Antonio Poggi
← All writing

Citizen science · Air quality

From Monitoring Taranto to Building Citizen-Owned Air-Quality Instruments

Seven years of dashboards, public alerts, civic evidence and one very geeky new experiment.

26 July 2026 11 min read Taranto to the sensor
Hourly benzene measurements at Via Machiavelli, including two readings above the project’s 27 micrograms per cubic metre reference level
A July 2025 public Omniscope view showing the evidence behind an automated benzene alert.

In January 2019, I published an article called “Air pollution: looking after my hometown”.

My hometown is Taranto, in Puglia, southern Italy: a beautiful city by the sea, living for decades beside one of Europe’s largest industrial sites and with the impossible argument of “health versus jobs” hanging over its citizens.

I was living in London, building Omniscope with the team at Visokio, but I kept thinking about what I could contribute using the tools and skills I had.

The question I asked was simple:

Could data, analytics and visualisation help ordinary citizens understand what was happening to the air around them?

That question became a project. The project became a public monitoring system. And, seven years later, it has led me all the way from published environmental data to open hardware and a physical sensor that sends measurements directly into Omniscope.

This is the continuation of that story.

Making public data genuinely accessible

Arpa Puglia, the regional environmental protection agency, publishes validated air-quality measurements from monitoring stations across the region.

The data was public, but “publicly available” and “easy for the public to use” are not the same thing.

Downloading files, combining dates, understanding pollutant fields, comparing stations and building charts creates a barrier. Most citizens should not need to become data engineers before they can ask whether pollution is changing in their neighbourhood.

I used Omniscope to build an interactive public dashboard for Taranto. It allowed people to:

  • see a summary of pollutant concentrations;
  • compare monitoring stations and areas;
  • explore the trend of an individual pollutant;
  • inspect all pollutants measured by a selected station;
  • compare distributions over different periods;
  • filter and drill into the underlying observations;
  • access the raw data behind the visualisations.

I later extended the history back to January 2016 and published an Italian version so the report would be useful to the people it was intended for.

This was my practical interpretation of data democratisation: not merely releasing a file, but making the information understandable, explorable and open to questions.

The missing-data moment

While building that first dashboard, I remembered a photograph I had taken near Taranto’s industrial area at 10:30pm on 2 November. Orange clouds were glowing against the night sky.

I wanted to inspect the corresponding PAH measurements from the Cokeria and Tamburi monitoring stations.

The observations were missing between 2 and 4 November.

That did not prove why the data was absent, and I did not claim that it had been deliberately withheld. But it exposed something important: without accessible information, citizens and institutions had lost the opportunity to examine what had happened.

The dashboard was already doing more than displaying charts. It was making both evidence and gaps in the evidence visible.

From a dashboard to a sentinel

The project continued to evolve.

I added automated data refreshes and an hourly report focused on benzene measurements from the Via Machiavelli station in the Tamburi district. I then built an automation that published an alert when a reading exceeded the acute reference level of 27 µg/m³ used by the project.

The alert included the measurement, a screenshot and a link to the interactive Omniscope report. Anyone seeing the warning could move from the social post to the underlying evidence and explore it.

The public channel has changed over time. The original automation posted on X; more recently I moved the alerts to Facebook. The purpose has not changed:

Do not leave an important measurement buried in a dataset, waiting for somebody to go looking for it.

An alert creates attention. The interactive report provides context.

This distinction matters. A screenshot alone can be dramatic but static. A report allows people to examine the preceding hours, compare stations, inspect historical patterns and decide whether an event is isolated or part of something larger.

Omniscope report summarising benzene peaks and trends across Taranto monitoring stations from 2023 to 2025

By December 2025, the monitoring had expanded into a multi-station view of benzene peaks and average concentrations across Cokeria, Meteo Parchi, Via Orsini and Via Machiavelli. Source: my original post on X.

From citizen information to civic evidence

The work found an audience beyond my own experiments.

Local activists at PeaceLink began using the Omniscope dashboard to monitor trends, investigate worrying events and communicate environmental data. PeaceLink described the work as an experiment in environmental data democratisation and citizen-led monitoring.

The collaboration eventually led to a historical analysis of more than 80,000 hourly benzene measurements from the Via Machiavelli station between 2013 and 2023.

The analysis identified 63 observations over 27 µg/m³:

  • 32 occurred in 2023;
  • 31 occurred across the preceding ten years combined.

PeaceLink presented the findings during a hearing at the Italian Senate concerning ILVA. The official slides filed by the Senate credit me for the data analysis and its processing with Omniscope, alongside the wider group that contributed to the work. The results were also reported by ISDE and other environmental and news organisations.

I am careful not to overstate what a dashboard can achieve. Software does not remove pollution, make policy or replace scientific and institutional responsibility.

But it can make evidence easier to interrogate. It can shorten the distance between a measurement and public awareness. It can help citizens and activists ask better questions, preserve a historical record and hold decision-makers to account.

That is meaningful.

Monitoring is more than detecting peaks

The project also grew beyond individual warnings.

Recent reports have compared PM10 concentrations in Tamburi and Talsano, examined the relationship between falling steel production and benzene levels, and followed changes between 2023 and 2025 across Cokeria, Meteo Parchi, Via Orsini and Via Machiavelli.

These analyses show why history matters.

Omniscope PM10 comparison between Tamburi and Talsano from 2017 to June 2025

PM10 in Tamburi and Talsano, 2017–June 2025. The report combines averages, relative differences, monthly changes and longer-term trends. Source: my original analysis on X.

Benzene observations at Cokeria and Tamburi Via Orsini from 2022 to 2025

Benzene observations at Cokeria and Tamburi–Via Orsini, showing a falling trend while preserving the full distribution of measurements. Source: my original post on X.

A peak demands attention, but a trend helps us reason about causes and consequences. A single number can be alarming; a comparison can reveal whether an area experiences a systematic difference. A decline can be welcome while still leaving unresolved structural risks.

Good data communication must be capable of holding more than one thought at once.

The latest experiment: moving closer to the source

Recently, I bought an AirGradient ONE, model I-9PSL, to understand the air around me. It measures CO₂, particulate matter, TVOCs, NOₓ, temperature and humidity.

At first, the device answered an immediate question: what is the air like now?

Then the familiar question arrived:

What has it been doing over time?

I did not want to send these measurements through a vendor cloud. I did not want an always-on computer polling the monitor in my home. I did not want to open a router port, install an MQTT broker or operate a separate ingestion service.

AirGradient publishes open firmware and provides a local API. The monitor already had the sensors, data, Wi-Fi and a programmable ESP32-C3 microcontroller.

It did not need to be polled. It could push.

I built:

  1. a private Android app for local measurements and configuration;
  2. a customised version of the AirGradient firmware;
  3. a direct, authenticated HTTPS integration with the Omniscope Workflow API;
  4. an Omniscope workflow that parses each measurement, adds a server-side UTC timestamp, appends it to history and feeds an interactive report.

Once every minute, the monitor now sends its current measurement directly to Omniscope.

There is no AirGradient cloud in that data path, no middleware, no local collector, no inbound access to my home network and no separate Python service.

AirGradient ONE
      |
      | one outbound HTTPS request every minute
      v
Omniscope Workflow API
      |
      +--> parse JSON
      +--> add UTC timestamp
      +--> append history
      +--> update the interactive report

The phone controls the device locally. The device publishes remotely. Omniscope receives, processes, stores and presents the history.

Omniscope workflow receiving and organising AirGradient measurements before the report

The receiving Omniscope workflow: HTTP input, schema definition, field organisation, historical data table and Report block.

Interactive Omniscope report showing particulate matter, VOC, NOx, temperature, humidity and carbon dioxide

The latest experiment working end to end: minute-by-minute AirGradient measurements available for interactive exploration in Omniscope.

From democratising the data to democratising the instrument

This latest build operates at a different scale from the Taranto project.

An AirGradient in my room is not a regulatory monitoring station. A personal, low-cost sensor should not be described as a substitute for official, validated environmental measurements.

But the underlying democratic principle is the same.

In 2019, I started by making institutional data easier to explore.

The automated monitoring then made that data active, bringing significant measurements into public conversation rather than waiting for people to discover them.

Now I control the complete path from my own physical sensor to the analytical record:

air → sensor → firmware → HTTPS → workflow → history → report

This suggests a wider possibility.

Any open air-quality monitor with sufficient resources, Wi-Fi, accessible measurements and support for outbound HTTPS could use a similar architecture. A school, community building, local organisation or citizen-science network could own its data path while still benefiting from centralised processing, history, analytics and visual exploration.

The goal is not to pretend that every citizen sensor is a scientific reference instrument. It is to give people more ways to observe their environment, understand their own data and participate in the questions that affect them.

Why Omniscope belongs in this story

As COO of Visokio, I spend much of my professional life thinking about the journey from raw data to useful understanding.

Omniscope is often seen at the end of that journey, where the charts and interactive reports appear. This project demonstrates a broader role.

The workflow is also the operational API. It receives the device message, parses the JSON, timestamps it, organises it, appends it and makes it available for analysis.

The distance between a physical measurement and an explorable report is one outbound request.

This tiny project touches the same concerns found in much larger systems: integration boundaries, security, configuration, ingestion, data quality, history, automation and human interpretation.

And yes, it was also enormous fun to make a box of sensors on my wall call a workflow on a Linux VM and watch a new point appear in a chart.

Publishing the complete work

I am publishing the implementation so others can inspect it, reproduce it and adapt it:

I am also creating a public repository that will document the complete timeline of my air-quality work: the original Taranto dashboard, the automated monitoring and social alerts, the historical analyses and this latest AirGradient experiment.

Open hardware made the latest chapter possible. Open data made the first one possible. Open documentation is how the work can continue beyond me.

Still looking after the air

In 2019, I hoped that people would spend some of their time looking after our planet.

Years later, I am still working on the same idea.

The tools have evolved. The dashboard became an automated sentinel. The sentinel contributed to civic evidence. And now an open sensor can send its measurements directly into the analytical environment my team and I have spent years building.

There is still much more to explore: multiple sensors, schools, community deployments, indoor/outdoor comparisons and new forms of citizen-owned monitoring.

But the principle remains clear:

Data becomes power when people can collect it, understand it, question it and retain control of it.

Sense locally. Control locally. Push privately. Explore openly.

For science and, inevitably, for the geeky pleasure of making the whole chain work.

References