14.4 Automated Stratification Methodology
14.4.1 Stratification Method
NASS’s traditional area frame stratification process involves the visual interpretation
of satellite imagery or aerial photography to subjectively determine PSU stratum
definitions based on percent cultivated land in a PSU boundary. The automated
stratification process replaces this traditional approach of visual interpretation with a
procedure that automatically and objectively determines area frame PSU stratum
definitions based on percent cultivated land within a PSU boundary by using
available geospatial land cover information. The data required for this stratification
study included a state-level area sampling frame with stratum specific PSUs, a 2010
state-level NASS CDL, and a 2010 JAS segment file with segment-level percent
cultivation calculated. The NASS traditional state-level area frames were created
using visual interpretation and used operationally to select the 2010 JAS sample.
Although the NASS traditional area frames were built prior to 2010, they were all
used operationally in 2010 and depended upon to reflect land cover conditions based
on percent cultivation. The 2010 CDLs included as many as 50 categories and a wide
variety of crops. For the purpose of stratification, a cultivated data set or “layer” was
first generated by recoding CDL pixels of crop categories into “cultivated” and
non-crop pixels into “noncultivated” (Boryan et al. 2012).
The detailed steps for the automated stratification process are given as follows:
1. Derive a state-level cultivated layer from a current CDL by grouping all crop
categories into one cultivated (crop) category and assigning the corresponding
pixels with a value of “1” while grouping the remaining categories into one
non-crop category and assigning the corresponding pixels with a value of “0.”
2. Load an individual area frame PSU boundary.
3. Load a CDL-based cultivated layer.
4. Overlay an area frame PSU boundary on the cultivated layer.
5. Compute percent cultivation of each area frame PSU by counting the total
number of pixels with value “1” (cultivated) and the total number of all pixels
within the PSU boundary. The percent cultivated is given by the number of “1”
pixels divided by the total number of pixels.
6. Determine the PSU stratum by checking the stratum definition lookup table to
map the computed percent cultivation to a defined stratum, and label the PSU
with a corresponding stratum number as a PSU boundary attribute.
7. Determine stratum definitions for all PSUs in the state by repeating steps 2–6 for
every PSU.
Figure 14.3 (left) illustrates PSU percent cultivation overlaid on and calculated
from the cultivated layer. With the calculated percent cultivation, each PSU could be
labeled into a stratum category as shown in Fig. 14.3 (right) based on the statespecific stratum definitions as given by Table 14.2. Figure 14.3 (right) shows the
resulting CDL-based stratification for the same area frame.
14 Geospatial Land Use and Land Cover Data for Improving Agricultural Area. . .
275
14.4.1 Stratification Method
NASS’s traditional area frame stratification process involves the visual interpretation
of satellite imagery or aerial photography to subjectively determine PSU stratum
definitions based on percent cultivated land in a PSU boundary. The automated
stratification process replaces this traditional approach of visual interpretation with a
procedure that automatically and objectively determines area frame PSU stratum
definitions based on percent cultivated land within a PSU boundary by using
available geospatial land cover information. The data required for this stratification
study included a state-level area sampling frame with stratum specific PSUs, a 2010
state-level NASS CDL, and a 2010 JAS segment file with segment-level percent
cultivation calculated. The NASS traditional state-level area frames were created
using visual interpretation and used operationally to select the 2010 JAS sample.
Although the NASS traditional area frames were built prior to 2010, they were all
used operationally in 2010 and depended upon to reflect land cover conditions based
on percent cultivation. The 2010 CDLs included as many as 50 categories and a wide
variety of crops. For the purpose of stratification, a cultivated data set or “layer” was
first generated by recoding CDL pixels of crop categories into “cultivated” and
non-crop pixels into “noncultivated” (Boryan et al. 2012).
The detailed steps for the automated stratification process are given as follows:
1. Derive a state-level cultivated layer from a current CDL by grouping all crop
categories into one cultivated (crop) category and assigning the corresponding
pixels with a value of “1” while grouping the remaining categories into one
non-crop category and assigning the corresponding pixels with a value of “0.”
2. Load an individual area frame PSU boundary.
3. Load a CDL-based cultivated layer.
4. Overlay an area frame PSU boundary on the cultivated layer.
5. Compute percent cultivation of each area frame PSU by counting the total
number of pixels with value “1” (cultivated) and the total number of all pixels
within the PSU boundary. The percent cultivated is given by the number of “1”
pixels divided by the total number of pixels.
6. Determine the PSU stratum by checking the stratum definition lookup table to
map the computed percent cultivation to a defined stratum, and label the PSU
with a corresponding stratum number as a PSU boundary attribute.
7. Determine stratum definitions for all PSUs in the state by repeating steps 2–6 for
every PSU.
Figure 14.3 (left) illustrates PSU percent cultivation overlaid on and calculated
from the cultivated layer. With the calculated percent cultivation, each PSU could be
labeled into a stratum category as shown in Fig. 14.3 (right) based on the statespecific stratum definitions as given by Table 14.2. Figure 14.3 (right) shows the
resulting CDL-based stratification for the same area frame.
14 Geospatial Land Use and Land Cover Data for Improving Agricultural Area. . .
275