Hourly predictive artificial neural network and multivariate regression tree models of Alternaria and Cladosporium spore concentrations in Szczecin (Poland)

Int J Biometeorol. 2009 Nov;53(6):555-62. doi: 10.1007/s00484-009-0243-2. Epub 2009 Jun 14.

Abstract

A study was made of the link between time of day, weather variables and the hourly content of certain fungal spores in the atmosphere of the city of Szczecin, Poland, in 2004-2007. Sampling was carried out with a Lanzoni 7-day-recording spore trap. The spores analysed belonged to the taxa Alternaria and Cladosporium. These spores were selected both for their allergenic capacity and for their high level presence in the atmosphere, particularly during summer. Spearman correlation coefficients between spore concentrations, meteorological parameters and time of day showed different indices depending on the taxon being analysed. Relative humidity (RH), air temperature, air pressure and clouds most strongly and significantly influenced the concentration of Alternaria spores. Cladosporium spores correlated less strongly and significantly than Alternaria. Multivariate regression tree analysis revealed that, at air pressures lower than 1,011 hPa the concentration of Alternaria spores was low. Under higher air pressure spore concentrations were higher, particularly when RH was lower than 36.5%. In the case of Cladosporium, under higher air pressure (>1,008 hPa), the spores analysed were more abundant, particularly after 0330 hours. In artificial neural networks, RH, air pressure and air temperature were the most important variables in the model for Alternaria spore concentration. For Cladosporium, clouds, time of day, air pressure, wind speed and dew point temperature were highly significant factors influencing spore concentration. The maximum abundance of Cladosporium spores in air fell between 1200 and 1700 hours.

MeSH terms

  • Air Microbiology*
  • Air Pollutants / analysis*
  • Algorithms*
  • Alternaria / isolation & purification*
  • Cladosporium / isolation & purification*
  • Computer Simulation
  • Models, Statistical*
  • Multivariate Analysis
  • Neural Networks, Computer*
  • Poland
  • Regression Analysis
  • Spores, Fungal / isolation & purification

Substances

  • Air Pollutants