Using distributed computing to accelerate the calculation of a solar exposure map of Christchurch

Abstract

Following the Christchurch Earthquakes an opportunity to identify preferred characteristics of a rebuilt city was created through the “Share-an-idea” process organised by Christchurch City Council. One characteristic that received a large amount of support was that the city should be “light and airy”. In an effort to determine the nature of pre-Quake Christchurch a solar exposure map was proposed. The map would show the amount of time that a location could receive sunlight and the amount of time that a location was in shadow. Such solar exposure calculations have previously been done at individual locations to determine placement of solar panels and house orientation but never on a city scale.

To determine the solar exposure for the entire city an automated process was required. Using images from Google StreetView it was possible to calculate the solar exposure for a location. To determine the available images for the city and to perform the calculations is a time consuming process. Each location requires approximately 30 seconds retrieving the image and performing the solar exposure calculations. For the 150,000 locations this would be a prohibitively time consuming task. To reduce the time to completion, a HTCondor Grid with 8 slots was used. A two stage process was implemented, the first stage “walked the city” to determine to available locations. The second stage retrieved the image for the location and performed the solar exposure calculations for a year at 15 min intervals resulting in approx. 35000 data items for each location. The challenges of this type of computing will be discussed including ensuring independence but allowing data sharing, detecting failure, verifying data and data storage.