NCL San Francisco - Update

Following the SFO aNCL, an update to the earlier San Francisco map was needed. The slight redesign of the map islands with changed key and added graphs of temporal elements. The exclusive islandification of the area is enforce with a new frame. This provides more visual stability, allowing for stronger vertical orientation.



Regarding the temporal elements, San Francisco is an evening city. The lefthand rose shows the twitter activity over the days of the week, starting with Sunday at the top, going clockwise. The rose on the right shows the tweet activity per hour of the day, starting at 00:00 at the top. Here we are showing local time in San Francisco. Hence the characteristic dip between three and five o'clock in the morning. San Francisco is a typical evening city with more activity in the late hours of the day. There is a lot more activity in the later hours of the day than there is in the mornings. Comparing to Bogotá for example, a morning city, SFO has this lag with a first peak around noon and then high activity between 20h00 and 23h00. After that it drops off quite quickly, being considerably lower around midnight reaching the base around 3 am.



San Francisco New City Landscape



Image by single-blogs using the GMap Image Cutter / San Francisco New City Landscape - Use the Google Maps style zoom function in the top right corner to zoom into the map and explore it in detail. Explore areas you know close up and find new locations you have never heard of. Click HERE for a full screen view. The maps were created using our CASA Tweet-O-Meter, in association with DigitalUrban and coded by Steven Gray, this New City Landscape represents location based twitter activity.

Tweet Times

When do people actually tweet? The data we used to generate the twitter maps New City Landscapes suggests that there are big fluctuations in terms of activity over the course of twenty four hours.
The Tweet-O-Meter by Steven Gray gives extrapolated numbers based on five second intervals as to how many tweets are sent in a specific urban area.
The New City Landscape maps became popular on twitter, third most tweeted topic over the day on tweetmeme on the 30th June, with Steven Fry picking it up and in the media with the Metro and the Daily Mail.

The data for the maps comes in via the Twitter search API where a spatial search is requested, in which the urban area is defined. For the maps however we can only use a fraction of this data set. Only the messages sent from o mobile client that actually includes a lat/long position with the original tweet are of interest. With those ones we have accurat location. There is also some geocoding done by Twitter, based on the profile location, but this is not true location data in a positioning sense.
So if someone makes the simple calculation of tweet numbers based on peak hour tweets per minute, this will not reflect the numbers used for the maps.
The landscapes are generated from a week worth of data. Looking at the temporal dimension of the data, shows the fluctuations in amount of activity there is.



Image by single-blogs / TimeRoses for each city and in comparison.

There are dramatic differences between the hours of the day but also between the days. In London and New York there is a peak on two days. A number of news topics coincide there. Those days are the 28th and the 29th of April 2010. There is the BP oil spill, just a few days old and also the last of the three UK leaders debates taking place on the 29th in the evening.


Image by single-blogs using Wordl / Topics in New York over the period recorded for the New Landscape Maps.

Looking at the # tags used in London the #leadersdebate clearly leads the twitter discussion, where are in New York the #dearsomeone leads. At both places the #nowplaying tag is very active.


Image by single-blogs using Wordl / Topics in London over the period recorded for the New Landscape Maps.

New City Landscapes - Paris on Time

The New City Landscapes have been introduced earlier as a visualisation of tweet activity in the urban context. The maps are derived from data sent via a mobile client and including location information.
The rising mountains and dropping valleys remodel the density of messages as a temporary urban landscape. Earlier coverage on this topic can be found HERE and HERE.
A more detailed series we start now looking at the different places New York, London, Munich and Paris individually. This time the focus is on Paris, Ille de France. In timeRose diagrams the temporary aspect of the data is developed with a visual means line to indicate characteristics of individual units. This method allows graphical analysis, highlighting the important aspects.


Image by single-blogs / New City Landscape of Paris, France. A topography map generated from twitter activity around the Ille of France.

The 'Dents des Halles' mark the highest point on the map, being a location of high tweeting activity. It is quite an important meeting point for people of all ages. It is a place to hang out, to stand around with no specific activity at hand. This seems to be an important condition for high twitter activity. Counter the assumption important places wil stand out, usually the less expected places close by will have the peak. Take the 'Tour Eiffel' for example it made it only as the 'Flanc Tour Eiffel' at the bottom of the 'Colline d'Champ-Elysees'. The mix is more complicated and I am guessing that everyday location combined with routine activities actually float on the top, over one of activities. However to make the peak it obviously needs a combination.
From the Tour Eiffel up to the 'Cime Excelmans' down the 'Flac des Princes' across the 'Carriere Marnes-la-Coquette', one reaches the 'Aiguille du Chesnay', The peak next to Versailles. This another example of lower activity than expected.
A group of three peaks to the north-east marks the airport Charles de Gaulle a dent that would follow the logic.


Image by single-blogs / TimeRose analysis of the tweeting activity in Paris, France over the period of one week. The means line helps to classify the information.

Looking at the activity over time of the individual weekdays the pattern between weekends and weekdays is quit obvious. The visualisation here is a timeRose where the 24 hours are plotted around the circle, with the amount of tweeting plotted radial.
The means line is used to mark the highest activity peaks, with the angle of it indicating the fraction of the day covered. A steep line means late morning and late night, representing the tendency on weekends. Whereas a flat line points to early morning and early evening activity, as it draws on weekdays.
There is a problem with the data from Wednesday, this is due to the fact that witter was down and we do not have data for this period. However the drop off's on both sides suggest a similar pattern as we find on the other weekdays.
The usual pattern is a three peak blob, representing morning, lunch and evening. With flater means the morning merges in to the lunch peak and a shift towards later times takes place. This shift starts to build up already during the week starting from Thursday.


Image by single-blogs / The island of twitter land Paris in the digital see of information. Generated from tweet density send form mobile devices in Paris, France.

The Other cities wil follow as blog posts very soon, stay tuned.
Thanks forsupport with the development of this to Annick Labeca at Urban Lab Global Cities