or functional polycentricity. However, selecting a proper set of polycentricity indices for cities in a particular region or country still needs prior expert knowledge. This study demonstrates that whole graph embedding, as a novel and efficient computational tool, can model the city polycentricity in an integrated manner without much prior knowledge. The new method can further support visual analytics and classification very well.
Spatio-racial distributions in major US cities change on the timescale of a single decade. Here we describe a methodology to forecast such changes a decade ahead. First, we transform the data from population counts to a grid of categorical population types. Then, we build an empirical model of past change using supervised machine learning and extrapolate it into the future to make a prediction. The model uses only statistics of population categories as features, there are no ancillary variables. To account for the non-stationarity of the change we use a synthetic training dataset based on past transitions and estimated future frequencies of these transitions. The methodology is described and validated by training a model on 1990-2000 data and using it to predict spatio-racial distributions in 2010. This prediction is then compared to the actual spatio-racial 2010 distribution. We have found that a highly accurate model of change can be constructed using this methodology. Extrapolating such models to the future results in some loss of accuracy, but the method still yields satisfactory predictions.
This study builds a bridge between the literatures for geographically weighted regression (GWR) and compositional data analysis (CoDA). GWR allows the modeling of spatial heterogeneity in regression models and is increasingly used in various fields. CoDA provides unique and useful tools for compositional data, which are restricted by a constant-sum constraint. Although compositional data are common in many scientific areas, it is not until recently that increasingly sophisticated statistical methods have been deeply investigated. Many types of spatial models based on geostatistics, spatial statistics, and spatial econometrics for compositional data have been proposed. However, there is less attention to both spatial heterogeneity and the constant-sum constraint. In this study, we propose a GWR model for compositional data. This allows us to model spatially varying relationships while considering the constant-sum constraint. We applied this model to analyze household income compositions at the county level in the US. The interpretational usefulness of the results of spatially varying compositional semi-elasticities is empirically performed.
Even if we are not aware, our emotions can influence and interplay with our navigation and use of mobile navigation aids. A given map display can make us feel good by reminding us of pleasant past experiences, or it can make us feel frustrated because we are not able to understand the information provided. Navigation aids could also make a given landmark emotionally charged, and thus more salient and memorable for a navigator, for example, by using an auditory narrative containing emotional cues. By storytelling, it would also be possible to provide details about a given landmark and connect proximal landmarks to each other. But how do navigational instructions in the form of emotional storytelling affect spatial memory and map use? Results from a preliminary study indicated that a video presentation viewed from a first person perspective is looked at more often than an abstract map, and this evidence becomes even stronger when instructions are emotionally laden. We discuss results in the context of place meaning and how emotions’ role in navigation should be further assessed, in particular to increase spatial learning from navigation aids.
