Adverse effects of transport noise on birth outcomes have been reported but limited evidence is available for exposure to aircraft and railway noise. This retrospective population-based cohort study aimed to investigate the association between aircraft and rail noise and birth weight. Exposure misclassification was explored through a noise measurement study, which aimed to develop statistical models to predict indoor noise using measured outdoor noise. Using Office for National Statistics birth records, 637,445 live singleton births were identified within the London Heathrow 50 dB aircraft noise contours area 2006-2015. Associations between aircraft and rail noise with birth weight were analysed using linear and logistic regression models. Confounders included road traffic noise, air pollution, maternal risk factors and socioeconomic status. Simultaneous indoor and outdoor noise measurements were conducted at 49 residential dwellings in London located close to major roads, railways, flight paths or any combination of these for three consecutive days. There were small significant reductions in birth weight associated with highest aircraft noise levels in partially adjusted models. After adjustment for area-level South Asian ethnicity, the associations became null. Sensitivity analyses restricted to moderate and higher noise levels showed significant adverse effects at highest night noise levels (Lnight). Railway noise was associated with a small significant increase in odds of low birth weight at the highest noise categories. Mean indoor-outdoor noise differences were 19.3 dB for LAeq,16h and 21.9 dB for Lnight. Models with outdoor noise explained 29% of the variation in daytime LAeq,1h and 31% of the variation in night-time LAeq,1h. Variance explained was doubled when other predictor variables were added. The findings suggest possible adverse impacts of aircraft and railway noise on birth weight at highest noise categories. Additionally, outdoor noise is a moderate predictor for indoor noise. Further studies using a dataset with individual level ethnicity is needed.