Bias in Ponded Infiltration Estimates Due to Sample Volume and Shape

with estimates increasing as much as a factor of 20 when the scale of measurement increased from 0.1 to 2 m (e.g., Estimates of saturated and unsaturated water flow in soil are imporGarbesi et al., 1996; Tidwell and Wilson, 1999). tant for predictions of infiltration, runoff, and solute transport. PreviMeasurement bias related to sample size has someous research indicates that ponded infiltration estimates are influenced by the volume or cross-sectional area of the measurement. Our study times been attributed to smaller samples having less probcompared quasi-steady infiltration measurements made using 20-, 30-, ability of encountering spatially infrequent, high magniand 45-cm-diameter cylinders driven 25 cm deep into 56 field plots tude portions of the soil (Iverson et al., 2001; Starr et al., under diverse agricultural management practices. Mean infiltration 1995). As pointed out by Sisson and Wierenga (1981), rate increased from 50, to 81, to 95 mm h 1 as diameter increased. however, this would be a violation of the central limit Standard deviation and range also increased with diameter. All three theorem, since the mean of randomly selected samples diameters produced lognormal data distributions. These results indishould be centered on the same value regardless of their cate that increasing the sample area is not equivalent to pooling of sample volume. If a sufficient number of small-volume many smaller samples, which would have produced the same mean but samples are used, the mean should be equivalent to that with a lower variance. Follow-up experiments with a double-ring configuration or a divider placed in the center of a 45-cm cylinder demonof larger samples. In terms of the variance, large-volume strated that adding vertical barriers reduced infiltration even when samples should be equivalent to pooling small samples, the total infiltration area was unchanged. A pulse of dye introduced unless the measurement technique creates artifacts de10 min before removing the ponded water showed an extensive netpendent on the volume or dimensions of the sample. To work of dyed flow pathways in all but the slowest infiltration situaaccount for the decrease in infiltration measured with tions. The pathways were not associated with visible macropores. smaller infiltration rings, Shouse et al. (1994) suggested Careful consideration should be given to the dimensions of samples that a stagnation zone (an artifact of the measurement used to estimate saturated and possibly unsaturated flow from infiltechnique) was introduced when partitions were driven tration experiments. into the soil to contain and measure infiltration. Many soil parameters, such as denitrification rate and water infiltration, are lognormally distributed (e.g., ParI rate measurements are key components kin and Robinson, 1992; Grigal et al., 1991). If samples of many soil, hydrogeologic, and environmental inare pooled before measurement, or their individual meavestigations. Infiltration rate is often used to estimate surements are averaged, the distribution of the means saturated hydraulic conductivity (Ks), a critical paramshould be more normal than the distribution of individeter in numerical models for surface hydrology, variably ual samples (Quinn and Keough, 2002). For a soil varisaturated flow, and contaminant transport. able that is randomly distributed at the spatial scale in As a matter of convenience, typical soil water flow question, measuring a larger volume of soil is equivalent measurements are made using small soil samples, often to pooling smaller samples (Parkin and Robinson, 1992). intact soil cylinders of 10-cm diameter. Some investiBoth the large and small volume sample data are then gators have compared measurements made on different sampling the same population and will center on the size samples and found substantial differences in mean same mean, but data from the larger samples will proinfiltration rates. Youngs (1987) found a lower variance, duce a lower variance and be closer to a normal, Gausbut a higher mean infiltration rate when the diameter sian, distribution. Data distributions from experiments was increased from 3.5 to 91 cm for rings driven 2 to by Shouse et al. (1994) and Sisson and Wierenga (1981), 5 mm deep. Sisson and Wierenga (1981) showed that 5-, however, show no tendency toward normality with in25-, and 127-cm-diameter rings gave similar mean infilcreased sample size. tration rates, although the mean increased from 6.25 to Methods for measuring saturated flow are designed to 8.48 and 8.51 cm d 1. Davis et al. (1999) compared conrestrict infiltration to a technically feasible soil volume. stant-head well permeameter, 6by 7-cm cores, and 20by While below-surface flow paths are not explicitly part 30-cm cores, and found that the largest core size gave of an infiltration rate estimate, methods that either reestimates from one to three orders of magnitude greater strict or enhance flow will affect the infiltration estithan the other two methods. Air permeability measuremates. It is known that detached cores can have much ments have been similarly shown to be scale dependent, higher infiltration rates than those measured in situ (Lauren et al., 1988). This is because detached cores may have Soil Scientist, USDA-ARS, Columbia Plateau Conservation Research vertical pores that are open on the bottom, thus allowing Center, P.O. Box 370, Pendleton OR 97801. Received 29 Dec. 2004. water to freely exit the core, whereas flow in the field *Corresponding author ([email protected]). may be restricted below the sampling depth. Methods Published in Vadose Zone Journal 4:1183–1190 (2005). that permit subsurface lateral flow outside the ponded Original Research surface area may overestimate the infiltration capacity doi:10.2136/vzj2004.0184 at the field scale, while methods that restrict lateral flow © Soil Science Society of America 677 S. Segoe Rd., Madison, WI 53711 USA may underestimate the infiltration rate. 1183 Published online November 16, 2005