Data description
The dataset is composed by the daily measurements of different metereological attributes acquired in Bergen over a period of 11 years, from the 1st of January 2000 till the 31st of December 2010, for a total of 4018 entries.The set of measured attirbutes consists of:
- Date: day, month and year of the acquisition
- Temperature: minimum, maximum and average value over the whole day; a few missing values have been obtained by linear interpolation
- Pressure: min, max and average, measured both at station level and sea level; a few missing values have been obtained by linear interpolation
- Precipitation: the amount of daily precipitation; no measurements are present for this attribute before January 2004, and, even after that, a lot of value are missing, just 1 663 out of 4 018 expected values are present
- Precipitation type: discrete value representing the type of precipitation; measured from 2000 to 2004, only 240 / 4 018 values are present
- Snow depth: should represent snow depth but it is always 0
- Snow cover: boolean value which represents the presence of snow on the ground; measured from January 2005, only 1 618 / 4 018 values are present
The last 4 attributes are not reliable at all, but I have tried to include at least "Precipitation" and "Snow cover" just to show some possible analysis.
SimVis
A framework for 3D data visualization and analysis, in this case the geometrical component has been simply discarded.It's plugin-based, which means it can be easily (more or less) extended.
It already implements some basic tool (histograms, 2D scatterplots, parallel coordinates, 3D rendering context) and some more advanced concepts, like view-linking and fuzzy degree-of-interest (doi)
Data overview visualization
Default SimVis features can be used to get a fairly good overview of the data and the dependences between different attributes
However it takes some time to setup the different views and it's hard to visualize multiple attributes altogether.
I implemented a "graph view" trying to give a more immediate overview of the dataset:
- time evolution of the attributes
- min, average and maximum values for temperature and pressure (station and sea level)
- attribute bar colored according to a transfer function
Advanced analysis
With so many different attributes, it would be interesting evaluating the correlation between them. Linked views and histograms/scatterplots are very effective in this context...
- Transfer functions together with scatterplots can be used to show the relationships
between 3 attributes simultaneously.
- 3D scatterplot potentially correlates up to 4 different attributes but should be used
carefully since they're quite cluttered and of difficult interpretation.
- Derived attributes, like ratios, differences, derivatives, etc. can show aspects of the data which are not immediately clear
