Directive 2008/50/EC builds its whole air-quality assessment framework around fixed sampling points, with the macroscale and microscale siting criteria set out in Annexes III and VIII. A reading taken from a sensor mounted on a moving vehicle does not fit that scheme: it has no stable location, no averaging period tied to a site, and it produces a value for every coordinate it crosses. The data is there, it is dense, and it sits outside the frame that ought to give it legal meaning. This is the governance problem that mobile monitoring puts on the table, and I am running into it while working around the RIPEG system.

Context

RIPEG (the project acronym, funded under the Tuscany Region POR FESR programme) is a portable system that detects ultrafine particulate, ozone, carbon monoxide and nitrogen oxides in real time, installed on vehicles driven through the urban area. What comes out is not a time series for one site but a trajectory: a sequence of readings, each with a timestamp and a position. Against the fixed network, what changes is the nature of the data, not the quantity.

Ultrafine particulate (diameter below 0.1 µm) shows this clearly. Directive 2008/50/EC regulates PM10 and PM2.5 as mass concentrations, because the reference method is gravimetric (Annex VI points to EN 12341 and EN 14907). The ultrafine fraction, though, weighs almost nothing: its signal is the number of particles per cubic centimetre, not the mass. You measure it with condensation counters, not a filter sampler. 2009 work on bicycle platforms — Berghmans et al. in Mol, Boogaard et al. across eleven Dutch towns — had already shown that ultrafine number concentrations change a lot within the same street and that their peaks do not line up with PM2.5 peaks. The quantity the vehicle measures well is exactly the one the fixed network does not see, and for which there is no regulated limit value.

Architecture

On the software side the system splits into three responsibilities. An on-board unit pairs the sensor readings with the position fix and transmits them. A server layer receives, normalises and stores them (PostgreSQL as the database, with a geospatial layer to index coordinates). A visualisation layer shows the trajectory on the map in real time.

Visualisation is where the 2010 technical choice is most exposed. The panel runs on HTML5 and Google Maps, without plugins: the browser fetches the data over AJAX and draws the measured points over the cartography. In mid-2010 the ground is still moving under your feet. The W3C Geolocation API is still at its latest Last Call Working Draft and has not yet reached Candidate Recommendation; the canvas element and the other HTML5 pieces are Working Drafts, not closed standards. Building the interface on HTML5 now is a bet on the direction of travel, not the adoption of a settled base. It is a defensible choice — it avoids the proprietary plugin — but it should be stated for what it is.

The critical point

The knot is not collecting the mobile reading; it is making it readable to someone who was not on the vehicle. A concentration of 30,000 particles/cm³ measured at one point, on its own, says almost nothing: it depends on the sensor and its calibration, on the vehicle’s speed at that instant (which sets the residence time in the air cell), on the delay between intake and reading, on how good the position fix is. A fixed station holds many of these factors constant and documented once and for all. The vehicle makes them vary along the trajectory.

So the raw reading, alone, cannot be governed. It can be governed only if every reading carries its own acquisition context: sensor identifier and calibration state, instantaneous speed, fix accuracy, measurement-chain delay. Without these metadata attached to the individual row, downstream reuse — comparing campaigns, aggregating over a street, building maps — turns out numbers of opaque provenance. With a mobile system this is not an optional detail: it is the condition for the data to mean anything outside the moment it was taken. Designing the data model, here, is a choice about governing the measurement before it is one of engineering.

Implications

The real weight falls on downstream analysis. To build synthetic indicators (street averages, exposure maps, aggregated series) on top of mobile data you need aggregation rules that are explicit and traceable: how many readings per cell, with which speed threshold, which imprecise fixes to discard. Geospatial business-intelligence tools can compute these indicators, but an indicator is worth as much as the rule that generates it, and the rule must be versioned alongside the data. A number on a map that cannot be traced back to the readings that compose it and the filters applied is a claim no one can verify, however polished the graphics. The control panel noze is building for RIPEG — collection server, HTML5 viewport on Google Maps and geospatial BI tools — addresses exactly this knot: https://www.noze.it/en/insights/ripeg-expo-shanghai/.

There is then the relationship with the regulated network. Mobile monitoring does not replace the fixed stations of Directive 2008/50/EC — it has neither their reference methods nor their averaging regimes. It sits beside them, and covers the fine spatial variability the fixed network, by the way it is built, does not resolve. Keeping the two planes distinct is in itself an act of data governance: treating a mobile trajectory as if it were a fixed-station series means giving the number a legal validity it does not have.

The system is among the projects selected for the “Italy of Innovators” showcase at the Shanghai Expo, open from 1 May to 31 October 2010 under the theme “Better City, Better Life”. The Expo’s theme is the city; air quality measured street by street is one possible description of it, provided the numbers stay reconstructible.

Limits

What is above describes a problem; it does not close it. At least three points stay open. The first is sensor drift: without a protocol for periodic calibration and for recording its state, the calibration metadata I called on stay incomplete. The second is the stability of the HTML5 base, today not closed: an interface built on Working Drafts inherits the uncertainties of those specifications. The third is the absence of a regulatory reference for ultrafine particle number: the quantity the vehicle measures best is the one for which there is no limit value to anchor it to. On none of the three does technology, alone, have the last word.

https://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2008:152:0001:0044:EN:PDF https://www.w3.org/TR/2009/WD-geolocation-API-20090707/ https://www.w3.org/TR/2010/WD-html5-20100304/ https://en.wikipedia.org/wiki/Expo_2010

Cover image: The Italy Pavilion at Expo 2010 Shanghai at night, its large glass facade lit from within, with a few visitors in the foreground — photo by 凌智, CC BY 2.0 — https://commons.wikimedia.org/wiki/File:Italy_Pavilion_of_Expo_2010.jpg