Evaluation of high-resolution forecasts with the non-hydrostaticnumerical weather prediction model Lokalmodell for urban air pollutionepisodes in Helsinki, Oslo and Valencia

[摘要] The operational numerical weather prediction model Lokalmodell LM with 7\,kmhorizontal resolution was evaluated for forecasting meteorologicalconditions during observed urban air pollution episodes. The resolution wasincreased to experimental 2.8 km and 1.1 km resolution by one-way interactivenesting without introducing urbanisation of physiographic parameters orparameterisations. The episodes examined are two severe winterinversion-induced episodes in Helsinki in December 1995 and Oslo in January 2003,three suspended dust episodes in spring and autumn in Helsinki and Oslo, anda late-summer photochemical episode in the Valencia area. The evaluation wasbasically performed against observations and radiosoundings and focused onthe LM skill at forecasting the key meteorological parameters characteristicfor the specific episodes. These included temperature inversions,atmospheric stability and low wind speeds for the Scandinavian episodes andthe development of mesoscale recirculations in the Valencia area. LMforecasts often improved due to higher model resolution especially inmountainous areas like Oslo and Valencia where features depending ontopography like temperature, wind fields and mesoscale valley circulationswere better described. At coastal stations especially in Helsinki, forecastgains were due to the improved physiographic parameters (land fraction, soiltype, or roughness length). The Helsinki and Oslo winter inversions withextreme nocturnal inversion strengths of 18°C were not sufficientlypredicted with all LM resolutions. In Helsinki, overprediction of surfacetemperatures and low-level wind speeds basically led to underpredictedinversion strength. In the Oslo episode, the situation was more complexinvolving erroneous temperature advection and mountain-induced effects forthe higher resolutions. Possible explanations include the influence of theLM treatment of snow cover, sea ice and stability-dependence of transfer anddiffusion coefficients. The LM simulations distinctly improved for winterdaytime and nocturnal spring and autumn inversions and showed good skill atforecasting further episode-relevant meteorological parameters. Theevaluation of the photochemical Valencia episode concentrated on thedominating mesoscale circulation patterns and showed that the LM succeedswell in describing all the qualitative features observed in the region. LMperformance in forecasting the examined episodes thus depends on the keyepisode characteristics and also the season of the year with a need toimprove model performance in very stable inversion conditions not only forurban simulations.

[学科分类] 大气科学