In short: Each New York cab journey is a point, the distance from the center is the distance of the journey and the position around the circle is the direction of travel (as if the circle were a compass). Color is added when its different during the day or night.
Description: Despite its appearance this image is another radial representation of the cab journeys taken at the beginning of 2015. Essentially it is the same as Study 1 but with both components relating to the journey itself in terms of distance and direction. Where Study 1 was a clock, this is a compass. The average time-of-day is represented by the blue channel. What appears are mirror images of common start points overlapping at common end points. The center of the image, the shorter distances, show the journeys within Manhattan. The result is that this image does not appear radial at all and seems almost out-of-focus. This is a great example of how increasing the dimensions shown do not always improve the understanding of what is being shown. What does start to appear are the echos of the areas in NYC that rely on yellow cabs the most and identifying the trips being taken is much less intuitive and, initially at least, really skews the viewers understanding of what they are looking at. This image would pay dividends in the longer term, however.
Point of interest: Aside from the obvious duplicates of Manhattan and Midtown, there are some shadows of street patterns that are not as easily related to any particular area. For a pattern to emerge in this way quite a strong signal in the data must exist and must be relatively unique in this context for it not to be washed out.
Technical: 6 months of yellow cab trip data for (Jan-Jun 2015) available at https://nycopendata.socrata.com/. Each journey is a point plotted radially with the distance from the center relating to distance and the angle being the direction. The blue channel is an average of the sum of journeys and an average time of day. The geographical scale is linear but the summation scale is a combination of Log and Linear scales to bring out the contrast.