https://hal-ineris.archives-ouvertes.fr/ineris-00971181Beauchamp, MaximeMaximeBeauchampINERIS - Institut National de l'Environnement Industriel et des RisquesMalherbe, LaureLaureMalherbeINERIS - Institut National de l'Environnement Industriel et des Risquesde Fouquet, ChantalChantalde FouquetGEOSCIENCES - Centre de Géosciences - Mines Paris - PSL (École nationale supérieure des mines de Paris) - PSL - Université Paris sciences et lettresA pragmatic approach to estimate probabilities of exceeding limit values in air quality : application to PM10 and O3HAL CCSD2013GEOSTATISTICSKRIGINGAIR POLLUTIONEXCEEDANCES OF AIR QUALITY LIMIT VALUESNUMBER OF EXCEEDANCES[SDE] Environmental SciencesCivs, Gestionnaire2014-04-02 15:59:112022-10-22 05:09:562014-04-02 15:59:11enConference papers1European legislation on ambient air quality not only defines limit values on time aggregates (hourly, daily, annual means) but also cumulative limit values. We propose a pragmatic methodology applied to PM10 number of exceedances of the daily limit value (50 µg/m3) and to AOT40, sum of differences between O3 hourly concentrations between 8 a.m and 8 p.m (CET) that exceeds 40 ppb (i.e. 80 µg.m-3), over the period may-september. A probabilistic frame based on preliminary hourly/daily mean concentrations mappings on a 1km resolution grid is used. Measurements at the monitoring stations are not sufficient to describe the area exceeding the limit value. Atmospheric concentration fields are then estimated by an external drift modelling that takes into account surface monitoring observations and outputs from the CHIMERE chemistry transport model. The kriging variance is also used in a conventional gaussian framework to define the error distribution. For PM10 and O3, the issue is simplified and concentrations are considered as independent since dependency is implicitly described by the drift. A combinatory approach and its computation by fitting a probability distribution is used for PM10. Expectation and variance of the conventional benefit in a gaussian framework are developed to model the AOT40 computation. Data from the french monitoring network are used. Comparison are done with useful variables and bootstrapping validation enables to check out modelling efficiency.