The purpose of this paper is to evaluate the errors induced by differe
nt hypotheses used to elaborate reduced kinetic mechanisms of troposph
eric chemistry. To do that, a reference chemical kinetic scheme was de
veloped: it includes a limited number of representative primary specie
s (6 alkanes, 4 alkenes, 2 aromatics) for which up to date kinetic con
stants and mechanisms were used. This chemical scheme constitutes a re
ference against which the various reduction methods were tested. The t
ests were performed for three different scenarios characterised by var
ious VOC/NOx ratios (5,10,20). These scenarios are assumed to correspo
nd to the chemical state of the atmosphere in urban areas and in rural
situations. In a first step, the high NOx approximation (i.e. a chemi
cal scheme without treatment of peroxy + HO2 and peroxy + peroxy react
ions) was tested. The results show that this scheme does not produce s
ignificant error on the simulated concentrations for NO concentrations
above 2 ppb. In a second step, three successive reduction methods wer
e applied to the reference mechanism: (1) use of the chemical operator
concept to treat the organic peroxy chemistry, (2) loss of informatio
n on the organic peroxy class, (3) lumping of secondary organic specie
s into surrogate species. The use of chemical operators provides a sat
isfactory representation of the organic peroxy chemistry for NO concen
tration down to 100 ppt. The scheme obtained after the loss of informa
tion on the organic peroxy class increases only slightly the errors co
mpared to the reference scheme. The lumping of secondary VOCs into sur
rogate species does not generate significant errors on most of the key
species (O-3, NOx, OH, etc.) but induces a significant overestimation
of PAN. At this step, the final reduced scheme involves 71 species an
d 150 reactions. These reduction methods seem to be relevant for most
of the diurnal situations encountered in the lower continental troposp
here. Copyright (C) 1996 Elsevier Science Ltd