Page 293 - Special Topic Session (STS) - Volume 2

STS493 Sofie d.B. et al.
concerning their fields. The number of generally used systems is limited to 3
or 4. The input for these systems can be both manual (H2S) and digitally (S2S)
covering many subjects NSIs are interested in. These include: parcel and crop
registration, pesticide usage, fertilization and harvest yields.
Even though this platform still has to be explored, harvest registration
systems show great potential to be used as an alternative to surveys, as they
are a rich source of information. Furthermore, the systems already have a
build-in export function. A function that is being used to pass on information
to stakeholders and a number of regulatory agencies alike. Here the farmers
remains in full control which data are shared and which are not, acting as a
gatekeeper. As for NSIs, these systems offer great opportunities to close the
data circle by S2S gathering of data and returning valuable statistical input
back into the systems.

5. Discussion
We contribute to existing literature in two ways: We propose a set of
criteria to support cost-benefit assessments of sensor measurements and
sensor data and we illustrate the criteria for two real-life cases. The criteria are
constructed from three viewpoints. The first viewpoint is from the perspective
of the survey itself; does the survey contain topics or questions that may
benefit from automated measurements. The second viewpoint is that of the
sensor: What are accuracy and costs of the sensor options. The final viewpoint
is the respondent: How does the respondent react to a request for sensor data.
One side remark is in place. Our assessments of the survey, sensor and
respondent criteria are subjective. This is partly for the very reason that they
are new and have not been tried in practice. It is very hard, for example, to
predict willingness to provide sensor data (or to consent to sensor data
linkage) independently of the context. Another reason is that sensors and
wearables by themselves show variety in accuracy and costs, even within
mobile devices, so that it is hard to judge about quality. It is imperative that
the assessments are made more rigorously by consulting multiple experts.
Also this exercise will be part of an in-depth follow-up paper.
In our inventory, we distinguish sensor measurements initiated by the
survey institute and existing sensor data. Although the data may originate
from the same type of sensors, the context of the two is very different. Self-
initiated sensor data require a data collection infrastructure and lead to direct
data collection costs. Obviously, other aspects such as data processing, data
storage, privacy and legislation are very different as well. In this paper, we
consider secondary data as complementary to survey data, i.e. we ask
respondents to consent to linkage. Consequently, a hybrid form of data
collection arises. However, such a combination of data sources is no goal by
itself; secondary data may provide the sole source of data for specific topics.

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