Quasi-continuous measurements of NOx, CO and C5}C12 hydrocarbons made during the MEDCAPHOT-TRACE experiment in a street canyon with heavy tra$c load were used to estimate the CO/NOx and 36 individual NMHC/NOx tra$c emission ratios in the Athens basin. A tra$c emission inventory has been compiled for Athens and aspects of this inventory were tested against measurements. The results indicate that although the main features of the 9:00 to 15:00 variations of the NMHC/NOx and CO/NOx inventory emission ratios are in agreement with observations, during the rest of the day the "ne structure of the variations of these ratios cannot be accurately predicted by the inventory. Comparison of pollutant emission ratios derived from ambient measurements with emission ratios predicted by existing inventories for Athens reveals serious discrepancies. Further, the experimental results and theoretical onsiderations indicate that the speciation of evaporative emissions changes with increasing ambient temperature in favour of the most volatile HC species, thus changing the speciation of tra$c emissions during the course of the day. This is an aspect that is not taken into account in present urban photochemical modelling inventories.