search

schwilklab / skyisland-climate

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

reconstruct climate summaries: we should clip to subset of landscapes #37

Closed dschwilk closed 7 years ago

dschwilk commented 7 years ago

We should not bother reconstructing climate summaries over the full rasters as these include buffer regions that were only included so that some topographic variables could be calculated. The PCA decomposition, modeling and reconstruction steps currently work on the full landscape rasters, but we should clip these first to a subset (eg portion of landscape that includes trees only). Do we have polygons that fit this description? If so, we should clip before running the reconstructions.

hpoulos commented 7 years ago

If we are going to go the rout of clipping to canopy cover, I would suggest we use the nlcd (national land cover dataset) for tree canopy cover. I can change the models in model builder to clip the DEMs first and generate the new data layers. This seems the most objective way to delineate canopy cover. We will likely have to threshold the canopy cover layer to some minimum value and then use the mask to clip the other layers. I am downloading the nlcd database now, but it's supposed to take another 2 hours before I can look at it.

dschwilk commented 7 years ago

I think all discussion of this issue moved to #40. I will close that issue to move discussion back here. Outcome of that discussion seems to be we can use the intersection of the watershed uits and the canopy cover approaches to limit our prediction landscapes to a more reasonable area of inference.

dschwilk commented 7 years ago

@hpoulos : please implement this clipping. I'd like to run the model reconstructions and I do not have enough hard disk room anywhere to reconstruct the current landscapes (but I ahve solved the memory porblem by doing this in stages).