Substantial progress has been made in developing sensor-based proximal phenotyping systems for cotton (Gossypium hirsutum L.), but research is needed to improve in-season prediction of lint yield and to improve accuracy in monitoring crop water stress using such a system. Here, we report on results of a 2-yr field study in which a proximal remote sensing system (measuring canopy height, spectral indices [normalized difference vegetation index, NDVI], and canopy temperature) was deployed every 2 wk over plots of eight cotton varieties at three rates of ET replacement (0, 45, and 90%). As expected, NDVI was an excellent predictor of canopy and biomass traits, including canopy height and leaf area index (LAI). The strength of correlations between in-season sensor measurements (NDVI and the canopy-to-air temperature difference [Tc – Ta]) and lint yield ranged from poor to fair when analyzed by irrigation rate and excellent when analyzed across all rates. Correlations were weaker in the drier of the two years tested and NDVI was a better and more consistent predictor of yield than Tc − Ta, though multiple linear regression integrating both variables improved results by up to 9%. Combining the Tc − Ta data with onsite atmospheric weather station data allowed calculation of empirical crop water stress index (CWSI) values. Using the CWSI metric, relative differences in crop water stress were clear among ET replacement levels once the canopy width reached the threshold for focusing the infrared temperature sensor on the canopy. This indicated that cotton water stress can be successfully monitored using a proximal phenotyping system, which can be quickly and easily deployed across many plots in research or production settings.