Management strategies that minimize P transfer from agricultural land to water bodies are based on relationships between P concentrations in soil and runoff. This study evaluated such relationships for surface runoff generated by simulated sprinkler irrigation onto calcareous arable soils of the semiarid western United States. Irrigation was applied at 70 mm h' to plots on four soils containing a wide range of extractable P concentrations. Two irrigation events were conducted on each plot, first onto dry soil and then after 24 h onto wet soil. Particulate P (>0.45 !Lin) was the dominant fraction in surface runoff from all soils and was strongly correlated with suspended sediment concentration. For individual soil types, filterable reactive P (<0.45 !Lin) concentrations were strongly correlated with all soil-test P methods, including environmental tests involving extraction with water (1:10 and 1:200 soil to solution ratio), 0.01 M CaCl 2, and iron strips. However, only the Olsen-P agronomic soil-test procedure gave models that were not significantly different among soils. Soil chemical differences, including lower CaCO3 and water-extractable Ca, higher water-extractable Fe, and higher pH, appeared to account for differences in filterable reactive P concentrations in runoff from soils with similar extractable P concentrations. It may therefore be possible to use a single agronomic test to predict filterable reactive P concentrations in surface runoff from calcareous soils, but inherent dangers exist in assuming a consistent response, even for one soil within a single field. P HOSPHORUS TRANSFER in runoff from agricultural soils to water bodies can contribute to blooms of toxin-producing cyanobacteria (blue-green algae) and other water quality problems associated with eutrophication (Foy and Withers, 1995;Leinweber et al., 2002). Accumulation of P in soils occurs following long-term application of manure or mineral fertilizer in excess of crop requirements, and a growing number of studies show strong correlations between concentrations of extractable soil P and filterable reactive P in runoff (for recent reviews see Haygarth and Jarvis, 1999;Sims et al., 2000). Linear relationships were observed in surface runoff (Pote et al., 1996(Pote et al., , 1999, while studies of subsurface drainage under natural rainfall can display nonlinear relationships, whereby filterable reactive P concentrations increase markedly after exceeding a threshold or "change point" in extractable soil P (e.g., Hesketh and Brookes, 2000). Understanding these relationships will contribute to the development of models and man-