Update LULC with NLUD-NLCD-tree-cover integration dataset

[davemfish]

The raster is forthcoming from Chris, but the associated lookup table is here https://drive.google.com/file/d/1OgC-4lrPvWOSk3QNskaXsGgnjfSpnfFW/view?usp=share_link

Updating the app means:

• [ ] hosting the new raster
• [ ] updating UI components in the scenario builder that reference specific landcover types. The new raster will likely have so many unique classes that a redesign of scenario-builder features will be required.

[davemfish]

After exploring the table linked above, here are some notes...

The lucode is the combination of, 20 NLCD classes 85 NLUD classes 4 Tree cover classes

That makes 6800 unique codes, but many will not exist in raster (e.g. "Open Water" + High tree cover). I wonder how many unique codes exist in the greater San Antonio region, or in a given "ecoregion". Also, how many of these classes actually have differences in their biophysical parameters for invest? If we know there won't be a different outcome from models between say, "Commercial - Retail/Shopping" and "Commercial - Entertainment", then we probably shouldn't make people choose between them and expect different outcomes.

So first I'm interested to know how many different categories (with respect to the raster & the biophysical parameter tables) actually exist. And then, assuming it is still a lot, we have some interesting scenario-builder design choices to make,

• Wallpapering can continue to work as-is
• Paint/Fill selection could happen with a hierarchal (tree-like) selection, where
  1. the user first chooses 1 of 20 NLCD classes,
  2. then is presented with only the NLUD classes that exist (in the entire raster? within the ecoregion?) for that NLCD class. Presumably this is a small enough list to be manageable?
  3. then chooses 1 of 4 Tree cover classes.

Although I'm not sure that NLCD -> NLUD -> Tree cover is a meaningful hierarchy. There are also "Tiers" within the NLUD, and it looks like the combination of Tier2 + Tier3 makes all 85 unique NLUD classes.

• We could have additional tree-cover options, like change Tree cover to <NONE | LOW | MED | HIGH> for all pixels in study area (but leave the NLCD & NLUD classes as-is). Though this potentially lets people create non-sensical combinations of LULC & Tree cover.

[davemfish]

FYI @chrisnootenboom I added some of my thoughts on how we'll use the new NLUD-NLCD raster to this issue. Let's continue the conversation soon!

[davemfish]

Tree canopy data is a work-in-progress, and does not yet have nationwide coverage, so fully integrated (NLUD - NLCD - Trees) dataset only exists for some places, like San Antonio.

Also, we have biophysical tables that work with the NLCD-only raster, so if we stick with NLCD for now, we can at least get models up and running.

[davemfish]

We have some new data from Chris and I've started exploring it in a notebook on my fork in branch exp/new-lulc.

Chris is going to double-check some of the logic for creating the raster, as there seems to be small numbers of pixels with unrealistic combinations of NLCD and NLUD types (e.g. "Open Water" + "Residential" and a number of others).

[davemfish]

Done in #123