Interactive Spatio-Temporal Data Alignment
Problem
In Earth Sciences, different variables are available for a given location at a given time, such as temperature, precipitation, soil water volume, land usage, or Köppen climate zones.
Oftentimes, locations and times of variables are not perfectly aligned, i.e., temperature might be available at point (x,y) at 9:32am, while precipitation is available at 3:58pm on the same day and 15km away and at 10:00am 40km away. One variable might have locations on a raster, while the other is irregularly located with varying density.
Domain scientists then come up with rules how to align spatio-temporal data, e.g., select temperature as reference and pair it to the next available precipitation measurement that is no more than 6h in the past and <30km away. Or collect all precipitation measurements within 12hrs, and transform them to a distance-weighted average.
However, visual support is needed s.t. domain scientists can see how their alignment rules impact uncertainty and missing data in the aligned dataset over space and time.
Aim
Depending on the type of work (SE, MA, BA, PR), the aim includes:
- literature review of interactive (spatio-temporal) data preprocessing
- define alignment rules
- design & implement interactive specification of alignment rules
- design & implement supporting visualizations
- evaluation of implemented prototype with domain scientists