środa, 20 listopada 2013

Datatypes in Time

Data we gather has some time reference in it. It might me a moment or period which the data describes, time when data was collected, validity period or time when data got published. There's a range of situations when temporal attributes are understood in different ways.

Events happening at different places and times, traffic data (Google),
changes in tube stations log-in's/out's (Oliver O'Brien), crowdsourced data (irevolution.net)

Spatiotemporal data representing various phenomena can be analysed in few ways, depending on it's spatial and temporal characteristics. If the location is fixed (when using fixed sensors or with statistical data assigned to administrative areas) it is simply a matter of comparing values in time but when records represent unique locations (crowdsourced data, events) this can get more tricky when you want to dig deeper into data than just viewing data in selected time sections.

GDP change forecast (BusinessInsider), Road works (TfL), moving trains (WatchDogs game)

GIS data comes in 3 geometry types - points, lines and areas which are described with any numer of attributes. Analysing this data in temporal dimension requires us to use some identity attribute allowing creation of continuity between records when one record is a previous/future state of another record, thus letting comparisons to be made. As mentioned, fixed location data is a straight-forward situation, same with data records having unique identity (like buses or trains). Working with unique location data, use of aggregation (in a grid) or snapping helps create continuity of temporal dimension in data.

Shopfront changes (red triangles), workforce transfers planning with geocoded locations,
crime aggregated in 25m grid, major links closure (black) effects on traffic

Working on my spatiotemporal analysis app I decided to follow principles outlined in previous post and delivered:
- 4 views showing four moments in time, within selected range
- 1 view showing data dynamics (change in analysed value between first and last moment)
- simple interface updating all map views with single click

Currently the app is running properly with all types of data but requires script manipulation to switch between data types and kind of analysis. Here it is with UK GDP data (go fullscreen to see annotations).

Features for future version:
- adding meaningful content to time view (global parameter? data lines for each spatial element?)
- allowing easy switches for loading different data types and applying different analysis
- exporting animations to KML and maintaining a linked component in Google Earth

Brak komentarzy:

Prześlij komentarz