search

schwilklab / skyisland-climate

Climate data and code for Sky Island project
2 stars 2 forks source link

DEM-terrain derivative "radiation" has errors #30

Closed dschwilk closed 8 years ago

dschwilk commented 8 years ago

The "radiation" layer has the same set of problems for each mtn range. Here is an example histogram of the values (CM):

radiation-issue-cm-example-hist

And here is a plot of radiation as a function of slope (showing that these high outliers occur on flat slopes:

radiation-issue-cm-example

Figures using data for the other two mountain ranges show the same distribution.

dschwilk commented 8 years ago

Checking with the new radiation grids emailed by @hpoulos to @dschwilk on 3/31/16:

  1. Outliers and overall problems with the distribution appear fixed.
  2. There are grid and linear artefacts. Here is an example from the GM: radiation-gm-zoom-artefacts
hpoulos commented 8 years ago

GM is the one that has this striping. The other two sites do not as far as I see in ArcMap. The DEMs do not have them for any of these. I just reran the radiation grid for GM using the following specifications, but got the same striping: latitude-31.94 (mean for the study area), whole year (2016) with a monthly interval on the 14th of each month, sky resolution 200 (default and appropriate for a monthly interval), 0.5 hour interval, calculations done in metric projection (NAD83) so z factor =1, horizon angles calculated for 32 directions (which is good for complex topography), diffuse proportion 0.3 (for generally clear sky conditions), transmissivity = 0.5 (for generally clear sky conditions).

dschwilk commented 8 years ago

Check others carefully. I think there was striping in CM although it was more subtle. But I may be wrong.

On 04/05/2016 02:35 PM, hpoulos wrote:

OK. I'm working on this now. GM is the one that has this striping. The other two sites do not. The DEMs do not have them for any of these.

— You are receiving this because you were mentioned. Reply to this email directly or view it on GitHub https://github.com/schwilklab/skyisland-climate/issues/30#issuecomment-205961068

hpoulos commented 8 years ago

I have looked into this on the web and there is no documentation that I can find. However, I think we discovered something important here in ArcGIS...when I run the radiation tool with the WGS184 unprojected data (decimal degrees), there is no striping in the GM output image

I had originally transformed everything to projected datums since all the other derivatives were based on meters. I will redo.

dschwilk commented 8 years ago

Are those white areas the maxed out values?

Weird that the arc gis tool does this when it is a projected raster since the documentation suggests using a meters projection! (http://help.arcgis.com/en/arcgisdesktop/10.0/help../index.html#//009z000000t5000000.htm) -- assuming that is the function you use. Oh well, as long as the z factor is right it should work in lat lon right?

hpoulos commented 8 years ago

I totally agree. Yes, if I specify the correct Z factor for the latitude it should be correct, but I would much prefer to use a projected, metric projection since that's what the tutorial says to do. You make a final decision and I'll make the grids however you think is best.

dschwilk commented 8 years ago

I'm a bit confused: if you use a metric projection you get grid artefacts so there is clearly something wrong with the algorithm then. We should be able to spot check this so as long as lat long works lets do that.

hpoulos commented 8 years ago

I really just wanted to the final OK from you. I agree. I have a bunch of deadlines this week, but this and the other dem-distance derivatives are first up on the plate for next week.

dschwilk commented 8 years ago

This all looks good as of the June 2, 2016 grid update. Closing.