The accuracy of time-composite sampling as a means to calculate phosph
orus loads from agricultural runoff was evaluated analytically and num
erically. It was shown that, when hydrographs and concentration graphs
were positively correlated, computations using a time-composite sampl
ing strategy resulted in load underestimation. Combinations of 11 synt
hetically generated and 12 actual hydrographs and concentration graphs
of agricultural runoff events were used for numerical analysis. The b
ias in load calculation with different sampling frequencies was determ
ined The influence of cross-correlation of hydrographs and concentrati
on graphs on the direction and magnitude of bias was demonstrated. Alt
hough flow-composite sampling is theoretically correct for load comput
ation, it is more expensive and, like other approaches, still involves
uncertainties in both flow and concentration measurements. The depend
ence of the coefficient of variation (CV) of load on CV of flow rate,
CV of concentration, and the cross-correlation between flow rate and c
oncentration is presented analytically and graphically. To achieve loa
d estimates comparable to the accuracy of flow-composite sampling, mul
tiple time-composite sampling methods were investigated This analysis
indicates that with eight concentration values from eight time-composi
te samples of a runoff event, a reliable load estimation can usually b
e made.