Mapping Irrigated Farmlands Using Vegetation
and Thermal Thresholds Derived from
Landsat and ASTER Data in an Irrigation
District of Australia
Mohammad Abuzar, Andy McAllister, and Des Whitfield
Abstract
An operational approach has been presented to demonstrate
how bilevel classes of satellite-derived surface temperature
and vegetation status can be jointly used to detect areas of
current irrigation with sufficient spatial detail for the po-
tential benefit of both growers and resource managers. An
iterative thresholding procedure to minimize within-class
variance was used for bilevel segmentation. The approach
was tested in an irrigation district of south-eastern Australia
in 2012/2013 crop year. The overall accuracy of identify-
ing farms with irrigated crops amounted to 88.4 percent.
Seasonal data of active irrigation and growing vegetation
helped define land cover classes for the better understand-
ing of current management practices of irrigated crops.
Introduction
Irrigated agriculture plays a vital role in meeting food and
fiber demands (Colaizzi
et al
., 2008). There has been an
added reliance on irrigated agriculture around the world after
dryland agriculture has been adversely impacted by recent
climatic changes (Aleksandrova
et al
., 2014). At the same
time, the irrigation sector is facing the challenges of compet-
ing demands for limited water resources. These conditions
have led to a sharp focus on how irrigation water and irrigat-
ed agriculture are managed, not only to achieve higher water
productivity but also to increase the sustainability of irrigated
agriculture (Santos
et al
., 2010). One key information require-
ment for effective management and decision making is up-to-
date and spatially-detailed information of irrigated farmlands.
This study presents an operational approach to map irrigated
areas with sufficient spatial detail by using the matrices of
satellite-based surface temperature and vegetation cover.
Satellite Remote Sensing has been widely used to map
cropland over a range of locations for the past few decades
(Ozdogan
et al
., 2010). However, operational approaches for
routine monitoring of irrigated agriculture are very rare, and
this is considered a knowledge gap (Thenkabail
et al
., 2012;
Thenkabail and Wu 2012). The current status of Remote Sens-
ing approaches to cropland mapping has been adequately rep-
resented by a range of studies published in the recent special
issues of two separate journals (
Remote Sensing
, 2010(2):9;
Photogrammetric Engineering & Remote Sensing
, 2012(78):8).
There have been far fewer studies specifically on irrigated ag-
riculture as compared to agriculture in general (Ozdogan
et al
.,
2010). This may be due to the limitation of optical sensors, a
very low number of thermal sensors onboard satellites, and/or
inappropriate spatial resolution of satellite images used to de-
tect irrigation. Many studies reported irrigated mapping only
within designated irrigation areas, or in areas commanded by
established irrigation infrastructure. This excludes irrigation
from alternative sources such as river/stream diversions and
farm dams. Reported changes in irrigation vary widely de-
pending on the scope of irrigation expansion and availability
of water (Brown and Pervez, 2014; Shahriar Pervez
et al
., 2014;
Thenkabail and Wu, 2012). In some cases there may be modest
changes in the overall irrigated area nationally, whereas the
regional patterns in gains and losses may vary significantly
(Brown and Pervez, 2014). Remote Sensing as a technology has
the potential to provide a spatially detailed irrigation mapping
for a more complete understanding of changes at local and
regional level (Thenkabail, 2010; Thenkabail and Wu, 2012).
Surface temperature has long been recognized as an indica-
tor of ground moisture and crop water (González-Dugo
et
al
., 2006; Idso
et al
., 1981). The differences in land surface
temperature can potentially be related to irrigation (Ozdo-
gan, 2008; Ozdogan
et al
., 2010; Thenkabail and Wu, 2012).
However, there is a lack of detailed studies using temperature
information for irrigation mapping (Ozdogan
et al
., 2010). In
this study, we introduce an approach to incorporate the rela-
tive differences of surface temperature to detect the areas of
irrigated agriculture. Irrigated areas are generally cooler than
the non-irrigated areas. Thermal sensors provide instanta-
neous temperature status on the surface, which is valuable in
assessing irrigation application.
The objective of this study was: (a) to detect irrigated areas
by using surface temperature and vegetation status derived
from satellite data, and (b) to achieve a land cover classifica-
tion by using the seasonal profile of irrigated areas. We tested
our approach in an irrigation district in south-eastern Austra-
lia. As a measure of validation for our mapped irrigated areas,
we have used information on irrigation water deliveries to
determine when farms were actually irrigated.
Irrigation is generally defined as the application of water
to crops in order to maintain and enhance crop growth. In the
context of irrigation management in an agricultural system,
important consideration is given to the amount of irrigation
Department of Environment and Primary Industries,
32 Lincoln Square North, Carlton, VIC 3053, Australia
(
).
Photogrammetric Engineering & Remote Sensing
Vol. 81, No. 3, March 2015, pp. 229–238.
0099-1112/15/813–229
© 2014 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.81.3.229
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
March 2015
229
