Detecting Impervious Cover with Artificial Lighting in Astronaut Photography from the International Space Station
Abstract
Impervious cover continues to pose a threat to flood-prone regions, especially the
ones in dense urban areas. Mapping the location of impervious surface becomes vital to
properly managing drainage and runoff in a city, along with the health of fluvial
ecosystems. Kotarba and Aleksandrowicz, in 2016, tested the ability of nighttime
astronaut imagery from the International Space Station (ISS) to detect impervious cover.
The artificial lighting emitted from a city’s nightscape is used as a proxy for
imperviousness. This paper expands on their research by focusing on the nightscape in
San Antonio, TX from December 2015, and by observing the effect the camera look
angle has on impervious surface detection. Analysis was done by comparing the ISS light
intensity imagery to 2016 National Land Cover Database (NLCD) Degree of
Imperviousness ground reference data on the basis of low, medium, and high urban
density. Difference images between reclassified ISS images and the NLCD reference data
for low, medium, and high imperviousness were calculated with up to 49% kappa
accuracy, a moderate agreement. ISS images’ overall accuracy increased with the growth
of the threshold for urban density although the kappa statistic was highest with the
smallest threshold for imperviousness. ISS photographs classified dense urban areas
correctly but failed to correctly classify poorly lit impervious surfaces such as rural
roadways, residential neighborhoods, and airport runways. ISS imagery had a high
producer’s accuracy, particularly with lower thresholds for imperviousness, meaning that
the likelihood is high that impervious cover detected by the ISS is actually impervious
cover. Overall, ISS imagery detects impervious surfaces moderately well and may be
more accurate due to the imperfect nature of the ground reference data.