I’ve been working with a group of other students on a student led project to make a solar suitability map for the whole state of Minnesota. A full writeup will come after the project is over, but here’s a quick look at what we’ve been doing.
We are using the MN State LiDAR data as the input data. There’s just under a terabyte of compressed LiDAR files.
We used Lastools to generate a raster Digital Elevation Model (DEM) data set. The DEM data set was another terrabyte of data.
The DEM was used as input to ArcGIS’s Area Solar Radiation tool.
We scripted both the DEM generation and the use of the Area Solar Radiation tool. Both scripts use a PostGIS database to store fishnets representing areas which need to be processed. The PostGIS database effectively acts as a job queue, which means we can run multiple instances of the scripts without re-processing the same area multiple times. This is important because the Area Solar Radiation tool is very slow.
The computer we are testing on is a Quad-core Xeon machine with 32 Gigs of RAM, so it’s a pretty powerful machine. If we ran a single Area Solar Radiation on that machine, it would take over 942 days to complete. With the script and the job queue, we can run 7 instances of the script concurrently (which uses up nearly 100% of the CPU) which brings us down to 134 days if we used just the one computer.
Fortunately, we are working with the Minnesota Supercomputing Institute on this project. With their resources we expect that we’ll be able to run the Solar Radiation tool on the whole state before the semester is over. At this point our scripts are running cleanly on our test computer. We should be running on the super computer in the next week or so.
In the mean time, here are some sample results of our testing. These images are actually measurements of how many kilowatts of solar energy any given point would receive during the year. Brighter points receive more sunlight during the course of the year.