Alexander Stewart Fotheringham

In his early career years after obtaining his PhD in 1980, he worked as a professor at University at Buffalo, becoming a full professor in 1988.[1] From 1991 to 1992, he held the position of Professor of Quantitative Geography at the University of Newcastle, setting the stage for his future endeavors.[9] From 1993 to 1994, Fotheringham worked as an Assistant Chair in the Department of Geography at the State University of New York.[9]

In 1994, he returned to the University of Newcastle as a professor of Quantitative Geography. Notably, during this time, he also took on the role of Director of the North-East Regional Research Laboratory, where his leadership played a pivotal role in advancing research efforts in the region. He remained in this position until 2004.[9]

Fotheringham expanded his horizons in 2004 by becoming a Visiting Research Fellow at the University of Leeds, where he remained until 2006.[9] Simultaneously, from 2004 to 2011, he assumed the SFI Research Professor and Director role at the National University of Ireland.[9]

Between 2011 and 2014, Fotheringham served as the Director of the Centre for GeoInformatics and held the Professor of Quantitative Geography position at the University of St Andrews.[9]

In 2014, Fotheringham began his tenure as a professor of Computational Spatial Science at Arizona State University, where he continues to make profound contributions to the field and influence the next generation of researchers.[9]

In addition to his academic roles, Fotheringham has received various prestigious recognitions. From 1995 to 1998, he was elected as the Chair of the Quantitative Methods Study Group of the Royal Geographical Society.[9] In 2009, he was appointed as Ireland's representative on the Governance Committee of the EU Joint Planning Initiative on Urban Europe, giving him an active involvement in shaping urban planning initiatives.[9]

In 2014, Fotheringham was selected as a member of the National Academy of Sciences’ Mapping Science Committee.[10][11] This committee seeks to organize research and inform on methods to use spatial data ethically to inform policy and benefit society.[10]

Fotheringham's most significant contribution to GIScience and spatial statistics may be his work in developing Geographically Weighted Regression (GWR).[4] GWR was first developed as a statistical technique in the 1990s by Fotheringham, Chris Brundson, and Martin Charloton.[5][20][21] Fotheringham has continued to be involved in researching expanding upon GWR, and its applications, in the years since.[21]

GWR is designed to address the limitations of traditional global regression models, such as Ordinary Least Squares (OLS), which assume that relationships between variables are global; that is, constant across space. GWR recognizes that relationships between variables are non-stationary; that is, they can vary from one location to another within a geographic area. In this way, it is comparable to other local spatial statistics, such as the univariate Getis-Ord Gi* and Local Moran's I, as compared to their global spatial statistic equivalents Getis-Ord General G and Global Moran's I.

In GWR, regression coefficients (parameters) are estimated locally for each geographic location or point, allowing for the modeling of spatial heterogeneity.[5] This means that GWR considers the spatial context of data and generates separate regression models for different locations, considering the varying relationships between dependent and independent variables. It provides insights into how these relationships change as you move across a geographic area, making it valuable for understanding spatial patterns and identifying areas with unique relationships.[5]

Geographically Weighted Regression is a cornerstone of GIS and spatial analysis, and is built into ArcGIS, as a package for the R (programming language), and as a plugin for QGIS.[22][23][24]

Time is recognized as significant to spatial analysis, with a substantial amount of literature within the discipline of time geography.[25][25] However, incorporating both space and time is a significant challenge for researchers. Fotheringham worked to address this problem in his 2015 paper titled "Geographical and Temporal Weighted Regression (GTWR)."[25] GTWR builds upon GWR by incorporating the dimension of time into the analysis.[25] This is accomplished by deriving both spatial and temporal bandwidths and using them to construct a weighted matrix.[25] GTWR is available as packages in R, such as GWmodelS.[26]

Multiscale Geographically Weighted Regression (MGWR) builds upon GWR by allowing for the comparison of variables at different spatial scales[7][27] This is accomplished by allowing for different neighborhood bandwidths for each variable.[7][27] MGWR is available both within ArcGIS, and as Python scripts published by a team of researchers including Fotheringham.[27][28][29] Fortheringham spoke at UCGIS on applying MGWR in a webinar titled Measuring the "Unmeasurable: Models of Geographical Context."[30]