Mobility data generated by mobile phones offer an opportunity to collect detailed, continuously updated information on human dynamics. This information has the potential to improve scientific understanding of human behaviour, and to inform evidence-based solutions to major societal challenges. Potential uses of mobility data include responding to infectious disease outbreaks and other natural disasters, improving transport infrastructure to address climate change, reducing social inequalities, or increasing economic opportunity. Despite their potential, mobility data have not proved to be a universally reliable source of evidence. This is primarily due to uncertainties in current systems for generating and using mobility data, where concerns about data provenance, technical limitations of mobility data collection, sampling biases, and the effect of privacy-preserving data transformations have limited the accuracy and reliability of conclusions drawn from these data. My PhD research aimed to resolve these sources of uncertainty and improve the use of mobility data by providing a detailed understanding of the sources and effects of different uncertainties, and by proposing effective approaches to address these uncertainties across a range of applications, data sources, and spatio-temporal contexts.