The sorption of 2,4-D and glyphosate herbicides in soil was quantified for 287 surface soils (0-15 cm) collected in a 10 x 10 m grid across a heavily eroded, undulating, calcareous prairie landscape. Other variables that were determined included soil carbonate content, soil pH, soil organic carbon content (SOC), soil texture, soil loss or gain by tillage and water erosion, and selected terrain attributes and landform segments. The 2,4-D sorption coefficient (Kd) was significantly associated with soil carbonate content (-0.66; P < 0.001), soil pH (-0.63; P < 0.001), and SOC (0.47; P < 0.001). Upper slopes were strongly eroded and thus had a significantly greater soil carbonate content and less SOC compared with lower slopes that were in soil accumulation zones. The 2,4-D Kd was almost twice as small in upper slopes than in lower slopes. The 2,4-D Kd was also significantly associated with nine terrain attributes, particularly with compounded topographic index (0.59; P < 0.001), gradient (-0.48; P < 0.001), mean curvature (-0.43; P < 0.001), and plan curvature (-0.42 P < 0.001). Regression equations were generated to estimate herbicide sorption in soils. The predicted power of these equations increased for 2,4-D when selected terrain attributes were combined with soil properties. In contrast, the variation of glyphosate sorption across the field was much less dependent on our measured soil properties and calculated terrain attributes. We conclude that the integration of terrain attributes or landform segments in pesticide fate modeling is more advantageous for herbicides such as 2,4-D, whose sorption to soil is weak and influenced by subtle changes in soil properties, than for herbicides such as glyphosate that are strongly bound to soil regardless of soil properties.