Remote sensing missions are crucial for contemporary human activity and our understanding of local and global Earth processes. Robotics-based methods and Unmanned Aerial Vehicles have been extremely successful in this effort. However, they still struggle to simultaneously cover an appropriate range of length and time scales, further requiring secondary manual sampling methods
to fully describe a system. This thesis covers the development and field-use of UAVs with the capability to interact with their environment, for remote acquisition of improved datasets (i.e. no secondary methods are required) in complex environments. In particular, I develop aerial-aquatic locomotion and sensor launching strategies with the objective of groundtruthing remote sensing data and extending UAV-based remote sensing beyond flight time.

Starting with underlying locomotion principles, I employ both numerical and experimental approaches in the design of different robotic solutions capable of interacting with their environment. When possible, these are also demonstrated in field use, targeting relevant sensing applications. In total, 4 different robots are developed and analysed: 1) An aquatic jet thruster is tested in cluttered aquatic environments, and demonstrates stability in water, jetting and gliding. 2) A hybrid flying-sailing vehicle (SailMAV) demonstrates long range autonomous sailing, and its transition to flight is investigated. In the field, I leverage the contact with water surface in sailing to measure the energy flux between lake-water and the atmosphere, and to perform acoustic biodiversity monitoring of underwater-life and bird populations. 3) A dual multicopter-submarine system (MEDUSA) demonstrates robust underwater sample recovery at depth. In the field, the underwater locomotion component enables me to take vertical profiles of Chlorophyll-a in freshwater lakes. 4) A multicoplter with integrated sensor launcher deploys sensors in cluttered, industrial and natural environments. In the field, optimal launch prediction allows me to deploy sensors in the forest high canopy and perform acoustic monitoring of bird populations, showing that current studies are highly biased towards the ground-level.

This thesis demonstrates a number of methodologies that extend the data acquisition capabilities of aerial robots. The presented methodologies can be employed towards other remote sensing applications, and the numerical and experimental results ultimately be employed in the development of similar methods.