3D Histogram

Tags:  

The following pictures show 152 years of temperature data recorded in Sydney, Australia by the Australian Bureau of Meteorology .  Maximum temperatures were recorded daily and this resulted in about 55,000 observations.

3D Histogram of Sydney Climate Data. Shows 152 years of data.

The largest image on the top shows a perspective view of the histogram. The image on the bottom left represents the original data set (55,000 observations). Its difficult to identify a trend .even with a large image. The middle picture on the bottom shows a top orthographic view of the histogram — a trend is emerging that temperatures during winter (the middle blue band) are getting less cold. The image on the bottom right uses polynomial regression to fit trend lines and then assigns colours to the resultant smoothed surfaces. Its clear that winter temperatures (June to August in the southern hemisphere) are getting less extreme.

Using Python, R, Matlab and Maya To Create the Histogram

Initially Python was used to pre-process and generally clean up the data.  R was used to then to further format the data.  For example, in order to make nice rectangular block of data, February 29th was excluded (38 observations) and data for 2012 was also excluded as this was the final year for the dataset and as a consequence data for the whole year was not available.

Visualising data using Maya

Maya is 3D animation application and is typically used in the film and television industry.  But with with WebGL becoming prominent, plug-ins like Sketchfab  allow 3D data to be displayed in a browser and then interactively manipulated by the user.  This facilitates a more tactile experience.

Github Code and Data

The code used to create this histogram is available on Github here.

OctocatSmall

Video Animation

Below you can see a 60 second video which demonstrates how winter in Sydney, Australia is slowly disappearing.  The data is from the Bureau of Meteorology and represents daily data over 150 years. The video shows the original point data and then how a mesh was fitted to these points.  The mesh was then assigned a colour; the higher the interpolated temperature, the ‘hotter’ the colour.