Assessment of high to low frequency variations of isoprene emission rates using a neural network approach

Abstract

Abstract. Using a statistical approach based on artificial neural networks, an emission algorithm (ISO_LF) accounting for high (instantaneous) to low (seasonal) frequency variations was developed for isoprene. ISO_LF was optimised using an isoprene emission data base (ISO-DB) specifically designed for this work. ISO-DB consists of 1321 emission rates collected in the literature, together with 34 environmental variables, measured or assessed using NCDC (National Climatic Data Center) or NCEP (National Centers for Environmental Predictions) meteorological databases. ISO-DB covers a large variety of emitters (25 species) and environmental conditions (10° S to 60° N). When only instantaneous environmental regressors (air temperature and photosynthetic active radiation, PAR) were used, a maximum of 60% of the overall isoprene variability was assessed and the highest emissions were underestimated. Considering a total of 9 high (instantaneous) to low (up to 3 weeks) frequency regressors, ISO_LF accounts for up to 91% of the isoprene emission variability, whatever the emission range, species or climate. Diurnal and seasonal variations are correctly reproduced for \\textit{Ulex europaeus} with a maximum factor of discrepancy of 4. ISO-LF was found to be mainly sensitive to air temperature cumulated over 3 weeks T21 and to instantaneous light L0 and air temperature T0 variations. T21, T0 and L0 only accounts for 76% of the overall variability. The use of ISO-LF for non stored monoterpene emissions was shown to give poor results.