Vegetation productivity metrics such as gross primary production (GPP) at the canopy scale are greatly affected by the efficiency of using absorbed radiation for photosynthesis, or light use efficiency (LUE). Thus, close investigation of the relationships between canopy GPP and photosynthetically active radiation absorbed by vegetation is the basis for quantification of LUE. We used multiyear observations over irrigated and rainfed contrasting C3 (soybean) and C4 (maize) crops having different physiology, leaf structure, and canopy architecture to establish the relationships between canopy GPP and radiation absorbed by vegetation and quantify LUE. Although multiple LUE definitions are reported in the literature, we used a definition of efficiency of light use by photosynthetically active "green" vegetation (LUE(green)) based on radiation absorbed by "green" photosynthetically active vegetation on a daily basis. We quantified, irreversible slowly changing seasonal (constitutive) and rapidly day-to-day changing (facultative) LUE(green), as well as sensitivity of LUE(green) to the magnitude of incident radiation and drought events. Large (2-3-fold) variation of daily LUE(green) over the course of a growing season that is governed by crop physiological and phenological status was observed. The day-to-day variations of LUE(green) oscillated with magnitude 10-15% around the seasonal LUE(green) trend and appeared to be closely related to day-to-day variations of magnitude and composition of incident radiation. Our results show the high variability of LUE(green) between C3 and C4 crop species (1.43 g C/MJ vs. 2.24 g C/MJ, respectively), as well as within single crop species (i.e., maize or soybean). This implies that assuming LUE(green) as a constant value in GPP models is not warranted for the crops studied, and brings unpredictable uncertainties of remote GPP estimation, which should be accounted for in LUE models. The uncertainty of GPP estimation due to facultative and constitutive changes in LUE(green) can be considered as a critical component of the total error budget in the context of remotely sensed based estimations of GPP. The quantitative framework of LUE(green) estimation presented here offers a way of characterizing LUE(green) in plants that can be used to assess their phenological and physiological status and vulnerability to drought under current and future climatic conditions and is essential for calibration and validation of globally applied LUE algorithms.