Few research works tackled the testing of noise annoyance models proposed in the literature by considering a new set of survey data. This lack contributes to limit the effect of endeavors in noise annoyance prediction. This lack is observed for both annoyance models built from field data and laboratory data. Different reasons might explain this. In the case of annoyance models based on psychoacoustic indices accounting for annoying auditory sensations, their relevance was highlighted in laboratory conditions as they might increase the part of explained variance of annoyance. But the use of these indices in field studies leads to issues. Actually, audio recordings are needed for their calculation, making too tedious their use in large-scale studies. This work proposed a methodology, and its testing, to estimate various psychoacoustic and noise indices from the A-weighted equivalent sound pressure level, LAeq, of different transportation noise sources. The transportation noise sources were urban and suburban road traffic noise, urban railway noise, and aircraft noise in cities close to French international airports. A database of index values has been built from various recordings of these transportation noise sources. From the database, relationships between the various indices and LAeq of the transportation noise sources were built, and satisfactorily tested by cross-validation. These relationships might be of interest for field studies dealing with transportation noise assessment. In the current study, the relationships were used to estimate variables of noise annoyance models based on different psychoacoustic and noise indices. The proposed relationships enabled the assessment of the prediction quality of these models also based on individual noise sensitivity. The comparison of their prediction quality with the one of noise annoyance models solely based on the day-evening-night level, Lden, and of a model based on both noise sensitivity and the day-night level, Ldn, highlighted an improvement of the correlation coefficients between the predicted and the field measured annoyance ratings with a ratio of 2 for the best enhancement. Such a ratio was observed as the correlation coefficient obtained for models solely based on Lden were weak. The results stated the interest of the proposed index relationships to be used in the investigation of improving noise annoyance models in future annoyance field studies.