Page 281 - Special Topic Session (STS) - Volume 2
P. 281
STS493 Stéphane D. et al.
being reinvested into the development of tools necessary to support collection
from alternative data sources such as scanners.
This new source of data required new processing tools to be created
outside the traditional system. The scanner data need to be pre-processed to
link the products to the CPI classification—machine learning is now being used
for this classification process.
Further developments are being considered, such as automated
substitution of products and the use of multilateral methods of index
calculation (i.e., the use of all data over time rather than a sample that mimics
field collection). Both of these developments are beyond the scope of the
three-year plan and would not be implemented before 2021.
Using sensors to collect information—satellite and telemetry
Satellite imagery is a key component of the Agriculture Statistics Program.
Statistics Canada has collected vegetation index values from satellite imagery
since the 1990s to support the Crop Condition Assessment Program, a web
mapping application that depicts crop and pasture conditions across Canada
in near real time. This data source, coupled with climatic data, is the foundation
for the crop modelling project. The results of this project have been accurate
enough to replace traditional collection methods for the September Field Crop
Survey since 2016, eliminating more than 9,000 phone interviews. In 2019,
Statistics Canada is looking to expand the modelling approach of the Field Crop
Survey to further reduce response burden.
In addition, Statistics Canada successfully used a combination of crop
insurance data and a crop classification map produced by Agriculture and Agri-
Food Canada with medium-resolution satellite imagery to estimate crop area.
Results were primarily used for validation at first, and this method is another
potential way to reduce response burden for the Field Crop Survey and future
censuses of agriculture. A 2019 pilot project is looking into updating crop area
and yield on a weekly basis, at the parcel level, as the growth season progresses
(in-season estimates) using near real-time satellite imagery, climatic data and
crop insurance data.
Statistics Canada is also working on an innovative project with the Canadian
Food Inspection Agency using the traceability data managed by the Canadian
Pork Council, which collects group movement data. Statistics Canada is using
data science and leading-edge methods to clean, process and use the data to
create real-time modelled pig inventories by location, along with the
probabilistic movements of each animal in the group. This partnership on
traceability in the pork industry is an excellent opportunity to benefit each
organization’s goals. Statistics Canada will have information on pig movements
and the inventory required for its statistical programs. The Canadian Food
270 | I S I W S C 2 0 1 9