# *translation:* Toepassingen en uitdagingen van sensordata [Applications and challenges of sensor data]

###### tags: `data`

Toepassingen en uitdagingen van sensordata [Applications and challenges of sensor data]
Joost Wesseling, Christa Blokhuis, Henri De Ruiter, Derko Drukker, Ernie Weijers, Hester Volten, Jan Vonk, Lou Gast, Marita Voogt, Peter Zandveld, Sjoerd Van Ratingen, Erik Tielemans

https://www.researchgate.net/publication/337167846_Toepassingen_en_uitdagingen_van_sensordata

OCTOBER 2019 NUMBER 3 MAGAZINE AIR

## Applications and challenges of sensor data

*Measuring together: where are we in 2019?*

Measurement of air quality has changed in recent years. Citizens are increasingly using new, cheap sensors. Sensor networks are created in cities and in the countryside. RIVM has responded to this and helped shape the development by testing and calibrating sensors and by creating a knowledge and data portal. In this article, we consider some of the questions that are currently pertinent. What challenges have we already overcome? Which ones await us?

### Measure innovations in the environment

The development of sensors makes it possible to measure air quality extensively and inexpensively. ==In 2016, RIVM therefore launched the Innovation Program for Environmental Monitoring within its activities for the Ministry of Infrastructure and Water Management. This is a five-year program with the aim of refining the national measurements and increasing the involvement of citizens in the living environment.== For governments and scientists, sensors are a cheap addition to official measurements and offer possibilities to measure with a higher frequency and resolution. In addition, sensors open doors for all kinds of innovative applications of measurements linked to GPS and / or mobile phones, bringing the measurement of individual exposure closer. An example is the 'Snuffelfiets' project of the province of Utrecht, in which residents from different cities get a particulate matter sensor on their bicycles to determine the air quality of bicycle routes.

The emergence of inexpensive sensors presents challenges for reference institutes. ==At first glance, it seems logical to test sensors and only start talking to citizens and governments if they meet official measurement criteria. However, RIVM has chosen to seek cooperation with citizens at an early stage.== However, in order to support sensor data from citizen measurements and to make the data usable, it is not only necessary to calibrate and apply the data, but also to set up infrastructure and communication channels. RIVM has given substance to this by developing the knowledge and data portal and being open to feedback from users in this process.

In light of the official use of measurements, a reference institute could decide to standardize sensor measurements quickly, by prescribing which sensor(s) to use. However, currently the technology is not yet fully developed and data from sensors is not of the same quality as that of official measurements. In addition, there are various citizen groups who experiment with sensors and share their findings. ==By imposing a standard in this phase, these smart citizens can feel limited in their freedom and initiatives that arise through city labs and citizens' associations can be inhibited.== There is also a risk that the data of these citizen groups will not be used, and that would be a shame.

In short, it is important to weigh the desire for standardization against the wishes of a wide group of stakeholders. This includes parties that collaborated less easily or intensively with RIVM, including citizens, municipalities and companies. The 2019 'Measuring Together' community has become a household name in the Netherlands in 2019. The facilities and the community are built up step by step and in consultation. The growing interest of citizens in air quality and the possibilities of current sensor technology formed a starting point for the knowledge portal (www.samenmeten.nl). Anyone who wants to measure themselves can find the necessary information here.

Where RIVM develops the 'Measure together' community in interna national consultations is one often amazed that such together operation between a reference institution and citizens exist.

*Figure 1: Display of the RIVM data portal on June 26, 2019. The particulate matter sensors show high concentrations of PM2.5 see in the southeast of the Netherlands. The official measurements show the same pattern.*

### Data portal

The data portal (https://samenmeten.rivm.nl) (Figure 1) is a central place where RIVM accesses the data from various sensors. ==All data that is made accessible via the data portal is publicly available.== To keep it that way, ==it is important that the data management is entrusted to an independent party.== By offering a platform for all interested parties or individuals, users can focus on applying their measurement results. In recent years, the data portal has grown into a place where more than ten projects and various individuals share their data. RIVM also collects data from Dutch participants in Luftdaten (https://luftdaten.info/nl). This citizen science project, originally from Germany, makes all their sensor data publicly available. Recently it has become possible to view the monthly averages of NO2 Palmes tubes in addition to real-time sensor data on the data portal.

> Data from sensors is not the same quality like that of official measurements.

The data on the RIVM data portal has recently been accessible via an Application Programming Interface (API). This makes it possible for everyone to download the desired amount of data from one or more sensors of their choice. A download function will be developed shortly for users who are not familiar with APIs. In addition to the current visualisations with interactive graphs on the data portal, RIVM is working on more ways to present the sensor data. Experiments are currently being conducted with time series, wind roses and summaries per week or month. Before the sensor measurements on the data portal are available, they must be sent from the measurement location to the server. In recent years, RIVM has tried various so-called Internet of Things techniques, with varying degrees of success. Connecting a sensor to the Internet via a Wi-Fi connection requires a number of steps from the user. This makes this technique especially suitable for more technically savvy people. ==A Wi-Fi connection sometimes drops out, which can lead to missing data.== LoRaWAN (Long Range Wide Area Network) is a form of internet communication that does not require Wi-Fi or 3G. It is offered by various parties and has been specially developed for longer distances. RIVM uses one of the non-commercial systems, “The Things Network” (TTN). ==Linking a measuring point to the LoRa network is fully automatic and is therefore easier than Wi-Fi. A disadvantage is that LoRa is mainly available in urban areas in the Netherlands and that it is not practical for RIVM to set up its own gateways in rural areas.== A third variant, Narrow Band IoT (NB-IoT), currently seems to be the most suitable solution for sending data from air quality sensors because it is not susceptible to interference and has national coverage.

*Figure 2: Maps of the dust cloud who passed over the Netherlands during Easter 2019. Use has been made of the 45 official reference stations and about 350 fine particles sensors. The dust cloud existed mainly from PM2.5. Legend: blue = 0 µ / m3; red = 130 µ / m3. The cards expire on April 20 23:00 (top left) to April 21 14:00 (bottom right), in steps from 3 o'clock.*

### Calibration of sensors

The current generation of sensors is characterized by varying performance, so that the data is not always immediately usable. RIVM is therefore experimenting with a number of calibration methods. For the nitrogen dioxide sensor B43F from Alphasense, a calibration function has been derived that includes temperature and ozone. A detailed explanation of the calibration of this NO2 sensor can be found at www.samenmeten.nl or in the article by Van den Elshout et al, 2019. Currently, about five hundred particulate matter sensors supply data to the data portal. This creates the possibility to combine the sensor data and the official measurements in the hourly maps for air quality (www.luchtmeetnet.nl), so that this map provides a refined picture of places where citizens measure. An experimental example was the measurement of particulate matter from the Easter fires in 2019. Most of the sensors on the data portal do not know under which circumstances the measurements take place, while ==it is known that particulate matter sensors are sensitive to air humidity==, among other things. That is why a 'on-the-fly' calibration is applied. RIVM is experimenting with two techniques for this.

In the first ‘on-the-fly’ calibration technique, a function is derived based on the relative humidity. It is believed that this function can be used for other sensors in nearby locations as groups of particulate matter sensors (type SDS011) provide similar results and humidity appears to be the most important factor in calibration.

With this information in mind, it is possible to use a map of measured relative humidity as a map with corrections for the particulate matter sensors. The real-time data from the KNMI is used for this. Many fine dust sensors are equipped with their own sensor for some meteo data such as pressure, temperature and relative humidity. ==The experience of RIVM is that these, also cheap, sensors break quickly and are therefore not a reliable source for calibration.== The KNMI measuring stations are spread over several dozen measuring points in the Netherlands. This means that the available humidity information at a specific location is usually approximate. ==The uncertainty this entails in the calibration is offset by the fact that the humidity changes only relatively slowly in space and time and that the accuracy and stability of the official meteorological data is very good.==

*Figure 3 Ratio between the hourly official measurements for PM2.5 at Amsterdam station Vondelpark and the average of 20 (Nova SDS011) sensors in the center of Amsterdam as a function of (100RH), where RH is the relative humidity in%. The graph shows the sensitivity of the sensors for high humidity and the corresponding correction factor (“over estimation”). There has been made use of data for the whole year 2018 (sensors and reference station) and the humidity contents of meteorological station Schiphol.*

The second ‘on-the-fly’ method uses the ratio between the measurement results of the official measurements and a group of nearby sensors. This makes the use of meteorological data unnecessary. Since the contribution to the concentrations of PM10 and PM2.5 from local sources in the Netherlands is on average low (often on the order of a few μg / m3), the measurement results of sensors for calibration can in most cases be compared with those of a reference station up to several kilometers away (Figure 1). The resulting ratios of concentrations at all locations with both sensors and official measurements can be interpolated for the whole of the Netherlands, again creating a map with corrections. The correction is applied separately for each individual sensor; the calibrated result is therefore not an average of the environment. RIVM is currently testing both methods and is also looking at the uncertainties of the methods.

The data portal shows the calibrated value of particulate matter sensors based on the second ‘on-the-fly’ calibration. RIVM will continue to work on improving sensor calibration in the coming years. Ultimately, it is expected that a selection can be made of sensors that can be (partly) applied in official monitoring and modeling.

> The correction is applied separately for each individual sensor

### Developments in sensor data

==The extent to which sensor data can be used for science and policy depends in part on the knowledge available about the possibilities and limitations of sensor technology.== As a reference institute, RIVM tries to keep up to date and to communicate about developments in this area. Rather than determining whether sensor data is good or bad, ==adding metadata helps users interpret sensor data. This is an important next step in the development of the data portal.== RIVM is also actively involved in international forums and working groups, such as the CEN Technical Committee, which are involved in the development of measuring standards for sensors. Recently, RIVM contributed to an article on the classification of data derived from sensor measurements, varying from raw data to highly processed data that also includes other data. (https://pubs.acs.org/doi/10.1021/acs.est.9b03950)

In addition to monitoring the qualitative, technical aspects of sensor data, it is at least as important for RIVM to keep an eye on the social significance of citizen science projects, as researched in the Boeren & Buren (Farmers & Neighbors) project (https://www.rivm.nl/boeren-en-buren). Here agricultural entrepreneurs, local residents and the municipality of Venray, under the guidance of the RIVM, jointly measure air quality. At a later stage, RIVM will publish more extensively about this project.

### Future vision

In recent years, RIVM, as part of the work for the Ministry of Infrastructure and Water Management, and in collaboration with citizens and others er parties, an acquaintance and data portal realised. There are various contributions to this supplied in the areas of sensor calibration and interpretation, and is the RIVM became part of a community of enthusiastic, self-measuring citizens. In the RIVM will update the data vapor in the near future broaden language with data from front sensors water quality and sound but also with more citizen sightings, for example about odor or other nuisance. Other next steps are integrating of satellite data in hourly maps and focus more on individual exposure as in the aforementioned project Sniffing bike. There lie with the data portal also opportunities to contribute to the implementation of the Clean Air Agreement, where citizen participation is important become part of. Furthermore, it provides RIVM to participate more often in projects like 'Farmers and Neighbors'. For reference institutes and comparable organizations it remains important to act as an independent act in the assessment of data, regardless of whether they were obtained with official measuring equipment or sensors. RIVM will remain in the coming years respond to developments in sensor technology and social questions that arise here.

*All authors are part of the 'Samen Meten' - team of the RIVM.*

This article is based on Wesselink et al., 2019. Development and Implementation of a Platform for Public Information on Air Quality, Sensor Measurements, and Citizen Science. Atmosphere. 10 (8), 445. https://doi.org/10.3390/atmos10080445.