The capacity of wheat to store water soluble carbohydrates (WSC) in the stem is regarded as a promising trait to buffer yield in environments with limited water availability. A high throughput, field-applicable, phenotyping technique would not only benefit agronomy/physiology applications but also help its quantification in wheat breeding programmes. The aim of this study was to evaluate if it was possible to estimate the concentration (WSCc, mgg−1) and amount (WSCa, gm−2) of stem WSC non-destructively and in situ using hyperspectral data obtained in wheat canopies, as opposed to currently available labour intensive laboratory methods. Hyperspectral reflectance data were obtained proximally at varying developmental stages from the canopy of wheat trials with a limited number of related genotypes growing under a range of management treatments, in two successive years. Data were calibrated, firstly independently for each year and then jointly, to provide a measure of stem WSC using partial least squares regression on wavelengths in the range of 350–1290nm. Pre-treated spectra (second derivative) enabled calibrations for the combined years with concentration (WSCc, mgg−1) (r2=0.90) and amount (WSCa, gm−2) (r2=0.88) of water soluble carbohydrate in the stems. In addition, from the same measurement, other canopy properties, leaf area index and canopy water content, could be simultaneously predicted. This study has shown that calibration models from canopy level data can robustly predict the dynamics of stem WSC throughout crop stages and treatments, while at the same time including variation in indices diagnostic of crop water and cover status, such as the Water Index and Enhanced Vegetation Index. Promising WSC prediction using spectral data below 1000nm needs to be investigated further, in order to harness the potential for impact using low cost silicon detectors.