DNB Terrain Correction
Performance
Steve Miller
23 May 2014
Approach
• Examine anthropogenic lights (cities) near sea-level and at
higher altitude.
• Select two adjacent orbits passes which provide i) an
eastern view, ii) a western view.
• To maximize possible parallax displacements, select
locations near scan edge.
• Cloud-free views to minimize ambiguity due to obscuration
and additional parallax effects.
• Compare east/west passes at each location for the original
(non-terrain-corrected) and new (terrain-corrected)
geolocation data.
• A consistent terrain correction will result in minimal feature
shift between the eastern and western views.
[Note: please run this powerpoint in slide-show mode in
order to toggle imagery and compare the east/west views]
Selected Domain of Analysis
Baku, Azerbaijan
(40.4N,49.9E,-28m)
Isfahan, Iran
(32.63N,51.65E, 1.59 km)
East and West Suomi NPP Passes
2143 UTC
2324 UTC
Baku
Isfahan
This pairing provides a nearly optimal comparison between a low and
high altitude light source at opposite viewing directions…
Baku (40.4˚N, 49.9˚E,-0.28 km)
No Correction
Terrain Correction
2143 UTC
2324
Baku is close to sea level: parallax displacement effects are minimal
 Terrain correction has little impact
 Some notable shifts to lights West/Northwest are evident of Baku, from
higher elevation towns nestled in the Greater Caucasus Range
Isfahan (32.63˚N, 51.65˚E, 1.59 km)
No Correction
Terrain Correction
2143
2324 UTC
At 5,217 feet (1.59 km) Isfahan parallax effects are substantial
 Terrain correction has significant impact (note dramatic shift of Isfahan in
the left-panel, in contrast to relative stability in the right panel.
Note of caution: clouds under-lit by cities will cause apparent shifts in city light
locations, since the terrain correction is done with respect to the terrain
elevation, not with respect to the cloud altitudes!!!
Summary
• The ~sea level site exhibits little shift for both terraincorrected and non terrain-corrected geolocation, as
expected.
• The high-elevation site shows dramatic improvements in
stability of signal for the terrain-corrected geolocation.
• Other examples (W. Straka, CIMSS) have shown the intraband consistency of terrain-corrected geolocation by
comparing DNB with M10 and M13 imagery of fires.
• There is some evidence of changes in lights between the
two passes, particularly for point sources—it has yet to be
determined whether these changes are the result of lights
physically changing (turn-on/off), artifacts of image
registration, or (more likely) a combination of both.
• Preliminary results are encouraging—should allow for
improved multi-spectral applications involving the DNB,
particularly with respect to surface lights at-elevation.