The GeoSphere dispersion model: what it shows and what it doesn't

GeoSphere Austria, commissioned by the Office of the Burgenland Provincial Government, produced a dispersion calculation for asbestos fibres from the four closed quarries Pilgersdorf, Bernstein, Rumpersdorf and Badersdorf (GRAMM/GRAL model, file number 2026/UM/000160, version 1.1 of 29 April 2026, 131 pages). We hold the document. For copyright reasons we do not pass it on, but we assess the methodology and quote from the freely describable passages.

This post is the technical long form of the short analysis on the Burgenland page. First, the central correction to our own earlier account: we initially read the 3-percent input value as the result of a TRGS 517 back-calculation to the total sample, and inferred a systematic underestimation from it. That was wrong. The value is a content in the dust, not a value back-calculated to the bulk rock. The arguments below therefore stay entirely within GeoSphere's own model logic.

What the model calculates

GRAMM/GRAL (Öttl 2015a/b, 2022b) is a validated Lagrangian particle dispersion model that GeoSphere Austria uses for regulatory immission forecasts. On the basis of a representative meteorological year (2023, 1,080 wind-field combinations), it calculates the PM10 concentration at defined receptor points and, in a second step, converts these into WHO fibre concentrations. The sources include extraction, processing, truck movements, wind erosion from open areas and spoil heaps; for each quarry, the year with the highest extraction volume of the last five years was used.

The key results at the nearest residents (annual mean AM / 95th percentile of the hourly values, in WHO fibres/m³):

Without quarry operation (only wind erosion of open areas), the annual means drop to a fraction: at most 26 F/m³ in Badersdorf, at most 23 F/m³ at the Rumpersdorf hunting lodge.

This is a methodologically clean first estimate, with the right tool and carefully chosen receptor points. It answers the question "how high is the direct fibre concentration from ongoing quarry operation?" Three points of context are needed to read the values correctly.

1. The 3-percent input value is set rather low

The conversion from PM10 to asbestos mass runs through a single factor: a maximum asbestos mass content of 3% in the inhalable dust (grain sizes below 0.1 mm). This value is the single highest of the four quarries, measured at Rumpersdorf, and is applied to all four sites. The report states this itself on page 80: the 3-percent value is the "maximum result from the Rumpersdorf quarry", used "to convert the simulated PM10 results to asbestos". It is thus a content in the dust, not in the bulk sample. Table 2-5 lists the maxima of the four quarries: Pilgersdorf 2.12%, Bernstein 1.9%, Rumpersdorf 3%, Badersdorf 2.1%.

Two points suggest that the value actually relevant for PM10 is higher.

The fraction mismatch. The 3% was determined in the inhalable fraction below 0.1 mm, that is, below 100 µm. But in the model it is applied to PM10, that is, to particles below 10 µm. PM10 is a subset of the coarser fraction. WHO fibres are by definition very thin (diameter below 3 µm, length above 5 µm, ratio greater than 3:1) and very probably enrich precisely in the finer PM10 fraction. The WHO-fibre mass content of PM10 therefore likely exceeds the 3% measured for the coarser overall fraction. The model carries a mean over a broader grain-size range onto exactly the fine subfield in which the fibres sit.

A maximum from few samples is not a ceiling. The report itself describes the WHO-fibre mass content in the inhalable dust as having a "very large range". From this widely scattered quantity, a small number of samples is drawn per quarry (the individual findings are in the associated quarry reports, not in the dispersion calculation), and the highest value is used as the input content. Statistically, however, the maximum of a small sample is not an extreme value of the underlying distribution. For a sample of size n, the expected rank of the maximum is n/(n+1). As a worked example: the maximum of six samples lies on average only at roughly the 85th to 86th percentile of the true distribution, not the 95th or 99th, and the fewer the samples, the lower it sits. With few samples the upper end of the distribution is not captured at all. Reading the maximum thus obtained as the "worst case" understates the real worst case of the distribution.

The separate methodological question of how the TRGS 517 extrapolation distributes the asbestos content between the dust fraction and the total sample is not the subject of this dispersion calculation. It is documented independently under Methodological assessment of the TRGS 517 extrapolation.

2. The model is linear, and the chosen value lies at the lower end of its own expectation

On page 24 the report names an expectation range that extends beyond the 3% used:

"Measurements in the quarries show on average 30–50 mass percent asbestos […]. This corresponds to an expected content of 2% to 5% WHO fibres in the dust."

The lower part of this range, 3%, was used. That would be uncritical if the model reacted weakly to this parameter. But it is linear: the asbestos mass content is applied as a scalar to the PM10 results only after the actual dispersion calculation (section 2.7: "the PM10 results are first scaled to 3%", after which the asbestos mass is converted into a fibre count via the mean fibre weight). Annual mean and 95th percentile therefore scale directly in proportion to the chosen percentage.

A worked example: at the Rumpersdorf hunting lodge RD-01, the modelled annual mean is 730 WHO fibres/m³ and the 95th percentile is 4,353. Using the upper end of the report's own expectation, 5% instead of 3% (factor 5/3), the annual mean rises to about 1,220 and the 95th percentile to about 7,255 WHO fibres/m³. Even the annual mean then exceeds the Taskforce guidance value of 1,000 fibres/m³, which is still just undershot in the model at 3%. The report does not examine the effect of such a variation of the input parameters; a sensitivity analysis is missing.

3. What the model structurally cannot see

GRAMM/GRAL calculates the instantaneous air concentration from the ongoing, direct quarry emissions. It does not represent three things:

• Fibres that settle on surfaces: roads, gardens, roofs, playgrounds.
• Material that accumulates over more than 30 operating years.
• Resuspension of this deposited material by wind, traffic or human activity.

This secondary reservoir is an independent exposure path that the model does not capture methodologically, especially for children near the ground. This is not a statement about whether the modelled primary values are too high or too low, but about a gap in the model's scope. Conversely, one of the model assumptions is conservative in itself: because no deposition is calculated, the fibres behave like a gas that does not settle out, which tends to over- rather than understate the direct plume. That is precisely why one cannot say blanketly that "all assumptions act in the same direction". The defensible, narrower statement is: on the axis of asbestos content the estimate of 3% is rather low, and the secondary path is missing entirely.

Why the guidance value is not a safety margin

The value of 1,000 WHO fibres/m³, often treated in the debate as a threshold, is a guidance value of the Burgenland Taskforce, not a health-based limit value. For asbestos, an IARC group 1 substance (carcinogenic to humans), no effect threshold is known below which there would be no risk; for asbestos the WHO assumes a mechanism with no safe lower bound. A value "below the guidance value" therefore does not mean "harmless"; it means only that the pragmatically set action threshold has not been reached. This is the yardstick against which the modelled values are to be read: several receptor points already lie, in the model under conservative assumptions, in the range of several hundred to over a thousand fibres/m³, and on the asbestos-content axis this estimate is the lower rather than the upper end.

Conclusion

The GeoSphere dispersion calculation is a good first tool, cleanly applied. Its values are not a secure ceiling of the real exposure. On the axis of asbestos content (fraction mismatch, sample maximum, lower end of its own 2-to-5-percent expectation) they are, if anything, an underestimate, and the secondary exposure path via deposited and resuspended material lies outside the model. A sensitivity analysis covering the range of the 3-percent parameter and of the fibre weights would make this uncertainty visible. Until then: the figures are robust as an order of magnitude, but not as a point estimate with a safety reserve on top.

Source: GeoSphere Austria, "Die Ausbreitung von Asbest, ausgehend von den Steinbrüchen in Pilgersdorf, Bernstein, Rumpersdorf und Badersdorf", 2026/UM/000160, version 1.1, 29 April 2026. Table 2-5 (p. 25), section 2.7 (pp. 24–26), results section 7.2, summary p. 80.