Page 292 - Special Topic Session (STS) - Volume 2
P. 292
STS493 Sofie d.B. et al.
getting high enough response rates for agricultural surveys has always been
cumbersome. Secondly the agricultural sector is innovative and developments
like smart and precision farming seem promising within the context of official
statistics (De Vlieg, 2018; Vonder, 2017; Thomas & McSharry, 2015).
In order to run his farm, in the context of precision farming the innovative
farmer uses many sensors throughout the year. During winter he scans his
fields for soil characteristics like humidity, sun coverage and nutrients. During
spring, new crops are planted in such a way that it matches the yield potentials
of field sections. In the summer, crop growth is carefully monitored, by
measuring soil humidity, nutrients in the plants, and the need for pesticides.
When the crops are harvested in autumn, the crops are weighted on the spot,
providing detailed insights in field sections with high and low yield.
The farmer’s data are stored in two different data platforms: a platform for
precision farming data, and the so called ‘harvest registration system’. The
precision farming platform is specifically tailored to the farmer’s needs and
houses the sensor-generated data. A first exploratory analysis of this database
shows that parts of questionnaires can be imputed using these data, while
other parts still have to be completed manually: data on fields and harvest
cover the requested information, while for the planting, plant treatment, and
number of employees parts of the requested data are missing in this database;
financial information requires other data sources altogether. As a result, we
concluded that parts of these sensor data are associated with some of the
concepts that are operationalised in questionnaires. However, before the data
could be analysed two steps had to be taken. First of all, a complete and
accurate description of the data was missing. The next step was data cleaning:
the data suffered from e.g. measurement errors (incorrect values), and
duplicate records. Some errors could be corrected by checking the data, but
in order to get a complete understanding of the content of the data the
farmer’s expertise about the metadata and the data generation process was
indispensable. While precision farming platforms such as this are promising, it
currently lacks the required data management and omnipresence necessary
to produce high-quality statistics. This of course is subject to change based on
developments in the industry. In addition, other challenges need to be
addressed when using these kinds of decentralized platforms: localizing them
in the first place can prove to be challenging, as well as getting access to the
data; this relates to issues like data ownership, data sharing, and trust, among
others. When getting data from various providers data harmonisation may
also be an issue: sensors produced by various manufacturers and used by
various farmers may be generating different kinds of data. These issues could
be overcome by the establishment of data service enters (DSCs). The second
data platform is a so called ‘harvest registration system’, which is used by many
farms in the Netherlands to import, store, interpret and share relevant data
281 | I S I W S C 2 0 1 9