Instrumentation & Measurement Magazine 24-7 - 47

asphalt are connected to cabinets which detect changes in the
impedance of the coil due to the vicinity of metallic objects, as
shown in Fig. 4a. These coils measure the velocity and length
of vehicles in each direction. The cabinets forward the traffic
measurements to a central server, in open access via a Rest API,
with a coarser aggregation: hourly average of both directions
and by vehicle length or type (bike or vehicle). Table 1-Table
1 shows an excerpt of the processed data, the hourly bidirectional
traffic at a measurement point, downloaded via the API.
Fig. 5a is a snapshot of a 4D Replay animation in the GDT displaying
the week-by-week comparison of hourly traffic. The
columns, placed at the position of the measurement stations
in the GDT, fluctuate dynamically according to the time series
data. The blue, red, and green columns represent traffic
before lockdown, during, and the difference between them,
respectively. In contrast, Fig. 6a depicts a decontextualized
2D representation of the time series at one of the measuring
points, suggesting a decrease of 30% of the traffic due to the
Covid-19 lockdown.
With respect to air pollution, the historic air quality data
curated by the Norwegian Air Research Institute (NILU) encompasses
the hourly means for Particulate Matter smaller
than 10 μm, or PM10, NOx, NO, and NO2
in the city centre
of Ålesund. Fig. 6b shows the timeseries for PM10. The Norwegian
maximum regulated mean daily value is added to
indicate that the threshold was reached only one day in March
2020. The value peaked in the first week, then levels began to
drop, and the values for PM10 stabilized around 12 μg/m3
before
the lockdown was announced (see the drop in traffic for
comparison in Fig. 6a). We cannot infer causal relationships
between air quality and the lockdown in this short period.
This indicates that air pollution is a multi-variable problem
which is not easily explained through a visualization for such
a small time window. The secure infrastructure supporting the
gathering of private energy consumption data and displaying
them in a privacy-preserving manner is detailed in Fig. 4b.
The transfer of the data from the consumers is handled by an
advanced measuring and control system (AMS). It is installed
at the premises and centralized in a nationwide datahub. The
data can be visualized as a heatmap, as shown in Fig. 5b, or a
time series, as plotted in Fig. 6c. While the time series allows
a better fragmentation of the data into aggregated categories
such as residential or industrial and commercial, thus identifying
a reduction in consumption in industries in the two later
weeks, the heatmap shows the data in a spatiotemporal context,
with industrial districts clearly identifiable with the red
zones compared to the residential areas in green.
Finally, the demographic data presented in this case study
was collected by the mobile telephony company Telia, using
a triangulated cell-phone position. Fig. 4a sketches the infrastructure
for gathering, storing, and presenting aggregated
data from complementary sources: the automatic passenger
counting from the public transport company (FRAM), the mobility
patterns from the mobile phone company (Telia), and
vehicle counts from SVV. These partially overlapping sets give
a holistic picture of mobility in the city. The data received for the
October 2021
study are aggregated per hour and per basic geographical unit.
To ensure complete compliance with GDPR, it is convenient to
aggregate data on a common basic geographical unit, defined
by the Norwegian central bureau of statistics, largely satisfying
a minimum of five persons per group. Fig. 6d, Fig. 6e, and Fig.
6f depict hourly scatterplots of the number of people in each
basic geographical unit in the few days before (blue) and during
(red) the weeks of lockdown in various districts. The data
suggest a dampened circadian variation during the lockdown
in the academic and residential districts, while the recreational
area experienced more daily variations during the weekdays.
Discussion
The cases illustrate the usefulness of visualization of measurements
and draw our attention to different facets of smart cities.
The Covid-19 pandemic has created natural experiments in
cities that would have been impossible to implement under
normal circumstances. The causal effect of the lockdown on
the traffic decrease is clearly identifiable. This underpins the
importance of understanding the measured data and the context
when interpretating them. Nevertheless, the case study
of air pollution showed the limitations of inspecting the data
in such a narrow time window and requires further investigations.
For example, using time series from past years,
predictive models could be compared with the real data and
visualized in the GDT.
Some of the challenges local governments face to win the
trust of their citizens are protecting their privacy and providing
them with accurate and transparent information. The
GDPR compliance of the system is achieved by aggregation,
filtering, and fragmentation of the storage. No critical data
leave the dedicated IT premises, and only privacy-compliant
data is exchanged with and stored on the 4D system. This introduces
a necessary step, and while it might lead to latency,
real-time information is not critical for the long-term planning
decision process.
The information accuracy is addressed in Table 3, which
lists the level of uncertainty announced for each measurement.
First, the electricity company reports that the AMS
under scrutiny present an accuracy in the regulated range,
with 0.2% for a 2% normative range. An AMS accuracy of 2%
seems very high, but when national annual electricity productions
of European countries are in the 10 to 100 TWh range,
this means a discrepancy of 0.2 to 2 TWh, which converts to a
yearly cost of millions of euros carried either by producers or
consumers.
Secondly, inductive loops and SIM card data accuracy are
not regulated by national directives in Norway, but their accuracies
are necessary for traffic estimation and calibration
of mobility simulation models as they often are deemed as
" ground truth. " The accuracy estimations reported by the
Swedish and Norwegian Road Authorities (Trafikverket and
SVV) are deemed appropriate. Finally, the European Standard
for gravimetric measurement method for the determination
of the PM10 and PM2.5 mass concentration of suspended particulate
matter (EN 12341:2014) requires an uncertainty of
IEEE Instrumentation & Measurement Magazine
47
Instrumentation & Measurement Magazine 24-7

Table of Contents for the Digital Edition of Instrumentation & Measurement Magazine 24-7
