Mapping up-to-date Wildfire Damage with ENVI Analytics
On the morning of Sunday March the 19th, a fire
was ignited in the heart of Sunshine Canyon Trail in Boulder, Colorado, causing
the evacuating of nearly 426 people and placing another 1,000 on high alert.
The fire burned from 1 a.m. early Sunday morning until late Monday night when
it was contained and then later died out. Nearly 70 acres of land and forest
trails were burned and left black. It was concluded that the fire was caused by
transient campers near the Sunshine Canyon Trailhead.
Being a Boulder resident and actually living in the pre-evacuation
zone, I did not take this matter lightly. While I was sweating it out, my mind
occupied itself with how we could map out this small disaster for government or
commercial use. With Harris having a close partnership with Planet, we were
able to quickly draw down some RapidEye 5 band satellite imagery of both before
and after the fire for immediate fire extent analysis. Below I will walk
through steps of how we were able to use our ENVI analytics with the help of
Planets RapidEye imagery to understand the extent of this wildfire only a short
time after the disaster occurred.
The first step in our workflow was to get some imagery for
before and after of the area that was effected by the fire. Below we have our
(Left) pre- and (Right) post- (Right) fire images that have not been processed.
As you can see in the two areas highlighted, we have some pre-processing to do.
Our area of interest (AOI) for the fire is circled below in blue.
The second step in our workflow is going to be the pre-processing to get rid of
some problem areas that may require atmospheric correction. For these datasets,
we ran the ENVI Radiometric Correction Quick Atmospheric Correction (QUAC) on a
subset of the image so we could focus our analysis only on our AOI. For more specific instruction of how QUAC
works, you can visit our help documentation online for a full overview of the
(PrefFire image is on left. Post-fire image is on right).
Next, after we corrected atmospheric distortions in the
subset of our data, we ran the data through the ENVI Image Change Workflow. What
this does is compares two images of the same geographic extent, taken at
different times, and identifies the difference between them based on either a
specified input band or on a feature index. With this ENVI workflow, the
results are given as either a “gain” or a “loss”. In this instance, we decided
to run our workflow on the ENVI Normalized Difference Vegetation Index (NDVI)
so that we could extract areas of “loss” due to the fire burning healthy
After thresholding our workflow to only include areas of
“loss”, and doing some post-processing cleanup, we were able to extract the
area of change or “loss” as a shapefile to overlay on top of our original
imagery. This gives us a look at the extent of the damage by comparing the
before and after imagery. Below we have our subsetted post-fire image with the fire
extent shapefile (Left), and subsetted our original pre-fire Image with the fire
extent shapefile (Right).