I’m currently sitting at Chicago’s O’Hare airport waiting for my flight back home to Germany. In an attempt to both not forget too much of it too soon and at the same time to keep me awake so I can sleep well on the plane I will now try to craft a wrap-up of my AAG 2015. I’ll start with some details about the sessions I visited and will finish with a more general recap.
Over the course of the five days (April 21-25, 2015) I visited a total of 18 sessions. This is a record for me, both because I didn’t join any field trips and because the weather outside was just too bad to randomly wander off like I did in previous years. As is always the case with conferences (and especially with the AAG) the sessions were rather mixed. I’ll not write about everything I heard (not that I would remember everything!) but will focus on the interesting bits.
It all began rather interesting in a session labelled “Travel satisfaction”, which had some good talks about one of the soft factors in the choice process for a travel mode. Since I’m now working for a transport research institute I naturally put a stronger focus on transport-related sessions and talks this year, and in one of our current projects we’re also increasingly trying to deal with these softer factors in some of our models.
After that it was already time for my very own talk “A Spatio-Temporal Betweenness Centrality Measure for the Micro-Scale Estimation of Pedestrian Traffic”! (Therefore I left the previous session a bit earlier and missed one talk.) My presentation went very well – especially considering the fact that I presented a model I had developed during my PhD research a considerable time ago and had to mostly “re-learn” and understand again what exactly I had done back then when I prepared the talk. Even so I finished well within my timeframe (not a common sight, unfortunately) and had enough time to take some questions and comments from the audience. The room was rather small but all in all I counted roughly 70 people. That’s a great turn-out for such a rather technical session, and especially since it was located in the far outlying venue of the AAG 2015, the University of Chicago Gleacher Center across the river from the Hyatt and the Swissôtel. I’m very happy to have made contact with Emmanouil and Daniel from the University of Birmingham, who had interesting things to say and suggest for my model and for my upcoming implementations in R and Python – Daniel is part of the PySal development group! Thanks again to Donna Peuquet from Pennsylvania State University and Diansheng Guo from the University of South Carolina for hosting and organizing the session.
The afternoon was dominated by two sessions of the “Spatiotemporal Symposium” and hence focusing heavily on time geographical applications and analyses. Mattias Hellgren from Linköping University (Sweden) showed once again the power of the TraMineR package for R – even though this was not the main topic of his talk. Tijs Neutens form Ghent University (Belgium) proposed an “index of public transportation gap”, basically the difference between an index of public transportation needs and provision. Sounds simple enough, but the results he showed were very interesting and I’ll make sure to find out more about it. After that Jiawen Yang from Peking University showed a congestion index generated from floating-car data (FCD), again something we’re currently trying to do ourselves, so it was interesting to see what other people do and where it’s getting them.
Since my doctoral research and the paper I published about it I’m very interested in quantitative methods for the estimation of populations. And so every year at AAG I go to a session called “Estimating Building Level Population Density” which consists entirely of presentations by the good people at Oak Ridge National Labs’ GIST. And every single year I’m disappointed, since they mostly regurgitate stuff from the years before. (Note to self: Skip this session next year!) The only interesting (since new!) talk was given by April Morton: “A Bayesian Model for Estimating Building Occupancy”.
Day two started with a double-session on “Geosimulation and Big Data”. Of special note here are presentations by Kira Kowalska, Martin Zaltz Austwick, and Andrew Crooks. Kira showed a bold topic modelling idea of using a “bag of words” approach and Latent Dirichlet Allocation (LDA) on street segments in order to identify optimal routes for police patrol cars. Martin generated a Monte Carlo-based Agent Based Model (ABM) to simulate rush hour bicycle traffic in Madrid from travel survey data. Andrew talked about ideas to leverage crowdsourced data for spatial analyses, focusing not so much on the classic Volunteered Geographical Information (VGI) but more on Ambient Geographical Information (AGI), data that is not necessarily provided on purpose but can be derived from crowdsourced data.
The first afternoon session “Mobility, Health, and the City” provided a good introduction for me to the wide field of “active transport modes”, i.e. walking and biking – and by Harvey Miller’s definition also public transport.
After that a session in what must have been one of the smallest conference rooms at AAG if not in the northern hemisphere, focused on the use of electric vehicles, its economic and societal effects and also mode choice behavior.
In the morning I attended two session of the “Human Dynamics in the Mobile Age” series which had interesting case studies of various analyses of tracking datasets in China, mostly actively tracked mobile phone data and call-detail records (CDR) but also smart card data from public transport systems. I was especially intrigued by Ziliang Zhao from University of Tenessee, Knoxville, and his hierarchical flow clustering method, a modification of a method by Diansheng Guo, who was also in the audience – a very interesting discussion! I’ll have to investigate deeper into both their methods and a new kernel smoothing method for aggregated movement data by Guo, which he had published just days earlier. Shih-Lung Shaw, also from University of Tenessee, Knoxville, posed the question “How Big is Big Enough in the Big Data Era?” in his title, but unfortunately did not provide an answer, yet. His analyses look promising, though, so there are lots of interesting results to be expected soon.
In the afternoon I attended two more sessions of the “Spatiotemporal Symposium” where, amongst others, Mark Horner from Florida State University presented an extension to his time geographic density estimation (TGDE) method. I do see some application for it in my research, so I will try to read up on it. Jiaying He talked about trajectory clustering methods for time diary and GPS tracking data to analyse patterns of human behavior, also something I’m interested in investigating more for my own research. Laure Charleux from University of Minnesota, Duluth, presented some ideas regarding a topic that I’m currently dealing with in my job: the issue of optimising transit waiting times – should you make them as short as possible or as enjoyable and useful as possible? You can’t do both and the behavioral and spatial determinants are manifold. It was great meeting Laure and talking about this non-trivial problem – it for sure was not the last time we talked about it! (Or at least I should hope so…)
In between I attended a session called “Transit Landscapes” which had a brilliant and engaging talk by Stephen Rijo from University of Denver about economic impacts of improved bicycle infrastructure in Denver. One of the best talks I saw at this year’s AAG!
The evening was then characterized by a rather nondescript event that nevertheless was a great deal for me. The (anonymous) reviewers of the Hazards, Risks and Disasters Specialty Group selected me as runner-up and honorable mention for the Gilbert F. White Award for my dissertation. I’m obviously very proud and humbled by that, and at the same time very happy that there are now more people than just me and the members of my defense committee that have read my thesis…
[… four-week interruption in writing this article …]
I spent Friday morning in two sessions called “Big Data for Urban and Regional Analysis”. Amongst others it held some interesting presentations by UCL CASA’s James Cheshire, who showcased the (relatively) new web platform called DataShine, Daniel Arribas-Bel and Emmanouil Tranos from University of Birmingham presented two approaches of analyzing mobile phone data (CDRs) in Amsterdam, showing what does and doesn’t work in deriving spatial structures and human behavior from such data.
In the afternoon I then attended two sessions with some of the greatest names in spatial analysis: Arthur Getis introduced CHaOS, a novel method for dissecting spatial variance, Stewart Fotheringham shared his ideas about big models vs. big data, Sergio Rey spoke about local indicators of mobility analysis, and Luc Anselin about conditional permutation inference and its computational issues in regard to local spatial autocorrelation statistics.
The two highlights of the final conference day were a spontaneous HOTOSM mapathon after the massive earthquake that struck Nepal the night before, initiated by Alan McConchie of stamen design, and a session called “The Geographies of ‘Back to the Future’ (1885-2015)”. The latter is actually a serious (well, not that serious) research project dealing with various geographical aspects of the famous movie franchise: cultural geography, political geography, religious geography, gender geography, media geography, you name it … A very nice and entertaining way to end this conference on a high note! Check the panel session’s organizer’s report here (including a Twitter quote by yours truly…)
Overall I’d say that for me the conference started great and then became more and more mediocre with every day – except for said two highlights on the last day. The variance in the quality of the presented papers is a well-known problem at AAG, mostly due to the fact that the papers are not peer-reviewed. This results in a great number of master course students presenting their research papers mixed with luminaries of their fields. On the one hand I think it’s great that the AAG can be a forum for everybody to present their studies, findings, and ideas, but on the other hand this can hamper the overall impression of the conference significantly.
In terms of organization this AAG was definitely the best I have attended so far – mind you that this was my fourth time! There was actually WiFi in most conference locations, it was actually usable (!) and the registration process went very smooth and professionally. My only gripe would be the locations of the conference rooms, spread out not only over three venues (Hyatt Regency Chicago, Swissôtel Chicago, University of Chicago Gleacher Center) but also at quite significant distances, which sometimes made it difficult or even impossible to move quickly from one session to another – especially during sessions. The trouble one had to undergo to reach the Gleacher Center was almost ridiculous…
Chicago was a very nice host city, at least for me as a first-timer, even though the weather was rather cold and unpleasant… But hey, we’re there for work, right? Nevertheless I had a great deal of fun there and met many awesome people. Now I’m looking forward to San Francisco for the 2016 AAG Annual Meeting.