Geographically Weighted Spatial Interaction (GWSI) Modelling
Project lead: Maryam KordiDealing with spatial heterogeneity remains an ongoing problem in spatial interaction modelling. It has been shown that local calibration of spatial models yields useful information about the actual situation in any particular part of the study region. This is in contrast to global models, where the spatial variation in the relationships is assumed to be constant. The aim of this research is to develop a framework for localising spatial interaction models in which the spatial heterogeneity can be detected and visualised over space. As a possible approach, a set of localised spatial interaction models is developed using the concept of geographical weighting. A Journey-to-work dataset is used as a case study to demonstrate the applicability of the localised geographically weighted spatial interaction models.
Flexible-bandwidth Geographically Weighted Regression (FGWR)
Project Lead: Wenbai YangGeographically weighted regression (GWR) has been widely used to model spatially varying processes in the fields of human geography, ecology, social science, and so on. The essential idea of GWR is that observations near to a model calibration point have more influence in the estimation of regression coefficients than observations farther away do. The standard GWR model employs a single bandwidth to control the distance-decay in this influence. In practice, such a uniform bandwidth may not be sufficient in reflecting complex spatial variations in relationships between variables. In an attempt to produce a more realistic model to reflect a multivariate process that may vary across different spatial scales, an extension to standard GWR is developed, allowing different bandwidths to be defined for different coefficients. In this way, different scales of nonstationarity in data relationships are addressed in the model. Experiments suggest that this flexible-bandwidth GWR can provide an improvement over standard GWR. It can act as a generalisation to simpler models including global linear regression, standard GWR and semi-parametric GWR.
Enhancing Volunteered Geographical Information (VGI)
Project Lead: Katarzyna Sila-NowickaCollecting, representing, and understanding human mobility patterns is becoming increasingly important for research, public policy, and private sector businesses. As we continue to generate mass amounts of spatially referenced data at increasingly fine temporal and spatial resolutions, an explicit focus on GPS traces, personal paths, human mobility-based behaviour, spatial interaction, and the various global/local scales of these data is essential. The main aim of this research will be modelling the determinants of spatial interaction based on GPS data. Additionally this research will be used to explore ways of adding value to GPS traces in terms of inferring various trip-making behaviours via spatial and analytical methods. This work is part of GEOCROWD’s Initial Training Network, under FP7 – People – Marie Curie Actions by the European Commission: “Creating a Geospatial Knowledge World”. The specific contributions of this project to the larger program will include transforming VGI data into meaningful chunks of information obtained with simplicity and speed, comparable to that of Web-based search.
An Evaluation of Visualisations for Geographically Weighted Statistical Methods
Project Lead: Tommy BurkeGeographically Weighted Spatial Statistical Methods are employed in a wide range of disciplines to analyse and interpret data where they are used to detect significant patterns or relationships. One of these methods, Geographically Weighted Regression (GWR), is used to examine processes that vary over space. There is little variation in the types of visualisations which are used to analyse the results of GWR. 2D Univariate maps, statistical summary tables and graphs of residual values are primarily used. Consequently, it is unclear whether other visualisation methods could be more effective for displaying the results. The research we are conducting focuses on evaluating different visualisation techniques for geographically weighted spatial statistical methods. We explored different visualisation techniques through user trials to ascertain their effectiveness for a given set of tasks. The goal of this experiment was to discover the most appropriate way to facilitate interpretation and analysis of Geographically Weighted Regression. Now, further research will be carried out based on the results of this experiment, including work on the effects of geographic scalability of geographic datasets.
