Wheat yields are stagnating or declining in many regions, requiring efforts to improve photosynthetic (source) traits such as biomass accumulation, light conversion efficiency (i.e. radiation use efficiency, RUE) and photosynthetic capacity. These are key traits that link light capture and primary metabolism with yield, but their study is quite labor intensive in the field. High-throughput phenotyping (HTP) was used in a population of spring wheat grown in three field seasons with contrasting RUE and photosynthetic rates. We built predictive models of RUE, biomass and photosynthetic traits using hyperspectral data collected during the vegetative and grain filling growth phases. Using linear models with vegetation indices, RUE was predicted with 70% accuracy. Water and greenness indices were the best indices to predict RUE during the vegetative stage, while indices related to gas exchange and leaf senescence produced better results for the grain filling stage. For photosynthetic traits partial least squares regression modelling was used to build models for the top, middle and bottom layers of the canopy and a combined layer model approach was also tested. Our findings show that the combined layer approach produced better predictions than models from individual layers for photosynthesis (R2 = 0.11, 0.34, 0.07 for top, middle and bottom layer vs R2 = 0.48 for combined layers). Predictions of photosynthesis (R2 = 0.48, RMSE = 5.24 µmol·m-2·s-1), stomatal conductance (R2 = 0.36, RMSE = 0.14 mol·m-2·s-1) were in the same range as previous studies using hyperspectral data with partial least squares regression. The multilayer modelling approach was used to predict RUE, and statistically significant correlations between photosynthesis predictions and RUE were found (R2 = 0.28, P<0.05 for vegetative period and R2 = 0.3, P<0.01 for grain filling period). Using HTP allowed us to increase the phenotyping capacity 40 times compared to conventional growth analysis and 30 times compared to conventional gas exchange measurements. The approaches presented in this study can be used to screen breeding progeny and genetic resources for source traits and improve our understanding of wheat physiology by incorporating different canopy layers to physiological models.