Landsat Orbital Repeat Frequency
and Cloud Contamination:
A Case St
United States
Samuel N. Goward, Tatiana V. Loboda, Darrel L. Williams, Chengquan Huang
Abstract
A primary Landsat constraint is cloud contamination. The
goal of this study is to evaluate how increasing Landsat
temporal repeat frequency might achieve weekly cloud-free
observations. This is a case study which examines eastern
United States observations, as a typical representation of the
temperate forest biome.
A comparison between simultaneous Landsat and
MODIS
observations shows that
MODIS
observations provide a
reasonable approximation of observed Landsat cloud cover
conditions. Daily
MODIS
cloud observations are examined,
using image compositing, to evaluate how increasing Landsat
temporal repeat frequency might reduce weekly Landsat
cloud contamination.
Results suggests that for these eastern US locations weekly
clear views are best achieved with daily satellite repeat
frequency, bi-weekly clear views with 2-day repeat, monthly
clear views are achieved with 4-day repeat and seasonal clear
views with 8-day repeat. To more fully understand the global
impact of cloud contamination on Landsat observations
similar studies are needed for the Earth’s other major biomes.
Introduction
Over the last half century, the Earth science community has
adopted satellite remote sensing as a means to document land
surface dynamics. Landsat and the
NOAA
Advanced Very High
Resolution Radiometer (
AVHRR
) observatories provided the
foundations for this work and stimulated the development
of systematic land dynamics monitoring both within and be-
tween years (Dethier 1974, MacDonald and Hall 1980, Masek
et al.
2013, Townshend 1995). This work continues today
with moderate resolution sensors on Landsats-7 and -8 and
the European Space Agency Sentinels-2a and -2b as well as
other similar satellites and sensors (Hagolle
et al.
2010, Ver-
mote
et al.
2016).These satellite observations have not only
revealed the complexity of land surface dynamics, but also
how evaluation of these dynamics can be disrupted by other
variables including sensor and satellite operations, varying
atmospheric conditions and, particularly, the impact of cloud
contamination.
There have been many studies which have explored how
to detect and mitigate cloud contamination in satellite optical
remote sensing measurements (Irish
et al.
2006, Platnick
et
al.
2003, Shenk and Salomonson 1971, Stowe
et al.
1987).
However, few studies have examined how regional and global
cloud cover dynamics interact with satellite observation
parameters such as those employed for Landsat-class obser-
vations. Such studies have been constrained by the lack of
empirical observations suitable for the specifics of a Landsat-
type system (i.e., sun-synchronous polar orbit, mid-morning
overpass, 16-18 day repeat cycle).
In this study we take advantage of the synergy between
Landsat and Earth Observing System (
EOS
) Terra Moderate
Resolution Imaging Spectroradiometer (
MODIS
) morning
overpasses to explore the impact of cloud contamination on
Landsat observations.
Background
In 1967, when the US National Research Council “Woods
Hole” working group proposed a Global Land Use (
GLU
) satel-
lite, similar to the Earth Resources Technology Satellite (
ERTS
aka Landsat) system launched in 1972, they struggled with
defining the observation repeat cycle of this system because of
the problem of clouds (National Research Council 1969):
The synoptic requirements of
GLU
pose some problems
in the acquisition of imagery for the land surface of the
earth, because of cloud cover. Preliminary estimates
show that though some parts of the world are cloud
covered most or part of the year, 20 to 30 overflights over
the course of a year should produce 90 percent or better
coverage. Cloud cover becomes a serious problem in ob-
taining imagery for disaster information, crop reporting,
and other uses requiring data at a specific time.
Only a simple understanding of global cloud dynamics
over land existed in the early 1960s, including; average global
cloud cover is near 50% and different zonal patterns exist in
the tropics, mid-latitudes and polar regions (Sellers 1965). In
addition, it was generally believed that cloudiness was lowest
at sunrise and increased as the sun heated the ground. This
led to the placement of the Landsat satellites in a mid-morn-
ing orbital pattern – a tradeoff between cloud contamination
and solar radiant intensity.
The original
ERTS
, renamed Landsat-1, was designed with
an 18-day repeat cycle which provided ~20 repeat observa-
tions per year, at least over the United States
1
. This repeat
Samuel N. Goward, Tatiana V. Loboda and Chengquan Huang
are with the Dept. of Geographical Sciences, University of
Maryland, 2181 LeFrak Hall, 7251 Preinkert Drive, College
Park, MD. 20742, (
), (
),
(
)
Darrel L. Williams is with Global Science & Technology,
7855 Walker Drive, Suite 200, Greenbelt, MD. 20770,
Photogrammetric Engineering & Remote Sensing
Vol. 85, No. 2, February 2019, pp. 109–118.
0099-1112/18/109–118
© 2019 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.85.2.109
1. From the beginning Landsat only acquired continuous repeat
observations over the continental United States, while other regions
of the world were observed less frequently depending upon available
power, on-board storage and, occasionally, predicted cloud conditions.
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
February 2019
109
