Operators in surveillance activities face cognitive overload due to the fragmentation of information on several screens, the dynamic nature of the task and the multiple visual or audible alarms. This paper presents our ongoing efforts to design efficient audio-visual alarms for surveillance activities such as traffic management or air traffic control. We motivate the use of congruent cross-modal animations to design alarms and describe audio-visual mappings based on this paradigm. We conclude with the design of a study to validate our designs as well as future research directions.