Vector Radar

Matthew Gougeon -

Background

The Weather Network is Canada’s foremost weather reporting service, however the precipitation presentation techniques it uses are outdated, difficult for users to understand and low resolution. Current methods use pixel based images to display data on the map, this causes loss of resolution at high zoom levels and also creates a choppy movement when forecasting because of large changes between images.
Radar display technologies are in need of an overhaul. The current design of weather radar was established during the 1980s and has seen very little change in terms of presentation or usability. Improvements in technology have allowed the possibility of advanced methods for displaying the data and The Weather Network wanted to exploit these possibilities.

Goals

The Weather Network assigned the research and development department with the task of redesigning weather radar to be more user friendly and interactive. The goal of the project was to create a system which fetched existing data and displayed it in using a user friendly and appealing approach.

Tools and Technologies

  • Python
  • OpenCV
  • PostgreSQL
  • PostGIS
  • C#

Solution

The solution created by the R&D team was a polygon based radar display technology, which used polygon overlays to represent precipitation data on a map. Image processing techniques were used in order to analyze the raster-based radar images and store the polygon data in a geospatial database. The database held polygons for each different precipitation intensity level, and a web API was built to serve this data to clients in a GeoJSON representation. In order to encode the weather movement data a .PNG image was used with the red and green channels representing latitudinal and longitudinal movements of the radar data over time. An optical flow algorithm was used to generate the image based on the last 60 minutes' worth of radar data.

Resulting from our efforts was a prototype system which displayed radar precipitation information in a user friendly manner with smooth radar shapes and fluid motion of precipitation over time.

The prototype is now in a public beta phase and can be found at theweathernetwork.com.