Here on planet Earth, there are unique and complex desertenvironments, like White Sands National Monument in NewMexico, where the use of robot for data collection could help further scientists’ understanding of desertidentification through the provision of new types of datathat are available in real time. We propose to develop a team of cognitively-compatiblerobot assistants that can flexibly support human exploration and mapping of arid desert environments with both low-leveland high-level human feedback, enabling human-robot collaborative reasoning and search strategy adaptation in response to dynamic incoming observations.