gadm36

Contains gridded IOAPI-compliant rasterized versions of the global administrative boundaries as defined by the gadm project version 3.6.

ioapi/gadm_<DOM>.IOAPI.nc: files have a time-independent raster where each grid is assigned an ID derived from gadm ID_<x> fields ioapi_frac/gadm_4US1_<DOM>.ID_<x>.IOAPI.nc: files have a raster files where each time is a single feature and the TFLAG YYYYJJJ identifies the ID_<x> values.

Globally Unique IDS

gadm v 3.6 level 1 and level 2 IDs were not unique, only unique within Level 0, which led to duplicate counties within the 12US1 domain. Some in Canada, some in Mexico, some in US.

To make the levels unique, I modified the codes:

• ID_0 = ID_0_orig * 10000
• ID_1 = ID_0 + ID_1_orig
• ID_2 = ID_0 + ID_2_orig

This applies to ioapi/gadm36_<DOM>.IOAPI.nc files.

Fractional 12US1 assignments

I also created a set of files using fractional area overlap. I started with 4US1, created masks (0: off; 1: on), and averaged neighboring grid cells (3, 3). This will obviously be limited to 9ths of a 12k cell, but it may be better than binary -- particularly for counties at 12k. For example, there is at least one county whose maximum cell coverage is 1/9.

Each country/state/county present within the domain. There are 7 countries, 110 states, and 3690 counties. I stored each as a separate TSTEP and the YYYYJJJ is actually the ID. Definitely violates the JJJ part for ID_0. Because I am using TSTEP to store separate shapes, it was best to store each variable separately.

This applies to ioapi_frac/gadm36_4US1_<DOM>.ID_<x>.IOAPI.nc

Making Custom Regions

There is a script to make custom regions. ./customregions.py -h will give details.