High interannual variability of forage production in semi-arid grasslands leads to uncertainties when livestock producers make decisions such as buying additional feed, relocating animals, or using flexible stocking. Within-season predictions of annual forage production (i.e., yearly production) can provide specific boundaries for producers to make these decisions with more information and possibly with higher confidence. We use a recently developed forage production model, ForageAhead, that uses environmental and seasonal climate variables to estimate the annual forage production approximated by remotely sensed vegetation data. The model uses observed seasonal climate data from winter and spring as an input together with summer climate scenarios (e.g., extremely warm and dry, moderately cool and wet) that are based on historic observations. Using these summer climate scenarios we were able to capture the interannual variability in the spatial extent of abnormally low and high biomass production in the western portion of the United States for years 2000-2018. The ForageAhead predictions captured similar spatial patterns of forage anomalies as another similar model (Grass-Cast). This method can be made available in a user-friendly automated system used by livestock producers and rangeland managers to inform within-season management decisions, which can be especially valuable for flexible stocking as it provides a range of possible annual forage production scenarios by the end of May.