Contents
Methodological reference
How a report becomes data
This page describes how a report is collected, structured and linked to automatic data, and which reference frameworks this approach relies on.
Version 1.6 — last updated on 27 September 2026
Version 1.6 — last updated on 27 September 2026 — Odour.Live, Method for collecting and characterising odour reports, version 1.6, 27/09/2026, https://odour.live/en/method
1. Purpose of this document
This page describes what is asked of a person who reports an odour, how their report is structured, which data are automatically attached to it, and which reference frameworks this approach relies on. It also states what the platform does not measure.
It is meant to be checked. If you are handling a case, disputing data or preparing a formal complaint, you must be able to know exactly where the figures put to you come from. That is the purpose of this document.
2. The path of a report
From the perception of an odour to its appearance in a report, a submission goes through six stages. None of them involves human judgement on the content of the report.
Perception
A person perceives an odour and opens the form. No account is needed and no cookie is set. The report can be submitted on the spot from a phone, or later from a computer.
Report
Three pieces of information are required: where the odour was perceived, its perceived intensity and the type of odour. Everything else is optional. This is a deliberate choice: a long form is an abandoned form, and a minimal report remains usable as long as it carries an intensity, a place and a time.
The reporter can add a free description, a perceived duration and a hedonic tone, and can correct the date and time when reporting after the event.
Location and timestamp
The position is given in one of three ways: device location, address search, or a point placed manually on the map. It is immediately blurred before any public display — see §7.
The time recorded is that of the observation, not of the submission. It is entered in local time and converted to Coordinated Universal Time for storage, so that clock changes do not create gaps in the time series.
Weather matching
The weather conditions for the reported place and time are automatically attached to the report. The reporter enters nothing and cannot influence anything.
This is what makes the analysis hard to dispute: the wind data is not reconstructed afterwards, at the moment someone tries to prove something. It is attached to the report as soon as it is created, before any analysis. Details in §6.
Structuring
The report becomes a record comparable to all the others: an intensity on a single scale, a type chosen from a closed list, a point, an instant, a state of the atmosphere. Two reports made by two people who do not know each other can be added together.
This is the key difference from a complaints register: a register accumulates stories, a database accumulates measurements of perception.
Aggregation and reporting
Reports are cross-referenced — by hour, day, area, type and wind direction — to produce the indicators and charts. No report is weighted, discarded or corrected.
What never happens to a report
It is not validated, moderated on its content, or weighted according to its author. It is not deleted at the request of a third party, whatever their status. It is not delayed: it appears in the statistics as soon as it is recorded. A reporter with an account can request the withdrawal of a report they submitted themselves — only they have this right. The report is immediately hidden: it is no longer displayed publicly but is still counted in the statistics. An administrator then decides either to detach it from the account, in which case it remains as an anonymous observation that is no longer personal data, or to delete it permanently.
3. Why five factors
An odour is not a nuisance. Five factors make the difference.
The problem with odour is that it cannot be measured like noise. There is no device you can install at a resident's home that continuously records what they are exposed to. Odour concentration is measured in a laboratory, on a sample taken at the source, by a panel of assessors: this tells you about the emission, not about what is experienced three hundred metres away on a Tuesday evening when the wind changes.
Hence a recurring difficulty: residents describe a real nuisance, the operator produces compliant measurements, and both are right. What is missing is a common language to describe the nuisance as it is perceived.
Environmental agencies have converged on five factors, grouped under the acronym FIDOL. They were not invented by the platform: they appear, in similar wording, in the guidance of the Environment Protection Authority of Victoria, the New Zealand Ministry for the Environment, the UK Environment Agency and the New South Wales EPA. Their value is that they explain why two people exposed to the same odour do not experience the same nuisance.
Frequency
Once a month or three times a day: repetition turns an inconvenience into a nuisance. This is the factor that distinguishes an incident from an established situation.
Intensity
The strength of the perception. The only factor that comes close to a measurement, and the only one for which there is a standardised scale usable by an untrained person.
Duration
A thirty-second whiff and four continuous hours produce the same line in a complaints register, and two completely different realities.
Offensiveness
At the same intensity, the smell of warm bread and the smell of putrefaction do not cause the same annoyance. This is the most often forgotten factor, and the one that explains the most disagreements.
Location
The context of exposure. An odour that is acceptable at the edge of an industrial estate is not acceptable in a child's bedroom or in a garden on a Sunday.
A single report says nothing about frequency and little about nuisance. The framework only makes sense when applied to a series. That is why the platform is of little use for documenting a single episode, and of great use for documenting a situation that persists.
A useful clarification, as it is often a source of confusion: these factors describe exposure and perception. They do not measure a concentration and do not allow a regulatory exceedance to be concluded. They make it possible to characterise a nuisance, which regulatory limit values precisely do not do.
4. How each factor is obtained
Not all factors are obtained in the same way, and it is important to know which ones are declared by a person and which ones are calculated.
| Factor | Origin | How |
|---|---|---|
| Frequency | Calculated | By aggregating reports over a period and an area. It is not asked of the reporter, and cannot be: a single report does not know its own frequency. Frequency only exists at the level of a series. |
| Intensity | Declared | Six-level categorical scale, aligned with VDI 3882 Part 1, required. Details in §5.1. |
| Duration | Declared | Duration class chosen from a list, including the case of intermittent whiffs, optional. Details in §5.6. |
| Offensiveness | Declared | Hedonic tone on a nine-point bipolar scale, aligned with VDI 3882 Part 2, optional. Details in §5.7. |
| Location | Declared | Point of perception, required, blurred before any public display. Details in §5.3. |
Two data items that are not FIDOL factors
The timestamp and the weather conditions are not part of the framework. They are not used to characterise the nuisance but to look for its origin: cross-referencing them is what makes it possible to link episodes to a wind direction, a time slot or an operational activity. The framework says what is experienced; these two data items help establish what causes it.
5. The reporting framework
Complete list of what is asked, what is optional and what is added automatically. Three fields are required; all the others can be left empty.
5.1 Perceived intensity, required
Six-level categorical scale, from weakest to strongest:
| Level | Label | VDI 3882 Part 1 equivalent |
|---|---|---|
| 1 | Very weak | 1 — sehr schwach |
| 2 | Weak | 2 — schwach |
| 3 | Distinct | 3 — deutlich |
| 4 | Strong | 4 — stark |
| 5 | Very strong | 5 — sehr stark |
| 6 | Extremely strong | 6 — extrem stark |
The VDI 3882 Part 1 standard defines seven levels, from 0 to 6, level 0 corresponding to no perception. Level 0 is deliberately not offered: people do not submit a report to say that they smell nothing. The six levels used therefore correspond exactly to levels 1 to 6 of the standard, without recalibration or interpolation.
The value of a standardised categorical scale is that it remains stable from one person to another and from one month to the next. Intensities become comparable between reporters, can be added up over a period and averaged, which a free assessment never allows.
5.2 Odour typology
Version 3.2 — 7 families, 52 active descriptors, updated on 02/09/2026.
The typology is built by combining sector-specific odour wheels, and offered in two input modes for the same framework: a simple mode (up to 3 families, with as many descriptors as needed in each) and a detailed mode (full list). Each family below groups the active descriptors offered in the reporting form. This list is generated automatically from the same module as the form: it cannot differ from what is actually offered to the reporter.
Agriculture / spreading
Manure / slurry / liquid manure · Animal / barn · Soil / earth / compost · Mushroom / mould · Chemical fertilizer
Waste / Sewage / Composting / Biogas
Sewer · Putrefaction / rotten · Rotten egg / sulphurous · Ammonia / urine · Household waste · Compost · Biogas / methanisation · Rendering / dead animals
Food service / Agri-food
Bakery / sweet · Frying / grease · Cooked meat · Peppery / spicy · Grilled / burnt · Fish / seafood · Roasted / coffee
Solvents & hydrocarbons
Solvent (paint, varnish, glue) · Petrol · Hydrocarbons (diesel, gasoil) · Kerosene · New plastic / PVC · Rubber · Resin / glue / epoxy · White spirit
Chemical industry / petrochemical
Ammonia (industrial) · Chlorine / chlorinated · Acetic acid / vinegary · Sulphur / sulphurous gas · Natural gas · Disinfectant / bleach · Medicinal / pharmaceutical · Cooked / fermented cabbage · Garlic · Cat urine · Rancid butter · Mentholated / fresh mint · Almonds
Industrial processes
Asphalt / bitumen / tarmac · Hot or burnt plastic · Hot metal / welding · Machine oil / industrial lubricant · Paper / wood pulp · Mineral dust (cement, concrete)
Smoke / combustion
Smoke (generic) · Smoked wood / smoking · Burning wood / wood heating · Burning plastic / tyre · Green waste fire
Descriptors removed from the form are never deleted: the reports that carry them remain readable and usable.
Change log
| Version | Date | Change |
|---|---|---|
| 3.2 | 02/09/2026 | One sub-descriptor added and three sub-descriptors removed, across three families. |
| 3.1 | 29/06/2026 | One sub-descriptor added and one sub-descriptor removed, in two different families. |
| 3 | 08/06/2026 | Reference list overhaul: one family split into two (including “Solvents and hydrocarbons”), bringing the total to eight families. |
| — | before 08/06/2026 | Initial development version, consolidated. |
The full history of the dictionary, version by version, is available on request.
5.3 Location, required
Three input methods: device location, address search, or a point placed manually on the map. The recorded point is where the odour was perceived, not the reporter's home nor the presumed source.
The exact position is never published: it is blurred by a random offset, within a maximum radius of 200 metres, before any public display. Details of the process in §7.
5.4 Date and time of the observation, prefilled, editable
By default, the time of submission. The field can be edited to report after the event, for example the next morning for an odour at night. It is the time of the observation that is recorded, never that of the submission.
The time is entered in local time and stored in Coordinated Universal Time. Hourly series therefore remain continuous when summer time begins or ends.
5.5 Free-text fields, optional
| Field | Limit | Use |
|---|---|---|
| Description | 500 characters | Circumstances, probable origin of the odour, consequences on daily life (windows kept shut, disturbed sleep, garden unusable…). Complements the typology without replacing it. |
This field feeds text analyses (word cloud, recurring terms) but is never automatically converted into a numerical indicator.
Health data excluded from this field
A “Comments” field inviting people to describe effects they experienced (headaches, nausea, irritation…) existed until 8 September 2026. It was removed from the form: collecting health data, a special category under the GDPR, requires a dedicated framework that has yet to be defined. Reports submitted before that date may still contain this information, which is kept under the same rules as the rest of the report (see §7).
5.6 Perceived duration, optional
Duration class, chosen in a single tap from six options:
- In whiffs, intermittently
- Less than 5 minutes
- 5 to 15 minutes
- 15 minutes to 1 hour
- 1 to 3 hours
- More than 3 hours
The “in whiffs” option is not a refinement: the French regulations applicable to certain classified installations explicitly distinguish odours “perceived in whiffs or continuously”. Offering it produces data that can be transposed directly into the vocabulary of a classified installation case file. This option deliberately has no numerical value and is always counted separately in totals, never converted into minutes.
This class makes it possible to calculate an estimated cumulative exposure time over a period, shown on the analysis pages (Area summary, Character and duration) and in reports.
5.7 Hedonic tone, optional
Bipolar scale with nine points, aligned with VDI 3882 Part 2, from −4 to +4, zero marking neutrality. No value is preselected: as long as the reporter has not touched the scale, nothing is recorded.
| Value | Label | What the reporter expresses |
|---|---|---|
| +4 | Extremely pleasant | Clearly enjoyable odour, sought after. |
| +3 | Very pleasant | Distinctly enjoyable. |
| +2 | Pleasant | Enjoyable, nothing more. |
| +1 | Slightly pleasant | Rather enjoyable, faintly. |
| 0 | Neutral | Neither pleasant nor unpleasant. Perceived, without judgement. |
| −1 | Slightly unpleasant | Marginally bothersome. |
| −2 | Unpleasant | Distinctly bothersome. |
| −3 | Very unpleasant | Strongly bothersome, one tries to get away from it. |
| −4 | Extremely unpleasant | Unbearable, causes an immediate recoil. |
Not to be confused with annoyance
Hedonic tone is a property attributed to the odour itself: slurry is unpleasant to almost everyone, regardless of the moment. Annoyance is the effect on a person, and depends on the time, the activity under way, repetition and habituation. Only the former corresponds to VDI 3882 Part 2.
Optional field. Making it required would lengthen the form for a marginal gain: the typology already carries an implicit hedonic charge. A partial completion rate is more than enough to produce a usable distribution.
5.8 Identity of the reporter, three levels to choose from
| Level | What is collected | Consequence |
|---|---|---|
| Anonymous (default) | Nothing. | No acknowledgement, no follow-up possible, no possibility of later correction. |
| Contact details | An email address. | A copy of the report is sent to the reporter, who can forward it to the authorities. |
| Citizen account | Free account. | Personal history of reports, optional neighbourhood alerts. |
The chosen identity level changes nothing in how the report is processed: an anonymous report carries exactly the same weight as an identified report in all analyses.
6. Weather data
Each report is associated with the state of the atmosphere at the reported place and time. This association is automatic: the reporter enters no weather data and cannot influence its value.
This is what makes it possible, across a series, to relate the occurrence of episodes to a wind direction, and therefore to a geographical area of origin.
| Item | Answer |
|---|---|
| Provider | Open-Meteo — free and open-source service, no API key required. The request is made when each report is created. |
| Source depending on age | Two different sources depending on whether the report is recent or older. Less than 5 days: forecast endpoint api.open-meteo.com/v1/meteofrance. More than 5 days: archive endpoint archive-api.open-meteo.com/v1/archive. |
| Model used (recent, < 5 days) | No model is explicitly imposed in the request: Open-Meteo applies its “Best Match” mode, which automatically selects the most suitable model for the requested point. In France, this is most often AROME (Météo-France, resolution ~1.5 to 2.5 km, hourly), falling back to ARPEGE Europe (~11 km) or ARPEGE World (~25 km) depending on location and lead time. The model actually used depends on the point and lead time, according to the selection rules published by the provider. |
| Model used (archive, ≥ 5 days) | Same principle: “Best Match” on the archive side, which combines ECMWF IFS HRES (~9 km, data from 2017 to today), ERA5-Land (~11 km, 1950-2016) and ERA5 (~25 km, before 1950). For the vast majority of reports on the platform (after 2017), the effective source is therefore IFS HRES at ~9 km. |
| Time resolution | Hourly in both cases. The weather value closest to the reported time is chosen within a ± 30-minute window. |
| Variables stored with the report | Temperature, relative humidity, wind speed, wind direction, gusts, surface pressure, and a qualitative description derived from a weather code (clear sky, rain, thunderstorm…). Precipitation and cloud cover are not stored with the report — they exist in a separate module used only for back-trajectory analysis. |
| When matching happens | When the report is created, before it is inserted into the database. |
| Immutability | Fixed. The value written at creation is never recalculated. |
| Missing data | The weather request is protected: if it fails, the report remains valid without associated weather data and simply cannot be used in wind cross-analyses. |
| Storage and export | The values appear in CSV exports (extraction for organisation accounts, export of one's own reports for a citizen account) in the stated units: degrees for direction, kilometres per hour for speed and gusts, degrees Celsius for temperature. |
Limits to interpreting wind data
Wind data from a model describes the airflow at the scale of its grid cell. It does not account for local terrain effects, building turbulence or night-time temperature inversions that concentrate odours at ground level. It is quite sufficient to establish a statistical relationship between wind direction and the occurrence of reports; it does not allow conclusions to be drawn about an isolated episode.
7. Anonymisation and data protection
The platform is designed to produce usable data without building a file of complainants. This is a design choice, not a configuration option.
No account required
Anonymous mode is the default and asks for nothing.
No tracking cookies
No cookies are set for tracking purposes.
Blurred position
The point of perception is randomly offset within a maximum radius of 200 metres before any public display. The exact position is never published or exported.
No weighting
An anonymous report counts as much as an identified report.
| Item | Answer |
|---|---|
| Blurring process | Random offset (angle and distance drawn at random) within a maximum radius of 200 metres, calculated once when the report is created and then fixed: it is never recalculated at each display, which rules out any attempt to recover the position by cross-checking several views. |
| Retention of the exact position | The exact coordinates of the perception are kept in the database for internal purposes (weather calculations, aggregations) but are neither published nor included in exports. Only the blurred position is shared. |
| Retention periods | Reports: five years from collection, then anonymised and aggregated archiving with no time limit. Accounts: until they are deleted, at the holder's request from their profile, validated by an administrator. Reports are not deleted with the account: the personal data attached to them (link to the account, contact details, comments where applicable) are erased, and each report remains as an anonymous observation. Technical logs: 90 days, with truncated IP address; the full IP address is kept for 30 days, only in the event of a proven security incident. |
| Hosting | Provided by Infomaniak Network SA, in Geneva. Switzerland benefits from an adequacy decision of the European Commission adopted on 15 January 2024; data transfers take place on that basis. |
| Exercising your rights | The rights of access, rectification, erasure, portability and objection can be exercised by writing to contact@odour.live. From their profile, an account holder can also export their reports, and request the deletion of their personal information (keeping their account) or of their account; these requests are validated by an administrator. A report submitted anonymously cannot be linked to anyone: it can therefore be neither corrected nor withdrawn. |
| Content of the export | Identifier, date and time, intensity, odour type, description, perceived duration and hedonic tone in plain text, blurred position, municipality / département / region, source of the report, status of the reporter's account, weather conditions (temperature, wind). No name, no email, no exact position. The description is only provided to the author of the report (and to the administrator); for other accounts it is empty, like the free Comments field, because these texts may contain personal information. Only keywords are extracted for anonymous text analysis. |
The author's right to withdraw a report is detailed in the privacy policy. Privacy policy.
8. What the platform does not measure
Odour.Live does not measure an odour concentration: it is neither dynamic olfactometry (EN 13725) nor standardised field inspection by a panel (EN 16841). It documents the real perception of residents, continuously and over time, precisely what a one-day campaign cannot capture.
It does not produce values comparable to a regulatory threshold, and does not allow non-compliance to be concluded. One consequence, regularly misunderstood in both directions: regulatory compliance and the absence of nuisance are two distinct notions. An installation can comply with its permit conditions and still cause odour perceptions; the reverse is also true.
It is precisely because this scope is clearly defined that the data produced are usable.
9. References
| Reference | Subject |
|---|---|
| EPA Victoria — Guidance for assessing odour | FIDOL factors, odour nuisance assessment. |
| Ministry for the Environment (New Zealand) — Good Practice Guide for Assessing and Managing Odour | FIDOL factors, assessment methodology. |
| DEFRA and Environment Agency (United Kingdom) — odour management guidance | FIDOL factors, odour management under an environmental permit. |
| EPA NSW (New South Wales) — Managing odour | FIDOL factors. |
| VDI 3882 Part 1 | Determination of odour intensity, scale used in §5.1. |
| VDI 3882 Part 2 | Determination of hedonic odour tone, scale used in §5.7. |
| EN 13725 | Dynamic olfactometry, cited in §8 as outside the scope. |
| EN 16841 | Determination of odour in ambient air by field inspection, cited in §8 as outside the scope. |
| Open-Meteo | Provider of the weather data automatically attached to reports, see §6. |
Odour.Live relies on these reference frameworks without being assessed against them by any body.
10. Version history
| Version | Date | Change |
|---|---|---|
| 1.0 | 28/08/2026 | First publication. |
| 1.1 | 07/09/2026 | Complete overhaul: path of a report, details of the FIDOL factors, complete field framework, documented weather sources, numbering and table of contents, enhanced PDF export. |
| 1.2 | 08/09/2026 | Removal of the “Comments” field (effects experienced) from the reporting form, as collecting health data requires a dedicated framework yet to be defined. The “Description” field is renamed and supplemented with an instruction explicitly excluding any health data. |
| 1.3 | 08/09/2026 | Clarification of the export content (§7): the free Comments field is no longer listed as included in the standard export for other users — it remains reserved for the author of the report and the administrator, with only keywords extracted for anonymous text analysis. |
| 1.4 | 20/09/2026 | Hedonic tone with no preselected value (§5.7); scope of exports clarified (§6, §7): the description is only provided to the author and the administrator, and a citizen account only exports its own reports. |
| 1.5 | 26/09/2026 | Withdrawal of a report and deletion of an account (§2, §7): requests are validated by an administrator; a withdrawn report is hidden (not displayed publicly, still counted in the statistics), then detached from the account or, by decision, deleted; personal data are erased, and detached reports remain anonymous observations. Removal of a mention of a count of detachments, which was not implemented. |
| 1.6 | 27/09/2026 | Retention period of technical logs clarified (§7): 90 days with truncated IP address; the full IP address is kept for 30 days, only in the event of a security incident (intrusion attempt, mass scraping of reports). |
Any change to the form (new descriptor, modified scale, change of weather source) results in a new dated version of this page.
Looking for something else?