Multi-Spatial Resolution Satellite and sUAS
Imagery for Precision Agriculture on
Smallholder Farms in Malawi
Brad G. Peter, Joseph P. Messina, Jon W. Carroll, Junjun Zhi, Vimbayi Chimonyo, Shengpan Lin, and Sieglinde S. Snapp
Abstract
A collection of spectral indices, derived from a range of
remote sensing imagery spatial resolutions, are compared
to on-farm measurements of maize chlorophyll content and
yield at two trial farms in central Malawi to evaluate what
spatial resolutions are most effective for relating multispectral
images with crop status. Single and multiple linear regres-
sions were tested for spatial resolutions ranging from 7 cm to
20 m using a small unmanned aerial system (
sUAS
) and satel-
lite imagery from Planet,
SPOT
6, Pléiades, and Sentinel-2.
Results suggest that imagery with spatial resolutions nearer
the maize plant scale (i.e., 14–27 cm) are most effective for re-
lating spectral signals with crop health on smallholder farms
in Malawi. Consistent with other studies, green-band indices
were more strongly correlated with maize chlorophyll content
and yield than conventional red-band indices, and multivari-
able models often outperformed single variable models.
Introduction
Precision Agriculture and sUAS
Precision agriculture has become a prominent subject of
research for remote sensing of cropping systems since the ad-
vent of individual-use small unmanned aircraft systems (
sUAS
)
(Zhang and Kovacs 2012). While national governments have
been monitoring agricultural production via satellite since
the 1970s (Macdonald 1984), coarse spatial resolutions, cloud
cover, and infrequent collection has rendered much of these
image data unfit for the needs of smallholder farmers (Mulla
2013). Now, the use of quadcopters and fixed-wing drones
for precision farming continues to grow as
sUAS
technologies
become increasingly cost effective and programmatic opera-
tions are developed that streamline the process of collecting
imagery and generating map outputs (Floreano and Wood
2015). Many
sUAS
, such as the senseFly eBee fixed-wing
aircraft with eMotion software, no longer require manual pi-
loting and can be launched with preprogrammed flight plans.
Moreover, associated software such as
Pix4D
photogrammetric
software (
Pix4D
SA
2017) can handle many of the previously
arduous image preprocessing tasks, such as image mosaicking,
georeferencing, orthorectification, and radiometric calibration.
The autonomous acquisition and processing of remote
sensing data via
sUAS
, coupled with attainable fine spatial
resolutions, are rapidly transforming the landscape of remote
sensing science. Lippitt and Zhang (2018) offer a conceptual
perspective on the historical and future use of
sUAS
in the field
of remote sensing, as well as the technological and theoretical
challenges faced; similarly, Pajares (2015) offers an exhaustive
synopsis of technologies, applications, sensors, and methods in
sUAS
-based remote sensing. Automation advancements in
sUAS
have made precision agriculture accessible to users with mini-
mal to no experience with aerial vehicle piloting and remote
sensing-based crop analytics. While the abundance and afford-
ability of
sUAS
may be a net positive for monitoring agriculture,
the emergent limitations, challenges, and potential misuse are
not completely understood, and given the continually chang-
ing nature of
sUAS
, optimal operational frameworks are not
always present or relevant (Mesas-Carrascosa
et al.
2015).
Multispectral Imaging Capabilities of sUAS
The standard agricultural-use
sUAS
multispectral camera, such
as the Parrot Sequoia used here, can acquire images across five
spectral wavelengths; onboard sensors include red, green, red
edge, near-infrared (
NIR
), and red, green, and blue (
RGB
) (Parrot
Drones
SA
2017). More sophisticated cameras with a broader
range of spectral bands can record shortwave infrared (
SWIR
)
and thermal infrared (
TIR
) (Saari
et al.
2017; Stark, McGee, and
Chen 2015), which are critical for monitoring crop water and
heat stress (Ceccato
et al.
2001). With the visible spectrum (
RGB
)
and
NIR
, a broad range of crop status indices can be calculated.
Most notable is the normalized difference vegetation index
(
NDVI
) (Tucker 1979). This metric is widely used as a measure
of crop health—a healthy plant will absorb visible light (espe-
cially blue and red), while the fortified leaf structure will reflect
a high amount of
NIR
.
NDVI
calculation is simply (
NIR
– red)/
(
NIR
+ red), returning values between
-
1 and 1, ranging from
nonvegetated (water or barren) to healthy crops and plants with
a high leaf area index (
LAI
) (Jiang
et al.
2006). Beyond being a
metric for crop health,
NDVI
is frequently correlated with crop
production and yield, thus many farmers find value in map-
ping
NDVI
across their fields (Wall, Larocque, and Léger 2008).
Brad G. Peter is with The University of Alabama, Department of
Geography, Farrah Hall, Tuscaloosa, AL 35405 (
).
Joseph P. Messina is with The University of Alabama, College
of Arts & Sciences, Clark Hall, Tuscaloosa, AL 35487.
Jon W. Carroll is with Oakland University, Department of
Sociology, Anthropology, Social Work, and Criminal Justice,
Varner Hall, Rochester, MI 48309.
Junjun Zhi is with the Anhui Normal University, School of
Geography and Tourism, Wuhu, 241002, China.
Vimbayi Chimonyo is with the University of KwaZulu-Natal,
School of Agricultural, Earth, and Environmental Sciences,
Scottsville 3290, Pietermaritzburg, South Africa.
Shengpan Lin (deceased) was with Michigan State University,
Social Science Data Analytics Initiative, East Lansing, MI
48824 at the time of research.
Sieglinde S. Snapp is with Michigan State University,
Department of Plant, Soil, and Microbial Sciences, Plant and Soil
Sciences, Michigan State University, East Lansing, MI 48824.
Photogrammetric Engineering & Remote Sensing
Vol. 86, No. 2, February 2020, pp. 107–119.
0099-1112/20/107–119
© 2020 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.86.2.107
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
February 2020
107
