Decision-making on animal welfare issues requires a synthesis of information. For the assessment of farm animal welfare based on scientific information collected in a database, a methodology called 'semantic modelling' has been developed. To date, however, this methodology has not been generally applied. Recently, a qualitative Risk Assessment approach has been published by the European Food Safety Authority (EFSA) for the first time, concerning the welfare of intensively reared calves. This paper reports on a critical analysis of this Risk Assessment (RA) approach from a semantic-modelling (SM) perspective, emphasizing the importance of several seemingly self-evident principles, including the definition of concepts, application of explicit methodological procedures and specification of how underlying values and scientific information lead to the RA output. In addition, the need to include positive aspects of welfare and overall welfare assessments are emphasized. The analysis shows that the RA approach for animal welfare could benefit from SM methodology to support transparent and science-based decision-making.