A Neural Network model for the estimation of bioclimatic indexes

Abstract

Many researches have highlighted the influence of climate on mortality, showing a high increase in mortality in the summer time during "heat waves", periods of very high temperature and humidity levels. The bioclimatic indexes are used in urban climate studies to describe the level of thermal sensation that a person experiences due to the modified climatic conditions of an urban area. The index provides a meaningful and realistic indicator that can not only be used as an information as to how hot it feels, but also as a readily identifiable warning for individuals subject to the physiological dangers of heat exposure. The authors have developed a methodology that, by means of the Neural Network (NN), permits one to predict the values of meteorological data and then the calculation of the bioclimatic indexes. The meteorological data required for the calculation of the bioclimatic index concerning hourly values of air temperature, relative humidity, wind speed have been used according to the records of the meteorological station of the Pergusa Lake (EN), for the year from 2003 to 2006 NN-estimated, the bioclimatic indexes values were compared with coincident bioclimatic indexes values obtained from air temperature and relative humidity observations recorded at standard meteorological stations. Statistical analysis showed a good agreement between the NN-estimated and the station-observed bioclimatic indexes values, with a root mean square error (RMSE) < 1%. The proposed methodology demonstrates the potential of using the NN for defining the bioclimatic indexes and its feasibility as an indicator to pre-alert authority to heat related risks and dangers of heat waves. © 2010 WIT Press.