Land surface states play important roles in the turbulent exchanges between ecosystems and their overlying atmosphere. Field methods to estimate turbulent fluxes have time‐variable source areas, while land surface observations are typically obtained at single plots with a smaller measurement scale. In this study, we characterize land‐atmosphere interactions in two semiarid ecosystems in the southwestern U.S. At each study site, we combine the eddy covariance method with a distributed network of soil moisture and temperature sensors, high‐resolution imagery of the spatial distribution of vegetation and soil patches, and novel spatiotemporal analyses to characterize the turbulent flux footprint analytically and identify the soil moisture, temperature, and vegetation conditions underlying the eddy covariance measurements. Four methods for aggregating the land surface observations to the scale of the daily flux footprint are tested. Our results reveal a large degree of spatial variability in the footprint, with stronger variations in soil moisture than in soil temperature. Single plot measurements are less reliable than the distributed network in capturing footprint conditions, particularly for soil moisture. Furthermore, a marked improvement is observed in the relations between turbulent fluxes and land surface states for methods capturing the footprint variability. We also identify that the composition of vegetation and soil patches in the time‐variable source area affects the relative magnitudes of the turbulent fluxes and the partitioning of evapotranspiration. Our study points to the importance of monitoring the spatial distribution of land surface states (e.g., soil moisture and temperature) and vegetation and soil patches when assessing land‐atmosphere interactions.
