The response of plants to water stress (dry soil and/or dry air) is a first-order control on the water and carbon cycles, and plays a key role in wildfire, crop productivity, and forest mortality.

Plants respond most immediately to water stress by adjusting their stomatal conductance (the amounts of carbon let in, and water let out, through pores on leaves); the degree of this adjustment under different conditions varies greatly among different plants. While water stress response has been studied extensively in individual plants, ecosystem-scale data and models are largely missing, limiting our ability to understand plant water stress effects at large spatial scales. This dissertation helps address that gap. In Chapter 2, I present the first field experiment directly testing the sensitivity of microwave radiometry (a type of remote sensing) to plant water potential in a forest. I show that vegetation optical depth derived from radiometry mirrors diurnal and seasonal changes in tree leaf water potential.

In Chapter 3, I use a simulation experiment to investigate how plant traits describing water stress response could be estimated by combining microwave radiometry with a land surface model. In this setting, I examine various satellite orbit configurations with different time-of-day sampling patterns. When simulated data from two satellites in different orbits is used to constrain the model, predicted evapotranspiration during drought has 40% less root mean square error than when only one satellite is used. Encouragingly, using two satellites similar to those already in Sun-synchronous orbits yields similar accuracy to using a geostationary satellite observing at all hours of the day, which has been proposed but would be much more expensive in practice.

In Chapter 4, I introduce a new model of stomatal response to water stress in which plants maximize time-integrated carbon gain that is discounted over a given time scale. This model is more biologically realistic than most current models in which only instantaneous carbon gain is maximized. By fitting the new model to eddy covariance data, I show that an ecosystem's stomatal discounting time scale is empirically linked with that location's typical time interval between rainfall events. Overall, the results presented in my dissertation can help see the forest for the trees - that is, help understand how water stress affects entire ecosystems. In the future, this work could be extended to predict vegetation responses to the increased stresses of climate change and identify hot spots of vulnerability or resilience around the world, among other applications.