# Estimating tree cover (or other vegetative cover) over impervious cover using NLCD data

Has anyone attempted to estimate tree cover or general vegetative cover over impervious cover using the NLCD data sets? The provided layers (land classification, tree cover, impervious cover) could be combined to produce an estimate - and I currently have an algorithm that does this - but I'm wondering if anyone else has tried.

As an example of what I am talking about, one grid cell might be classified as medium density residential in the original data set (type = 23), with an impervious cover percentage of 70% and a percent tree cover of 40%. Obviously there is some overlap of canopy and impervious cover, since they sum to be greater than 100%. So, based on the fact this cell was classified as a medium density residential cell, we use a set of simple mathematical equations to estimate tree cover over impervious.

Link to NLCD data sets: http://www.mrlc.gov/nlcd2011.php

• Interesting. Is "over land of canopy and impervious cover" meant to be "overlap of..."? Also, you might want to add a link to NLCD, since this is the only question tagged with that. Jan 16, 2015 at 23:38
• Good catch. Edits made. Jan 16, 2015 at 23:59
• +1 Just curious why you would want to deal with NLCD to do this when there are much better datasets that would yield higher accuracy for the land cover types you are talking about?
– Aaron
Jan 17, 2015 at 0:16
• @Aaron, Do any of these other datasets have national coverage? Jan 17, 2015 at 1:29

As @Aaron mentioned, NLCD data is one option, but a rather course resolution one.

If you want to get vegetation cover or tree canopy cover using free national coverage data at a fairly high resolution (1-2 meter) I would recommend these two options:

Vegetation cover

• Use NAIP aerial imagery data that has 4th NIR band

• Perform NDVI analysis on image to classify between vegetation and impervious land cover

Canopy cover

• I was hoping for a national data set that was already processed. I don't have the time to process such a data set myself. This leads me to believe that the NLCD dataset is the current best option. Feb 18, 2015 at 5:28