Drought in the western US is linked to climate trends in temperature, precipitation, ecosystem water use and atmospheric water demand, but the mechanisms underlying climate variability and ecosystem response are poorly understood. Interannual and sub-seasonal variability in carbon uptake is strongly correlated with moisture limitation in the western US, and plants in turn can weaken or intensify drought conditions through seasonal compensation effects such as increased spring evapotranspiration resulting in a drier summer, or direct phenological changes such as earlier leaf-out. Plant physiological responses to increasing atmospheric carbon dioxide can also result in changes to stomatal conductance and plant water use that mitigate drought stress. Current drought indices do not take detailed plant-based carbon-water cycle interactions into account. Improvements in modeling drought events requires a better understanding of ecosystem processes related to plant stress responses and the underlying parameters that control carbon and water cycle simulations in land surface models. In this study, we investigate the selection and sensitivity of land model parameters that control land surface-atmosphere exchange of carbon dioxide (CO2) and water within the Simple Biosphere Model v4.2 (SiB4). We focus on parameters related to stomatal conductance and photosynthesis and on plant types most prominent in the western US. We construct ensembles of parameter choices and evaluate these against site-level data (e.g. eddy covariance towers), where available, as well as against atmospheric measurements of carbon cycle trace gases (CO2, carbon-13 isotopes of CO2, and carbonyl sulfide) through forward modeling simulations that take atmospheric transport into account. The use of precise and well-calibrated atmospheric trace gas measurements from NOAA’s Global Greenhouse Gas Reference Network provides a novel way to constrain biosphere parameter selection.