The statistical power of post-fire soil-hydraulic property studies: Are we collecting sufficient infiltration measurements after wildland fires

Infiltration and associated soil-hydraulic properties, such as field-saturated hydraulic conductivity (K-fs), sorptivity (S), or saturated soil-water content (theta(s)) are measured after wildland fires to assess risks of water-related hazards and water supply impairment. Yet the statistical power of these data to distinguish between burned and unburned infiltration is seldom assessed, despite the typically small number of total measurements (N). Here, parametric and nonparametric statistical approaches are used to assess the primary influences on the statistical power and whether the N is sufficient to achieve a representative power of 0.9. About half of the studies examined did not achieve a power >= 0.9 at the as-sampled N, suggesting the need for post-fire infiltration studies to collect more measurements. Effect size was the primary control on power, and therefore the N needed to achieve a given power. Because effect size tends to decrease with time since fire, N generally needs to increase with time since fire in infiltration studies. Power estimates were similar for parametric and nonparametric analyses. Comparisons between infiltration methods suggested that steady state rainfall simulation produced a greater power for a given N, relative to both tension infiltrometers and constant or falling head permeameters. The greater power of the steady state rainfall simulation method was the result of smaller variability, compared to the tension infiltrometers and constant or falling head permeameters. The substantial variability of post-fire tension infiltrometer measurements may complicate using these data in meta-analyses. This work further suggests that synthesis or meta-analysis studies of post-fire soil-hydraulic properties may be more robust if data are weighted by statistical power. The differences in statistical power between infiltration measurement methods require further demonstration for a broader range of datasets before extrapolation beyond the datasets considered here.

Automated summary

More measurements are needed in post-fire infiltration studies; effect size is the primary control on statistical power; steady state rainfall simulation method has greater statistical power compared to other measurement methods.