While technological advances allow a larger collection of environmental data, e.g data can be more densely recorded in space and time, basic statistical issues remain in the analysis of extreme values based on these data. Which data are relevant? What models and methods are appropriate, for which tasks? How do we get realistic uncertainty quantification? The analysis of extreme values presents special challenges and these will be discussed generally, and more specifically by reviewing some modelling issues raises by participants of the August 2017 CRM Workshop on Environmental Risk Modelling and Extreme Events.