Page 265 - Special Topic Session (STS) - Volume 4

STS583 Yakob M. S.

However, it is evident that these improvements come with additional costs
and therefore having a clear understanding of the cost efficiency would be
vital.
Remote sensing costs can be broadly divided into two categories: a) image
purchase and b) data treatment (purchase and maintenance of hardware and
software, recruitment of staff, and training etc.).
The cost-efficiency of using remote sensing in agricultural statistics can be
evaluated by comparing the gains obtained (usually expressed as a reduction
inf sampling variance) to the additional costs involved (cost of imagery, data
analysis, staff training, and investment in hardware and software). Hardware
and software costs have drastically decreased in recent years due to open-
access software that are now widely available. With cloud-based image
analysis now a standard, low-cost personal computers and disk storages allow
for the analysis of very large image data sets. However, staff availability and
competence needs special attention (Latham, 2017). Multi-disciplinary
expertise in Geographic Information Systems (GIS), image analysis, statistics,
yield modelling, agrometeorology, soil science and crop science will be
required and therefore the bulk of costs will be incurred in this respect.
Thanks to initiatives undertaken by the National Oceanic and Atmospheric
Administration (NOAA), the U.S. Geological Survey (USGS) and the European
Space Agency (ESA), vast real-time freely accessible depositories allow for
downloading or online processing thereby tackling what the United Nations
Security Council considers the Big Data challenge (2015). Currently, high-,
medium- and low-resolution imagery is freely available in raw format and as
derived products, such as geometrically (RMS 1.5 pixels) and radiometrically
(top of atmosphere) rectified imagery, vegetation indices, regional or country
mosaics, and periodic cloud-free coverage. However, the VHR imagery with a
ground sampling distance (GSD) lower than 5 m have to be still purchased.

2. Methodology
In this paper, the cost benefit for the use of remote sensing in agricultural
statistics in relation to its applicability in optimizing the sampling design of
agricultural surveys, improving estimators and crop monitoring and yield
forecasting will be discussed.

3. Result
3.1 Optimization of sample design
During agricultural censuses or agricultural surveys, the primary activity is
to have a clear delineation of the primary sampling units, usually coinciding
with Enumeration Areas. An efficient method to define EAs is the use of
imagery (having a resolution from 0.5 m to 2 m) in a GIS environment, seeking
to subdivide the entire territory into entities with physical limits corresponding

254 | I S I W S C 2 0 1 9