What's the best projection for modeling incoming solar radiation?
I know that my question is about analysis and not necessarily cartography... but it deals with map projections, and I've seen some other "Ask a Cartographer" answers that come close to what I'm after. So here goes...
I am interested in calculating incoming solar radiation (insolation) for hundreds of 1 degree by 1 degree DEM tiles, covering the western U.S. Because the overall scope of my project is very large, I'm generally using a standard USGS projection for all my geospatial data (Albers conic equal area, central meridian = -96, standard parallels = 29.5 and 45.5, latitude of origin = 1, units = meters). As such, true north is up to 17 degrees off of grid north as I get out toward the west coast.
To do the solar calculations, I'm using the ArcGIS Solar Analyst extension. This program calculates slope, aspect, and viewshed for each DEM cell, but doesn't give any ability to account for differences between grid north and true north. In documentation, the authors state the following: "For viewshed calculations, the Solar Analyst assumes DEMs are in a projection that preserves direction. Most DEMs are available in projections that do not exactly preserve direction. For some commonly used projections (e.g. UTM), slight rotation of the study area to register correctly with true north can be sufficient to improve accuracy."
It seems to me that if I want accurate results, I need to project each DEM tile into a projection where grid north equals true north before feeding it into Solar Analyst (then reproject the outputs back to my standard projection). From others posts on here, it seems like Mercator, Miller Cylindrical, transverse Mercator (centered in the middle of each tile), or Plate Caree would all be options that might achieve what I want. What do you think? Do you have any recommendations?
Also, one thing to note... I will be doing this all with python scripting, so adjusting parameters like central meridians or standard parallels to best fit each tile would be easy to do.
Thanks very much for your help!
Mapping Center Answer:
You correctly surmise that the best way to do this would be to reproject each 1 degree by 1 degree DEM tile, perform the analysis, and reproject back to the original projection. If you need to use a projection that preserves direction, instead of the projections you suggested, try the Lambert Conformal Conic projection -- directions are reasonably accurate at medium to small scales.
To modify the projection for each tile, set the central meridian to halfway between the east and west extents of the tiled area and set the standard parallels to the latitudes that are 1/6 of the way up from the southern-most extent of the tiled area and 1/6 of the way down from the northern-most extent of the tiled area.
For example, for Oregon, the northern border is at about 46 degrees north and the southern border is at about 42 degrees north, so the total range is about 4 degrees of latitude. Divide that range by 6 – in the Oregon example, that is 0.6. Add 0.6 to the smaller latitude, and subtract it from the larger latitude. Finally, set the standard parallels down one and up one from the top and bottom edges, respectively. In this example, that would be 42.6 and 45.4.
If you would like to post a comment, please login.