Producers rely on subjective visual assessments to estimate forage mass in their pastures, which often are inaccurate and lead to poor stocking decisions. The objective of this trial was to compare five nondestructive sampling techniques for predicting forage mass in three alfalfa–tall wheatgrass [Medicago sativa L.; Thinopyrum ponticum (Host) Beauv.] pastures in the southern High Plains. Procedures included canopy height measured with a pasture ruler and rising plate meter (RPM), percentage of green pixels from ImageJ analyses, percentage of green points from photo point count in PowerPoint, and normalized difference vegetation index (NDVI). Height and RPM were linearly regressed on measured forage mass while the remaining were linearly regressed on the natural log of measured forage mass. Considering their limitations, stepwise regression was used to find the best combination of digital and physical procedure (Height + ImageJ) that reduced model error. Calibration models were then applied to external data to determine predictive ability of each procedure. The PowerPoint model was the best, most precise option if restricted to a single sampling procedure, whereas the combined model possessed the superior combination of high R2 and low model error. The combined and PowerPoint models possessed the highest R2pred, but the combined model would be more applicable since it did not saturate when measured forage mass exceeded 1200 kg DM ha–1. The use of ImageJ with canopy height measurements for forage mass prediction requires simple equipment, lends itself to automation, and is adaptable to various forage systems.