nspectus of the
ijing, pp. 8-12.
Remote sensing
Mountain, Acta
ons in China: a
100, pp.34 -37.
ed deforestation
northeast China.
and patterns of
n the Changbai
idscape Ecol.12
transitions and
analysis of land
J Remote Sens,
ind Fundamental
26305), National
Young Scholars
editors for their
DEVELOPMENTS IN MONITORING RANGELANDS USING REMOTELY-SENSED
CROSS-FENCE COMPARISONS
Adam D. Kilpatrick **, Stephen C. Warren-Smith ® John L. Read “, Megan M. Lewis ^, Bertram Ostendorf *
* School of Earth and Environmental Sciences, The University of Adelaide, Adelaide, SA 5005, Australia -
(adam.kilpatrick, megan.lewis, bertram.ostendorf, john.read)@adelaide.edu.au
® Institute of Photonics and Advanced Sensing, The University of Adelaide. Adelaide, SA 5005, Australia —
stephen. warrensmith@adelaide.edu.au
*Corresponding author
Commission VIII, WG VIII/8
KEY WORDS: Agriculture; Environment; Land Use; GIS; Landsat; Land Cover; Monitoring
ABSTRACT:
This paper presents a new method for the use of earth-observation images to assess relative land condition over broad regions, using
a cross-fence comparison methodology. It controls for natural spatial and temporal variables (e.g. rainfall, temperature soils,
ecosystem) so that we can objectively monitor rangelands and other areas for the effects of management. The method has been tested
with small and large scale theoretical models, as well as a case study in South Australian rangelands. This method can also be applied
in other systems and experiments such as field trials of crop varieties as a robust spatial statistic.
METHODS
Cross-fence sample pairs are often used in field ecology to
control spatio-temporal variation allowing a direct comparison
of grazing pressures. In our method, we extract a very large
number of cross-fence pairs of pixels from a vegetation index
(or land cover) raster (figure 1). The average ratio of vegetation
cover from this cross-fence pair is determined for each fenceline
in a system, and this cross-fence ratio is used to generate an
equation for the value of each paddock. The resulting series of
homogenous linear equations is solved in order to rank each
paddock objectively.
Paddock —4——-—*
Ec Sample
pair
pO
à à
Ceo]
D-H0
zs
:
Fence——s|
Figure 1. Cross-fence sampling layout
RESULTS
We have tested this methodology successfully using small and
large model simulations, showing that it reproduces expected
rankings in those scenarios, and that weighting factors have
their desired influence if applied. Our theoretical models show
that the expected land-cover rankings are accurately predicted
for large systems of paddocks and that grazing gradients as a
result of the piosphere effect have an influence on paddock
rankings. We have applied this monitoring methodology to
Landsat TM images 6 years apart in a region of pastoral and
mining leases in arid South Australia: changes in land condition
rank over time conform to those expected as a result of
documented changes in management over the study period, as
per expectations documented in various published ecological
studies of the region.
CONCLUSIONS
This methodology is a significant breakthrough in the analysis
of remotely sensed data in order to monitor fenced rangelands.
It has the potential to be applied as the mainstay monitoring
methodology to detect both good management and management
leading to overgrazing in millions of square kilometres of
rangelands in Australia and internationally. Its key attribute is
its ability to rank paddocks against each other, allowing
comparisons with paddocks of known management to inform
decision making. Following on from this research, several new
avenues are being explored, including spatial models of the
effects of wind and distance to water on grazing distribution of
animals in rangelands, methods of defining appropriate scales of
experimental design in cross-fence studies, methods to automate
the detection of fencelines and pseudo-fences from imagery and
the potential to apply this type of analysis to other scenarios,
such as in field trials of crops using field data.