The land surface plays a central role in shaping weather and climate extremes. Through evapotranspiration, soil moisture, and vegetation dynamics, the land both influences and responds to atmospheric variability. Yet many of these feedbacks remain poorly understood, limiting our ability to predict extreme events. In this talk, I will show how we can move from pixels to processes: using remote sensing, in-situ networks, machine learning and physical model to diagnose land–atmosphere feedbacks across scales. I will highlight advances in understanding droughts and vegetation stress, and discuss how Earth observation can help improve models and resilience in a changing climate.