MOVING LOCAL INDICATORS OF SPATIAL ASSOCIATION FOR TIME GEOGRAPHY

Abstract

Time Geography provides an elegant theoretical framework for examining movement as information. The Time Geographic perspective conceptualizes observed movement as a signal encoding an emergent description of the mover. An actor’s movements are shown to capture latent behavioral responses to pressures in the environmental context and to the presence of other movers. Considered as a set of spatiotemporal relocations constrained by what is accessible or desirable to the mover, Time Geographic methods allow us to extract evidence of the mover’s evolving state and agency from trajectory data. This thinking has proven applicable in studies of animal movement, individual access in transportation systems, mover identity and the lived experience of a varied social space, psychiatric healthcare outcomes and tourism.Despite its established and fast-growing literature, Time Geographic innovation is focused on the action of discrete, thinking, living individuals. Displacement or relocation in the action of non-point, invisible or distributed moving phenomena is understudied in this literature. Time Geographic treatments for questions surrounding the movements of aerosol plumes, aggregate or censored disease and crime event counts, samples of invasive plant presences and neighborhood-level price responses to rent control policies are not, for example, available in the literature. This limitation is due in part to the discipline’s epistemological construction of the mover and in part to instrumentation. Individual movement captured by location-aware technologies readily translates to the common trajectory data structure facilitating Time Geographic analysis. However, corresponding routines characterizing the displacements of non-point, spatially indistinct or invisible phenomena as trajectory data are not available. Confronting the notion of a non-point or invisible mover raises immediate questions about their reliable delineation and the location of their representative space-time anchors or location fixes within the trajectory structure. Additionally, Time Geographic conceptualizations of the mover depend, at least in part, on a grasp of the mover’s agency, their behaviors, preferences and lifespans. An ontological bridge between the prevailing conception of an individual’s movement and the metric displacements of a distributed actor has eluded the attention and perhaps the imagination of Time Geographers. We offer an invitation to consider the ways such an entity might move. This dissertation contributes directly to the building of this ontology with the introduction of an adapter capable of producing space-time trajectories from change observed in area-aggregated spatiotemporal observations. Termed Moving Local Indicators of Spatial Association (Moving LISA), this novel probabilistic method avails a new class of subject matter to the existing traditions of Time Geographic analysis, to movement ecology and mobility science. Essentially, the Moving LISA constructs an entity-linking over space and time for the appearance and persistence of extreme observations, or locations clustering in a single variable. Trajectories are generated from the lineages of these clusters and returned to the user for further analysis and comparison. Chapter 2 expands on the research gap addressed by Moving LISA, along with its theoretical and mechanistic procedure as a sequencing and spatiotemporal entity-linking of cross-sectional LISA results. A prototypical formulation for the Moving LISA procedure is discussed, along with relevant pseudocode reproducing the method. The approach is further illustrated across both a contrived simulated dataset and empirical US Census data for the State of Oklahoma. Chapter 3 engages with Moving LISA results as a motivating visualization accompanying a series of corresponding, cross-sectional spatially autoregressive model results concerned with the proliferation of specialty vape stores in Oklahoma, from 2013-2023. This chapter represents Moving LISA as a companion to existing methods from spatiotemporal public health research, while simultaneously delivering the first known temporal examination of specialty vape store presence and socioeconomic vulnerability conducted for the state of Oklahoma. Over the course of the study period, a statewide dispersal in clusters of specialty vape store presence were observed alongside evolving associations between the expectation for vape store counts and socioeconomic covariates at the Zip Code level. Covariates capturing dimensions of social vulnerability (example, Households With SNAP Benefits Last 12 Months, Percent Latino of Any Race, and Percent 25-34 Years) were shown as significant positive effects on the expected count of specialty vape stores at the Zip Code level over the 3 study periods. Chapter 4 demonstrates the application of existing analytics from movement ecology to output Moving LISA trajectories directly, with an investigation of the emergent gentrification process in the Oklahoma City Metropolitan Statistical Area for 2009-2023. Recent theoretical and empirical developments from Urban Geography offer a parsimonious identifier for the onset of gentrification, where median home prices and incomes are reframed obtaining a signal locating newly gentrifying areas. Drawing from this development, I have analyzed the respective Moving LISA trajectories of the covariates as a dyad; an interacting pair of movers relevant for questions developed in movement ecology. Where respective time-series of median home price hotspots and median income cold spots were reconstructed as Moving LISA trajectories, I found confirmation in a set of theoretical expectations for the temporal character of the gentrification process. With the application of trajectory comparison metrics from movement ecology, attraction or cohesion in dyadic movement among aggregate movers representing high home prices and high median incomes was observed. A corresponding repulsion between high home prices and low median incomes was also identified. Further, in concert with a theoretical expectation where these trends would intensify within urban areas, non-parametric tests for difference in pseudomedian location in dyadic movement metric results showed significant intensification of these respective attractive and repulsive movements, for trajectories found inside versus outside of the Oklahoma City limits. Chapter 4 represents the first direct analysis of Moving LISA trajectories and suggests for the procedure’s utility in further analyzing the movements of distributed spatial phenomena. This dissertation is directed in large part to expanding the set of phenomena subject to Time Geographic analysis. The notion of action in a non-point or invisible spatial process and its construction as a moving object forms a nascent literature within Time Geographic thought, dealing foremost with the method gap for its proper characterization. The Moving LISA functionally bridges this epistemological and methodological gap and further avails new research questions to examination in the Time Geographic tradition. This dissertation is meant as a primer for a growing branch of geographic inquiry. It provides a theoretical narrative, tooling, and a few example applications. Future research would call first for a protocol assisting the incorporation of domain knowledge into Moving LISA applications. Forthcoming studies might directly address the method’s sensitivities, explain caveats and the quasi-agency and pseudo-biology of aggregate movers in terms of impact to analysis as fully qualified Time Geographic objects.