The objective of this study was to evaluate the spatial variability of soil attributes and the use of modeling to estimate forage production and carrying capacity (CC) in a semiarid region. Data were analyzed using geostatistical methods, including semivariograms analysis and mapping of each soil chemical attribute. Forage production was simulated at 99%; 95% and 90% guarantee levels, considering areas suitable for mechanized forage production, native pasture areas, irrigable areas and, ephemeral wetland areas. The exponential model best fit the attributes of organic matter, potassium, phosphorus, and pH, while the spherical model was optimal for base saturation, stoniness index, slope index, and general index. The Gaussian model provided the best fit for the cost index. Phosphorus had the lowest range (235 m) and demonstrated a strong spatial dependence (<25%). The highest forage production occurred in irrigable areas, with yields of 112,270.00, 178,661.00, and 215,455.00 kg year-1 at the 99%, 95%, and 90% guarantee levels, respectively. The 90% guarantee level enabled a 31% higher CC than the 99% level, with the highest CC observed in mechanized areas-accounting for about 71.8% of the property's total CC due to greater forage production. Modeling effectively quantified areas capable of producing forage, with lower guarantee levels supporting higher forage production and carrying capacity.