München 2019 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
UP: Fachverband Umweltphysik
UP 2: Air pollution and tropospheric chemistry
UP 2.5: Vortrag
Dienstag, 19. März 2019, 12:30–12:45, HS 22
Evaluation of the ENFUSER air quality model — •Oliver Mehling1,2, Leena Järvi2, Lasse Johansson3, and Pauli Paasonen2 — 1Physikalisches Institut, Albert-Ludwigs-Universität Freiburg, Germany — 2Institute for Atmospheric and Earth System Research, University of Helsinki, Finland — 3Finnish Meteorological Institute, Helsinki, Finland
The quantitative evaluation of air quality models dates back to the 1980s, but, until today, most commonly only basic metrics like the Pearson correlation coefficient r2 are reported.
In our work, a broader approach is taken: through a multitude of evaluation metrics, model bias is pondered against actual meteorological variables such as temperature, relative humidity and atmospheric stability. Partial correlation of model bias and subsets of key meteorological variables, conditioned on the remaining variables, is used to identify probable “drivers” of bias. In addition, the use of hourly instead of daily averages allows to resolve how well a model predicts diurnal concentration patterns of air pollutants. A cluster analysis is used to group diurnal patterns of bias into different categories, for which interpretations of recurring over- and underestimations are given.
These methods are used to perform a comprehensive evaluation of FMI-ENFUSER, a novel air quality model applied to the Greater Helsinki Area, which combines dispersion modeling, information fusion and statistical approaches to forecast NO2, O3, PM10 and PM2.5 concentrations. We discuss how a strictly statistical evaluation helps to draw real-world conclusions to foster future model improvements.