An adaptation of the covariance modeling for large scale geostatistics estimation in Air Quality
Abstract
The geostatistical estimation is used to interpolate Air Quality data on large scale regular grids, from both observations at monitoring sites and deterministic models outputs. But the sources of pollution and therefore the data typology are highly variable (rural background, urban and suburban background, traffic-related pollution), making false the assumption of stationarity. At large scale (regional, national, continental), a first idea is to divide the total concentration in two terms: a first rural term and a second linked to the pollution brought by urban areas. A second modeling introduces space-NOx emissions random fields to relax the constraint of stationarity. The results are compared to classical kriging estimators such as external drift modeling and the application in an operational context is discussed.